*Article* **Household Tenure and Its Associations with Multiple Long-Term Conditions amongst Working-Age Adults in East London: A Cross-Sectional Analysis Using Linked Primary Care and Local Government Records**

**Elizabeth Ingram 1,\*, Manuel Gomes 1, Sue Hogarth 2, Helen I. McDonald 3, David Osborn 4,5 and Jessica Sheringham <sup>1</sup>**


**Citation:** Ingram, E.; Gomes, M.; Hogarth, S.; McDonald, H.I.; Osborn, D.; Sheringham, J. Household Tenure and Its Associations with Multiple Long-Term Conditions amongst Working-Age Adults in East London: A Cross-Sectional Analysis Using Linked Primary Care and Local Government Records. *Int. J. Environ. Res. Public Health* **2022**, *19*, 4155. https://doi.org/10.3390/ ijerph19074155

Academic Editor: Paul B. Tchounwou

Received: 8 February 2022 Accepted: 29 March 2022 Published: 31 March 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

**Abstract:** Multiple long-term conditions (MLTCs) are influenced in extent and nature by social determinants of health. Few studies have explored associations between household tenure and different definitions of MLTCs. This study aimed to examine associations between household tenure and MLTCs amongst working-age adults (16 to 64 years old, inclusive). This cross-sectional study used the 2019–2020 wave of an innovative dataset that links administrative data across health and local government for residents of a deprived borough in East London. Three definitions of MLTCs were operationalised based on a list of 38 conditions. Multilevel logistic regression models were built for each outcome and adjusted for a range of health and sociodemographic factors. Compared to working-age owner-occupiers, odds of basic MLTCs were 36% higher for social housing tenants and 19% lower for private renters (OR 1.36; 95% CI 1.30–1.42; *p* < 0.001 and OR 0.81, 95% CI 0.77–0.84, *p* < 0.001, respectively). Results were consistent across different definitions of MLTCs, although associations were stronger for social housing tenants with physical-mental MLTCs. This study finds strong evidence that household tenure is associated with MLTCs, emphasising the importance of understanding household-level determinants of health. Resources to prevent and tackle MLTCs among working-age adults could be differentially targeted by tenure type.

**Keywords:** multimorbidity; multiple long-term conditions; comorbidity; social determinants of health; housing; household tenure; data linkage

#### **1. Introduction**

The co-occurrence of multiple long-term conditions (MLTCs) within a single individual is a major public health challenge both globally and in the UK. The nature and extent of MLTCs is influenced by social determinants of health (SDoH) [1]. The role of individualand area-level social determinants has been widely reported—prevalence and incidence of MLTCs are greater with increasing age, for women, for ethnic minorities, and those living with greater socioeconomic deprivation [1–6]. Yet recent evidence suggests that household-level SDoH (such as household tenure) are often overlooked as determinants of MLTCs despite comparatively large effect sizes for household compared to area-level SDoH [7]. In their landmark report, the Academy of Medical Sciences (AMS) concluded that most evidence focuses on "population or individual-level" determinants and that "it will be valuable to consider whether factors that operate at the household-level can also influence MLTCs" [1]. In addition, exploring these relationships amongst working-age

adults has received little attention [1,7,8]. This is despite recent evidence that suggests the median age of onset of MLTCs decreased from 56 years in 2004 to 46 years in 2019 [9].

Household tenure—whether someone privately rents their home, rents from social housing, or owner-occupies—is widely considered a SDoH [10]. In recent years, homeownership in England has increased amongst older adults and decreased in mid-life, with the private rental market increasingly housing working-age adults [11]. First introduced in 1980, the UK Government's Right to Buy policy and its future iterations enabled some more wealthier social housing tenants to legally buy their properties at a discount, resulting in tenure types more segregated by economic status and social class [12]. Different tenure types are thought to influence health through differences in exposure to various household- and area-level stressors, such as household overcrowding and access to green space [13–15]. However, studies examining associations between household tenure and MLTCs report mixed results, have not explored associations in the English context and have not examined interactions between tenure and other household-level sociodemographic circumstances [7]. This is important as evidence suggests that context-specific factors such as degree of homeownership, and supply and conditions of rented housing may profoundly influence the meaning associated with residing in different tenures across geographies and over time [7,14,16,17].

Using an innovative dataset linking data from local government, health and social care, this study aimed to examine and quantify associations between household tenure and MLTCs amongst working-age adults residing in a deprived borough of East London.

#### **2. Materials and Methods**

#### *2.1. Study Design, Data Source and Participants*

This cross-sectional study uses the Care City Cohort, which links administrative health and social data across local government services, health providers, and health commissioners for residents of Barking and Dagenham (LBBD) [18]. Data are linked at both individual and household levels. LBBD is a deprived, outer borough of East London, with approximately 211,988 residents and a younger and more ethnically diverse population compared to the rest of England [19]. See Appendix A for an overview of the dataset and data linkage steps. This manuscript was prepared following the RECORD checklist [20].

This study used a cross-section of the primary care and local government data taken on 1st April 2019. Individuals were included if they were of working age (between 16 and 64 years old, inclusive) [21], identified as residents of the borough by Mayhew and Harper's Residents' Matrix [22], and were not living in a residential home.

#### *2.2. Outcome Measures: MLTCs*

MLTCs status was determined based on the presence or absence of 38 long-term conditions recorded in a participant's primary care record. Flags of these conditions were derived using publicly available code lists [23].

This study operationalised three definitions of MLTCs in consultation with patients and clinicians:


The third definition was operationalised as conditions originating from different bodily systems are thought to be harder to treat due to different origins and/or treatment plans [1,24]. See Table A2 for the 38 conditions, how these conditions were grouped by bodily system and their distribution across the study cohort. Binary variables were created to indicate the presence or absence of each MLTCs outcome for each participant.

#### *2.3. Main Exposure: Household Tenure*

Individuals were defined as "owner-occupiers" if living in an owner-occupied household (outright or with a mortgage), "private renters" if living in a privately rented property, or "social housing tenants" if living in a socially rented household (from local government or a housing association). A fourth "unknown" category was created to account for missing data. Data on tenure were extracted from the council's housing data systems.

#### *2.4. Covariates*

Data on age and sex were extracted from primary care records. Eight categories were created to code individuals' ages in years (<16, 16–29, 30–44, 45–54, 55–64, 65–74, 75–84, and 85+). Sex was coded as male or female. Data on ethnicity were extracted from council records and coded into five categories: "White", "Black", "Asian", "Other" and "Unknown". Data on BMI and smoking status were extracted from primary care records. BMI was coded into five categories defined by the NHS as follows: underweight (below 18.5), healthy (between 18.5 and 24.9), overweight (between 25 and 29.9), obese (between 30 and 39.9) and morbidly obese (over 40), with a sixth "unknown" category to account for missing data. Smoking status data were coded into four categories: non-smoker, smoker, ex-smoker, or "unknown".

Data on household welfare benefits, occupancy and household type were extracted from council housing records. Households receiving welfare benefits to support rental payments ('housing benefit') were classified by whether eligibility was based on receipt of other welfare benefits and, if so, the type: Employment Support and Allowance (ESA), Pension Credit, Income Support or Job Seeker's Allowance (JSA). Two further categories reflecting households solely in receipt of housing benefit or in receipt of no benefits were created. Occupancy data were recorded into four categories to reflect 1–2, 3–5, 6–10 and 11 or more people within a household. Data on household type captured households as six types: adults with children, adults with no children, single adult with children, single adult, older adults with no children, and three generations.

To provide a marker of overall deprivation in each participants' residential area relative to other areas in the borough, borough-specific Index of Multiple Deprivation (IMD) quintiles were calculated for each small geographical area (Lower Super Output Area; LSOA) using 2019 IMD scores [25]. Each LSOA comprised a maximum of 3000 residents and 1200 households [26].

#### *2.5. Main Data Analysis*

Multilevel logistic regression modelling was used to explore associations between household tenure and MLTCs prevalence amongst working-age residents with complete data (see Table 1 and Figure A1). To assess the relative impact of adjusting for individual compared to household-level covariates on the association between tenure and MLTCs prevalence, we built three distinct models for each outcome. First, an unadjusted model with no covariates included. Second, a model adjusted for individual-level sociodemographic characteristics available in the dataset and found to be associated with both MLTCs prevalence and household tenure in previous literature [17,27–29]. These covariates were age, sex, ethnicity, BMI and smoking. The third and final model for each outcome additionally adjusted for household benefits receipt, occupancy and type to control for potential household-level factors correlated with both household tenure and MLTCs (see covariates above). We chose to adjust for household benefits receipt as it was the best proxy measure available in the dataset for other important covariates such as employment. We chose to adjust for household occupancy and type as a previous systematic review examining household- and area-level social determinants of MLTCs found these factors were associated with MLTCs prevalence in some contexts [7]. Model fit was assessed using Akaike's Information Criteria (AIC). We considered multilevel models to account for the potential clustering of individuals within geographical areas, as individuals are likely to be more similar in terms of individual, household- and area-level factors if residing in the same

areas than if residing in different areas. All models included random effects at the Lower Layer Output Area (LSOA) level to account for clustering within areas. Models were estimated using the lme4 package in R, using restricted maximum likelihood [30]. The 95% confidence intervals were calculated using the Wald test [31].


**Table 1.** Characteristics of study participants (N = 132,296).


**Table 1.** *Cont.*

Note: the denominator for all characteristics (individual and household) is the number of individuals. OR = odds ratio; 95% CI = 95% confidence interval; ESA = Employment Support and Allowance, and JSA = Job Seeker's Allowance. \* Calculated for Barking and Dagenham based on raw Index of Multiple Deprivation scores (2019) [25].

#### *2.6. Subgroup and Sensitivity Analyses*

Three interaction terms were separately added to the final model for each outcome to evaluate potential interactions between household tenure and other household factors. We assessed interactions with receipt of benefits, household occupancy and type (see covariates above) as these are most likely to modify the association between housing tenure and MLTCs, and they also act at the household-level. Any differences in these household-level characteristics by tenure type can be found in Table A3.

#### **3. Results**

#### *3.1. Participant Characteristics*

Of the 232,671 participants whose primary care and local government records were successfully linked, 132,296 participants were eligible for inclusion in this study. A total of 78,379 records (33.7%) were excluded as individuals were not of working age, 21,847 records (9.39%) were excluded due to unconfirmed resident status and 95 were excluded due to living in a residential home (0.04%) (see Figure A1).

The 132,296 study participants resided in 59,535 households and 110 LSOAs. Table 1 gives an overview of the study participants. A total of 86,770 participants (65.6%) were between the ages of 16 and 44 years old and 68,004 (51.4%) were female. A total of 69,611 (52.6%) were of White ethnicity and 68,631 (51.8%) were overweight, obese or morbidly obese. A total of 54,324 participants (41.1%) were owner-occupiers, 39,885 (30.1%) were private renters and 35,776 (27.0%) were social housing tenants. Crude prevalence of basic, physical-mental, and complex MLTCs was 17.9% (23,683/132,296), 4.7% (6269/132,296) and 6.0% (7931/132,296), respectively.

The number of participants with missing data on tenure, ethnicity, BMI, and smoking status were 2311 (1.75%), 514 (0.39%), 25,415 (19.2%) and 13,054 (9.87%), respectively. A total of 102,430 participants had complete data across all variables and were included in analyses (see Figure A1).

#### *3.2. Household Tenure and MLTCs*

After adjusting for individual-level characteristics (age, sex, ethnicity, BMI, and smoking), social housing tenants were more likely to have basic MLTCs (OR 1.90; 95% CI 1.83–1.98), physical-mental MLTCs (OR 2.60, 95% CI 2.43–2.79,) and complex MLTCs (OR

2.23, 95% CI 2.10–2.37) when compared to owner-occupiers (Table 2). For private renters, there was no evidence of a difference in the odds of basic MLTCs compared to owneroccupiers (*p* = 0.630). Conversely, private renters were more likely to have physical-mental and complex MLTCs when compared to owner-occupiers (physical-mental MLTCs: OR 1.29, 95% CI 1.19–1.40; complex MLTCs: OR 1.16, 95% CI 1.08–1.25).

**Table 2.** Estimated odds ratios of multiple long-term conditions (MLTCs) with household tenure for working-age adult residents with complete data (N = 102,430).


Complex MLTCs = the co-occurrence of three or more long-term conditions affecting three or more different body systems within a single individual. \* OOC = owner-occupied. <sup>a</sup> Model 1—an unadjusted model with no covariates. <sup>b</sup> Model 2—model adjusted for individual-level covariates: age, sex, ethnicity, BMI and smoking status. <sup>c</sup> Model 3—model adjusted for model 2 covariates plus household benefits receipt, household occupancy and household type.

After additional adjustment for household-level characteristics (benefits receipt, occupancy, and household type), social housing tenants were still more likely to have MLTCs compared to owner-occupiers, but associations were weaker for all three definitions of MLTCs (basic MLTCs: OR 1.36, 1.30–1.42; physical-mental MLTCs: OR 1.46, 95% CI 1.35– 1.58; complex MLTCs: OR 1.34; 95% CI 1.25–1.44). On the other hand, private renters were less likely to have basic MLTCs (OR 0.81, 95% CI 0.77–0.84), physical-mental MLTCs (OR 0.85, 95% CI 0.78–0.93) and complex MLTCs (OR 0.81, 95% CI 0.74–0.87) (Table 2). IMD quintiles were not included in final models for the three MLTCs outcomes as adding these resulted in poorer model fit.

#### *3.3. Subgroup Analyses*

Our subgroup analyses suggest subgroup effects according to household benefits receipt, occupancy and household type (see Tables A4–A6). The odds of MLTCs for private renters (compared to owner-occupiers) were considerably stronger for households in receipt of benefits compared to those not receiving benefits. For example, odds of basic MLTCs were 76% greater for privately rented households where someone was in receipt of ESA compared to households not receiving ESA (OR 1.76, 95% CI 1.35–2.29). There was no evidence of an interaction between living in social housing and household benefits receipt (see Table A4). The odds of MLTCs for both social housing tenants and private renters (compared to owner-occupiers) were higher for single-adult households compared to households with adults and children. For example, the odds of basic MLTCs for social housing tenants compared to owner-occupiers were 31% greater for single-adult

households (OR 1.31, 95% CI 1.15–1.50). Evidence for subgroup effects for other household types were weaker, with most interactions not statistically significant (see Table A6).

#### **4. Discussion**

#### *4.1. Summary of Study Findings*

Risk of MLTCs amongst working-age residents of a deprived East London borough was greater for social housing tenants and lower for privately renters, when compared to owner-occupiers. These associations remained significant after adjusting for a range of individual- and household-level characteristics and were consistent across different definitions of MLTCs. Other household-level variables—household benefits receipt, occupancy, and type—were important modifying factors, with associations between tenure and MLTCs greater for individuals in single-adult households and households in receipt of certain benefits.

#### *4.2. Comparisons with Existing Literature*

Our prevalence estimates are in keeping with previous estimates for this age group [6,9,32,33]. Prevalence of MLTCs was greater with increasing age and for females, consistent with previous literature [1,6]. However, prevalence was lower for ethnic minority compared with White participants, which contradicts many studies and may be an age-related effect [1,27,32]. In this study, participants lived in a deprived borough in East London where older and younger individuals tend to be White and ethnic minorities, respectively.

We found that social housing tenants exhibited greater risk of MLTCs compared to owner-occupiers, aligning with findings from Northern Ireland yet contradicting those from a Hong Kong-based study [34,35]. This supports the idea that associations between household-level SDoH and MLTCs may be context specific, influenced by housing policy, supply and conditions of social housing, stigma and other household circumstances such as benefits receipt (see Table A3) [7]. In the UK specifically, social housing tenants may be exposed to various "hard" (material) and "soft" (psychological) factors that interact to cause or exacerbate MLTCs [14]. Evidence suggests social housing tenants in the UK have higher levels of C-reactive protein, a biomarker of inflammation associated with various long-term conditions [17,36]. In addition, social housing tenants have less control over the condition of their property and their built environment, and are less able to leave their property, whilst owner-occupying affords ontological security—the sense of security and control afforded when owning your home [37,38]. On top of this, the UK Housing Act (1998) requires social housing to be allocated based on certain criteria, one of which is ill health. As such, MLTCs may be a qualifying characteristic for eligibility for social housing, which may explain our estimated associations.

The lower risk of MLTCs found for private renters compared to owner-occupiers contradicts previous research from the US and Northern Ireland [32,35]. Our analyses adjusted for variables not adjusted for in these studies—household benefits receipt, occupancy, and household type. Our findings suggest these were important explanatory factors for the association between tenure and MLTCs, but they did not explain all of the additional risk experienced by social housing tenants, nor the decreased risk for private renters. In the UK, the private rental market is expanding considerably, and private renters are an increasingly heterogenous group in terms of their demographic, social and economic circumstances [11]. As such, more longitudinal, causal analyses are needed to unpick the complex relationships between different tenure types and MLTCs, taking into account the influence of other household characteristics.

We found that the association between tenure and MLTCs was greater for individuals in single-adult households and households with one or two occupants when compared to higher numbers of occupants. However, previous research examining associations between living alone and MLTC prevalence presents mixed results [7]. In our context, a deprived borough of East London, single-adult households may have less social support and be more financially uncertain than households with multiple occupants, increasing their vulnerability to any adverse effects imposed by their tenure [39]. We also found that the association between tenure and MLTCs was greater for individuals in households where someone was in receipt of certain benefits. Only one previous study has explored subgroup effects in the relationship between tenure and MLTCs and they similarly found that household financial burden mediated this relationship, albeit with a small effect [34]. Our findings support this work, and, again, suggest further research should capture data on, and account for, other household-level characteristics when examining relationships between tenure and MLTCs.

Differences in the risk of MLTCs with tenure type were not explained by commonly used area-level deprivation measures as most areas in our study are amongst the most deprived nationally [7]. These findings further demonstrate the importance of capturing data on, and understanding, household-level SDoH as this information could support service planning when area-level deprivation measures are unable to capture enough variation to model socioeconomic inequalities in MLTCs. In addition, our findings were consistent across different definitions of MLTCs, illustrating the importance of household tenure as a risk factor for MLTCs.

#### *4.3. Strengths and Limitations*

This is the first study to explore associations between household tenure and MLTCs in England. Our findings add to the current literature, and our analyses would not have been possible without the innovative linkage of primary care and local government data. We operationalised three definitions of MLTCs that captured different types of MLTCs with different degrees of complexity. We used publicly available code lists to determine the presence of each condition.

Our study was conducted in one deprived borough in East London and, whilst our findings could be generalisable to other urban areas, they may not hold in contexts that are less deprived, more rural and have different tenure profiles [7,40]. We restricted our analyses to complete cases, which assumes that any differences between individuals with missing and complete data are explained by differences in observed individual and household characteristics included in the regression models. We recognise that there may be other variables associated with the missing data that we have not adjusted for. However, this is unlikely to have significantly changed the results due to the limited role that BMI and smoking status have in the association between tenure and MLTCs prevalence [41]. We did not account for disease severity or symptom burden on the patient, or other dimensions of MLTCs such as frailty. We may have misclassified households where owner-occupiers privately rented rooms, which may have biased estimates towards the null if private renters who co-resided with their owner-occupying landlords differed systematically in their health compared to private renters who did not. In addition, our measure of household benefits receipt did not capture eligibility for benefits, and we could not adjust for other important factors such as education. The cross-sectional study design did not allow us to explore temporal relationships between tenure and MLTCs. We adjusted for household benefits receipt, occupancy, and household type as potential confounders, but also demonstrated important subgroup effects according to some of these characteristics. It is possible these variables may modify the relationship between tenure and MLTCs. More longitudinal analyses are needed to determine how these factors interact over time to impact MLTCs.

#### *4.4. Implications for Practice and Policy*

Most interventions for MLTCs focus on retired, older adults, yet our findings indicate that working-age adults are an important population to consider when aiming to address MLTCs. There is currently a gap in models of care or interventions aimed at working-age adults, for whom there may be greater opportunity for prevention of MLTCs through addressing SDoH than amongst older adults [1]. Initiatives that target preventative resources

at working-age adults with MLTCs who live in social housing could slow the progression of MLTCs and improve health outcomes, ultimately saving future costs [8].

#### **5. Conclusions**

This study finds strong evidence that risk of MLTCs amongst working-age residents of a deprived East London borough was greater for social housing tenants and lower for privately renters when compared to owner-occupiers. Associations were consistent across different definitions of MLTCs, which emphasises the importance of understanding and addressing household-level determinants of health. Our findings suggest that resources to prevent and tackle MLTCs could be differentially targeted by tenure type and that workingage adults are an important population to consider in preventative strategies. Further research should employ longitudinal research methods to assess temporal relationships between household social determinants and MLTCs.

**Author Contributions:** Conceptualisation, E.I., M.G., S.H., H.I.M., D.O. and J.S.; data curation, E.I.; formal analysis, E.I.; funding acquisition, J.S.; investigation, E.I.; methodology, E.I., M.G., H.I.M. and D.O.; project administration, E.I.; software, E.I.; supervision, M.G., S.H., H.I.M., D.O. and J.S.; writing—original draft, E.I.; writing—review and editing, E.I., M.G., S.H., H.I.M., D.O. and J.S. All authors have read and agreed to the published version of the manuscript.

**Funding:** This study is independent research funded by the National Institute for Health Research School for Public Health Research (Grant Reference Number PD-SPH -2015-10025) and the National Institute for Health Research Applied Research Collaboration (ARC) North Thames. The views expressed in this publication are those of the authors and not necessarily those of the National Institute for Health Research or the Department of Health and Social Care. The APC was funded by the National Institute for Health Research School for Public Health Research. DO is also supported by the National Institute for Health Research (NIHR) Biomedical Research Centre (BRC) at University College London Hospitals (UCLH).

**Institutional Review Board Statement:** This study was conducted in accordance with the Declaration of Helsinki. The study protocol was approved on 13th March 2020 by Care City's formal process for data access (no project identification code provided).

**Informed Consent Statement:** Patient consent was waived as this work uses data provided by patients and collected by the NHS as part of their care and support. Only anonymised data were released.

**Data Availability Statement:** Restrictions apply to the availability of these data. Data were obtained from Care City and no applicable data are available without their permission. The study protocol is available on request.

**Acknowledgments:** The authors would like to thank Jenny Shand, Simon Lam and Phil Canham for their support with data access and their help with understanding the origins of the data. We would also like to thank Melvyn Jones, the Care City Community Board and the NIHR ARC North Thames Research Advisory Panel for their advice and expertise when developing our definitions of multiple long-term conditions.

**Conflicts of Interest:** The authors declare conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **Appendix A. Overview of the Care City Cohort and Data Linkage Steps**

In 2017, the leaders of Barking and Dagenham Council, North East London NHS Foundation Trust (NELFT) and Barking and Dagenham, Havering and Redbridge Clinical Commissioning Group (BHR CCG), and their Caldicott guardians (a senior person within each organisation who is responsible for protecting the confidentiality of people's health and care information and making sure it is used properly), signed data sharing agreements to create a dataset that linked administrative data for the population of Barking and Dagenham (B&D) between 1st April 2011 and 31st March 2017. Since its creation, the dataset has been updated on an annual basis. It is hosted in the Barking and Dagenham, Havering, and Redbridge NHS Accredited Data Safe Haven, with governance and oversight provided by the Barking and Dagenham, Havering, and Redbridge Information Governance Steering Committee.

The dataset was created as part a larger research programme of work [18]. It contains routinely collected administrative health and social data across local government services, health providers, and health commissioners. Data are linked at the individual and household levels using linkage keys (replacing NHS numbers and Unique Property Reference Numbers; UPRNs). The data are pseudonymised and include information on sociodemographic characteristics, health variables, household variables and data on health and social care service utilisation. Data on all sociodemographic and health variables for each cross-section are taken as a snapshot on 1st April 2019 to account for in-year changes in variables. The dataset is not currently publicly available but was made available to the wider research community in Autumn 2020.

More information on the dataset can be found here [42] and here [43]. More information on the codes and algorithms used to classify variables as part of the creation of the Care City Cohort can be found at this reference [18].

This study used data from the 2019/20 cross-section of the Care City Cohort. We requested access to pseudonymised sociodemographic and health variables extracted from primary care data, and resident data extracted from local government data. We did not have access to other data available within the Care City Cohort, such as data on health and care service utilisation.

Data were provided unlinked with linkage keys, i.e., with the identification codes generated to replace NHS numbers and UPRNs. We used these to link the data at the individual and household levels. First, we linked the individual- and household-level local government data on Household\_ID (the household-level identification code created by Care City to replace UPRNs). Second, we linked the individual-level primary care data to the linked local government data on Patient\_ID (the individual-level identification code created by Care City to replace NHS numbers). Third, we linked a fourth dataset provided by Care City that detailed care homes in Barking and Dagenham and their Household\_IDs. We linked this to the cohort data on Household\_ID. Finally, we linked a fifth dataset from ONS that contained area-level deprivation data from 2019. We linked this dataset to the data on LSOA code (a unique number identifying each small area/LSOA in England). All linkages were conducted in R software using the merge function from the R base package. Figure A1 illustrates the results of the linkages of the separate primary care and local government datasets. A total of 232,671 individuals were linked across primary care and local government datasets (84.0% of the original primary care records).

To assess whether there were any potential selection biases in the linkage results, we calculated standardised differences in key variables for matched and unmatched primary care records [44]. Standardised differences of 0.2, 0.5, and 0.8 indicate small, medium and large effect sizes, respectively [44]. We were not able to assess potential biases in social variables extracted from local government records (i.e., in the household tenure variable and other household variables) as, by definition, unmatched primary care records did not have corresponding local government data. However, the number of unmatched local government records was considerably low (N = 369). Table A1 presents the results of analyses conducted to assess potential biases in the linkage results for matched and unmatched primary care records. These results indicate that selection biases were not introduced in selected variables originating from primary care records as a result of the success of data linkages, which is in keeping with previous analyses of this data [18].

**Figure A1.** Results of data linkages. \* Number of participants with missing data on each variable sum to greater than 29,866 (132,296 minus 102,430) as some participants had missing data across more than one variable.


**Table A1.** Results of analyses to assess potential biases in the linkage results for matched (N = 232,671) and unmatched (N = 44,269) primary care records.

MLTCs = multiple long-term conditions. \* Variable taken from primary care records, unlike in the study analyses.

**Respiratory N (%)** Asthma (currently treated) 6551 (4.95) Bronchiectasis 143 (0.11) Chronic obstructive pulmonary disorder 1325 (1.00) Sensory Blindness and low vision 905 (0.68) Chronic sinusitis 1617 (1.22) Hearing loss 4726 (3.57) Psoriasis or eczema 812 (0.61) Cardiovascular Atrial fibrillation 451 (0.34) Coronary heart disease 1446 (1.09) Heart failure 302 (0.23) Hypertension 14,518 (11.0) Peripheral vascular disease 169 (0.13) Endocrine Diabetes 8728 (6.60) Thyroid disorders 4403 (3.33) Cancer Cancer (in last 5 years) 1157 (0.87) Musculoskeletal Painful conditions 7417 (5.61) Rheumatoid arthritis (or other inflammatory polyarthropathies and systematic connective tissue disorders) 2871 (2.17) Mental health Alcohol problems 1170 (0.88) Anorexia and bulimia 820 (0.62) Anxiety (and other neurotic, stress-related and somatoform disorders) 3935 (2.97) Depression 9055 (6.84) Dementia 58 (0.04) Psychoactive substance misuse 1451 (1.10) Schizophrenia and bipolar 8624 (6.52) Neurological Epilepsy (currently treated) 750 (0.57) Learning disability 905 (0.68) Migraine 331 (0.25) Stroke and transient ischaemic attack 844 (0.64) Multiple sclerosis 177 (0.13) Parkinson's disease 54 (0.04) Genitourinary Chronic kidney disease 444 (0.34) Prostate disorders 666 (0.50) Gastrointestinal Chronic liver disease and viral hepatitis 1341 (1.01) Constipation (treated) 741 (0.56) Diverticular disease of intestine 893 (0.68) Irritable bowel syndrome 3914 (2.96) Inflammatory bowel disease 718 (0.54)

**Table A2.** The 38 long-term conditions grouped by 10 bodily systems and their distribution across the study cohort (N = 132,296).

Peptic ulcer disease 760 (0.57)


**Table A3.** Household benefits receipt, occupancy, and household type, by tenure for complete cases (N = 102,430).

Note: the denominator for all variables is the number of individuals rather than households. +ESA = Employment Support and Allowance; JSA = Job Seeker's Allowance.

**Table A4.** Estimated odds ratios of basic, physical-mental, and complex MLTCs with household tenure when the final models tested for interactions between tenure and household benefits receipt for working-age adults residing in B&D in 2019/20 (N = 102,430).



**Table A5.** Estimated odds ratios of basic, physical-mental, and complex MLTCs with household tenure when the final models tested for interactions between tenure and household occupancy for working-age adults residing in B&D in 2019/20 (N = 102,430).


**Table A6.** Estimated odds ratios of basic, physical-mental, and complex MLTCs with household tenure when the final models tested for interactions between tenure and household type for workingage adults residing in B&D in 2019/20 (N = 102,430).



#### **References**


### *Protocol* **A Protocol for a Mixed-Methods Process Evaluation of a Local Population Health Management System to Reduce Inequities in COVID-19 Vaccination Uptake**

**Georgia Watson 1, Cassie Moore 1, Fiona Aspinal 2, Claudette Boa 3, Vusi Edeki 3, Andrew Hutchings 4, Rosalind Raine <sup>2</sup> and Jessica Sheringham 2,\***


**Abstract:** Population health management is an emerging technique to link and analyse patient data across several organisations in order to identify population needs and plan care. It is increasingly used in England and has become more important as health policy has sought to drive greater integration across health and care organisations. This protocol describes a mixed-methods process evaluation of an innovative population health management system in North Central London, England, serving a population of 1.5 million. It focuses on how staff have used a specific tool within North Central London's population health management system designed to reduce inequities in COVID-19 vaccination. The COVID-19 vaccination Dashboard was first deployed from December 2020 and enables staff in North London to view variations in the uptake of COVID-19 vaccinations by population characteristics in near real-time. The evaluation will combine interviews with clinical and non-clinical staff with staff usage analytics, including the volume and frequency of staff Dashboard views, to describe the tool's reach and identify possible mechanisms of impact. While seeking to provide timely insights to optimise the design of population health management tools in North Central London, it also seeks to provide longer term transferable learning on methods to evaluate population health management systems.

**Keywords:** population health management; data linkage; population health; inequalities; inequities; process evaluation; protocol

#### **1. Introduction**

In many countries, health policy has moved towards greater integration between different organisations that plan, commission, and deliver health and care [1]. In the latest stage of policy reforms in England, for example, all areas were statutorily required to form integrated care systems (ICSs) by April 2021 that include hospital, mental health, and community trust healthcare providers; primary care providers; clinical commissioning groups; and local authorities, which have a lead role for public health [2].

Sharing patient information across organisations is recognised as a key part of health system integration. An evaluation of four international integrated care systems conducted by the Nuffield Trust describes 'informational integrative processes' as one of the six key factors in the success of, or difficulties in, these programs. The challenges of data sharing across organisations, both in the United Kingdom and internationally, have been well documented [3]. However, a number of data sharing systems are now being developed and deployed. For example, the Whole Systems Integrated Care (WSIC) system in North

**Citation:** Watson, G.; Moore, C.; Aspinal, F.; Boa, C.; Edeki, V.; Hutchings, A.; Raine, R.; Sheringham, J. A Protocol for a Mixed-Methods Process Evaluation of a Local Population Health Management System to Reduce Inequities in COVID-19 Vaccination Uptake. *Int. J. Environ. Res. Public Health* **2022**, *19*, 4588. https://doi.org/10.3390/ ijerph19084588

Academic Editor: Paul B. Tchounwou

Received: 28 February 2022 Accepted: 4 April 2022 Published: 11 April 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

West London, England was set up in part to facilitate the journey towards integrated care systems [4].

It is increasingly accepted that data sharing in itself is not sufficient to drive integration, and in turn improve population health outcomes [5]. As Scott et al. argue, 'data alone does not save lives. It is knowledge derived from data analysis and applied in practice that saves lives' [6]. Population health management (PHM) is an emerging technique used by local health and care partnerships in England. It uses data to help practitioners to understand their population, and then to use this understanding to inform practice [7]. It involves linking and analysing health and care data from different organisations to understand the health of a local population and their current service need, and to predict what local people will need in the future. In a PHM approach, this information is then used to inform decisions on the design and delivery of services in order to improve the health and wellbeing of the population and reduce inequities [8].

There is a lack of evidence about how (or whether) PHM can achieve its aims. There is some evidence that the use of data sharing platforms could influence patient care, but this is mainly from case studies, where evaluation has often been conducted in-house [9,10]. Before considering evaluating the effectiveness of such tools on population health, we need to know more about how PHM data sharing platforms may enable a population health management approach, i.e., to inform decisions to improve the health and wellbeing of their population and reduce inequities. This information is needed to optimise the design of PHM platforms and programs and inform impact evaluations in the future.

Process evaluations—which seek to describe how, and why, a program or intervention works—are often used alongside impact evaluations [11]. For emerging interventions, they can also serve a useful purpose to understand key aspects of delivery of an intervention under development. This protocol describes a process evaluation of a population health management system in London, England using quantitative and qualitative methods. It has dual aims:


#### *1.1. Context: The Population Health Management Innovation*

This evaluation is centred on North Central London's Integrated Care System (North London Partners, NLP), which provides care to 1.5 million people across five boroughs of London. North Central London is comparatively well advanced in its deployment of a near real-time population health management tool that integrates health and care data from across the system.

NLP uses HealtheIntent, a PHM platform developed by the digital provider, Cerner, which combines data across the 28 local authorities and health and care organisations in the integrated care system. It links and standardises data from across the health and social care system (such as general practices and hospitals) and re-presents these data—as 'registries' and 'dashboards'—to staff based in the constituent organisations. An aim of NLP's PHM program is to identify and reduce health inequities and, consequently, elements of HealtheIntent are specifically designed to assist users to identify segments of the population with unequal access to care to inform the development of targeted responses [7]. The design of the tools is underpinned by several core principles, including the relevance of intersectionality (i.e., recognising that social characteristics, such as ethnicity, gender, and socioeconomic circumstances, are interconnected and can create distinct, and sometimes amplified, experiences of disadvantage), the conceptualisation of health inequities as existing on a gradient rather than as a binary (present or absent), and the roots of inequities being in material and psychosocial factors upheld by political and economic structures.

One of the first HealtheIntent tools used in NLP was a COVID-19 vaccination dashboard. Nationally, COVID-19 vaccinations were rolled out in phases. In the first phase, starting from 8 December 2020, the target was for all adults over 65 years of age, those in care homes, NHS and care staff, and clinically vulnerable people to have been offered a first vaccine dose by 15 February 2021 [12]. The second phase of the vaccination rollout, from 13 April 2021, covered the population aged 18–64 years and maintained priority by age and clinical risk [13]. As with many vaccination programs, there was concern that inequalities in uptake would result in inequalities in the risk of COVID-19 infections and serious sequelae. The Scientific Advisory Group for Emergencies (SAGE) reported that, in previous roll outs of national vaccine programs, there was lower uptake in minority ethnic populations [14]. Given the ethnic inequities in COVID-19 death rates, there were specific concerns about ethnic inequities in COVID-19 vaccine uptake [15].

The HealtheIntent COVID-19 vaccination Dashboard (referred to in the rest of the paper as the Dashboard) was developed at the end of 2020 (Figure 1). It sought to enable staff to view variations in COVID-19 vaccination uptake almost in real time. It became available to end users (NLP staff) in December 2020 and continues to be developed, updated, and improved in response to changing requirements. This has led to many iterations of the Dashboard, but, at the time of this evaluation, the Dashboard contained an overview page, a page describing uptake by eligibility cohorts, several demographic and equalities pages, data quality pages, a case-finding tool, and a user guide.

Users have access to different versions of the Dashboard depending on their staff role and type. All users, including non-clinical staff, can see anonymised, aggregated data, but only those with permission, such as primary care staff, can access individual patient data. An Overview page describes overall vaccination uptake. An Equalities and Demographics page segments (i.e., enables users to stratify) the population by gender, ethnicity, IMD quintile, first language spoken, age, and geography. An example of segmentation is by the level of deprivation experienced. The Dashboard stratifies the population into five deprivation quintiles, in line with evidence about health inequalities existing on a gradient [16]. While this design cannot guarantee that users of the tool focus on the middle quintiles as well as the lowest quintile, it does provide users with the capability to do so and respond to findings.

The Dashboard tool also allows users to tailor what they view by providing filters (i.e., restricting the view to specific sub-populations). The Dashboard's many filters include the user view (where users can limit what they see to their own care team type or organisation), COVID-19 information (vaccination eligible cohort, number of doses received, vaccine manufacturer), and health and care information (carer status, known to adult social care, long term conditions, number of long term conditions, mental health conditions, homelessness, bedbound and housebound status). The demographic variables displayed in charts on the Demographics and Equalities pages can also be used to filter data. The system is designed to prevent presenting data in numbers so small that identification might be (theoretically) possible to those without permission to access identifiable data.

#### *1.2. Objectives*

In order to address our first aim of understanding how a population health management system is used, we have proposed two objectives:


To address our second aim of building capacity for wider evaluation of PHM systems, there are two objectives to equip public health practitioners, working as embedded researchers in NLP, with the skills to undertake with supervision both the qualitative and the quantitative arms of the study.

To reflect on the suitability of our methods, in particular we work closely with and train locally embedded researchers to determine the extent to which this model is a workable model for future evaluations of population health management.


**Figure 1.** Illustrative screenshot of the HealtheIntent COVID-19 Vaccination Dashboard. Copyright: North London Partners.

#### **2. Materials and Methods**

This study will combine qualitative methods to identify potential mechanisms of impact of the Dashboard and quantitative analysis of Dashboard usage and reach.

To support part of the study's capacity building aim, university researchers will be working in collaboration with public health practitioners who are seconded part-time to a research role, funded by a grant intended to build capacity for public health research in local authorities. The seconded practitioners' main public health roles are within public health teams in local and regional government. The practitioners will gain transferable skills in research and evaluation through access to specific courses and seminars (e.g., in evaluation methods) and through undertaking all stages of the research process, from submission for ethical review to dissemination of findings, with supervision and guidance from university researchers. If this objective is fulfilled, it will equip the seconded practitioners with the experience and skills to undertake evaluation of NLP's PHM tools in the future. It will also advance our understanding of how such collaborations could be used to conduct evaluations of other PHM systems.

The evaluation is designed to be relatively rapid (i.e., completed within 6–12 months) to ensure the findings are timely enough to influence future local population health management innovations. Therefore, we will incorporate the following elements of rapid evaluation approaches: multiple researchers collecting data concurrently; and sharing interim findings with stakeholders to shape interpretation and analysis, and to sustain their involvement and support [17].

The study was granted approval by UCL Ethics Committee, ref: 2037/005. We started activities for the evaluation in September 2021. We envisage completing most stages of the evaluation by the end of June 2022, though further analysis of the dataset may be undertaken after this date.

#### *2.1. Proposed Qualitative Data Collection and Analysis*

We will undertake semi-structured interviews (*n* ≈ 20) online using MS Teams with a purposive sample of staff who have responsibility for an aspect of COVID-19 vaccination planning or delivery.

Study population: We will interview staff at different levels of seniority, in clinical, strategic, commissioning, and analytical roles across different organisations in NLP (primary care, hospital or mental health providers, social care, public health), and will seek to ensure we capture experiences across all five North Central London boroughs. The sample size is approximate because some individuals will have more than one role, and thus will be able to cover more than one of our desired attributes.

Interviews will explore staff experiences of using the Dashboard and how variations in vaccination uptake shown in the Dashboard informed their actions to address inequities. We have developed a topic guide (an example provided as Supplementary Data) informed by normalisation process theory, which provides guidance for exploring the perceptions of staff and the actions that staff take when a new product or innovation is introduced into an organization [18]. In line with normalisation process theory, the interviews will explore the following:


To develop the guides and to develop consistency between interviewers, an exercise of 'concept mapping' was undertaken by the lead interviewer (G.W.) with supervision from F.A., whereby, for each topic covered by the guide, a short 'concept' description was developed to guide interviewers on what the question was seeking to obtain. This led to revisions of the interview schedule and enabled each interviewer to tailor their own guide to their own language and style. The guide was initially 'soft' pilot tested with a colleague and then, after refinements, was piloted with three staff working in North Central London. No significant changes were made to the guide at this point, so these interviews will be included in the final dataset.

Interviews will be conducted by several individuals (G.W., C.M., V.E., C.B.) working within and external to NLP, to expedite data collection. All interviewees will be asked to sign a consent form before being interviewed. Participants' names and roles will not be disclosed and all data will be anonymised to minimise the risk of identifying participants. All interviews will be recorded and transcribed in full by a transcription service. The

transcribers will remove any identifiers such as names and organizations before securely returning transcripts to the researchers. Researchers will read and further redact transcripts if any potentially identifiable information remains in the text. To expedite analysis, interviewers will note key points from their interviews immediately after conducting them.

Transcripts will be analysed using the Framework Method using Excel by G.W. and C.M. with reading of selected transcripts by J.S. and F.A [19]. A preliminary coding framework drawn from the topic guide was developed by G.W. and C.M. in discussion with J.S. and F.A. to expedite initial descriptive analysis. Further codes and overarching themes and refinements to the analytical strategy will be generated inductively. Discussions with the wider team will take place to discuss emerging findings and resolve discrepancies in coding and interpretation of the data.

Documents and correspondence about the Dashboard, including descriptions of the rollout of vaccination in NLP and iterations of the Dashboard, will be examined to provide contextual evidence for the interviews and to build a timeline of key events in the program to inform both qualitative and quantitative analysis (see Combining Qualitative and Quantitative Data).

#### *2.2. Proposed Quantitative Data Extraction and Analysis*

The proposed quantitative data collation and analysis part of the study will use anonymised staff usage data already stored within HealtheIntent to describe variations in usage of the COVID-19 vaccination Dashboard since its launch in December 2020. It seeks to capitalise on the extensive data automatically generated about usage whenever these population health management tools are used. A request for anonymised data has been submitted to the HealtheIntent service desk. This request includes the numbers of staff by organization and over time that are registered to use the Dashboard. It also includes a request for figures on the actual use of the Dashboard, both in terms of logins and activities while on the Dashboard.

Initial descriptive analysis of usage will be undertaken in Stata and will involve two components [20]. First, the analysis will seek to enumerate the denominator population (i.e., the number of accounts of individuals that were registered to use the COVID-19 Dashboard) and its characteristics (e.g., organisation and geographical area). Second, the proportion of those using the Dashboard among those registered will be generated in key time periods (informed by the timeline constructed, see Proposed Qualitative Data Collection and Analysis section). Usage will be examined for any part of the Dashboard. Where possible, usage will be examined for specific equalities pages of the Dashboard and among specific groups of users, defined by organisation, staff role type, and geographical area.

#### *2.3. Combining Qualitative and Quantitative Data: Proposed Approach*

As described above, we have planned to use the qualitative and documentary data to construct a timeline of key events that will inform the intervals for the quantitative analysis. Interim findings from the qualitative and quantitative data will be shared within the study team at regular intervals, to inform the interpretation of findings from each method, and potentially to prompt further analysis. For example, interview data that reports barriers to, or motivations for, usage may be used to support interpretation of quantitative data showing variations in usage patterns. The extent to which it is possible to combine qualitative and quantitative findings will depend on the data obtained. We will seek to use both qualitative and quantitative findings to support the development of candidate program theories by which population health management could achieve its intended outcomes that could be used in future impact evaluations.

#### **3. Discussion**

This protocol describes a process evaluation of a specific population health management tool within one geographical area of England. It will combine qualitative and quantitative methods to describe staff usage of a specific tool, the COVID-19 vaccination

uptake Dashboard, and how it informs their decisions and ways of working to reduce inequities in vaccine uptake. Working with colleagues based in North Central London means that any learning gained even in the earliest stages of the process evaluation can be rapidly fed back to inform continuing Dashboard development and new population health management tool development and rollouts. The findings from the study also have the potential to have wider significance in advancing methods for evaluating population health management, and thus could build capacity for further evaluations of population health management programs.

#### *Strengths and Limitations*

A key strength of this evaluation is the collaboration between academia, local public health, health care, and regional public health teams. The use of embedded local researchers combined with senior sponsorship promises to ensure the evaluation remains grounded in local service priorities and serves to build local evaluation capacity. The timeliness of the evaluation, and sharing of preliminary findings, aligns with the principle of continuous learning and improvement underlying NLP's PHM program and, more specifically, its use of linked data to support health and care providers addressing inequities. Regional public health input has brought a wider policy perspective and academic input brings independence and objectivity to the evaluation and provides methodological rigour.

The evaluation is subject to some important limitations or challenges. It is taking place in 2021 and 2022, at a time of considerable uncertainty owing to the COVID-19 pandemic. Therefore, it is possible this will affect access to interview participants and access to quantitative data. Strategic input from internal project sponsors will be sought to address barriers and encourage participation to reduce the risk of the project stalling due to other priorities. In the ongoing qualitative data collection and analysis, rapid evaluation approaches were chosen to enable timely findings and feedback to NLP and will also be subject to further in-depth analysis.

We anticipate two major challenges in the quantitative aspect of the evaluation. Access to data held within the population health management system by an external partner, such as a university-employed researcher, would require extensive information governance procedures, reducing the timeliness of the evaluation. However, all organisations within NLP contribute data to the system and are designated data controllers. This designation enables all partners access to non-identifiable data, which makes internal evaluation a possibility. To make use of the opportunity for internal analysis, a local analyst in a funded embedded researcher role will undertake the analysis with the support of external quantitative expertise from ARC North Thames. In addition, the data on staff usage have not previously been subject to evaluation or monitoring. It is thus not well understood what information is feasible to extract from the system, what processes are required to make this information suitable for analysis, or how best to do this. Therefore, the evaluation also seeks to clarify the range of data available, the processes for data extraction, and management before analysis. We will also iterate what data we request and develop a more detailed analysis plan as our understanding of the data evolves.

#### **4. Conclusions**

This protocol describes an evaluation that seeks to understand how staff use a specific population health management system tool to inform decisions and ways of working that reduce inequities in vaccine uptake. In the short term, achieving this aim should serve the local health and care system by providing useful insights to inform future population health management activities. The evaluation also aims to develop the capacity for wider evaluation of PHM systems. We will have met this aim if our evaluation equips local practitioners with the skills to conduct further evaluation and if it generates transferable learning about the methods for evaluating such programs in collaboration with local health and care professionals in the future.

**Supplementary Materials:** The following supporting information can be downloaded at https: //www.mdpi.com/article/10.3390/ijerph19084588/s1, S1: topic guide for interviews.

**Author Contributions:** Conceptualization, G.W., C.M., F.A., A.H. and J.S.; methodology, G.W., C.M., F.A., A.H. and J.S.; investigation, G.W., C.M., V.E. and C.B.; writing—original draft preparation, G.W., C.M. and J.S.; writing—review and editing, A.H., F.A. and R.R.; funding acquisition, J.S. and R.R. All authors have read and agreed to the published version of the manuscript.

**Funding:** This report is independent research funded by the National Institute for Health Research ARC North Thames (NIHR200163), Clinical Research Network North Thames capacity building for public health, and Agile Workforce Funding (UCL004). The views expressed in this publication are those of the author(s) and not necessarily those of the National Institute for Health Research or the Department of Health and Social Care.

**Institutional Review Board Statement:** The study was granted approval by UCL Ethics Committee, ref: 2037/005.

**Informed Consent Statement:** Informed consent was obtained from all subjects involved in the study.

**Data Availability Statement:** Not applicable.

**Acknowledgments:** The authors would like to Amy Bowen, Director of System Improvement, North London Partners in Health and Care, and Sarah Dougan, Director of Population Health Management, North London Partners in Health and Care & Consultant in Public Health, London Boroughs of Camden & Islington, for their vital shaping of the study design and ongoing sponsorship of the evaluation. The authors would also like to thank Fola Tayo and Nira Shah, and other members of the ARC North Thames Research Advisory Panel for their advice on the study design and on communicating population health management to a lay audience.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results.

#### **References**


### *Article* **Development of Public Health Core Outcome Sets for Systems-Wide Promotion of Early Life Health and Wellbeing**

**Liina Mansukoski 1,\*, Alexandra Albert 2, Yassaman Vafai 1,3, Chris Cartwright 3, Aamnah Rahman 3, Jessica Sheringham 4, Bridget Lockyer 3, Tiffany C. Yang 3, Philip Garnett <sup>5</sup> and Maria Bryant 1,6,\***


**Abstract:** We aimed to develop a core outcome set (COS) for systems-wide public health interventions seeking to promote early life health and wellbeing. Research was embedded within the existing systems-based intervention research programme 'ActEarly', located in two different areas with high rates of child poverty, Bradford (West Yorkshire) and the Borough of Tower Hamlets (London). 168 potential outcomes were derived from five local government outcome frameworks, a community-led survey and an ActEarly consortium workshop. Two rounds of a Delphi study (Round 1: 37 participants; Round 2: 56 participants) reduced the number of outcomes to 64. 199 members of the community then took part in consultations across ActEarly sites, resulting in a final COS for systems-based public health interventions of 40 outcomes. These were grouped into the domains of: Development & education (N = 6); Physical health & health behaviors (N = 6); Mental health (N=5); Social environment (N = 4); Physical environment (N = 7); and Poverty & inequality (N = 7). This process has led to a COS with outcomes prioritized from the perspectives of local communities. It provides the means to increase standardization and guide the selection of outcome measures for systems-based evaluation of public health programmes and supports evaluation of individual interventions within system change approaches.

**Keywords:** early life health; core outcome set; public health interventions; systems approach

### Accepted: 24 June 2022 Published: 28 June 2022

Received: 23 May 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Citation:** Mansukoski, L.; Albert, A.; Vafai, Y.; Cartwright, C.; Rahman, A.; Sheringham, J.; Lockyer, B.; Yang, T.C.; Garnett, P.; Bryant, M. Development of Public Health Core Outcome Sets for Systems-Wide Promotion of Early Life Health and Wellbeing. *Int. J. Environ. Res. Public Health* **2022**, *19*, 7947. https:// doi.org/10.3390/ijerph19137947 Academic Editor: Paul B. Tchounwou

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

#### **1. Introduction**

#### *1.1. Background and Objectives*

Core outcome sets (COS) are "an agreed standardized collection of outcomes" used in evaluations of intervention research [1]. The use of COS has been promoted to harmonize the outcomes used and to ensure that key stakeholders are consulted on the relevance of what is being measured in evaluations [2]. No existing core outcome set has been adapted specifically for the systems-wide promotion of early life health and wellbeing in public health research in the UK, two widely used outcomes frameworks are the Public Health Outcomes Framework (PHOF) and the NHS Outcomes Framework [3,4]. Though an important resource to highlight key indicators to measure the success of some early life interventions, the most widely used existing framework for public health, the PHOF, was not developed to ensure the use of a minimum set of outcomes to be used across studies to facilitate comparisons. Most COS in the pediatric literature, on the other hand, focus on a specific illness or disease, not on public health outcomes [5].

We sought a COS to support the evaluation of a UKPRP-funded programme of research called ActEarly. ActEarly is a large research consortium aimed at promoting health and wellbeing in early life in two different areas with high rates of child poverty: Bradford in West Yorkshire and the Borough of Tower Hamlets in London [6]. Living in an area with high levels of child poverty often coincides with exposure to other economic, physical, cultural, learning, social and service environmental risk factors, which can predispose children and their families to poorer mental and physical health outcomes. In 2019, ActEarly was launched to address these issues with the aim of creating testbeds of upstream interventions within 'whole system city settings' (i.e., understanding and addressing the interconnectedness of distal and proximal determinants) [6,7]. The programme is a partnership between academics, local governments, the NHS, Bradford Institute for Health Research, community and third sector organizations and staff/students at affiliated universities (University of York, Leeds, Bradford, Queen Mary University London, University College London, London School of Hygiene and Tropical Medicine). The ActEarly programme combines interventions with citizen science and the co-production of research with local communities across the two study sites [8,9].

As ActEarly is a system-wide intervention, it necessitates system-wide outcome sets that incorporate multiple aspects of health, well-being and the physical and social environment in which the families and children of Bradford and Tower Hamlets live. The COS was deemed essential, not only to ensure consistency and comparability in what is being measured by planned project evaluations within ActEarly, but to facilitate a system-wide meta-evaluation of the whole ActEarly programme, including planned long-term economic modelling [10]. The lack of an agreed set of core public health outcomes specific to early years and childhood health and well-being that takes a whole-systems perspective was identified as a key gap in our evaluation work in this area. Rather than providing a wider selection of outcomes (i.e., similar to the PHOF framework), the COS presented here was intended to represent the 'minimum' required set of outcomes (though not necessarily excluding the inclusion of other outcomes). Thus, we aimed to develop the public health 'Core Outcome Set for Early Years (COS-EY)'. The specific objectives of this COS development were to:


#### *1.2. Scope*

To define the scope of the COS development, we followed the Core Outcome Measures in Effectiveness Trials (COMET) guidance [2]. However, rather than targeting a specific health condition, we extended our scope to include outcomes that would be deemed important across the whole system. Given the intended breadth of this work, we therefore anticipated that we would develop a series of combined COS within domains such as: Social environment, Physical health, Poverty, etc. Thus, although our goal was to develop an overarching systems-based COS, we also anticipated developing domains, and that each of the domains would generate a separate sub-COS consisting of a smaller set of outcomes (~three to seven).

#### *1.3. Interventions*

The development of the COS-EY was guided by ActEarly's three themes (Healthy places; Healthy livelihoods; and Healthy learning) and four cross-cutting themes (Food & healthy weight; Play and physical activity; Co-production and Citizen science; and Evaluation). Each theme consists of multiple projects located across the two study sites. Examples of ActEarly projects include an evaluation of the Healthy School Streets programme in both

Bradford and Tower Hamlets; the Join Us: Move. Play (JU:MP) local delivery pilot which aims to test and learn more about what helps children aged 5–14 years to be active; and co-production of the Horton Park regeneration project in Bradford (for further details of these and other ActEarly projects, see [11]). There are no constraints placed on potential study designs and there is a great variety of approaches taken within ActEarly to achieve the overall goal of early promotion of good health and wellbeing. This means the process to develop the COS needed to be flexible and fit for purpose to accommodate different study designs, populations and evaluations.

#### **2. Materials and Methods**

Guided by the principles set out in the COMET (Core Outcome Measures in Effectiveness Trials) Handbook [2], we designed a modified Delphi study consisting of two rounds of a consensus survey administered to our panel of experts and stakeholders, followed by a face-to-face public consultation with community members using 'dot voting' (details below). The Delphi method was first developed by the RAND corporation and is commonly used to create consensus by asking participants to answer questions across multiple rounds. After each round, responses are fed back to the participants [2,12]. The decision to start the process with the expert and stakeholder consultation, followed by the community consultation, was taken because of their knowledge of interventions and the whole system changes needed to be seen.

#### *2.1. Registration*

The COS development was registered on the COMET website (#1910) and the reporting of the study is in line with the COS-STAR Statement [13,14].

#### *2.2. Participants*

The populations that are the targets for the application of the COS-EY in the first instance were children and families living within the ActEarly study areas: Bradford Metropolitan Area in West Yorkshire and the Borough of Tower Hamlets in London (Figure 1).

**Figure 1.** Maps of ActEarly study areas: Bradford Metropolitan District (**left**) and the London Borough of Tower Hamlets (**right**).

Stakeholder groups who were involved in the COS development included: ActEarly researchers, community and council partners and community members in Bradford and Tower Hamlets. This wide consultation allowed us to consider the viewpoints and expertise of academics, as well as affiliated local government and public health professionals. In addition, it was considered vital that the communities in which ActEarly operates were consulted to prioritize the evaluation of changes in factors that were important and meaningful to the families and children living in each local area.

For the first round of the survey, anyone within the immediate or wider ActEarly team, including academics, practitioners, local government, voluntary sector organizations and community representation, was eligible to take part (due to the snowball sampling, it is not possible to provide a precise sample size of how many people were invited to take part in the Delphi surveys but we estimate that the link to the survey may have reached anywhere between 70 to 100 people).

For round two, the eligibility criteria stayed the same, but we extended our promotion and reach in an attempt to get wider participation. At this point, the project had grown in size and reach and we felt it was important to ensure individuals who had newly joined, or newly become collaborators, had the opportunity to contribute to the COS development. Potential participants were identified from the activity logs of the ActEarly projects and by asking ActEarly theme-leads to signpost key collaborators and partners, local government links and members of the communities associated with ActEarly and other related projects.

The eligibility criteria for participation in the community consultations were purposefully left open and included any adult attending any of the events at which the consultations took place. To widen the reach of the consultation, we conducted all three consultations in open, public areas. In Bradford, this included Horton Park and Peel Park. Both parks held free entry events that were visited by local children and families over the summer of 2021. In Horton Park, the event was an Eid celebration aimed at local families. In Peel Park, the event was a council-funded Play Bradford event. We estimate that each event was attended by 100+ local families but do not have exact figures. We did not collect demographic, social or health information from the families but most participants arrived at the events on foot from the surrounding neighborhoods. In Tower Hamlets, the consultation was conducted in collaboration with the Bromley by Bow Centre who identified the Old Ford Road Summer Fun Day event at Butley Court as suitable for the consultation.

#### *2.3. Information Sources (Development of the Minimal Dataset)*

The initial list of potential outcomes was derived from existing local sources including: the Bradford Key Indicators set; Tower Hamlets key indicators; Tower Hamlets 'I' statements (publicly derived framework); the Tower Hamlet common outcomes framework; ActEarly community survey codes; and individual suggestions from stakeholders at previous ActEarly workshops (Figure 2). This process involved collating all outcomes from each of these local sources, in which the words and presentation of text were retained. Outcomes which were repeated by more than one source (e.g., childhood obesity) were only included once in the minimal dataset. However, those deemed to be 'similar', but not identical, were retained as separate outcomes (e.g., 'mental health' and 'mental well-being'). The listed outcome sources were developed locally and are regularly updated (thus links cannot be provided).

**Figure 2.** Process to reduce the number of outcomes.

#### *2.4. Consensus Process*

#### 2.4.1. Surveys

The outcomes in the surveys were based on a collation of everything gathered from the activities in the 'information sources (development of the minimal dataset)' paragraph. Potential participants in the consensus surveys learnt about the study via email or word of mouth and snowballing of these (e.g., via existing groups/teams). The purpose of the study was summarized to participants on the first page of the survey to give context. The survey was completed using the online survey platform Qualtrics [15]. Invited participants received reminder emails. This survey asked participants to rate the importance of each outcome on a scale of 1–9 (from 1 "Not important at all" to 9 "Very important"). After all outcomes were rated, participants were asked to suggest any new outcomes not yet included. Our Delphi process did not include the collection of identifying information, but survey respondents were asked to state their stakeholder role (i.e., Academic, Clinical academic, Local government, Voluntary sector, Community representative, National/regional government, Commercial sector, Other).

The shortened Round 2 survey was also sent using Qualtrics. As in Round 1, invited participants received email reminders about the survey. In addition to asking participants to rate the importance of each survey, the Round 2 survey presented the group-average results of the first survey and encouraged participants to review these results before re-rating the outcomes. At the end of the Round 2 survey, there was an option to request outcomes that had been excluded after Round 1 to be re-introduced, as well as space to leave any other comments or suggestions.

#### 2.4.2. Community Consultation

The final part of the consensus process was undertaken after the second survey had been analyzed (and the number of outcomes was hence reduced) with community members, that is, local families with children (Figure 3). In consensus methods, consultation with patients, or community members, is recommended when there is no clear consensus among the experts and it can ensure that outcomes are included that are important to community members [2]. The community member consultation was conducted using 'dot voting' and by utilizing principles of the nominal group technique, which facilitates quick, structured decision making [16–18]. In dot voting, participants are given colored dot stickers that they can use to indicate their votes in priority setting and consensus exercises. In addition to the 'dot voting,' we facilitated a play activity that children could engage in, whilst adults were asked to contribute to the core outcome consultation.

**Figure 3.** Community consultation in Bradford.

To make the process of voting as easy as possible, participants were asked to select and rank three outcomes they considered to be most important by placing their colored stickers on posters that included all the outcome names (green sticker for most important, yellow for second most important and orange for third most important outcome). The consultation facilitators (researchers) were present to answer any questions that arose and help explain the project and the outcomes that were voted on.

#### *2.5. Analysis*

#### 2.5.1. Outcome Scoring/Feedback

Survey items were scored on a 9-point Likert scale (where 1 was "Not important at all" and 9 was "Very important"). Although no definitive recommendation exists on the optimal number of points for a Likert scale in COS development, a 9-point Likert scale has been proposed for use in consensus processes to reduce the number of outcomes, before face-to-face consultations are taken to reach a final consensus [19]. The scores generated from Round 1 and Round 2 of the consensus surveys were analyzed using descriptive statistics (mean and median score, standard deviation, range) and by calculating expert agreement to identify which outcomes participants agreed were less important, outcomes for which there was good agreement for prioritizing and outcomes about which participants were uncertain.

The proportion of experts/stakeholders (details of participants in Table 1) agreeing was calculated as:

*Proportion in agreement* <sup>=</sup> *N o f experts scoring an item within a speci fied range Total N o f experts*

**Table 1.** Participants who took part in the Delphi surveys.


<sup>1</sup> This category includes people who identified their participant group as being 'Other' and defined it as: regional sport's charity, clinical commissioning group, think tank, research manager and community researcher.

The 'proportion in agreement' is sometimes referred to as the agreement index and multiplying the index by 100 results in the % of experts who agree with a given outcome based on our criteria set above.

All statistical analyses were performed using Stata 16 [20].

#### 2.5.2. Consensus Definition

To define consensus, we used 'proportion within a range'. This definition of agreement is widely used in Delphi studies [21]. Agreement was defined as more than 80% of the panel scoring an item within a specified range on the 9-point Likert scale. Commonly, items scored as 1–3 are considered to indicate the outcome is of limited importance, items scored 4–6 are considered to be important but not critical and items scored 7–9 are deemed to be critical [2].

As recommended by the literature, we selected our agreement threshold of 80% in advance [22,23]. 80% is above the median threshold reported in the literature for the determination of consensus, which is 75% [21]. This slightly stricter threshold was selected due to the relatively large initial number of items in the Round 1 survey (N = 168), which needed to be reduced considerably to arrive at a feasible number of core outcomes. Disagreement was defined as <80% of the panel scoring an item within the specified range.

Thus, our process for keeping or removing outcomes applied the following rules:


This procedure was repeated with the Round 2 data following Round 2 survey implementation; however, we applied a less stringent inclusion cut-off (>70% of experts scoring 7 or higher) at this stage to provide members of the public in both communities with a large range of potential outcomes to consider. Missing observations (where an expert did not score a given outcome) were excluded from analyses.

#### 2.5.3. Community Consultation—Analysis

Following the dot voting process, outcomes were ranked by the number of votes by each study site with the aim of creating a 'top 10 ranking for each site. Each dot was given a score of 1 (dot color was not considered), and these were summed for each outcome. Outcomes ranked in the top 10 for each site were included in the final COS, even if the expert consensus on the given outcome was below the 80% cut-off (>80% of experts scoring the item 7 or higher) to signify the importance of public opinion.

#### *2.6. Ethics*

The University of York Department of Health Sciences Research Governance Board approved the study (reference: HSRGC/2021/458/E). Survey participants were asked to consent to take part. Community consultation did not collect any personal or identifiable information about the participants beyond the dot votes, and no informed consent was obtained.

#### **3. Results**

#### *3.1. Participants*

37 participants completed the Delphi questionnaire in Round 1 and 56 in Round 2. Due to us using snowball sampling when sending out the survey, we could not estimate how many of the people receiving the survey chose to participate in it. Participant stakeholder representation for the Delphi surveys is provided in Table 1, indicating that most respondents were academics or representatives from local government. A total of 199 members of the community took part in consultations (135 in total for the two events held in Bradford and 64 in total for the one event held in Tower Hamlets, London).

#### *3.2. Outcomes Considered at the Start of the Process (Minimal Dataset)*

The lists of outcomes from existing sources from both localities were reviewed and presented in our surveys using the same text/format as the original source. Unless they were described using identical terms (e.g., more than one source including 'childhood obesity'), all outcomes were included even if they appeared to be measuring similar constructs (e.g., 'Speech/language/communication' and 'vocabulary'). This resulted in a minimal dataset of N = 168 outcomes (Figure 2; Supplementary Table S1). The outcomes were subsequently grouped into eight draft COS domains by the immediate study team (Connectedness; Crime and safety; Development and education; Health behaviors; Mental health; Physical environment; Physical health; and Poverty, Social mobility and inequalities (Supplementary Table S1).

#### *3.3. Delphi Studies*

Following Round 1, 28 out of the 168 outcomes met the 80% threshold for automatic inclusion and were automatically included in Round 2 of the survey. According to our prespecified criteria, no outcomes could be automatically excluded following Round 1 as none had more than 80% of participants who scored 3 points or lower (=considered to be of limited importance). Overall, we noted that all outcomes received relatively high scores and were considered important by our experts (range in mean scores 5.4–8.2). This meant that to reduce the number of outcomes, while also ensuring that there were enough outcomes left across the different domains, we had to adapt our approach to include outcomes that did not meet the automatic inclusion threshold. To achieve this, we decided to include any outcome that achieved higher than 70% agreement (=% of experts

giving a score of 7 or higher), rather than 80% agreement, in the second round following discussions within the research team. Additionally, we refined our list, including removing three outcomes representing the same construct as other outcomes provided responses were not dissimilar (e.g., self-confidence, removed due to presence of self-efficacy). One outcome was moved from the Physical environment domain to Development and education (language acquisition), and one outcome label was changed (from maternal physical activity to parental physical activity). These changes were made based on the expert feedback received in Round 1. Finally, one outcome domain name ('Connectedness') was changed to 'Social environment' and included outcomes from the Connectedness category, as well as four outcomes previously included under Physical environment (Figure 2).

Round 2 of the survey included 74 outcomes across 8 outcome domains. 36 outcomes were scored 7 or higher by >80% of the Delphi survey respondents and were automatically included in the community consultation. As in Round 1, no outcomes achieved the threshold for automatic exclusion. There was a discussion within the research team to decide which of the remaining outcomes should be taken forward to the next stage of the consensus process. As in Round 1, it was agreed that outcomes for which there was some consensus, but which did not reach the automatic inclusion threshold, would be included (=agreement > 70%). In addition, we chose to add back in any outcome where three or more stakeholders had suggested re-introducing an outcome that had been deleted following Round 1.

In total, 64 outcomes were taken forward for review within the community consultation. After summing up the community votes for each outcome, we found that several outcomes that ranked highly had the same number of votes. Thus, rather than having our intended 'top 10 community-ranked outcomes', we had 11 in Bradford and 14 in Tower Hamlets. Despite the overall similarity between the sites, some highly ranked outcomes in Tower Hamlets were considered of less importance in Bradford, and vice versa. For instance, participants in Tower Hamlets saw housing, traffic and air quality as key issues, whereas in Bradford, mental health outcomes and access to high-quality health services were brought up by many.

A comparison between the outcomes rated highly by the community and the expert agreement scores revealed that four of the most highly rated outcomes from the community consultations had not achieved 80% agreement from the experts. As planned, these outcomes were included in the final COS-EY (educational attainment, traffic, traffic levels outside schools and child weight). The remaining outcomes that were included in the community top rankings were consistent with those ranked by the experts (all achieved over 80% expert agreement) and therefore met the criteria for automatic inclusion. A total of 24 remaining outcomes that were ranked less frequently by members of the public, and where expert agreement was <80%, were removed.

#### *3.4. Final COS-EY*

To formulate the final COS, we once again reviewed the outcome labels and domains for clarity, including considerations of outcome hierarchy, as recommended by some of our stakeholders. An example of this is the outcome called 'traffic', which until this point was separate from another outcome called "traffic levels outside schools". In the final COS-EY, these two are captured by the higher-level outcome label 'traffic'. Overall, this process resulted in five outcomes being combined with an existing outcome, and one outcome being split into two outcomes. We reduced the number of domains from eight to six, to ensure each domain had a balanced number of outcomes (Table 2). The final COS-EY consisted of 40 outcomes, divided into six domains: Development & education; Physical health & health behaviors; Mental health; Social environment; Physical environment; Poverty & inequality (Table 2).


**Table 2.** Final COS-EY.

#### **4. Discussion**

This study has resulted in the development of a public health COS with six domains which can be used collectively or individually to support the evaluation of system-wide programmes designed to promote health and well-being at a population level. The COS-EY provides a set of outcomes that we recommend other evaluators adapt to align with their stakeholder priorities. We developed the COS using the ActEarly consortium as an exemplar and to support the ActEarly evaluation. There were no published COS available that were suited to our purpose, and overall, there are relatively few COS specifically designed to be used in public health interventions, particularly those delivered across a whole city [5]. There was high stakeholder agreement on the final 40 ActEarly core outcomes and the final decision on which outcomes to include was based on a large community consultation. We recommend that going forward, the COS-EY is considered for adaptation for evaluation research in this area. For ActEarly, the next step is to identify existing data sources and to decide on precise measures to assess each outcome. This work will utilize routine data collected across both study sites and aligns with the ongoing efforts to link different routine data sources [24].

#### *4.1. Comparisons with Existing Outcomes Frameworks and Literature*

There is a significant overlap between the COS-EY and the PHOF, which may relate to at least some of the stakeholders being aware of the existing framework; therefore, they may have used it as a point of reference when thinking about core outcomes for public health. It is important to note that, whereas the PHOF is a tool to highlight key indicators to consider, the COS-EY is a minimum set of outcomes to include in the evaluation of system-level interventions in early years and childhood settings. The overlap between the COS-EY and the PHOF means that there are publicly available data for many outcomes, including, for example, parental and child obesity, physical activity, child development, air pollution, (self-reported), well-being and homelessness. Similarly, outcomes that are included in the key indicator frameworks used by the two ActEarly local governments, (the Bradford Metropolitan District Council and the Borough of Tower Hamlets), achieved high expert consensus and are included in the COS-EY. Examples include housing, poverty and employment. Taken together, the six domains that our outcomes are categorized under (Development & education; Physical health & health behaviors; Mental health; Social environment; Physical environment; Poverty & inequality) highlight the system-wide factors that underpin early years health and well-being. The inclusion of outcomes such as 'access to opportunity' and 'children get best start in life' can be considered unique in that as far as we are aware, the existing frameworks do not include them, but both were considered highly important in our consensus work. One of the partners of ActEarly, the Bromley by Bow Centre in Tower Hamlets has further investigated the meaning of the 'children get best start in life' outcome and found that key elements contributing to this outcome for the Tower Hamlets community were: how families inhabit the environment and space around them; the role of play and activities for children; the stability and security needed for a firm family foundation; and the connection and support within families' wider networks [25]. The final point raised by the communities, "connection and support within a family's wider network", can be understood as a systems-level outcome in that no singular measure can be expected to capture it.

There were a few unexpected exclusions that resulted from the consensus process. Breastfeeding, a key indicator in early years health research, and one of the outcomes in the PHOF that is relevant to ActEarly, was not included in the final COS-EY. Similarly, healthy life expectancy at birth, infant mortality and adverse childhood experiences (ACE), were removed. Life expectancy and infant mortality are globally tracked and are reported summary indicators that are thought to capture the overall quality of the early life period [26–28]. These outcomes were removed following the community consultation after failing to reach either a stakeholder consensus that was high enough for automatic inclusion, or a high priority ranking from the community. ACE were also not included in the final COS-EY despite the growing body of evidence that ACE scores are a risk factor for later-life physical and mental health outcomes, and as such, could be thought a key outcome to include in any early life research [29,30]. It is not known to us why the listed outcomes did not achieve the consensus threshold, but it could be that stakeholders felt that the interventions included in ActEarly are unlikely to result in changes in these markers of early life circumstance, or stakeholders were not familiar with the ACE concept. For community members, we think these outcomes may have felt intangible or far removed from their everyday experience—unlike other outcomes that were highly ranked (e.g., traffic). The interactions in the dot voting process are short, which means that there was not time for extensive discussions about each outcome. It is worth highlighting that our outcome sets are the minimum outcomes advocated in this area of research; thus, this does not preclude others from adding in outcomes that are deemed of high relevance even if they are not within the COS-EY.

The development of existing local government or public health outcome frameworks should include interaction with the public, rather than solely consultation with professional bodies, though this is not always done. In our community consultations, we found there was a great interest in providing researchers with feedback on what measures were meaningful to the community. The consultations further highlighted how preconceived notions held by the researchers (e.g., regarding what the most pressing public health issues are) may not reflect the lived experiences of community members. In Bradford, this became evident in the high priority given to wider, structural outcomes such as happiness and mental health, access to high-quality healthcare, employment and poverty, compared to outcomes related to diet, exercise and obesity prevention, which are some of the most pressing national and international public health priorities [31,32]. In addition, safety at home and domestic abuse were of importance for members of the public at one of our study sites and were raised despite the stigma that is commonly associated with discussing these issues [33]. It could be that the anonymity provided by our consultation method may have helped community members feel confident to give their votes to these outcomes, compared to, for example, focus groups or interview methods where the researcher knows the identity of the participant [34].

In Tower Hamlets, an important focus of discussions and responses was around housing and the issue of overcrowding (particularly during lockdown) was mentioned. Another key issue was around traffic, in particular parking and the tensions arising from it. The differences we observed between the two sites suggest that it may be advisable that researchers wishing to use and adapt the COS-EY for their own purposes start with the list of outcomes provided here, followed by some consultation with key local stakeholders to ensure they are fit for purpose.

#### *4.2. Limitations*

This study did not aim to achieve consensus on what the best measures or data sources are for each outcome and work needs to be undertaken before the COS-EY can be used in practice. Furthermore, some outcomes are relatively ambiguous and could be understood to mean multiple things, e.g., "access to opportunity". There is also some overlap between different outcomes—it could be argued that diet and physical activity are very closely related to obesity and therefore not all three need to be included separately. On the other hand, obesity is a complex and multifaceted issue which does not only relate to food and physical activity (which in themselves, contribute to many things beyond an individual's weight status), therefore, we chose not to combine these outcomes.

The Delphi process is dependent on the expert knowledge of stakeholders that are consulted [35]. This means that it cannot be considered an objective 'truth'. With our chosen sampling strategy, it was not possible to estimate a response rate for the surveys, and therefore, we do not know who chose not to take part in the consultation and why. This was mitigated by us contacting and identifying researchers and local authority staff who were already involved with ActEarly and have a stake in selecting appropriate and meaningful outcomes. A limitation of the dot voting process was that community members only spent a few moments reading and reviewing the potential outcomes and understanding the voting process before voting and moving on. There was less time and space for more indepth interactions and conversations with the researcher. This means there was a trade-off between low participant burden (and ease of access to the consultation) and the depth of the information we could collect from the community. In many settings in the UK, an additional barrier to community consultations can be language barriers. We mediated this in Bradford by involving a bilingual researcher in the team that collected the community consultation data. This meant that families and individuals who did not speak fluent English could ask questions about the project and the consultation in their native language. Finally, it is worth highlighting that the exclusion of outcomes from the final COS-EY does not necessarily mean that they are not of value or should not be considered by researchers. Importantly, choosing to employ the public health COS-EY in any intervention evaluation should not be taken to mean that other outcomes should not be considered, and we recommend adapting this COS to the local context wherever appropriate. For instance, if the focus of a future project was more specific than that of ActEarly, e.g., solely around the physical environment, the researchers may wish to explore that specific subset of outcomes in more depth and

consult local communities on whether some of the excluded outcomes should be brought 'back in'.

#### *4.3. Next Steps*

The COS-EY outcome selection to date has not been driven by what can be measured, which means that for now, we cannot be sure that all the outcomes included in the COS can be reported in a meaningful way. Therefore, and before the COS-EY can be fully implemented, we recommend that further work is undertaken to confirm the definition of each outcome, prior to deciding on the most appropriate measures or data sources. This process may be quite challenging for outcomes that cannot be captured by a single metric, but at this stage in the work, we do not think this should mean the automatic exclusion of such outcomes from the COS. Rather, we encourage further research into the area to tackle the issue of defining measures for outcomes such as 'children get best start in life', as clearly, these are priorities for stakeholders and communities alike. One solution to this may be the development of short lists of outcome measures that would represent each core outcome (with e.g., varying degrees of data collection burden or depending on local data availability, such as long and short versions of a questionnaire; or household vs individual level data). These lists could then be used as a starting point by investigators. While not offering perfect standardization the way a single measure per outcome would do, the process of creating lists of appropriate measures would be a step towards better standardization of public health outcomes across studies. Another avenue for future work would be to explore the relevance of the COS-EY from a policy and practice perspective and consider to what extent this work may be useful outside the research context. For instance, the COS-EY could be used by local authorities when making decisions about routine data collection practices and availability.

#### **5. Conclusions**

The public health COS-EY represents an initial attempt at system-wide core outcome sets developed to evaluate interventions that promote early life health and well-being, in consultation with local communities. Our chosen approach resulted in a comprehensive list of 40 outcomes, and highlighted important differences between expert knowledge and lived experience across Bradford and Tower Hamlets. Our aim was to use the COS-EY in the evaluation of the ActEarly research program in the first instance, but the COS could be applied to other settings where there is interest in evaluating early life health and well-being from a 'wider determinants' of health perspective.

**Supplementary Materials:** The following supporting information can be downloaded at: https:// www.mdpi.com/article/10.3390/ijerph19137947/s1, Table S1: Original list of 168 potential outcomes.

**Author Contributions:** Conceptualization, M.B. and L.M.; methodology, M.B. and L.M.; software, M.B., L.M. and Y.V.; formal analysis, L.M., A.A. and M.B.; investigation, L.M., A.A., A.R. and Y.V.; data curation, L.M., A.A. and Y.V.; writing—original draft preparation, L.M.; writing—review and editing, L.M., A.A., Y.V., C.C., A.R., J.S., B.L., T.C.Y., P.G., M.B; visualization, L.M.; supervision, M.B. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by UK Prevention Research Partnership (UKPRP), grant number MR/S037527/1. Individual projects within ActEarly receive funding from universities and directly from different funding agencies. The funders had no role in the design and planning of the present work. C.C. and A.R. were funded through the National Institute for Health Research under its Applied Research Collaboration Yorkshire and Humber [NIHR200166] and J.S. was part funded by The National Institute for Health Research under its Applied Research Collaboration in North Thames. This report is independent research supported by the National Institute for Health and Care Research ARC. The views expressed in this publication are those of the authors and not necessarily those of the National Institute for Health and Care Research or the Department of Health and Social Care.

**Institutional Review Board Statement:** The study was conducted in accordance with the Declaration of Helsinki, and approved by the Institutional Review Board (or Ethics Committee) of The University of York Department of Health Sciences Research Governance Board (protocol code HSRGC/2021/458/E, date of approval 9 July 2021).

**Informed Consent Statement:** Informed consent was obtained from all subjects involved in the study.

**Data Availability Statement:** The data presented in this study are available on request from the corresponding authors to ensure anonymity of participants.

**Acknowledgments:** We would like to thank all the participants for their invaluable input to this study.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


### *Systematic Review* **Workplace Interventions to Reduce Occupational Stress for Older Workers: A Systematic Review**

**Daniel Subel 1,\*, David Blane <sup>2</sup> and Jessica Sheringham <sup>3</sup>**


**\*** Correspondence: daniel.subel.20@alumni.ucl.ac.uk

**Abstract:** The working life of individuals is now longer because of increases to state pension age in the United Kingdom. Older workers may be at particular risk in the workplace, compared with younger workers. Successful workplace interventions to reduce occupational stress amongst older workers are essential, but little is known about their effectiveness. The aim is to evaluate current evidence of the effectiveness of interventions for reducing stress in older workers in non-healthcare settings. Four database searches were conducted. The search terms included synonyms of "intervention", "workplace" and "occupational stress" to identify original studies published since 2011. Dual screening was conducted on the sample to identify studies which met the inclusion criteria. The RoB 2.0 tool for RCTs was used to assess the risk of bias. From 3708 papers retrieved, ten eligible papers were identified. Seven of the papers' interventions were deemed effective in reducing workplace stress. The sample size for most studies was small, and the effectiveness of interventions were more likely to be reported when studies used self-report measures, rather than biological measures. This review indicates that workplace interventions might be effective for reducing stress in older workers. However, there remains an absence of high-quality evidence in this field.

**Keywords:** intervention; workplace; occupational stress; older workers

#### **1. Introduction**

By 2040, it is predicted that one in every two people of working age will be aged 65 or over [1]. The global ageing population has resulted in government concerns regarding the future of the workplace [2]. The increase in life expectancy and the lack of equitable social resources available has been a catalyst for most European governments to increase the state pension age [3].

Prolonging the working life of individuals cannot be done without due diligence and needs to be medically supervised, as suggested by MISPA (Mitigating Increases in the State Pension Age) [4]. Before governments can continue increasing state pension age, it needs to be assessed how this can be conducted, without damaging or harming the health of workers affected by these changes–particularly workers in physically demanding and highly stressful occupations.

Older workers face greater or different hazards to their health than younger workers. Bravo et al.'s [5] review found that in 50% of the papers they reviewed, older workers were at a much greater risk of fatal workplace injuries, when compared to their younger counterparts. Older workers are more likely to have pre-existing long-term health conditions, which can affect their capacity to work or the kinds of work they are able to sustain. There is also evidence that they experience greater sickness absence [6].

Stress—adverse reactions to excessive pressure—and burnout are recognised as a major risk to the health of all workers [7,8]. There is conflicting evidence whether older workers are at a greater risk of stress than younger workers [9]. It is clear, however, that

**Citation:** Subel, D.; Blane, D.; Sheringham, J. Workplace Interventions to Reduce Occupational Stress for Older Workers: A Systematic Review. *Int. J. Environ. Res. Public Health* **2022**, *19*, 9202. https://doi.org/10.3390/ ijerph19159202

Academic Editor: Giulio Arcangeli

Received: 9 June 2022 Accepted: 26 July 2022 Published: 27 July 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

older workers are likely to face different stressors to younger workers, not just through pressures within the workplace, but also through additional caring responsibilities outside of work [4]. Moreover, there is agreement that, despite legislation to prevent it, there is evidence that older workers are subject to age discrimination [10]. Therefore, interventions to improve workplace health in older workers may well need to be different to those of younger workers because it cannot be assumed that the problems they face, or the mechanisms by which interventions work, will be the same.

Based on currently available research, very little is known about this topic. Evidence of the effectiveness of workplace interventions for older workers is lacking. Poscia et al.'s [3] systematic review found a paucity of high-quality evidence on workplace health promotion for older workers. There was a suggestion that active workplace interventions help improve the health of older workers, but included studies used small, convenience samples not representative of the working population.

Pieper et al.'s [11] more recent review of reviews on workplace health promotion interventions found that psychological interventions, such as stress management, cognitive behavioural therapy and mindfulness-based interventions have the ability to significantly reduce stress. However, Pieper et al. found few reviews specifically focusing on older workers and they reported that there was insufficient evidence to conclude that psychological interventions were the most successful and effective to reduce occupational stress amongst older workers. Interventions were predominately targeted towards white-collar workers, teachers, and healthcare providers. Interventions for healthcare providers may be of limited generalisability to other settings, given the specific nature of healthcare settings and healthcare work and the hazards that may present in this environment. When assessing previous studies on this topic, the overwhelming majority focus upon younger people employed in advantaged occupations, using small cohort sizes. Furthermore, they use inconsistent and haphazard outcome measures to assess interventions' successes, which results in studies being unrepresentative, difficult to replicate and unable to demonstrate the impact on increased state pension age for older workers.

Before policy makers can enact changes to state pension age, they must have access to a sufficient level of high-quality research which has outlined the impact on individuals working longer, as well as interventions used to retain older workers. This information must also be accessible to employers, so they are made aware of the most successful interventions in the workplace to reduce occupational stress and maintain their workforce. This article intends to provide policy makers and employers a review of the current literature and research in this field. This study also has the potential to provide union representatives, and workers themselves, with evidence for them to vouch to their employers for adequate, appropriate, and successful interventions in their workplaces.

This systematic review sought to answer the question:

"What is the evidence of effectiveness of workplace interventions for reducing occupational stress in older workers outside of the healthcare sector?"

The objectives were to:


#### **2. Materials and Methods**

#### *2.1. Search Strategy*

PRISMA guidelines for reporting systematic reviews were followed throughout the process of this review [12], see checklist in Supplementary Materials. Four database searches were conducted: OvidMedline, PsycInfo, Scopus and Web of Science. After an initial literature search, a PICO model was developed (see Table 1), which helped form the database search terms for the review [13]. Previously systematic reviews' search terms, including Pieper et al. [11] and Poscia et al. [3], helped to inform the search terms. The search term combinations were first applied in OvidMedline, which uses MeSH terms, and then modified and adapted for use in the other databases (see Table 2). In all the databases, the presence of key words was sought in "all fields", which would detect the terms in key words, titles, abstracts and full papers. Initially age terms were included in the search strategy, but this resulted in an improbably low number of results retrieved, so this term was dropped. The searches took place throughout the first week of August 2021, therefore only research published before 31 July 2021 were included in this review.

**Table 1.** The PICO model.


**Table 2.** Search Terms.


\* MeSH terms are indicated with a "/" after the search term. Programmes, setting and outcome search terms were combined with "AND" in each database.

#### *2.2. Inclusion/Exclusion Criteria*

Table 3 depicts the inclusion/exclusion criteria. Eligible papers had to report an intervention in a non-health sector workplace, specifically focusing on older workers. The papers had to have been conducted in an Organisation for Economic Co-operation and Development (OECD) country, to ensure findings have some relevance to the United Kingdom (UK) context [14]. Studies without a control group or baseline data, or without an aim of reducing workplace stress, were excluded. The authors did not set out to select papers which specified a specific control condition but sought papers which described what interventions were compared with. Qualitative papers, such as focus groups or interviews, were excluded from this review as quantitative papers were deemed to illustrate more objective results and are more likely to be conducted on a large number of participants.


**Table 3.** Inclusion/Exclusion Criteria.

The definition of an older worker was developed by adapting multiple definitions from various sources. Firstly, if the paper classified the intervention or participants as older workers, regardless of the mean age, these interventions were deemed to be focused on older workers. Secondly, for OECD countries, the average age at which an individual reached normal pension age in 2016 was 63.7 years old for women and 64.3 years old for men [15]. If the mean age of participants in a paper were within 15 years of normal pension age, it was concluded that older workers were included in this intervention.

#### *2.3. Study Selection and Screening*

Papers from the four databases were exported to Microsoft Excel. Title and abstracts of all papers screened by DS (author and reviewer) and a secondary reviewer (AH). Any papers which were unclear or resulted in polarized views, were then resolved by discussion with a third reviewer and co-author (JS). After the title and abstract screening, the remaining papers underwent a full-text screening.

Each paper that met the inclusion criteria on screening was carefully assessed for its relevance to older workers. Papers that were specifically focused on older workers were placed in the primary dataset. Papers where data on the effectiveness of the intervention on older workers was included, but without a specific focus on older workers, comprised the secondary dataset.

#### *2.4. Data Extraction and Critical Appraisal*

Data were extracted from all eligible papers used a data extraction form by DS, with a sample checked by JS (see Appendix A: Data Extraction Form) to cover features including: study design and employment setting; the age of participants; nature of the intervention; reported effectiveness. Interventions were coded into three categories–psychological interventions, educational interventions, and physical interventions. Outcome measurements were grouped by whether self-report or biological samples were used to measure stress.

The RoB 2.0 tool (Risk of bias in randomised trials) [16] was used by DS and JS in each paper to form a judgement about the risk of bias across six different domains. The RoB 2.0 tool was chosen as it enabled the reviewers to form their own assessment of an article's quality, in regard to its risk of different types of biases. If a domain or the overall judgement was deemed to have a high risk of bias, this meant that the reviewers believed that there was an issue with the paper that substantially lowered their confidence in the results. Some concerns of bias indicated that a paper included an issue which could potentially lower the reviewer's confidence in the results. If the overall judgement was that the paper had a low risk of bias, this meant that the reviewers were confident that the study results were valid.

The included studies were described, and the characteristics and methods for ascertaining stress levels were summarised. Based upon what was written in each paper, the effectiveness of the interventions was summarised, using quantitative data to assess the success of each intervention.

#### **3. Results**

#### *3.1. Characteristics of Included Studies*

From 3708 papers identified in the database searches, ten papers met the inclusion criteria (Figure 1). Five papers had a specific focus on older workers (the primary dataset). A further five papers did not have a specific focus of the research on older workers (the secondary dataset). As the mean age of participants in both datasets were similar (see Table 4), they are considered as one dataset in the rest of the paper.

Five papers were conducted in the United States [17–21]; the other five originated from Europe (Germany [22], the Netherlands [23], Finland [24], Italy [25] and Norway [26]). The number of participants ranged from 14–779, with three studies have less than 40 participants. Only one study included over 500 participants [24] (see Table 4).

Three studies were conducted with university faculty staff [17,18,21], and three in manufacturing or technical environments [19,20,22]. Two studies were conducted amongst police officers [23,25]. The remaining two studies were conducted with office workers [24,26].

The age of participants was described in two ways (Table 4). Six studies described the age range of participants in the intervention; the upper limit for the age range was between 57 and 68; the lower limit for the age range was 18 to 50. Three papers only included participants over the age of 40, with Calogiuri et al.'s [26] paper using only participants older than 50 years old. In the seven papers that documented the mean age of participants, mean age was over 40.9 years. Five papers had a mean age of over 48 years [17,18,21,23,24].

**Figure 1.** PRISMA Flow Chart (Adapted from Page et al. [12]) \* Papers were automatically excluded using filters in the search databases where they were outside of our date range and language of publication.



\* Median has been calculated by the researcher as the midpoint between the range. In Ojala's study, Public Sector. workers included construction and transport workers, office workers, food services and managerial specialists.

#### *3.2. Risk of Bias*

None of the papers had an overall high risk of bias (Table 5). Four papers were judged to have a low risk of bias. Some bias concerns were identified in six papers. In nine out of the ten papers there was a lack of detail on the randomisation of participants, which may have led to post-test reporting bias by participants exaggerating the effects of the intervention. Most papers showed a strong adherence to the intended intervention. Fischetti et al.'s [25] study showed a potential high risk of bias in the measurement of outcome. While the study used validated scales to assess stress, the score was high because of the study's pre-post evaluation design. It is possible that participants may be subject to bias in overestimating the effects of participation on their well-being.

**Table 5.** Results from the Risk of Bias Critical Appraisal.


Key for Table 5: 1 = Low Risk of Bias. 2 = Some Concerns. 3 = High Risk of Bias.

#### *3.3. Study Methods*

The most common form of intervention was psychological interventions (*n* = 8). Psychological interventions included mindfulness-based, cognitive behavioural therapy and stress management interventions. Three studies used physical interventions, which involved exercise, walking, weight training or circuit training programmes [20,25,26]. One paper included an educational intervention [17] focused on health education (Table 6).

Of the ten papers in this review, five papers [19,20,22,24,25] reported that the control group received no intervention during the research but were waitlisted to participate in the intervention at a later date. In Malarkey et al.'s [18] study, the participants in the control group received a lifestyle and educational intervention, compared with the mindfulnessbased intervention that the experimental group received. Hughes et al.'s [17] study control group received a light level of health education compared with the experimental group, who received the health promotion intervention. Hoeve et al.'s [23] control group received a regular education intervention, without any mindfulness training. Two papers' control groups [21,26] had either an indoor standard work break or indoor exercise, compared to the experimental groups whose interventions were conducted outdoors. No conclusive pattern emerged between which control condition was in place and the outcome of the intervention. Table 6 illustrates that of the five interventions [19,20,22,24,25] in which the control group received no intervention, three of these papers reported an effective intervention in the experimental group.

Six papers conducted their interventions in the workplace offices, two papers were carried out via online means in the workplace, and a further two papers took place outside of the workplace, in green areas and nature.

All papers in this review used self-reported questionnaires to collect data on stress. Three of these papers also collected cortisol levels, either from saliva samples or blood tests [18,22,26]. Four of the papers used the Perceived Stress Scale Questionnaire to assess the level of psychological stress perceived in participants.

The shortest intervention took place over the course of two weeks [26]. Three of the papers' interventions took place for over six months, including follow up time [17,22,24]. The longest duration for intervention was Hughes et al.'s [17] 12-month study.

#### *3.4. Study Findings*

Changes in stress levels as a result of each intervention are reported in Table 6. In seven out of the ten studies [19,21–26], there was improvement in at least one measure of self-reported stress levels. However, none of the three studies that measured cortisol levels [18,22,26] found any significant differences between the intervention and the control group's cortisol levels.

Three interventions [17,18,20] showed no evidence of effectiveness on any measure. There were no consistent patterns in terms of the intervention type (psychological, physical educational), workplace setting or delivery method between effective and ineffective interventions.



was said to be available via contact with the Journal or Author. After contacting both, no response was received.

#### **4. Discussion**

#### *4.1. Main Findings*

The evidence of the effectiveness of interventions to reduce stress in older workers was varied. Seven out of the ten papers reported some effectiveness in reducing selfreported stress in older workers as a result of interventions. Studies that measured cortisol levels did not report any reduction in stress. Most of the interventions were psychological in nature, but there was no difference in reported effectiveness between psychological, physical, or educational interventions. It should be noted that most of these interventions were only short-term, and therefore, longer-term impacts of these interventions are not clearly demonstrated.

#### *4.2. Methodological Considerations*

There were some important limitations in the studies included in the review. Firstly, the number of participants reported in each study was generally low, with three papers comprising studies of less than 40 participants and only one study with more than 500. Due to the low number of participants, it is difficult to generalise the results of these interventions to the broader population [27]. In most of the studies, participants had to volunteer to take part. In some studies, it was not clear how many employees that were eligible declined to take part so the acceptability of such interventions in the workplace cannot be concluded from this study.

Secondly, none of the ten papers observed the longer-term impacts of the interventions. Whilst papers stated or implied that the interventions were longer-term solutions to the problem of occupational stress amongst older workers, they provide no conclusive evidence of long-term benefit. The concern regarding the long-term effects associated with workplace interventions has also been discussed by others. Steenstra et al. [28] reported how the effect of interventions require a very long follow-up, which is extremely difficult to achieve and maintain. They concluded that the interventions' effect would most likely dilute over time and not result in any long-term benefits. Similarly, in this review, two out of the three papers which had interventions lasting more than 8 months were shown to have mixed or no effect on reducing workplace stress. This is suggestive evidence in support of Steenstra et al.'s conclusions that the impact of workplace interventions to reduce stress could have little to no long-term benefit if the intervention is not maintained in the workplace. It is also possible maintenance of a short-term intervention is not enough; workplaces may need different kinds of approaches to maintain reductions in stress levels in the longer term.

Thirdly, there was a range of self-report questionnaires used, which collected data on various aspects relating to stress, mental health, or other factors. When analysing the interventions, as different measurement outcomes are used, it can cause difficulties in understanding which intervention is the most effective.

There were some limitations also in the conduct of this review. Only papers published in English were included in the review. Studies which were written and published in other languages were removed at the first stage of screening. Whilst the majority of papers which were found in the database search stage of the review were written in English, those in other languages may have been beneficial to include in this review. Using free, online translation software to translate any non-English studies has become more common in academic reviews, and if this research was to be conducted again in the future, including non-English studies, and using translation software should be strongly considered. It was also beyond the scope of the review to include qualitative studies. Whilst these would not have definitely addressed questions of effectiveness, they could have provided useful insights into why intervention achieved their effects. The RoB 2.0 tool which was used for performing the critical appraisal does not prompt consideration of wider aspects of quality and relevance, for example, what the control conditions were. This could be seen as a potential limitation in several of the papers in the review.

A further challenge faced in this review was the ambiguity regarding the definition of an older workers. The initial search terms included specific terms and synonyms for "older worker". However, this resulted in a very small number of papers being retrieved. Therefore, this search term was removed and at abstract and full paper screening, the reviewers determined which papers focused on older workers and which did not. Eliminating "older workers" as a search term in the database search led to a potential risk that relevant literature, with a clear focus on older workers, may have been overlooked. However, "older worker" was hard to define partly because the classification of an older worker varies across countries. The ELSA (English Longitudinal Study of Ageing) and the JSTAR (Japanese Study of Ageing and Retirement) both describe workers over the age of 50 as "older workers", however, Kingston and Jagger [29] argue that cohort studies with the lower age limit of 50 to 65 years old, often have fewer very old people in the studies, therefore, are not fully representative of older workers. The nature of the risk of being an older worker varies in the context of workplace settings, occupations and job demands. For example, as seen in Fischetti et al.'s [25] and Hoeve et al.'s [23] research, police officers may be more at risk of injury at a younger age, due to the physical nature of their occupation. This may result in police officers aged 40 being deemed as older workers in their profession, although at their chronological age, in society and in other professional groups, they would be classified as younger. However, at this age, it is possible for some police officers that the nature of their work may change, to become more 'desk based'. In this case, the current workplace exposures may be more similar to office workers, but the prior exposures they faced from working in communities may have long lasting and distinct effects on their health that are not experienced by those who have spent their entire careers in office-based jobs. Due to the small number of studies identified, this review was not able to explore the differences in the nature of interventions across workplaces. This is needed in future because different causes of stress based upon a range of diverse types of employment may affect the sustainability and the effectiveness of interventions to reduce workplace stress.

#### *4.3. Interpretation of Findings and Comparison to Previous Studies*

Of the ten papers sourced for this review, only five reported a specific focus on older workers, demonstrating the lack of robust and available literature on this topic. This finding is consistent with older systematic reviews researching workplace interventions for older workers [3,11,28]. More than ten years ago, Crawford et al. [30] urged for more research to be conducted on health and safety management interventions for workers over the age of 50 in relation to the physical and psychological changes that occur when workers reach this age.

Interventions that used self-reported measures appeared more effective when compared to biological measures. However, it is important to note that self-reports and biological samples measure different things. Taking part in an intervention may improve subjective well-being in an individual, even if it has no biological effect. This does not imply that the intervention was unsuccessful or ineffective. Indeed, McDonald [31] suggests that gaining self-reported data from participants is the most logical way to learn more about an individual. Arguably, an individual's subjective well-being is what would keep them in the workplace.

Whilst it is understood that older workers might not always face more workplace stress compared to younger workers, they could be more at risk of specific stressors connected to responsibilities outside of work, age discrimination and physical health conditions that are more common in older age [32]. In the ten papers' interventions, there was not enough description regarding the extent to which specific stressors associated with older age were addressed. It is, however, significant that there were five studies that did not seek to focus on older workers, potentially overlooking distinct stressors. In these studies, it is also possible that the overall effectiveness could have been driven by higher effectiveness in younger populations but there was not sufficient data reported in these papers on effects by participant age to explore this. The context in which the interventions effect change may be important. Interventions in the workplace, which are promoted and supported by employers may encourage participants not only to take part in the intervention, but also to make changes to their lifestyle and behaviour, which, in turn, would ultimately improve their well-being and decrease their stress levels at work [33].

#### *4.4. The Significance of the Review and Public Health Implications*

Since Crawford et al.'s [30] review was undertaken, policy and demographic changes have lead to a higher proportion of older workers in many countries, increasing the importance of health and safety interventions for workers over the age of 50. The need for such research has not been addressed and the knowledge gaps that were present in the literature remain.

This review has demonstrated that there is still not sufficient research available for governments and policy makers to make an informed decision on the impact of increasing state pension age on the population. If they are determined to extend the working life of individuals, governments will need to ensure that there is no detriment to the health of older workers.

The lack of high-quality literature on this topic results in this review being unable to provide any definitive conclusions regarding the most effective and successful workplace intervention to deal with occupational stress. This systematic review can be updated to illustrate newly published literature about older workers' well-being in years to come. The significance of implementing a successful intervention to promote and maintain the health of older workers is vital for the longer-term wealth creation and sustaining of both the economy and health of the population [34].

This review has not shown an adequate amount of successful workplace interventions to support older workers' occupational stress to mitigate the public health implications of raising state pension age, as reported by MISPA [4]. More extensive and robust research is required to illustrate to both employers and policy makers that increasing state pension age will result in; increased morbidity and mortality rates for those in demanding occupations; overwhelming the already sparse healthcare services-both for occupational health and primary care; and, worsening the health for workers who are already ill. Careful considerations need to be made to ensure that older workers are not adversely harmed by increases to state pension age. It is fundamental that interventions, which have been proven successful for older workers, must be introduced into more workplaces to ensure a smoother transition for older workers who are now working longer.

#### **5. Conclusions**

As the population ages, and statutory pension age increase, the proportion of older workers will increase in the workplace. Older workers face distinct and sometimes greater risks to health and well-being compared with younger workers, which may place them at particular risk of stress. This review found some promising evidence that interventions in the workplace can improve self-reported stress in older workers in the short term. It also highlighted the paucity of studies with interventions specifically designed for older workers. Further studies are required to understand longer term impacts of workplace interventions on older workers and to elucidate what type of intervention is most likely to be effective in different workplace settings.

**Supplementary Materials:** The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/ijerph19159202/s1, PRISMA 2020 Checklist.

**Author Contributions:** The idea of the study was developed by D.S. and J.S., with input from D.B. D.S. developed the study protocol and searched the four databases. The title and abstract screening process was conducted by D.S. J.S. supported the screening process as a third reviewer in events where a decision could not be agreed upon. The Critical Appraisal was undertaken by J.S. and D.S. D.S. drafted the manuscript. All three authors contributed to subsequent drafts and agreed upon the final manuscript for submission. All authors have read and agreed to the published version of the manuscript.

**Funding:** Jessica Sheringham is supported by the National Institute for Health and Care Research (NIHR) Applied Research Collaboration (ARC) North Thames. The views expressed in this publication are those of the authors and not necessarily those of the National Institute for Health and Care Research or the Department of Health and Social Care.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Acknowledgments:** The authors wish to thank all who supported them throughout the project. Thank you to Hynek Pikhart and Martin Bobak at University College London (UCL) for their constructive feedback throughout the developmental stages. We also wish to give particular acknowledgement to both Anne Harte, who acted as the secondary reviewer during the screening and data extraction process and Jenny Lisle, who provided specialist input regarding the role of occupational health.

**Conflicts of Interest:** The authors declare no conflict of interest.


**Table A1.** Data Extraction Form from Primary Dataset.






**Table A2.** Data Extraction Form from Secondary Dataset.






#### **References**


### *Review* **Inequity in Access and Delivery of Virtual Care Interventions: A Scoping Review**

**Sabuj Kanti Mistry 1,2,\*, Miranda Shaw 3, Freya Raffan 3, George Johnson 4, Katelyn Perren 4, Saito Shoko 5, Ben Harris-Roxas <sup>6</sup> and Fiona Haigh <sup>5</sup>**


**Abstract:** The objectives of this review were to map and summarize the existing evidence from a global perspective about inequity in access and delivery of virtual care interventions and to identify strategies that may be adopted by virtual care services to address these inequities. We searched *MEDLINE*, *EMBASE*, and *CINAHL* using both medical subject headings (MeSH) and free-text keywords for empirical studies exploring inequity in ambulatory services offered virtually. Forty-one studies were included, most of them cross-sectional in design. Included studies were extracted using a customized extraction tool, and descriptive analysis was performed. The review identified widespread differences in accessing and using virtual care interventions among cultural and ethnic minorities, older people, socioeconomically disadvantaged groups, people with limited digital and/or health literacy, and those with limited access to digital devices and good connectivity. Potential solutions addressing these barriers identified in the review included having digitally literate caregivers present during virtual care appointments, conducting virtual care appointments in culturally sensitive manner, and having a focus on enhancing patients' digital literacy. We identified evidence-based practices for virtual care interventions to ensure equity in access and delivery for their virtual care patients.

**Keywords:** inequality; health equity; health services; virtual care; COVID-19; scoping review

### **1. Introduction**

Health inequities are referred to as those differences in health that are systemic, avoidable, unfair, and unjust [1]. A health equity approach recognises that not everyone has the same level of health or level of resources to address their health problems, and it may therefore be important apply different approaches in order to achieve similar health outcomes [2,3]. Health inequities are associated with a range of factors including age, gender, ethnicity, geographic location, and socioeconomic status [1,4]. A recent report has documented nine drivers of health inequity in relation to healthcare services: housing, income and wealth, health system and services, education, employment, social environment, transport, public safety, and physical environment [5]. Outcomes are determined by the dynamic interaction between service users and the health systems [6]. The World Health Organization (WHO) also identified gender, education, income, employment status, and ethnicity as the major factors associated with health inequity [7]. Health inequities are an established global phenomenon [8,9] and is a particular concern in a multicultural country

**Citation:** Mistry, S.K.; Shaw, M.; Raffan, F.; Johnson, G.; Perren, K.; Shoko, S.; Harris-Roxas, B.; Haigh, F. Inequity in Access and Delivery of Virtual Care Interventions: A Scoping Review. *Int. J. Environ. Res. Public Health* **2022**, *19*, 9411. https:// doi.org/10.3390/ijerph19159411

Academic Editors: Jessica Sheringham and Sarah Sowden

Received: 19 May 2022 Accepted: 27 July 2022 Published: 1 August 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

with a history of settler colonialism, such as Australia [10,11]. Evidence demonstrates that the determinants of health inequity often lead to adverse health outcomes in the form of morbidities and mortality among vulnerable and marginalised populations [4,12,13].

Virtual care can be defined as "any interaction between patients and/or members of their care team occurring remotely, using technology with the aim of facilitating or maximising the quality and effectiveness of patient care" [14]. It has been identified as an approach that may partially address health inequities through improving access and availability of health services [15,16]. However, there are also concerns that virtual care services could exacerbate existing health inequities if services are not accessible, available, and acceptable to vulnerable population [17,18]. Virtual care interventions received particular attention during the COVID-19 pandemic, as many health services rapidly transitioned to providing virtual care services as an emergency method of reaching their clients [19]. The restriction of in-person health services and the rapid implementation of virtual care has been driven by necessity but also presents a significant opportunity to develop and strengthen the provision of virtual care [20–22]. However, this expansion in virtual care services using healthcare technologies also created the potential for the widespread digital divide to act as a potent barrier in successful implementation of virtual care interventions and a cause of health inequities. The digital divide is defined as disproportionate access and utilization of health technology and internet among certain population groups, characterised by their geographical, social, and geopolitical criteria or other features [23]. There are suggestions of a "digital paradox" where the "population groups that could potentially benefit most from digital innovations are the ones that would experience the highest barriers to access" [24].

Recently, several studies were conducted on the expansion of virtual care interventions, particularly in relation to the COVID-19 pandemic [25–27]. Many of these studies considered virtual care as a way of minimising the risk of COVID-19 transmission [26,28], to triage during emergency responses [21] and monitoring patients within their homes [21]. One such intervention is the RPA Virtual Hospital (rpavirtual), launched in February 2020 as a new model of care that combines integrated hospital and community care with digital solutions. It was the first service to introduce virtual care for COVID-19-stable patients in isolation in New South Wales, Australia, and has been demonstrated to be widely accepted by patients [29]. However, the potential equity issues related to rpavirtual and other similar virtual care interventions have not yet been adequately explored and described. Therefore, this scoping review aims to map and summarize the knowledge about equity issues in the access and delivery of virtual care interventions and to identify strategies to address potential inequities that may be adopted by virtual care services.

#### **2. Materials and Methods**

This scoping review is reported following the guidelines of PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-analysis extension for Scoping Reviews) [30]. The review protocol is registered at the website of Centre for Primary Health Care and Equity, UNSW Sydney (https://cphce.unsw.edu.au/research/rapid-literaturereview-identify-equity-issues-access-and-delivery-virtual-care, accessed on 1 March 2022), and PRISMA-ScR is provided in Supplementary Table S1.

#### *2.1. Data Sources*

We searched for peer-reviewed articles in electronic databases: *Medline*, *EMBASE*, and *CINAHL*. Both medical subject headings (MeSH) and free-text keywords were used to search relevant articles in these databases that were published in the English language between January 2010 and January 2021. The detailed search strategy is presented in Table 1.

#### **Table 1.** Search strategy.


*2.2. Study Selection*

The articles yielded in the initial database searches were assessed by two independent reviewers in relation to the inclusion and exclusion criteria developed for this study (Box 1). All of the steps of study selection procedure were performed in Covidence (https://www. covidence.org, accessed on 1 March 2022). In the first stage, the title and abstract of the articles were assessed by two reviewers. The articles that passed this initial screening stage entered full text screening. The full texts of these articles were obtained, and more in-depth assessment was carried out against the inclusion and exclusion criteria. The reason for the exclusion for each of the articles was also noted at this stage. Any difference in assessment between the reviewers was resolved by discussion.

**Box 1.** Inclusion and exclusion criteria.

#### **Inclusion criteria**


#### **Exclusion criteria**


#### *2.3. Data Extraction*

The data were extracted from the included studies in a Microsoft Excel template developed by the authors. Information including country, study setting, study design, study participants, characteristics of the intervention/study, type of virtual care modalities, type of inequity issues identified/addressed, main findings, summary of the results, and relevance to virtual care interventions were extracted.

#### *2.4. Data Mapping*

As the objective of scoping reviews is to map and summarize the available evidence, we performed descriptive analysis, which involved frequency counting and basic thematic coding [31].

#### **3. Results**

#### *3.1. Search Results*

Database searches yielded a total of 3021 articles, from which 1990 underwent screening after removal of the duplicates. The assessment of the title and abstract of the articles resulted in the exclusion of 1901 articles, and 89 articles underwent full-text screening. Finally, 41 articles satisfied the selection criteria and were included in the review (Figure 1). The detailed characteristics of the included studies are presented in Supplementary Table S2.

**Figure 1.** PRISMA diagram of study selection.

#### *3.2. Study Settings*

Of the forty-one included studies, thirty-one were conducted in the USA, three were carried out in Australia [32–34], two in Canada [35,36], one in Italy [37], one in China [38], one in Germany [39], one in Norway [40], and one in Scotland [41]. The studies were carried out either in a community or in a clinical setting, such as a hospital or primary care.

#### *3.3. Study Designs*

A range of study designs were used in the included studies. Twenty-three of the included studies followed a cross sectional design [5,32,34,36–55], five studies carried out retrospective analysis of the collected data [56–60], six studies followed cohort design [61–66], two were randomised controlled trials [67,68], and two followed a mixed-method design [33,69]. One study followed a combination of retrospective analysis and cross-sectional study design, [70] while the study design was not clear in two studies [35,71].

#### *3.4. Type of Participants*

The participants in most of the studies were adults, often with chronic conditions such as diabetes [45], cardiovascular disease [39], and mental health problems [61]. Most of the studies considered both native English speakers and those speaking languages other than English. Only a few studies considered all participants speaking a language other than English, such as Spanish [56] or Chinese [34]. Several studies examined outcomes of specific cultural and ethnic minorities. However, since most of the included studies were conducted in the USA, the population groups were mostly Black, Hispanic, and African American [46,48,50,54,55,60,66–68].

#### *3.5. Virtual Care Modalities*

The included studies considered several modalities of virtual care interventions ranging from video conferencing [37,41,42,44,46,48,49,51,57,59–64,66,67,71], teleconferencing [34,35,46,48,49,51,53,54,56–58,60,62,63,66,71], messaging [42,45,50], emails [42], health apps [5,39,40,50], patient portals [58,61,68,70], personal health records [59,61], and eHealth service use via the Internet [32,40,47,69].

The majority of the studies described the use of virtual visits (either video or audio) in comparison to face-to-face visit or video visit in comparison to audio/tele visit. Several video conferencing platforms, such as Zoom (https://zoom.us/, accessed on 1 March 2022) or Microsoft Teams (https://www.microsoft.com/en-au/microsoft-teams/group-chatsoftware/, accessed on 1 March 2022), were used to perform video visits in the reported studies. Some of the studies reported on non-synchronous communication tools such as text messaging, health apps, patient portals, or eHealth service use. Text messaging, health apps, or patient portals were generally used to book appointments with service providers, access health information, track health outcomes, or communicate with health service providers. On the other hand, eHealth services were offered to promote online learning, counselling, and information sharing, and these aims were accomplished through browsing search engines, health apps, social media, and video services.

#### *3.6. Types of Inequity Issues Identified/Addressed*

#### 3.6.1. Cultural and Ethnic Inequities

Twenty-one studies [33,42,44–46,48,50,54,55,57,59–68,70] explored cultural and ethnic inequities in access to virtual care services and outcomes. The majority of these found that cultural and ethnic minorities, including those of African American, Black, Hispanic or Latinos, Asian American, Aboriginal and Torres Strait Islander, or Filipino descent, were less likely to access virtual care services compared to the White participants. For example, in their study, Schifeling and colleagues [60] found that non-White patients were less likely to have a video visit than White patients. Likewise, Walker et al. [68] found that African American patients used the patient portal less than White patients (40.4% difference, *p* = 0.004). However, four studies [42,44,50,65] reported a different result where

the likelihood of using virtual care services was higher among the cultural and ethnic minorities compared to White participants.

#### 3.6.2. Sociodemographic and Socio-Economic Inequities

Older people were identified as experiencing significant barriers to accessing and using virtual care services in most of the studies [5,32,36,38,39,41,43,45–50,55,60,61,63,65,68,70,71]. For example, Leng et al. [41] found that the patients under 60 years were over two times more likely to use video consulting (odds ratio (OR) 2.2, 95% CI 2.1–6.6). Nelson et al. [45] also pointed out that the probability of responding to texts tended to increase from about age 25 years until roughly age 50 years and then appeared to decrease with increasing age. Eberly et al. [64] further noted that younger participants were more likely to be engaged with video call appointments compared to telephone call. The only exception was reported by Pierce et al. [46], where age of 65 years and above was associated with higher odds of virtual care use (OR 1.21, 95% CI 1.05–1.40). It is also notable to mention that all nine studies [33,39,46,53,57,61,63–65] that explored the role of gender in accessing virtual care services found that females were less likely to use virtual care services compared to males. Two studies [63,65] also found that unmarried participants were less likely to access virtual care services. Meanwhile, Wegerman et al. [66] found that participants who were single or previously married (separated, divorced, widowed) had higher odds of completing a telephone appointment, while married participants were more likely to complete a video appointment.

Thirteen studies explored the use of virtual care in relation to the socioeconomic status of the participants, and all of these found that lower socioeconomic status was associated with lower use of virtual care services [5,32,33,38–40,47,48,50,51,61,63,67]. Alam et al. [32] reported that access to virtual care services was lower among participants from disadvantaged socioeconomic backgrounds. Likewise, other studies [5,33,38,40,48,50,51,61,63] also reported that low socioeconomic status was associated with decreased access to virtual care services. Not surprisingly, some of the included studies that explored the role of education in accessing virtual care services [5,32,33,38–40,47,48,50,67] also found that participants with lower education status were less likely to access the virtual care services.

#### 3.6.3. Inequity Issues Related to Digital/eHealth Literacy

Seven studies [32,38,39,41,45,56,69] reported a lack of digital/eHealth literacy among the participants as a significant barrier to accessing virtual care services. Ernsting and colleagues [39] found that mHealth app users had higher levels of eHealth literacy compared to non-app users. A study [69] also reported that eHealth literacy increase was associated with a 3% increase in the number of searches for health information on the internet (beta = 0.03, 95% CI 0.00–0.06). Meanwhile, Leng et al. [41] found that higher computer proficiency correlated with an increased willingness to engage in video consultations.

#### 3.6.4. Technological Inequities

Several studies [32,37,48,50] also found that improved access to digital devices and internet can increase the use of virtual care services. Arighi et al. [37] reported that issues such as a lack of devices (computers, phones, or tablets) with internet connection and poor internet connections were the main causes of failed virtual care. Alam et al. [32] pointed out that access to broadband internet services was associated with increased use of virtual care services.

#### **4. Discussion**

This review was conducted to explore inequity issues in relation to access to and delivery of virtual care interventions and to consider the international evidence of actions to address inequity issues that may be adopted by or provide learnings for rpavirtual and other similar virtual care interventions. The main drivers of inequity in access to virtual care identified in the literature review were relatively older age, unemployment, less income, lower education level, belonging to cultural or ethnic minorities, lack of access to digital devices or good internet connection, and lack of digital/eHealth literacy.

In recent times, due to the COVID-19 pandemic, virtual care interventions have been widely used due to restricted in-person health service delivery [25,26]. It has also been documented that the patient experience and their acceptance of virtual care during this pandemic has been generally good [72,73]. At the same time, it is also worth noting that the expansion of this digital innovation without due consideration of strategies to address inequity of access has the potential to increase health inequities due to poverty, digital health literacy, and lack of access to digital technology among some of the population [74].

Reviews carried out during the COVID-19 pandemic [75,76] also stressed the importance of virtual care interventions as an alternative to in-person health service delivery during a period of restrictions on face-to-face health service delivery. Doraiswami et al. (2020) [75] reported that virtual care could play a pivotal role in the health sector in future, but its feasibility and implementation in a resource-poor setting is challenging. In this regard, it is critical to mention that future virtual care interventions will be influenced by broader health and clinical governance agendas and directions in investment across systems.

While some recent reviews [77–79] have highlighted the effectiveness of virtual care as a way of delivering health care in a cost-effective way, with improved patient communication, outcomes, and satisfaction, the equity dimension of the virtual care interventions is not fully addressed in these reviews. The present review has helped to bridge the knowledge gap around inequity issues associated with virtual care and identified areas for further research.

The present review highlighted that access to virtual care services is particularly limited among patients from ethnic minorities, which suggests there is a need to carefully tailor services to ensure equitable access. Multilingual and culturally sensitive virtual care services can be of high value in this regard. For example, a culturally sensitive approach documented by Shaw et al. (2013) [34] could be to address cultural diversity in the developing of a virtual care intervention. This qualitative study was conducted among Chinese and Arabic patients and their carers to explore their willingness to take part in a telephone-based supportive-care intervention. The majority of the study participants supported the provision of a culturally sensitive intervention in their own language via an online platform. However, the participants identified that confidentiality of the clinical information was a concern and preferred an initial in-person appointment with patients to increase participation. It was also suggested that there should be the provision of an "on-call" support process initiated by patients to provide patients with access to assistance in times of high need between scheduled calls.

Access to virtual care services is linked to the level of digital literacy of the patients. For example, Ernsting et al. [39] and Guendelman et al. [69] strongly emphasised the importance of improving digital literacy of patients in order to address inequity of access to virtual care services. Older people and individuals with limited digital health literacy are less likely to access virtual care services and require targeted support. The present review indicates that availability of younger caregivers or caregivers with higher digital literacy can result in increased access to virtual care services [37].

Consideration of different levels of digital and health literacy across patients should be a part of routine planning for virtual care services. For example, an educational component can be incorporated in interventions to increase virtual care literacy among vulnerable patients. In addition, delivery methods can be updated, for example, by adapting portals to be comfortably used by less digitally literate patients or appropriately tailoring information or platforms to vulnerable patients.

Virtual care service delivery planning should consider the variances in service uptake between different socioeconomic classes. Access to digital resources influences a person's capacity to access and utilise virtual care. Research has also documented that the digital divide in terms of access to digital devices and strong internet connectivity is significant among people with lower level of education and lesser income [80,81]. When engaging patients with virtual care services, consideration should be given as to whether patients have access to appropriate devices and a reliable internet connection. rpavirtual and other similar virtual care interventions should include in their referral process that patients require devices and internet connection to access services.

This review was subject to some limitations. There are several synonyms used to represent inequity issues in the literature. While we were broad in searching the literature, we may still have missed some articles utilising different terminology. However, we explored both the MeSH terms and keywords to address this. We also limited our searches to three major databases, and there could be additional relevant articles available in other databases. We searched for only the peer-reviewed articles and therefore might have missed some grey publications.

We restricted our searches to English literature only, and therefore, we could miss relevant articles that are written in a language other than English. Furthermore, we only searched for studies published in last decade (January 2010–January 2021); therefore, we could miss some articles published before 2010 and after 2021.

#### **5. Conclusions**

This review highlights that while there is potential for virtual care to improve health service delivery, particularly during the COVID-19 pandemic, there can be widespread inequities in access to and delivery of virtual care interventions. These inequities are based on sociodemographic characteristics of the participants, such as age, gender, and ethnicity as well as other factors, such as access to appropriate digital technology, digital and health literacy, cultural acceptability, and trust and perceived quality of care. This review has identified several promising practices, such as the inclusion of young and educated caregivers, providing culturally sensitive interventions, and improving digital health literacy among patients. These strategies can be adopted by rpavirtual and other virtual care interventions to ensure equity in access and delivery of virtual care services. Future research should focus on how these promising practices can be implemented in clinical settings.

**Supplementary Materials:** The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/ijerph19159411/s1, Table S1: Preferred Reporting Items for Systematic reviews and Meta-Analyses extension for Scoping Reviews (PRISMA-ScR) Checklist; Table S2: Characteristics of included studies.

**Author Contributions:** Conceptualization, S.K.M., M.S., B.H.-R., G.J. and F.H.; methodology, S.K.M., B.H.-R., G.J. and F.H.; formal analysis, S.K.M., F.R., G.J., K.P. and S.S.; writing—original draft preparation, S.K.M., K.P., S.S. and F.R.; writing—review and editing, M.S., G.J., B.H.-R. and F.H.; supervision, M.S., B.H.-R. and F.H.; project administration, S.K.M., M.S., B.H.-R., G.J., K.P. and F.H. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received funding from Sydney Local Health District.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


### *Article* **Chronic Kidney Disease and Nephrology Care in People Living with HIV in Central/Eastern Europe and Neighbouring Countries—Cross-Sectional Analysis from the ECEE Network**

**Bartłomiej Matłosz 1,\*, Agata Skrzat-Klapaczy ´nska 2, Sergii Antoniak 3, Tatevik Balayan 4, Josip Begovac 5, Gordana Dragovic 6, Denis Gusev 7, Djordje Jevtovic 8, David Jilich 9, Kerstin Aimla 10, Botond Lakatos 11, Raimonda Matulionyte 12, Aleksandr Panteleev 13, Antonios Papadopoulos 14, Nino Rukhadze 15, Dalibor Sedlácek ˇ 16, Milena Stevanovic 17, Anna Vassilenko 18, Antonija Verhaz 19, Nina Yancheva 20, Oleg Yurin 21, Andrzej Horban <sup>2</sup> and Justyna D. Kowalska <sup>2</sup>**


**Abstract:** Chronic kidney disease (CKD) is a significant cause of morbidity and mortality among patients infected with human immunodeficiency virus (HIV). The Central and East Europe (CEE) region consists of countries with highly diversified HIV epidemics, health care systems and socioeconomic status. The aim of the present study was to describe variations in CKD burden and care between countries. The Euroguidelines in the CEE Network Group includes 19 countries and was initiated to improve the standard of care for HIV infection in the region. Information on kidney care in HIV-positive patients was collected through online surveys sent to all members of the Network Group. Almost all centres use regular screening for CKD in all HIV (+) patients. Basic diagnostic tests for kidney function are available in the majority of centres. The most commonly used method

#### **Citation:** Matłosz, B.;

Skrzat-Klapaczy ´nska, A.; Antoniak, S.; Balayan, T.; Begovac, J.; Dragovic, G.; Gusev, D.; Jevtovic, D.; Jilich, D.; Aimla, K.; et al. Chronic Kidney Disease and Nephrology Care in People Living with HIV in Central/Eastern Europe and Neighbouring Countries—Cross-Sectional Analysis from the ECEE Network. *Int. J. Environ. Res. Public Health* **2022**, *19*, 12554. https://doi.org/10.3390/

Academic Editors: Jessica Sheringham and Sarah Sowden

ijerph191912554

Received: 2 August 2022 Accepted: 27 September 2022 Published: 1 October 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

for eGFR calculation is the Cockcroft–Gault equation. Nephrology consultation is available in all centres. The median frequency of CKD was 5% and the main cause was comorbidity. Haemodialysis was the only modality of treatment for kidney failure available in all ECEE countries. Only 39% of centres declared that all treatment options are available for HIV+ patients. The most commonly indicated barrier in kidney care was patients' noncompliance. In the CEE region, people living with HIV have full access to screening for kidney disease but there are important limitations in treatment. The choice of dialysis modality and access to kidney transplantation are limited. The main burden of kidney disease is unrelated to HIV infection. Patient care can be significantly improved by addressing noncompliance.

**Keywords:** HIV; chronic kidney disease; Central and Eastern Europe

#### **1. Introduction**

Chronic kidney disease is a significant cause of morbidity and mortality among patients infected with the human immunodeficiency virus (HIV). The prevalence around the world in the HIV-infected population varies but, in most reports, it is estimated to be between 4.7% and 9.7% [1–5]. With effective antiretroviral therapy (ARV), life expectancy in individuals with HIV has increased. As a consequence, the spectrum of kidney diseases in people living with HIV has broadened, including not only HIV-related problems or drug toxicity but also renal damage from chronic noncommunicable diseases.

The Central and East Europe (CEE) region with a population of about 300 million consists of several countries with highly diversified HIV epidemics, health care systems and socio-economic status [6]. The data on kidney disease and health care in the region used to be scarce. In many reports regarding kidney disease prevalence and kidney care, the countries from the region are labelled as 'no data available' [5,7]. The availability of the data from the region is, however, improving and in the latest issue of the International Society of Nephrology Global Kidney Health Atlas, many of the blind spots were filled with data [8]. Still, the lack of national registries in the region remains a substantial obstacle to collecting reliable and comparable data. The ISN Global Kidney Health Atlas indicates 18 regional registries as the data source for Western Europe (out of 25 countries) and only 2 regional registries for the Central and Eastern Europe region (out of 19 countries). The resources available for kidney care seem to improve across the CEE region, although they still tend to lag behind other parts of Europe, especially in terms of low or very low rates of kidney transplantation [9].

As the data on renal disease in the population of people living with HIV are even more limited than in the general population, the present study aimed to investigate the state and limitations of care for chronic kidney disease and end-stage kidney disease in countries represented in the Euroguidelines in Central and Eastern Europe (ECEE) Network Group.

#### **2. Materials and Methods**

Euroguidelines in Central and Eastern Europe Network Group was initiated in February 2016 to compare and improve the standard of care for HIV infection in the region. Information on kidney care in HIV-positive patients was collected through online surveys sent to all members of the ECEE Network Group in November 2018. Respondents were ECEE members from 20 countries in the region (Albania, Armenia, Belarus, Bosnia and Herzegovina, Bulgaria, Croatia, Czech Republic, Estonia, Georgia, Greece, Hungary, Lithuania, Macedonia, Poland, Republic of Moldova, Romania, Russia, Serbia, Slovenia, Turkey and Ukraine).

The collected data were exported to the R statistical software. All analyses were performed using R software (version 3.6.2). The responses were based on real-world data, including centres' own databases. As all of the responders were practitioners actively involved in patient care, data regarding the availability of diagnostic tools and treatment

relayed on their real experience. Some answers had to be analysed by country (e.g., coverage from public funding). In such cases, data received from the same country were used for validation purposes and then aggregated for by-country analysis.

Survey questions regarded institutional data, different aspects of nephrology care screening and diagnostic tests for kidney disease, availability of specialized nephrology care, burden and causes of kidney disease and guidelines employment. All survey questions can be found in Appendix A. The study did not include individual patient data and did not require ethical committee approval.

#### **3. Results**

The survey was sent to 43 members of the ECEE Network in 20 countries. We received 18 responses from 16 countries. According to the World Bank classification, among the responding countries, there were three lower–middle-income countries, five upper–middleincome countries and seven high–income countries. The vast majority of responding centres were described as treating PLWH as well as other infectious diseases (61%) and both hospital and outpatient-based (83%). The median population treated in a single centre was 1415 (IQR 600–2514) patients.

In most ECEE countries, the population of PLWH is relatively young with only four centres where patients > 65 years account for at least 10% of the population. In nine centres (50%), the proportion of elderly patients is less than 5%. We did not observe any significant correlation between age and the frequency of CKD. MSM was the most common route of HIV infection in 10 centres (55.5%). Insix6 centres (33.3%), IVDU is a common route of infection (more than 25% of infected patients). In 11 centres (61.1%), the majority of patients receive ARV therapy within one year after HIV infection diagnosis.

All but one centre use regular screening for kidney disease in all HIV (+) patients irrespective of ARV therapy usually within 3–6 month intervals. Among basic diagnostic tests, creatinine and abdominal ultrasound are available in all centres. In a few centres, urinalysis and albumin-to-creatinine ratio are not available (11.1%, *n* = 2 and 27.8%, *n* = 5, respectively). NMR, biopsy, scintigraphy and cystatin are rarely available (less than 50% of centres). Only a few centres did not use eGFR (22%, *n* = 4) nor albumin (28%, *n* = 5) for chronic kidney disease screening (Figure 1). The most commonly used method for eGFR calculation is still the Cockcroft–Gault Equation, but the CKD-EPI Equation is used with a similar frequency (38.9%, *n* = 7 and 33.3%, *n* = 6, respectively). MDRD formula is used only in three centres (16.7%). In four centres (22.2%), the eGFR method is not known.

**Figure 1.** Availability of diagnostic tests.

Specialized nephrology consultation is available in all centres. In three centres, a nephrologist is available on-site on a regular basis. In nine centres, consultation is available on call, and in five centres, patients are referred for consultation to the external institution. Nevertheless, the waiting time for consultation in any of the centres is not longer than 1 month and in half of the cases, it is up to one week. In the majority of centres, nephrology care is provided without any fee, although in seven centres, patients have to partly contribute to costs.

The main causes of chronic kidney disease are not related to HIV infection. The majority of responders (55%) answered that the first most common cause of chronic kidney disease is comorbidity (i.e., hypertension and diabetes). As the second most common causes of CKD, antiretrovirals' and other drugs' nephrotoxicity were indicated most commonly (33% and 28% responses, respectively). The most common and second most common causes of chronic kidney disease are depicted in (Figure 2A,B).

**Figure 2.** (**A**) The most common causes of CKD. (**B**) The second most common causes of CKD.

The median frequency of CKD was 5%, although there was significant variability between countries ranging from 0% to 20% of HIV-positive patients. With 37 reported cases out of over 58,000 individuals in care, end-stage renal disease is rare (0.06%) and only a few patients require RRT in most centres. Additionally, death resulting from end-stage renal disease is rare with no or single cases reported from almost all centres.

The only modality of treatment for ESRD completely covered from public funding (no matter of HIV status) in all ECEE countries is haemodialysis. Respondents from four countries declared that neither living nor deceased donor transplantation is founded on public funds. In five countries, the treatment of CKD complications and peritoneal dialysis is not covered. Only seven centres (38.9%) declared that all treatment options for ESRD (haemodialysis, peritoneal dialysis, and kidney transplantation) are available for HIV+ patients. In 11 centres (61.1%), the only treatment option was dialysis, and among them, in 6 centres (33.3%), only haemodialysis was possible.

In most centres, treatment of chronic kidney disease is the responsibility of a nephrologist or interdisciplinary team (72.2%). In five centres (27.8%), infectious disease specialists are primarily responsible for kidney care in people living with HIV. The most commonly indicated barrier in kidney care was patients' noncompliance (six centres). Other issues (i.e., health service availability, nephrologist availability, and distance from care point) were also common.

The most commonly used guidelines were EACS guidelines: eight centres (44.4%) use only EACS and seven centres (38.9%) both EACS and national ones. Guidelines adoption was usually described as moderate (61%) centres.

#### **4. Discussion**

The reported prevalence of CKD varied from 0 to 20% (median 5%). The highest rates (more than 10%) were observed in Hungary, Poland and Serbia. A very low prevalence of CKD (below 2%) was observed in Macedonia, Georgia, Estonia and Lithuania. In the general population, the prevalence of chronic kidney disease worldwide is 1.5–21% and in fact, it is highly variable in seemingly similar countries such as in Europe [7,10]. Significant differences were observed even within one country [11]. Aumann et al. showed that in different regions in Germany, the prevalence may be higher by a factor of 2. In the Pomerania region (SHIP-1 study) in Northeast Germany and the region of Augsburg in Southern Germany's Cooperative Health Research Study (KORA F4), the prevalence was 5.9% and 3.1% accordingly. The same is true for HIV-positive patients in Europe [2,12]. The population of people living with HIV across the ECEE countries network is widely variable. Our results show huge disparities in epidemics across quite a small geographical region. There are countries with a high prevalence of patients infected through intravenous drug use as well as countries with MSM population dominance. The prevalence of elderly patients in the HIV-infected population may vary from as high as 20% in some countries to almost no such patients, as reported in four countries where less than 1% of patients are aged 65 or more. This may result in different risk factors for chronic kidney disease in different countries. Although many authors attribute rising chronic kidney disease to the ageing HIV (+) population, we were not able to see any significant increase in the reported CKD prevalence in centres with a high proportion of elderly patients [13,14].

According to EACS guidelines, screening for renal disease (eGFR and urine protein) should be carried out in every HIV-positive patient at least once a year (with a wide range of every 3–12 months) [15]. In patients with risk factors for established CKD, the frequency of eGFR monitoring should be increased. Additionally, in patients with decreased eGFR or proteinuria, abdominal ultrasound should be performed. In this survey regarding kidney care clinical practice in people living with HIV, we observed general good availability of diagnostic tests and treatment in chronic kidney disease. Most of the kidney function tests were available in all centres and specialized nephrology care was provided without delay. It is worth mentioning that some centres still do not utilize urinary albumin/protein for screening for kidney disease which is required by EACS guidelines [15]. Poorer availability

of some specialized diagnostic tests such as nuclear magnetic resonance imaging or kidney scintigraphy should not be a problem as almost all centres reported the possibility of patients' referral to a nephrologist for specialized care.

The definition of chronic kidney disease is based on the widely accepted KDIGO classification of CKD [16]. For GFR estimation, EACS guidelines recommend the use of the CKD-EPI formula, although it indicates that the abbreviated Modification of Diet in Renal Disease (MDRD) or the Cockcroft–Gault (CG) equation may be used as an alternative. KDIGO guidelines state that eGFR should be calculated with the CKD-EPI formula and an alternative creatinine-based GFR-estimating equation is acceptable only if it has been shown to improve the accuracy of GFR estimates compared to the 2009 CKD-EPI creatinine equation [16]. Almost all centres employ regular universal screening for kidney injury at 3–6 month intervals, which is even more frequent than required by EACS guidelines. The vast majority of the centres (67%) still use, for the estimation of GFR, equations other than the primarily recommended CKD-EPI equation. This fact may potentially have clinical significance as the CKD-EPI creatinine formula was validated in the HIV (+) population in many studies and seems to outperform other formulas by means of accuracy [17,18].

HIV may cause kidney injury in several ways. The classic kidney disease of HIV infection, HIV-associated nephropathy (HIVAN), causes rapid kidney function deterioration and, if not treated, usually leads to end-stage renal disease. The incidence of HIV-associated nephropathy decreased with the use of highly active antiretroviral therapy [19]. Additionally, a number of immune complex kidney diseases have been reported in patients with HIV infection, including membranous nephropathy, membranoproliferative and mesangial proliferative glomerulonephritis, and "lupus-like" proliferative glomerulonephritis [20,21]. However, the ageing cohort of HIV-positive patients may be at increased risk for kidney disease unrelated to direct HIV injury. Coinfections and comorbid or treatment-related diabetes and hypertension may play an important role [22]. The increasing role of traditional risk factors for CKD seems to be supported by the results of our survey. More than half of the centres indicated that comorbidity was the most common cause of CKD. Other commonly reported causes were nephrotoxicity of drugs and illegal substances. Only one centre reported HIV nephropathy as the main cause of chronic kidney disease. As the second most common causes of chronic kidney disease, ARV nephrotoxicity and other drugs' or illegal substances' nephrotoxicity were reported in the majority of centres.

Even in centres with a high prevalence of chronic kidney disease, end-stage renal disease was not a common problem. Only single cases of end-stage renal disease were reported. Only one country reported a significant number of patients requiring renal replacement therapy but not treated. The possible reasons include not only resource shortage but also patients' noncompliance indicated by many participating centres as an important obstacle in providing care. Noncompliance was also described in previous studies as a serious problem with more than 50% of patients not attending the scheduled consultations [23]. No country reported differences in kidney care for HIV and the general population. Some centres indicated that kidney transplantation is founded on the public health insurance system but there is no possibility for HIV (+) individuals for transplantation. This may indicate that access to care, in reality, is not truly equal. Another possible explanation is that in many ECEE countries, kidney transplant programs have a very limited capacity [24]. Thus, the availability of transplantation may be limited in general in those countries despite public funding.

In general, access to nephrology care for people living with HIV is seemingly good. All centres employ regular renal function screening and in the vast majority of centres, nephrology consultation is possible without a delay. Nevertheless, our findings showed the lack of treatment options for ESRD in HIV-positive patients in a substantial proportion of Central and Eastern Europe Countries. In the ECEE region, one country in four has no public funding for kidney transplantation (which is also true for HIV-negative individuals). The only renal replacement therapy reimbursed in all ECEE countries is haemodialysis. The fact that treatment of CKD complications is also not covered by public health insurance

in many countries may negatively influence the patients' prognosis. Additionally, in many centres, chronic kidney disease treatment is solely the responsibility of the infectious disease specialists which is not optimal and is against current guidelines.

All centres indicate that there are barriers to kidney care access for HIV patients. The most commonly indicated obstacle was patients' noncompliance, although all the other answers (distance from care, nephrologist availability, and healthcare system access) were also commonly indicated.

There are some important limitations to be discussed. This was an online survey-based study where we preselected respondents based on our best knowledge of expertise and up-to-date acquaintance with epidemiological and clinical data in their centres. Secondly, the source of information on coinfection prevalence varied from personal communication to detailed epidemiological surveillance; thus, the weight of the data presented may vary significantly across countries.

#### **5. Conclusions**

In the ECEE region, people living with HIV have full access to screening for kidney disease. The screening might be improved by the employment of albuminuria screening in all centres. There are some important limitations in access to renal replacement therapy both regarding the choice of dialysis modality and kidney transplantation. It must be stated that those limitations are also true for the HIV (−) population. The main burden of kidney disease in the ECEE region is not directly related to HIV infection and treatment but comorbidity and patient care can be improved by addressing noncompliance.

**Author Contributions:** Conceptualization and methodology—J.D.K. and B.M.; formal analysis— B.M. and J.D.K.; data collection and curation—S.A., T.B., J.B., G.D., D.G., D.J. (Djordje Jevtovic), D.J. (David Jilich), K.A., B.L., R.M., A.P. (Aleksandr Panteleev), A.P. (Antonios Papadopoulos), N.R., D.S., M.S., A.V. (Anna Vassilenko), A.V. (Antonija Verhaz), N.Y., O.Y., J.D.K. and A.S.-K.; writing— B.M. and A.S.-K.; supervision—A.H.; project administration—A.S.-K.; funding acquisition—A.H. All authors have read and agreed to the published version of the manuscript.

**Funding:** The study has been funded by a research grant issued by the Research Development Foundation in Hospital for Infectious Diseases.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results.

#### **Appendix A**

Survey questionnaire. Access to nephrology care for HIV infected patients in CEE countries. INSTITUTIONAL DATA

1. Contact info:






#### NEPHROLOGY CARE

	- a. Yes, in my clinic daily or weekly [ ] b. Yes, in my clinic monthly [ ] c. Yes, in my clinic on call when demanded [ ]
	- d. Yes, referral to the other center [ ]
		- e. No, nephrology consultation is not available in my clinic for HIV+ patients [ ]

\_\_\_\_\_\_\_\_\_ days/weeks/months/years

### 15. Is specialized nephrology care in your country:

	- a. Yes, it is the same [ ]
	- b. No, it is different (please describe \_\_\_\_\_\_\_\_\_\_\_\_\_) [ ]

#### CHRONIC KIDNEY DISEASE

	- a. HIV nephropathy,
	- b. glomerular disease,
	- c. ARV nephrotoxicity,
	- d. other drugs or illegal substances nephrotoxicity,
	- e. comorbidities (i.e., diabetes mellitus, hypertension),
	- f. unknown,
	- g. other (please specify) \_\_\_\_\_\_\_\_\_\_
		- most common \_\_\_\_\_\_\_\_\_\_
		- second most common \_\_\_\_\_\_\_\_\_\_
	- a. On dialysis \_\_\_\_\_\_\_\_\_\_

\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_




22. What options are routinely available for patients requiring renal replacement therapy:



	- a. Yes, it is the same [ ]
	- b. No, it is different (please describe \_\_\_\_\_\_\_\_\_\_\_\_\_) [ ]
	- a. Dialysis
		- i. Hemodialysis [ ]
		- ii. Peritoneal dialysis [ ]
	- b. Kidney transplantation
		- i. living donor [ ]
		- ii. deceased donor [ ]
	- c. Management of CKD complications (i.e., anemia, bone disease) [ ]
	- a. Geography (distance from care)
	- b. Nephrologist availability
	- c. Patient non-compliance
	- d. Healthcare system access
	- e. Other (please specify) \_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_

#### GUIDELINES

26. Do you have national guidelines on HIV care?


27. When were your national/local guidelines updated?

#### year \_\_\_\_\_\_\_\_\_

28. Does your national guidelines cover kidney disease management in HIV population?


29. What guidelines regarding kidney disease do you use at your centre?

	- a. Not optimal
	- b. Moderate
	- c. Optimal

If you have any other comments you can place it here:

#### **References**

1. Mocroft, A.; Kirk, O.; Gatell, J.; Reiss, P.; Gargalianos, P.; Zilmer, K.; Beniowski, M.; Viard, J.-P.; Staszewski, S.; Lundgren, J.D. Chronic renal failure among HIV-1-infected patients. *AIDS* **2007**, *21*, 1119–1127. [CrossRef] [PubMed]

\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_


### *Article* **Serious Injury in Metropolitan and Regional Victoria: Exploring Travel to Treatment and Utilisation of Post-Discharge Health Services by Injury Type**

**Jemma Keeves 1,2,\*, Belinda Gabbe 2, Sarah Arnup 2, Christina Ekegren <sup>3</sup> and Ben Beck <sup>2</sup>**


**\*** Correspondence: jemma.keeves@monash.edu

**Abstract:** This study aimed to describe regional variations in service use and distance travelled to post-discharge health services in the first three years following hospital discharge for people with transport-related orthopaedic, brain, and spinal cord injuries. Using linked data from the Victorian State Trauma Registry (VSTR) and Transport Accident Commission (TAC), we identified 1597 people who had sustained transport-related orthopaedic, brain, or spinal cord injuries between 2006 and 2016 that met the study inclusion criteria. The adjusted odds of GP service use for regional participants were 76% higher than for metropolitan participants in the orthopaedic and traumatic brain injury (TBI) groups. People with spinal cord injury (SCI) living in regional areas had 72% lower adjusted odds of accessing mental health, 76% lower adjusted odds of accessing OT services, and 82% lower adjusted odds of accessing physical therapies compared with people living in major cities. People with a TBI living in regional areas on average travelled significantly further to access all post-discharge health services compared with people with TBI in major cities. For visits to medical services, the median trip distance for regional participants was 76.61 km (95%CI: 16.01–132.21) for orthopaedic injuries, 104.05 km (95% CI: 51.55–182.78) for TBI, and 68.70 km (95%CI: 8.34–139.84) for SCI. Disparities in service use and distance travelled to health services exist between metropolitan Melbourne and regional Victoria following serious injury.

**Keywords:** serious injury; traumatic brain injury; orthopaedic injury; spinal cord injury; road trauma; access to healthcare; healthcare utilisation; geography

#### **1. Introduction**

Transport-related injuries are expected to become the third leading cause of disability worldwide by 2030 [1]. Despite advances in trauma care, people with orthopaedic injury, traumatic brain injury (TBI), and spinal cord injury (SCI) continue to experience long-term physical disability, psychological dysfunction, and interference from pain [2–4]. There is a need to understand whether long-term outcomes for people with serious transportrelated injury can be improved through a coordinated and revised approach to postdischarge healthcare.

Urban and regional disparities in access to care exist, with people living in regional areas travelling further to access post-discharge healthcare after major trauma [5]. Both people with serious injury and health professionals have reported limited availability and difficulties accessing necessary care as barriers to health service delivery following injury, particularly for people living in regional areas [6–8]. It is unclear whether these barriers to post-discharge care are more significant for people in regional areas as a result of regionalised trauma system design, which centralises higher-level trauma centres in inner metropolitan areas.

**Citation:** Keeves, J.; Gabbe, B.; Arnup, S.; Ekegren, C.; Beck, B. Serious Injury in Metropolitan and Regional Victoria: Exploring Travel to Treatment and Utilisation of Post-Discharge Health Services by Injury Type. *Int. J. Environ. Res. Public Health* **2022**, *19*, 14063. https://doi.org/10.3390/ ijerph192114063

Academic Editors: Paul B. Tchounwou, Jessica Sheringham and Sarah Sowden

Received: 6 September 2022 Accepted: 21 October 2022 Published: 28 October 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

<sup>1</sup> Department of Physiotherapy, Epworth Hospital, Melbourne 3122, Australia

Despite survivors of serious injury having long-term and complex healthcare needs, the level of specialised care provided beyond hospital discharge varies depending on the type of injury [4,5,9]. People with TBI and SCI are more likely to receive rehabilitation from specialised services due to the complexity of these injuries [10]. After an orthopaedic injury however, there is no clear pathway for rehabilitation once discharged from a major trauma centre [10]. Given the high prevalence of disability amongst trauma survivors, both with and without serious neurotrauma, consideration for the whole pathway of trauma care from acute management to specialised rehabilitation and community care is pertinent [10].

This novel study is the first to use geospatial analysis to clearly quantify the differences in travel to services and service use for people in different geographic areas by type of injury. The aim of this work was to understand how different injury populations use post-discharge health services across regional and metropolitan areas and explore the distances travelled to health services in the first three years following hospital discharge. Improving our understanding of post-discharge service utilisation is an important step in ensuring necessary services are accessible and available for people with transport-related serious injury.

#### **2. Materials and Methods**

#### *2.1. Study Design*

Our registry-based cohort study used linked data from the Victorian State Trauma Registry (VSTR) and Transport Accident Commission (TAC). Our study follows the Strengthening of Reporting of Observational Studies in Epidemiology checklist, see Appendix A [11].

Victoria is the second most populous state of Australia with a population of 6.46 million people, including over 2 million people residing outside the Greater Melbourne region [12]. Victoria has an inclusive trauma system consisting of two adults, and one paediatric, major trauma centres, which are located in metropolitan Melbourne.

The population-based VSTR collects data about all people with major trauma in Victoria, with major trauma defined as: (1) death due to injury; (2) an injury severity score (ISS; based on the abbreviated injury scale (AIS) 2005 version, 2008 update) >12; (3) admission to an intensive care unit >24 h; (4) or an injury requiring urgent surgery [13]. The registry has an opt-out rate <1% and includes data on prehospital care, pre-existing health conditions, injury characteristics and complications, and discharge information [13].

The TAC is Victoria's no-fault third-party insurer for people who have sustained a transport-related injury, covering medical treatment, rehabilitation, support services, and financial assistance. People are covered by the TAC if their injuries are sustained as a result of driving a car, motorcycle, bus, train, or tram. Cyclists injured in a collision with a moving or stationary motor vehicle (after 9 July 2014) are also covered by the TAC. Pedestrians are covered by the TAC when their injuries arise as a direct result of impact with a motor vehicle, motorcycle, train, or tram. Full details of eligible claimants and expenses covered by the TAC are outlined in the Transport Accident Act 1986 [14]. The TAC collect data pertaining to an individual claim, including detailed information regarding the post-discharge health services paid for by the TAC. These data include details of the date and type of service, service description, and where the service provider is located. The TAC provides these data to the VSTR, linked by claim number. A standardised and secure process is followed to ensure that no patient-level data are provided to the TAC by the VSTR.

#### *2.2. Participants*

Victorians who sustained major trauma from a transport-related event between 1 January 2006 and 31 December 2016 with a TAC compensation claim were identified within the VSTR. People were included if they sustained isolated orthopaedic injuries; a moderate to severe TBI or SCI; were aged 18 years or older at the time of injury; had three-years of TAC claims data; and resided in Victoria with a known residential address (Figure 1). The orthopaedic group consisted of people who had sustained an extremity injury with AIS score >1 and/or spine injury with AIS two or three and no other injury with AIS >1 [15]. Traumatic brain injury (TBI) cases were considered to be moderate or severe if they had a head injury with an AIS severity score >2 and the first recorded Glasgow coma scale (GCS) score <13, with or without other system injuries [15]. Mild traumatic brain injuries are not captured by the VSTR unless sustained with other system injures so were excluded from this study. Spinal cord injury was defined as an injury to the spine with an AIS severity score >3, with or without other injuries [15].

**Figure 1.** Flow diagram of inclusion criteria.

#### *2.3. Variables*

The three key outcomes of interest in this study were: service use, the number of trips per person and distance travelled to health services in the first three years following hospital discharge. Service use was defined as the percentage of participants who used a health service at some point within the study period. The number of trips per person refers to the number of times a health service was visited by service users. Distance travelled was the median trip distance per person from their residential location to the provider location, measured in kilometres.

Health services were categorised as: General Practitioner (GP), other medical professionals (e.g., neurologists, pathologists, psychiatrists, surgeons), mental health services (psychology, social work and case management), physical therapies (physiotherapy, exercise physiology, and hydrotherapy), and occupational therapy (OT). Speech pathology was excluded as this service was used almost exclusively by TBI participants.

#### *2.4. Data Measurement*

Demographic information, pre-existing health conditions, injury diagnosis and severity, and hospital length of stay and discharge status were extracted from the registry. Data relating to a TAC claim, client address, and service provider locations were provided by the TAC for all services funded between 1 January 2006 and 31 December 2019.

Each participant's residential address at the date of hospital discharge was mapped by their local government area (LGA) to the Accessibility/Remoteness Index of Australia 2016 (ARIA+) and categorised into major city, inner regional, or outer regional [16]. For analysis, the metropolitan group consisted of people living in 'Major Cities' and the regional group as people living in 'Inner Regional' or 'Outer Regional' areas (Figure 2). No participants were residing in 'Remote' or 'Very Remote' areas by the ARIA+ classification index. Socioe-

conomic status was categorised using The Index of Relative Socioeconomic Advantage and Disadvantage (IRSAD) according to LGA [17].

**Figure 2.** Boundaries for major cities (Greater Melbourne), dark grey, and regional. Victoria, light grey.

Geographic coordinates for each participant's address and service provider address were compiled through geocoding in RStudio version 3.5.1, and a random sample of 100 were manually checked using Google Maps [18]. Any incomplete provider address locations were entered manually using Google Maps to obtain coordinates. For a single provider it was possible for multiple locations to exist. We mapped the travel distances for all locations using Here Routing API (https://developer.here.com, accessed on 17 Febrary 2020) and used the shortest distance, assuming that people would visit the closest provider location to their homes.

#### *2.5. Statistical Methods*

Summary statistics were used to describe demographic information, injury characteristics, and service use outcomes. Medians and interquartile range were reported for skewed categorical variables and frequencies and percentages for continuous variables.

Regression models were used to provide estimates of the association between the outcomes of interest and region by injury type. Models were run for each outcome using region as an interaction term with injury type. All models were adjusted for the covariates of age group, sex, Charlson comorbidity index, ISS, and IRSAD based on factors known to impact healthcare utilization [19,20]. Multivariable logistic regression was used for service use (yes/no), and negative binomial regression was used for the number of trips per person, while a general estimating equation (GEE) was used to model distance travelled to services used. For the GEE, a Gaussian model and identity link was used, and an exchangeable correlation was assumed between trips to the same service within each individual. Adjusted odds ratios (OR) and incidence rate ratios (IRR), and the corresponding 95% confidence intervals were calculated for the logistic and negative binomial regression models, respectively. As distance travelled was positively skewed, the data were log transformed before modelling. Model fit was evaluated for concordance and discrimination using residual plots [21]. All analyses were completed in Stata Version 16.0 with the exception of the geospatial analyses which were conducted using RStudio version 3.5.1 [18].

#### **3. Results**

There were 9365 cases of transport-related major trauma identified from the VSTR; 17.1% (*n* = 1597) were eligible for this study (Figure 1). The characteristics of included participants for each injury group are presented in Table 1.


*n* = 19 missing [1]. *n* = 34 other or missing [2]. LOS = length of stay; IQR = interquartile range; CCI = Charlson comorbidity index, ARIA+ = Accessibility/Remoteness Index of Australia; IRSAD = Index of Relative Socioeconomic Advantage and Disadvantage.

Across all injury groups, most participants were men and the median age was 33 years (IQR 23–48). Thirty-five percent of participants resided in regional areas. Most participants were injured in motor vehicle or motorcycle crashes. The SCI group had the longest length of acute hospital stay. In the first three years following hospital discharge, the 1597 participants visited health services 159,090 times for GP services, other medical appointments, mental health services, physical therapies, and OT (Table 2).

Figure 3 provides a summary of the key findings from the multivariable regression analysis for service use, number of trips, and distances travelled to services. More specific results from the models for each outcome and injury group are reported in each section below.


**Table 2.** All services used by participants in the first three years post-discharge by injury type.




**Figure 3.** Summary of key findings for regional participants compared with participants in major cities.

#### *3.1. Service Use*

The adjusted proportions of people using GP services were higher for regional participants in all injury groups (Figure 4). Across all other services, the adjusted proportions for service use were greater for people living in major cities compared with people living in regional areas, except for people with TBI accessing mental health services.

**Figure 4.** Adjusted proportion of service use by injury group and region.

In the orthopaedic and TBI groups, participants in regional areas, compared with major cities, had 76% higher adjusted odds of seeing a GP but 56% and 57% lower adjusted odds of attending other types of medical services, respectively (Table 3). In the orthopaedic group, participants in regional areas, compared with major cities, had 37% lower adjusted odds of attending mental health services and 45% lower adjusted odds of attending occupational therapy services. In the SCI group, participants in regional areas, compared with major cities, had 72% lower odds of accessing mental health, 82% lower adjusted odds of accessing physical therapies, and 76% lower adjusted odds of accessing OT services (Table 3).

**Table 3.** Regional variation in service use and number of trips per person in the first three years following hospital discharge determined by multivariable regression analysis.


**Table 3.** *Cont.*


\* *p* value for the logistic regression analysis of service use. † *p* value for negative binomial regression analysis of number of trips, per person, for participants who used that service. *p* values in bold type are significant. IQR = interquartile range, OR = odds ratio, IRR = incidence rate ratio.

#### *3.2. Number of Trips*

For all injuries and service types, people in regional areas used fewer services than people residing in major cities after adjusting for covariates (Figure 5). Physical therapies were the most commonly used service across all injury groups. In the orthopaedic group, the mean number of trips for participants in regional areas, compared with major cities, was 29% lower for medical services and 24% lower for physical therapy services. In the TBI group, the mean number of trips for participants in regional areas, compared with major cities, was 35% lower for medical services. In the SCI group, the mean number of trips for participants in regional areas, compared with major cities, was 35% lower for GPs, 37% lower for physical therapy and 45% lower for OT services (Table 3).

**Figure 5.** Adjusted number of trips to services in the first three years post-discharge by service type and injury group.

#### *3.3. Distance Travelled*

In the TBI group, participants in regional areas travelled significantly further to access all post-discharge health services compared with participants in major cities (Figure 6). In the SCI group, however, participants in regional areas travelled further only to attend medical services (RGM 2.66, 95%CI 1.63–4.36) (Figure 6). In the orthopaedic group, participants in regional areas travelled 1.4 times further to see a GP (95%CI 1.06–1.88), 2.26 times further to attend other medical services (95%CI 1.76–2.89), and 1.7 times further to OT services (95%CI 1.06–2.62) compared with participants in major cities.

**Figure 6.** Ratio of geometric means for distance travelled by people in regional areas compared with major cities by injury group and service type.

For visits to medical services, the median trip distances for participants in regional areas with any injury type ranged from 68.70 km (95%CI: 8.34–139.84) to 104.05 km (95% CI: 51.55–182.78) (Figure 7). Comparatively, the median trip distances for participants in major cities with any injury type ranged from 9.44 km (95%CI 4.92–23.05) to 13.50 km (95% CI 6.65–25.59) (Figure 7).

**Figure 7.** Median and IQR of raw distances travelled to healthcare by service type and injury group.

#### **4. Discussion**

In this study, we compared health service usage and distances travelled by people with transport-related orthopaedic, brain and spinal cord injuries across regional and metropolitan Victoria in the first three years following hospital discharge. For most services and injury types, people in regional areas used fewer services but travelled further to access them than people in metropolitan areas. People with orthopaedic injuries and TBI in regional areas had greater odds of seeing a GP compared with their metropolitan counterparts. This research provides an important contribution to our understanding of how geography impacts healthcare utilisation following major trauma.

We found that regional participants with orthopaedic injuries and TBI had greater odds of attending GP services than metropolitan participants, despite having to travel further. This may be explained by people in metropolitan areas living closer to trauma centres with better access to specialised rehabilitation providers, therefore being less reliant on their local GPs [23–25]. Following major trauma, GPs play a critical role in providing ongoing community support, monitoring for secondary complications of injury and psychosocial issues, and assisting in the patient's return to work [26]. For people living in metropolitan areas, it is possible that these issues may be monitored by a specialised rehabilitation team, including allied health and specialist physicians. Our findings highlight the importance of regional-based GPs having adequate knowledge of injury complications and a network of specialists that may be able to carry out shared virtual consultations to ensure timely and effective management closer to home [27].

Consistent with previous research, our findings suggest that people with serious injuries living in regional areas use fewer health services than their metropolitan counterparts [28–31]. Having to travel further to access healthcare for people in regional areas may limit accessibility [5,8,24]. Compounding the challenge of distance, transportation difficulties [6,8,29,32] and a limited availability of skilled providers [7,33] have been reported as barriers to accessing necessary services for people with orthopaedic injuries, TBI, and SCI, particularly for those in regional areas. A key consequence of reduced service use is that people with serious injuries living in regional areas often report higher levels of unmet care needs [25,30,34–36]. Ensuring the availability of local infrastructure or alternate service delivery methods is essential for people with serious injury due to the chronicity of the condition [8,34,37].

In this study, we found that for all injury types, people living in regional areas travelled further than people in metropolitan areas to access all services. However, after adjusting for covariates, our findings were more nuanced. People living in regional areas with TBI travelled significantly further to all health services than those in metropolitan areas, whereas for people with SCI, a significant difference was only found for travel to medical services, which was based on region. Due to the complexity and long-term issues associated with SCI, people with SCI may choose to live in areas where they can access necessary services [24]. In comparison, given the varying degree of severity of TBI, some people with TBI may place less importance on ease of service access and availability when deciding where they want to reside. These novel findings reinforce the importance of specialised telehealth services and outreach clinics for people in regional areas with TBI to reduce travel burden and ensure access to adequately skilled healthcare services.

In addition to considering alternate service delivery modes, at a systems level, this research contributes a new perspective for post-discharge care coordination for people with serious injury. Policy makers involved in the planning of healthcare pathways across the continuum of care should consider the extra distances travelled by people in regional areas and possible travel burden, which may impact post-discharge healthcare utilisation. Further research to understand patterns of health service utilisation for other groups at risk of health inequities, such as older adults, Aboriginal and Torres Strait Islander People, and culturally and linguistically diverse communities, particularly in regional areas, will further aid in improving healthcare planning.

#### *Study Limitations*

This population-based cohort study provides novel insights into geographic variations in healthcare use following transport-related orthopaedic, brain, and spinal cord injury. However, a limitation of this work was that due to multiple service provider locations being provided, we assumed that an individual attended the closest facility to their home and used the shortest trip distance. This study also only included services that were centrebased; therefore, for people with TBI and SCI, who are likely to have received services in the community or at home, the number of services used may be underrepresented. This also includes care from the Spinal Community Integration Service, a Victorian program that provides people with SCI assistance with returning home and participating in their communities in the first 12 months following discharge. Due to the nature of how these services are billed to the TAC, it was not possible to ascertain specific details of what services were provided on exact dates and at specific locations. However, as this was the same for both regional and metropolitan participants, this is unlikely to have impacted the regional variation within groups. It was also assumed that participants all travelled by car to attend services. Due to the reimbursement available for taxi travel and motorised travel expenses for TAC patients, it is most likely that participants would choose one of these options over human-powered or public transport.

#### **5. Conclusions**

Health service use following traumatic orthopaedic, brain, and spinal cord injury is complex and continues for years following the initial injury. This research has identified disparities in service use and distances travelled to health services across metropolitan and regional Victoria following serious injury. With people in regional areas using fewer services, except for GPs, and attending these services less often, there is a risk of unmet service needs for these individuals. An increased travel distance to services is one factor that may be contributing to the inequality in access to healthcare in regional areas compared with metropolitan areas. These findings reinforce the need for a review of how specialised rehabilitation services are delivered to people residing in regional areas following major trauma and whether access to post-discharge services is available to everyone long-term, regardless of where they reside. Further research exploring whether there is an association between service use, distance travelled, and health outcomes is necessary to ensure postdischarge care is optimised for people with serious injuries.

**Author Contributions:** Conceptualization, J.K., B.G., C.E. and B.B.; methodology, J.K., B.G., S.A. and B.B.; validation, J.K., B.G. and B.B.; formal analysis, J.K. and S.A.; resources, J.K.; data curation, J.K.; writing—original draft preparation, J.K.; writing—review and editing, J.K., B.G., S.A., C.E. and B.B.; visualization, J.K.; funding acquisition, J.K. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded in part by an Epworth Medical Foundation, Translational Research Grant. JK was supported by an Australian Government Research Training Program Scholarship. CE was supported by a National Health and Medical Research Council of Australia (NHMRC) Early Career Fellowship (1106633). BG was supported by an Australian Research Council Future Fellowship (FT170100048). BB was supported by an Australian Research Council Discovery Early Career Researcher Award Fellowship (DE180100825).

**Institutional Review Board Statement:** The study was conducted in accordance with the Declaration of Helsinki, and approved by the Monash University Human Research Ethics Committee (HREC) (Project ID 18433, 10 April 2019). The VSTR has ethics approval from the Department of Health and Human Services HREC (reference\_11/14), Monash University and all trauma receiving hospitals.

**Informed Consent Statement:** Informed consent was obtained from all subjects involved in the study.

**Data Availability Statement:** Not applicable.

**Acknowledgments:** We thank the Victorian State Trauma Outcome Registry and Monitoring (VS-TORM) group for providing VSTR data and Sue McLellan for their assistance with data preparation. We also express our appreciation to the TAC for the provision of data and their assistance in data preparation.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results.

#### **Appendix A**

**Table A1.** Strengthening The Reporting of Observational Studies in Epidemiology (STROBE) statement checklist.



#### **References**


### *Review* **A Scoping Review of Approaches to Improving Quality of Data Relating to Health Inequalities**

**Sowmiya Moorthie 1,\*, Vicki Peacey 2, Sian Evans 3, Veronica Phillips 4, Andres Roman-Urrestarazu 1, Carol Brayne <sup>1</sup> and Louise Lafortune <sup>1</sup>**


**Abstract:** Identifying and monitoring of health inequalities requires good-quality data. The aim of this work is to systematically review the evidence base on approaches taken within the healthcare context to improve the quality of data for the identification and monitoring of health inequalities and describe the evidence base on the effectiveness of such approaches or recommendations. Peerreviewed scientific journal publications, as well as grey literature, were included in this review if they described approaches and/or made recommendations to improve data quality relating to the identification and monitoring of health inequalities. A thematic analysis was undertaken of included papers to identify themes, and a narrative synthesis approach was used to summarise findings. Fiftyseven papers were included describing a variety of approaches. These approaches were grouped under four themes: policy and legislation, wider actions that enable implementation of policies, data collection instruments and systems, and methodological approaches. Our findings indicate that a variety of mechanisms can be used to improve the quality of data on health inequalities at different stages (prior to, during, and after data collection). These findings can inform us of actions that can be taken by those working in local health and care services on approaches to improving the quality of data on health inequalities.

**Keywords:** health inequalities; health disparities; data quality; public health

#### **1. Introduction**

Health inequalities are often defined as "differences in health across the population and between different groups" [1]. The study of health inequalities aims to better understand factors that contribute to unfair differences in the status of people's health to address them and achieve fairer and more inclusive health care. Inequalities in health can arise because of differences in the care that people receive and the opportunities they have to lead healthy lives, including differences in health status (e.g., life expectancy), quality and experience of care, and wider determinants of health [1].

Data analysis to improve understanding of health gaps is an important exercise that contributes to an aspiration for fair and inclusive health. Good data is vital for understanding inequality in health service provision and health outcomes, and necessary for informing and evaluating attempts to improve care or reduce inequality. In the United Kingdom, health inequalities are identified by analysing data across socio-economic factors, geography, and specific characteristics including those protected in law such as sex, ethnicity or disability, and socially excluded groups. However, the quality of data underpinning these analyses can be improved [2–4]. Good-quality data are data that are fit for the purpose; therefore, criteria on what constitutes "good" can vary. Dimensions such as

**Citation:** Moorthie, S.; Peacey, V.; Evans, S.; Phillips, V.;

Roman-Urrestarazu, A.; Brayne, C.; Lafortune, L. A Scoping Review of Approaches to Improving Quality of Data Relating to Health Inequalities. *Int. J. Environ. Res. Public Health* **2022**, *19*, 15874. https://doi.org/10.3390/ ijerph192315874

Academic Editors: Jessica Sheringham and Sarah Sowden

Received: 17 October 2022 Accepted: 25 November 2022 Published: 29 November 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

completeness, accuracy, relevance, availability, and timeliness of data can be assessed to determine data quality [5].

Several policy reports released in the UK have highlighted the importance of improving the quality of data used for the identification and monitoring of health inequalities [6,7]. In particular, identifying and reducing inequalities linked to ethnicity are a key part of expectations in terms of improving NHS services [8]. Recommendations from these reports include ensuring consistent reporting and analysis of data on ethnicity, health, and health care and documenting and evaluating best practices [6]. The need for better data coverage across all age groups and allowing self-identification, particularly around ethnicity, has also been recommended [9].

Health inequalities have been increasing in England over the past 10 years and the COVID-19 pandemic has starkly highlighted inequalities that exist [10,11]. The pandemic has also demonstrated that collecting data at speed and using healthcare data in flexible and creative ways is possible [12]. This has renewed emphasis on the need for action to address inequalities at both national and system levels. This includes initiatives to improve data and make better use of data to address health inequalities [13,14]. A comprehensive understanding of the evidence base on how data quality can be improved and what has been shown to work is essential to inform the myriad of initiatives in the UK to address health inequalities.

The aim of this work was to identify and review the evidence base on approaches taken within the healthcare context to improve the quality of the data used for the identification and monitoring of health inequalities. The specific objectives were to describe the approaches that have been used or recommended to improve the quality (availability, completeness, accuracy, relevance, and timeliness) of data for identification of health inequalities, to describe the approaches that have been used or recommended to improve the quality of data for monitoring changes in health inequalities, and to describe the evidence base for the effectiveness of such approaches or recommendations.

#### **2. Methods**

#### *2.1. Article Identification and Selection*

A systematic literature search was conducted in the databases Medline (via Ovid), Embase (via Ovid), Global Health (via Ebscohost), Cinahl (via Ebschohost), and Web of Science (Core Collection) in September 2021 using search terms relating to data, data quality, and specific terms such as protected characteristics and tailoring them for each database (detailed search terms in Supplementary Materials). Search terms were deliberately broad to maximise the identification of relevant work, as more specific preliminary searches did not identify papers that were already known to be relevant. The searches were not date-limited initially. However, because legislation and guidance in the United Kingdom around recording data on health inequality characteristics has changed considerably in the last 10 years, we subsequently discarded reports published prior to January 2010. Papers that were published after 2010 but reported on data collected before 2010 were included. We undertook citation searching to identify other sources of information not identified in the database searches. In addition, a grey literature search was conducted using the advanced search function in Google. The search terms 'improving data quality health inequalities' were used; searches were restricted to .pdf file types. The first five pages of the search were examined for any documents that could be included. All results were limited to the English language.

#### *2.2. Inclusion and Exclusion Criteria*

Protocols for scoping reviews are not eligible for publication in PROSPERO; however, we have presented our findings as much as possible according to PRISMA guidelines [15] (Supplementary Materials). Peer-reviewed scientific journal publications, as well as grey literature, were included in this review if they described mechanisms to improve data quality relating to the identification and monitoring of health inequalities. Work that focused solely on improving the quality of data on health outcomes was not included, nor was work that simply evaluated data quality rather than presented attempts to improve data quality.

Two reviewers (SM and ARU) independently carried out primary screening of titles and abstracts to identify articles eligible for inclusion based on our eligibility criteria. A third reviewer (LL) reviewed all articles selected for inclusion and those marked as unsure and resolved any disagreements between the reviewers. The decision made by the third reviewer was final for inclusion or exclusion. Two reviewers (LL and SM) screened full-text articles for further assessment of eligibility for inclusion in this review. Discussion was undertaken to resolve any discrepancies. Figure 1 shows the search and selection outcomes for each stage of the review process.

**Figure 1.** Flow diagram of included and excluded studies.

#### *2.3. Data Extraction and Synthesis*

Data extraction was carried out using an iterative process. One author (SM) read all papers eligible for inclusion, grouped them into broad categories based on the main type of data that was discussed (ethnicity, gender/sexual orientation, social determinants, general, or other), and extracted relevant text from sections of the papers that provided recommendations, methods, or approaches to improving data quality. Papers were categorised as "general" if they did not specify a particular inequality dimension or were across several dimensions. All the papers were subsequently read by at least one other author (VP, SE, or LL) to confirm and supplement the extraction before coding and to ensure quality and consistency. Any discrepancies in data extraction were resolved through discussion. A thematic analysis was then conducted by one reviewer (SM) to identify themes, which were then summarised narratively and validated by another reviewer (LL). There was heterogeneity in the included reports in terms of the subject matter and approaches used. This precluded us from using traditional quality-assurance measures for critiques of the papers.

#### **3. Results**

The initial database search revealed 21,788 records. Following automatic de-duplication and removal of articles published pre-2010, 7830 articles were identified for primary screening. A total of 110 articles met the eligibility criteria for retrieving full texts after primary screening. A further 27 reports were identified by the grey literature search. Seventy-nine studies were excluded following assessment of full texts. The main reason for exclusion was a lack of discussion on mechanisms to improve data quality. A total of 57 publications were included in the review. Table 1 provides a summary of the characteristics of these reports. Most were peer-reviewed publications (*n* = 49) with the remainder being grey literature (*n* = 8). Many were reporting on data related to the dimension of ethnicity (*n* = 31) or were more general across indicators related to health inequalities (*n =* 15). A smaller number were identified that were focused on dimensions of sexual orientation and gender (*n* = 6), or on specific areas such as infectious diseases, learning disabilities, or cardiovascular care. Most were from the US or UK. None of the studies identified were high on the traditional hierarchy of evidence, and in most cases the approaches that were used for improving data quality had several elements that could not be disaggregated.


**Table 1.** Included studies and

characteristics.




#### *3.1. Distal Initiatives*

The mechanisms and approaches that were upstream of data collection and analysis, but which impacted on these, were grouped under the theme "distal initiatives". A total of 26 reports stated that policy and legislative imperatives such as mandating data collection led to improvements and consistency in data quality (Table 1). This is through making it a priority and incentivising data collection and leading to the creation of data systems that facilitate such efforts [4,53,58]. Reports also evidenced how data collection had improved since the introduction of mandates and the prioritisation of ethnicity data collection [4,19,31,42,43,45,47,58,65]. In the UK, the Equality Act 2010 and incentivisation under the Quality and Outcomes Framework (QOF) had a significant impact on the completeness of ethnicity data [45,47]. Mathur et al. (2014) [47] describe how the proportion of patients with a valid self-reported ethnicity record changed over time (1995 to 2011) in English hospital data and GP data (via the Clinical Practice Research Datalink, which covered 6% of all GP practices in 2012). The proportion of people with a usable ethnicity recording in Hospital Episode Statistics (HES) inpatient data jumped from 50% to just under 70% in one year between 2000 and 2001. Between 2008 and 2011, the proportion with a usable record also changed from around 20% to around 50% in the HES A&E and outpatient data. The authors do not discuss what lay behind the improvement in HES data quality. Collection of sexual identity, gender, and behaviour, whilst lagging behind, have also been impacted by legislation that is incentivising data collection [33,62]. Furthermore, given the sensitive and private nature of information such as ethnicity, disability, gender, and sexual orientation, legal safeguards to ensure nondiscrimination on the basis of this information are also important factors that impact on data collection efforts [42,45].

#### *3.2. Wider Actions to Enable Improvements in Data Quality*

While mandating data collection leads to improvements in data quality, it needs to be supported by wider actions to enable organisations to put in place mechanisms to improve data quality at source [4,23,31,32]. Of the included reports, 38 provided evidence that achieving senior-level buy-in [4,34,42,45,65], the development of staff training programmes [19,20,22,24,25,27,29,31,32,35–37,49,54,58,61,62], guidance on how to use data [19,29,34,37], engagement activities with citizens, patients, and communities [17,25,29,49,56,58,65], and training on analysis of source data all contribute to efforts to improve data quality [19,20,24,25,27,29,31,32,35,36,54,58,61,62].

Senior-level buy-in is needed to prioritise data collection and put in place systems, such as IT infrastructure, to enable data collection, as well as utilisation of the data for service improvement. Davidson et al. (2021) report that obtaining executive-level buy-in was crucial for recording and improving ethnicity data collection in NHS Lothian [34]. Reports have shown that this can be achieved by demonstrating the value of data collection and analysis [19]. Using the data to demonstrate how outcomes or experiences vary for different groups, while also recognising the limitations of the data, created an awareness and interest in inequalities. This should result in an improvement spiral, driving a demand for better-quality data that in turn creates more interest in the intelligence based on that data [19,29]. Several papers reported the deliberations and recommendations of multidisciplinary groups created specifically to address issues in data quality in specific areas such as disability [38], paediatrics [57], deaf communities [18], and COVID-19 and ethnicity [65], or more broadly [68]. These examples demonstrate the value of multidisciplinary groups in informing efforts and developing effective solutions for improving data collection and analysis efforts.

Staff reluctance was cited in many reports as a key factor that may hinder attempts to improve data quality [4,19,20,29,71]. This was due to a lack of knowledge about the importance and use of the data, combined with staff reluctance to offend patients by asking for sensitive information. Training programmes were able to address this barrier and also assuage concerns relating to the use of systems to collect such data [20,22,24,25,27,29,31,35,49,54,58]. In addition, the development of guidance on using data was cited as a mechanism to

improve data completeness and quality [4,34,43,68]. Training staff in communicating the rationale for data collection to the public and patients and on describing the parameters required was also a mechanism to improve data collection [34,60]. This was through building trust and openness between data collectors and providers [36,58]. One study suggested that ethnic matching could be one way of avoiding refusal during data collection [29].

In addition to staff reluctance, patients or the public may also be reluctant to provide data, or data collection instruments may not be appropriately developed for them. Several papers cited the importance of patient, public, or community involvement in initiatives to collect data or develop instruments such as surveys in data collection [27,58]. This involvement can help shape the questions that are asked and avoid marginalisation [17,36,38,56,63].

#### *3.3. Data Collection Instruments, Systems and Standardisation*

Many reports cited that data quality and granularity are impacted by the lack of standardised definitions. This creates pragmatic and logistical issues for data collection [19,21,71] through a lack of uniformity in data collection instruments such as surveys, as well as in IT systems that assign codes to different categories of data. Lack of standardised definitions and coding practices can cause major challenges when attempts are made to link data and in further analysis [63]. The introduction of standardised categories, or certain fields that are compulsory to complete as part of the design of IT systems, were mechanisms that were used to improve the recording and the quality of data [28,60–62]. Two papers recommended that consistency in coding and naming across different surveillance systems was also a way to enable consistency and more efficient linkage of sociodemographic data [23,25].

The importance of periodically revisiting these categories and ensuring their relevance was also shown to be an important activity [59]. Audit processes to monitor the completeness and accuracy of data and the methods used in data collection were discussed [20,70]. These processes allowed the assessment of data quality to put in place mechanisms for quality improvement [31]. One paper [16] reported on an instrument that could be used to compare and benchmark health information systems; however, it is unclear to what extent such tools are utilised or practical. Many grey literature reports in the UK recommended standardised protocols for collecting and recording ethnicity data as a mechanism to improve quality [4,65,67,70]. The importance of ensuring systems are in place to enable this was also discussed [31,38,63].

Improving the granularity and data fields available for individuals to self-assign their ethnicity or other characteristics was also shown to improve the completeness of data. For example, providing more options for self-reporting reduced the unknown ethnicity in certain studies [60]. This was achieved through providing more options (which are sometimes more relevant) to survey responders, resulting in less selection of the "unknown" category. Several reports used multidisciplinary groups to develop better understanding of the data that professionals from diverse disciplines thought should and can be collected [38,57,65,69].

#### *3.4. Methodological Approaches to Improve Data Completeness and Accuracy*

In addition to efforts to improve data at source, we also identified reports that described methods for improving data completeness and accuracy using statistical or other approaches (*n* = 27). This included data linkage, using proxy variables, or imputation through other methodologies [24].

Mathur et al. (2014) [47] found that when patients appeared in both the Clinical Practice Research Datalink (CPRD) and the Hospital Episode Statistics (HES) datasets with a usable ethnicity code in both datasets, the code was the same category in just 73% of cases. They found that when patients appeared in both datasets, completeness of usable ethnicity data in the CPRD increased from 78.7% to 97.1% once ethnicity data from HES was added. Knox et al. (2016) [45] looked at hospital admission rates by ethnicity in Scotland between 2009 and 2015, using the most recently recorded ethnicity to populate all admissions for that patient. This reduced the numbers of episodes with missing ethnicity from 24% to 15%, and the researchers completed the missing data for the remaining 15% by assigning those cases to ethnic groups in proportion to the distribution of known ethnicity by age and sex.

A number of imputation techniques can also be used to obtain more complete data; however, different methodologies have limitations and strengths [24,26]. Examples of the methods used include randomly assigning ethnicity, for example, on the basis of the distribution in the observed dataset or using a reference dataset [70], and using geographic location or probabilistic methods to infer ethnicity [35,50,51,58].

Several studies have investigated the use of algorithms to improve the completeness of ethnicity data by assigning ethnicity codes to individuals on the basis of their names, when self-identified data is missing [24,30,35,40,46,50,55,70]. The utility of this approach is recognised to differ considerably across countries because of significant variations in the composition of the population. Smith et al. (2017) [55] used the 'Onomap' software to categorise children and young people in the Yorkshire cancer registry as white, South Asian, or 'other' on the basis of their name, and also took ethnicity information from HES where this was recorded. Eleven per cent had missing HES ethnicity data and Onomap classified most of these patients. However, it is not clear whether these name-derived classifications were accurate, and these categories are very broad. The use of different methods to assign ethnicity did result in some different estimates of ethnic variation in cancer incidence, demonstrating the importance of accurate data.

Ryan et al. (2012) [50] also used Onomap and an additional name recognition software, Nam Pehchan [72], to predict the ethnicity of cases in a regional cancer registry who were missing this information following linkage with hospital inpatient data. They found that the software packages were accurate at predicting South Asian ethnicity but poor for other groups. They also looked at predicting ethnicity based on geographical area of residence but found this was also a poor predictor.

One paper also described the use of read codes to identify patients with learning difficulties (LD). NHS England issued guidance in October 2019 on improving the identification of people on the general practice LD register [69]. This required GPs to use a list of codes provided to check that all eligible patients were included on the practice LD register. The impact of this guidance on the numbers of patients on the register does not appear to have been evaluated. However, there was previous work evaluating the use of diagnostic read codes that found that this approach did identify small numbers of additional people who should have been on the register, and some further patients were found using specific descriptive codes [51]. The authors concluded that searching read codes to improve practice LD registers was quick and viable but not sufficient to capture most of the people eligible for inclusion, particularly those with milder learning difficulties. There does not appear to be evidence on how best to identify the remaining patients who could be included.

#### **4. Discussion**

Our scoping review identified a variety of mechanisms by which data quality in relation to health inequalities can be improved (Table 2). While the focus of many of the papers is on ethnicity data, many of the findings are also applicable to other dimensions of health inequalities because of the similarities in the issues that impact on data collection. There were relatively few papers that discussed improvements of data related to socioeconomic status; however, this might be because such data are collected through other means, rather than self-reporting, and the practice for collating this data is better established. There were also relatively few papers that discussed improvement of data relating to gender and sexual orientation or disability. In addition, while some included papers discussed the issue of intersectionality, the impact in terms of data analysis or data collection were often not fully explored.


**Table 2.** Summary of best practices.

We have classified the mechanisms that can be used to improve the quality of data on health inequalities as more distal or proximal to the source data. Distal factors that impact on data quality include legislation and policies that are in place to ensure and mandate collection of data to enable addressing health inequalities. While many countries recommend the monitoring of data related to equality and discrimination, the extent to which this is implemented and actioned for health varies. Much of this is due to the differing structures of health systems and legislation that are in place globally. These distal factors impact on the ability to collect data related to equality and discrimination. For example, in the UK, the duty of data collection falls with public bodies [42], whereas this is not necessarily the case in other countries. Nevertheless, several included reports evidenced the fact that legalisation and policy were key contributors to the success of high-quality data collection efforts. Mechanisms to enact these policies and enable data collection form the next series of mechanisms to improve data quality. Reports described a variety of mechanisms, such as senior-level buy-in, staff training programmes, patient and public involvement, needed to enable creation of data systems that take into consideration the purpose of data collection and are timely and relevant.

Data pertaining to health inequalities may be collected by different organisations involved in health and care provision. They may collect these data for different purposes, meaning that the granularity of information requirements may differ. In addition, definitions in relation to many protected characteristics such as gender and ethnicity vary and evolve over time. This is because these are composite social constructs, attempting to bring together a number of different elements. For example, ethnicity is a composite of cultural factors, language, and ancestry, amongst others. This is evidenced by reports from the UK [4] that do not make a strong distinction between race and ethnicity, though work from the US distinguishes between these concepts, particularly when considering people from a Hispanic/Latinx background. Furthermore, these concepts change over

time, meaning minority groups can change in size and new groups may become more prevalent. Many reports cited that redefinition of how populations are categorised in relation to characteristics related to health inequalities is needed over time [24,29,36]. For example, it is now more common to collect data that allows us to identify a subcategory of White Eastern European, or distinguish between Black African groups. Similarly, few would have included 'nonbinary' as a possible answer option to a question on gender five years ago. Thus, engagement across citizens, providers, and those creating data systems is needed to ensure the data that are collected are acceptable, relevant, and fit for purpose, and yet retain the ability to compare across time to monitor change and assess the impact of policies and interventions that aim to prevent and reduce health inequalities.

The report of improvements to data collected by NHS Lothian is a good example of the multi-layered approach that is needed to improve data quality [34]. The Scottish government and the Commission for Racial Equality requested the Scottish health boards to improve the recording of patient ethnicity data, and all boards were required to produce an action plan with progress measures. Davidson et al. (2020) report an impressive increase in the proportion of patients with a recording of ethnicity from 3% to over 90% in just three years (between 2008 and 2012). The authors attribute this improvement to several factors, chief amongst these being the decision to make ethnicity a mandatory field in the hospital data systems. Other important factors were thought to include the training of individuals responsible for data collection, awareness raising with relevant clinical and management staff and sharing a clear purpose and vision, and executive buy-in from senior clinical and management colleagues to ensure staff were able to prioritise recording these data. Making it clear to staff how ethnicity information is used was also important to maintain their motivation to collect these data. In this case, the data were used to demonstrate that rates of A&E use by ethnic minority groups did not appear to be linked to rates of registration in primary care. The progress made by NHS Lothian is in contrast to many other NHS Boards in Scotland where, over the same period, recording remained poor or improved much more slowly, despite an identical governance and legal context [73].

The importance of staff training is also evidenced by some older studies. A review by Iqbal et al. (2009) showed that staff training was the main intervention for which there was evidence of data quality improvements for patient ethnicity, followed by adequate resources to allow data collection and use [74]. Training should be tailored to the local context and explain why it is important to gather standardised data on patient ethnicity, what the data will be used for, and how to ask the questions and record responses. The review also recommended collecting self-reported ethnicity as routine during GP registration.

Self-reported data are the gold standard for certain data such as gender and ethnicity that can inform studies of health inequality. However, the work included for this review has identified a wide range of reasons why individuals may be reluctant to share personal data relevant to these characteristics. A paper from NHS Scotland points out that different settings can have substantially different rates of refusal (for ethnicity data reporting), which suggests different organisational approaches to asking for and recording the information [70]. High rates of refusal (or high use of an 'other' category) can be compared against peer organisations and could likely be brought down by learning from successful approaches elsewhere. Improving public and patient understanding of why this information is being collected and how it will be used can also encourage efforts to improve data collection and, therefore, quality. Nevertheless, there will likely always be some people who decline to give information on their ethnicity, or other personal information not perceived to be directly relevant to their immediate care, and it is important to recognise their right to decline to provide this.

It can take time to put in place a best practice that leads to the collection of goodquality data in relation to health inequalities. In addition, as evidenced by many of the reports, this may still lead to incomplete data with inaccuracies. Thus, mechanisms that can improve the accuracy, quality, and completeness of available data are also important. We identified studies that reported the use of methodologies such as linkage, imputation, and the use of proxy variables. However, there are several limitations to these methods. Using naming software or linked data to improve the completeness of ethnicity data takes considerable time and analytical expertise and is not ideal for producing useful up-to-date routine reports for health services [50,55]. However, the studies that examined the use of naming software took place at a time when recording ethnicity for hospital inpatients was much poorer. It seems likely that their findings have less relevance today when hospital data are much less likely to be missing data on patient ethnicity, given that both studies were of cancer patients (who are likely to appear in HES data). Using naming software to estimate ethnicity may still have utility when data cannot be linked to hospital or other data, but clearly this approach to filling ethnicity-data gaps needs caution. It is likely to struggle more with mixed-ethnicity individuals (an increasing proportion of the UK population) and is unlikely to be able to produce the detail necessary to distinguish between subgroups.

Data linkage has been evaluated for its utility in reducing missing data. If the same individual is identifiable in two datasets, information from one dataset can be used to check or complete the information in the other. Data linkage can be powerful for 'filling in the gaps' and has been used by NHS Digital to increase coverage of ethnicity data during the COVID-19 pandemic [75]. However, using data linkage to improve ethnicity data on a routine basis, so that it can be useful for producing near-real-time intelligence to inform services and policy, is challenging given the requirements for analyst capacity and time [24]. Improving data through data linkage also requires having a resource to link to that contains accurate self-reported ethnicity data and has high coverage across the population. In England, this resource could potentially be census data, HES data, or GP data, or death certification data for people who have died. However, there are issues with each of these sources. Census data is very sensitive and not easy to access and is only updated every 10 years. GP data is known to have patchy coverage. Recent HES data has better completeness for the people included in the dataset, but coverage is an issue because of the requirement that patients have been hospital users. Using ethnicity data from death certificates is also likely to bring accuracy issues as, of course, ethnicity cannot be self-reported in these cases and, in fact, often mismatches the data in hospital records. Even within the group of patients who appear in the HES data, using HES as a source of accurate ethnicity data may be inadequate.

This scoping review has some strengths in that we used a systematic approach to identify as many reports as possible discussing different mechanisms to improve data quality. Yet, it is likely that there are reports that we missed, especially in the form of grey literature, because of the broad nature of the subject matter. The majority of the reports were from the UK or US. This might be a result of our search terms not being optimal. Other factors include the extent to which health inequalities monitoring has been implemented and is a priority as part of healthcare delivery [76,77]. Nevertheless, this work identified evidence for several distal and proximal approaches that can be taken within the healthcare context of the United Kingdom to improve the quality of the data used for the identification and monitoring of health inequalities. Some of these approaches may be transferable to other healthcare contexts. However, given differences in definitions and drivers of health inequalities and provision of health care around the globe, they may not apply to the same extent.

#### **5. Conclusions**

Accurate and timely data are essential in identifying inequalities in health and care, in understanding where inequalities occur and which groups are affected, and in assessing the impact of interventions. Despite this, many health-related datasets either do not routinely collect important dimensions of inequality or are limited by poor-quality data. Where data are available, they may not always be used to the best extent. Our review identified that a variety of effective mechanisms are available and can be utilised to improve data quality. These include those that are distal and impact on data collection, or those that are more proximal to the source data and can aid in data analysis. Given the renewed emphasis on

the need for action to address health inequalities at both a national and a system level, it is important to understand how systems can easily implement the mechanism described in our review. This will likely require working with senior leaders, staff, and analysts to gain buy-in and identify effective ways to implement mechanisms to address issues with data quality. Further work is underway to understand how best to support health and care staff to act on the evidence identified in this review to improve the quality of data relating to health inequalities within their organisations and local systems.

**Supplementary Materials:** The following supporting information can be downloaded at https: //www.mdpi.com/article/10.3390/ijerph192315874/s1: Search Terms and PRISMA-ScR checklist. Reference [15] is cited in the supplementary materials.

**Author Contributions:** S.M.: V.P. (Vicki Peacey), S.E. and L.L. conceived the study objective and design. V.P. (Veronica Philips) carried out the searches. A.R.-U. contributed to screening the titles. S.M. and L.L. carried out the scoping review with feedback from S.E., C.B. and V.P. (Vicki Peacey) All authors contributed to the original drafting, reviewing, and editing of the manuscript. All authors have read and agreed to the published version of the manuscript.

**Funding:** SM: LL, SE, ARU are supported in part by the NIHR Applied Research Collaboration East of England (ARC EoE). NIHR Applied Research Collaboration East of England, grant number G104017. The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR, or the Department of Health and Social Care or Cambridgeshire County Council.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Data can be made available on request.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


### *Article* **Who Presents Where? A Population-Based Analysis of Socio-Demographic Inequalities in Head and Neck Cancer Patients' Referral Routes**

**Jennifer Deane 1, Ruth Norris 1, James O'Hara 1,2, Joanne Patterson <sup>3</sup> and Linda Sharp 1,\***


**Abstract:** Head and neck cancers (HNC) are often late stage at diagnosis; stage is a major determinant of prognosis. The urgent cancer referral pathway (two week wait; 2WW) within England's National Health Service aims to reduce time to diagnosis. We investigated factors associated with HNC route to diagnosis. Data were obtained from the English population-based cancer registry on 66,411 primary invasive HNCs (ICD C01-14 and C31-32) diagnosed 2006–2014. Multivariable logistic regression determined the likelihood of different diagnosis routes by patients' demographic and clinical characteristics. Significant socio-demographic inequalities were observed. Emergency presentations declined over time and 2WW increased. Significant socio-demographic inequalities were observed. Non-white patients, aged over 65, residing in urban areas with advanced disease, were more likely to have emergency presentations. White males aged 55 and older with an oropharynx cancer were more likely to be diagnosed via 2WW. Higher levels of deprivation were associated with both emergency and 2WW routes. Dental referral was more likely in women, with oral cancers and lower stage disease. Despite the decline over time in emergency presentation and the increased use of 2WW, socio-demographic variation is evident in routes to diagnosis. Further work exploring the reasons for these inequalities, and the consequences for patients' care and outcomes, is urgently required.

**Keywords:** head and neck cancer; routes to diagnosis; socio-demographic inequalities; healthcare inequalities; emergency presentation

#### **1. Introduction**

The UK lags significantly behind other European and high human development countries with regards to cancer outcomes [1]. Evidence suggests that this is due, in part, to later-stage diagnosis [2], including relatively high proportions of cancers which are diagnosed on emergency presentation [3].

In general, cancer survival rates are strongly associated with stage at diagnosis; the earlier the stage the better the chance of survival [4]. Late-stage cancer at diagnosis may be the result of delays at various points across the diagnostic pathway; these delays can be in presentation (time from symptom onset to first presentation to primary care), primary care (time from first presentation to referral for specialist assessment), and secondary care (time from specialist referral to diagnosis) [5].

#### *1.1. Routes to Diagnosis in Cancer*

The Urgent Cancer Pathway, known as 2 Week Wait (2WW), was established in the English National Health Service (NHS) in 2000 [6]. A target of 14 days from the point of referral for suspected cancer symptoms, to the point of first assessment with a specialist at the hospital, was put in place. Whilst, in part, this pathway was intended to reduce patient

**Citation:** Deane, J.; Norris, R.; O'Hara, J.; Patterson, J.; Sharp, L. Who Presents Where? A Population-Based Analysis of Socio-Demographic Inequalities in Head and Neck Cancer Patients' Referral Routes. *Int. J. Environ. Res. Public Health* **2022**, *19*, 16723. https://doi.org/10.3390/ ijerph192416723

Academic Editor: Paul B. Tchounwou

Received: 31 October 2022 Accepted: 9 December 2022 Published: 13 December 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

anxiety around waiting for investigations into a possible cancer diagnosis, it was also hoped that it would shorten the primary care interval, allowing identification of cancers at an earlier stage, widening treatment options, and improving survival. A decline over time in the proportion of cancers which present as emergencies has been attributed, in part, to the introduction of this pathway [7,8].

#### *1.2. Head and Neck Cancers*

Head and neck cancer (HNC) is an umbrella term for malignant tumours arising in the oral cavity, larynx, pharynx, nose and salivary glands. HNC is now the 8th most common cancer and is responsible for 3% of all cancer diagnosis in the UK [9]. No effective, organised, HNC screening is in place (although there are country-specific and international events designed to increase awareness among healthcare practitioners (HCP) and the public such as head and neck cancer awareness weeks). Therefore, patients are generally diagnosed due to the presence of symptoms. Symptoms vary and include ear pain, persistent sore throat, a neck lump (enlarged lymph node), persistent mouth ulcers, and airway obstruction. Due to tumour location, patients may present symptomatically at a variety of different healthcare settings, including the GP practice, community Dental practice, a Dental hospital or, less commonly, a hospital emergency department [10].

#### *1.3. Inequalities in HNC*

Equity in healthcare systems is a marker for healthcare quality [11]. Care should be provided in a way that does not vary in quality due to sociodemographic or socio-economic status (SES). There are multiple inequalities relating to HNC in the UK. Incidence is strongly socio-economically patterned, with rates around 2–4 times higher in those resident in more deprived, compared to less deprived areas [9]. Around 60% of HNCs are diagnosed at a late stage [12] and the proportion diagnosed early is lowest in the most deprived areas [12]. Moreover, survival is also worse in those resident in more deprived areas [13,14].

Current knowledge on the route(s) patients take to receive a HNC diagnosis is limited; improved understanding of whether there are socio-demographic inequalities in this could help to highlight areas for improvement in service provision. We therefore undertook a population-based study investigating socio-demographic inequalities in HNC routes to diagnosis in England. Specifically, this study set out to establish whether there are sociodemographic inequalities in HNC patients diagnosed via (i) emergency versus primary care routes; (ii) 2WW versus any standard primary care routes; and (iii) dentists versus all other non-emergency routes.

#### **2. Methods**

#### *2.1. Study Design and Setting*

Registrations for all patients with a primary invasive HNC (ICD C01-14 and C31-32) diagnosed in England between 2006 and 2014 were abstracted from the National Cancer Registration Database (NCRD). Ethnical approval was obtained from the Yorkshire and the Humber South Yorkshire Research Ethics Committee on 16th November 2017 (Ref number 206040), and this population-based study is reported according to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidance [15].

#### *2.2. Data Sources and Linkage*

The NCRD is a population-based cancer registry that seeks to systematically identify and record information on all newly diagnosed tumours in patients resident in England. The registry receives data from across the NHS which approximates to around 300,000 malignant tumour diagnoses annually [16]. Reporting of NHS hospital cancer data is mandatory. Each NCRD record is linked, using patient NHS number, to UK NHS Hospital Episode Statistics (HES) data to provide information on comorbidities and cancer treatments.

A "route to diagnosis" was assigned, by the National Cancer Registration and Analysis Service (NCRAS) to each cancer registration using a combination of data from the NCRD, HES, Cancer Waiting Times, and Cancer Screening programmes. The route to diagnosis refers to a sequence of interactions between the patient and the healthcare system which leads to the diagnosis of cancer [17]. Each registration is assigned one of eight main routes to diagnosis codes: GP referral; 2WW; emergency presentation; other outpatient, screen detected (not relevant for HNC); inpatient elective; death certificate only (DCO); and unknown. Within several of these main routes, there are (sub) routes which can be used to distinguish patients who were referred from different types of practitioners (e.g., 2WW (GP), 2WW (dentist), 2WW (other)). This route to diagnosis dataset has been used to document the diagnostic route for a range of cancers [18], but it has not been previously used to compare which HNC patients are present and are diagnosed through which routes.

#### *2.3. Population*

The population of interest was patients with an incident primary invasive HNC (n = 70,334). In instances where a patient had records for multiple primary tumours in the head and neck (n = 1308), a hierarchy determined which tumour record to retain for analysis. This was as follows: (i) the earliest diagnosed tumour; (ii) the earliest tumour referral date; (iii) the tumour marked as potentially positive for human papilloma virus (HPV), based on proxy information (morphology and subsite); and (iv) selected at random from the remaining tumours. Childhood tumours in patients aged <20 years old were excluded from the analysis. Cases with missing routes to diagnosis (n = 2243) and cases diagnosed only at the time of death (n = 98) were also then excluded. This left an analytical cohort of 66,411 patients (Figure 1).

#### *2.4. Explanatory Variables*

Explanatory variables of interest were as follows: age at diagnosis, sex, cancer site, deprivation category, period of diagnosis, ethnicity, urban/rural category, stage, grade and comorbidities. Age at diagnosis was categorised as 20–54, 55–64, 65–79 and 80+ years. Cancer sites were grouped as oral cavity (C02-C06; including palate), oropharynx (C01, C09, C10), larynx (C32) and other HNC (nasopharynx C11; hypopharynx C12, C13; salivary glands C07, C08; other sites C05, C07-C08, C11-C13; and non-specific sites C14, C31). Deprivation was an area-based measure of the income domain of the Index of Multiple Deprivation (IMD) [19] Quintile 1 includes the people resident in the least deprived and quintile 5 those resident in the most deprived areas; these refer to quintiles of the general population. Deprivation was used as a SES proxy measure. Period of diagnosis was grouped into 3-year time bands (2006–2008; 2009–2011; and 2012–2014). Ethnicity was classified as white, non-white (other ethnic group) and unknown (missing and unknown ethnicity). Urban/rural categorisation was based on areas of residence at diagnosis and was collapsed to either rural or urban [20]. Cancer summary stage was assigned using the TNM staging system (I–IV or other (unknown/missing)). Tumour grade was classified as 1 (low grade, undifferentiated)—4 (high grade, differentiated) and unknown (unknown/missing). A weighted comorbidity score based on the Charlson Comorbidity Index [21] reported the number of in-patient hospital admission for different relevant comorbidities recorded in the period 3 to 27 months before diagnosis (with the index cancer disregarded). Comorbidities were classified as none, 1 and 2+.

#### *2.5. Outcome Variables*

The outcome variables of interest were route to diagnosis. For the purpose of this analysis, the NCRD operationalisation of (sub)route to diagnosis was categorised as follows: (i) emergency presentation (comprising (sub)routes: A&E, emergency GP referral, emergency transfer, emergency admission or attendance); (ii) all primary care routes (that is, all routes which would have been initiated in primary care: GP referral, inpatient referral, outpatient (dentist and other referral), 2WW (dentist, GP and other)); (iii) 2WW (all

urgent cancer referral routes: dentist, GP and other); (iv) standard care routes (that is, all non-urgent non-emergency, cancer referral routes: GP referral, inpatient referral, outpatient (other and dentist referral)); (v) dentist (all routes which started with a dentist: outpatient and 2WW); (vi) and all other non-emergency routes (referral routes which did not start with a dentist: GP referral, inpatient, outpatient (other referral) and 2WW (GP and other)) (Supplementary Figure S1).

**Figure 1.** Flow diagram of the analytical cohorts for Analyses 1–3. Abbreviations: DCO: Death certificate only; GP: General practitioner; HNC: Head and neck cancer; 2WW: Two week wait.

#### *2.6. Statistical Analyses*

Three analyses were undertaken to explore the role of socio-demographics on route to diagnosis (Figure 1; Supplementary Figure S1). Analysis 1 included the whole analytical cohort (emergency presentation) and considered whether there was a difference in those patients presenting through the emergency route compared with all primary care routes (i.e., comparing categorisations (i) and (ii) above). Analysis 2 considered only patients coming through primary care routes (category (ii) above), and compared 2WW referral versus standard primary care-initiated routes (i.e., (iii) vs. (iv)). Analysis 3 again considered only patients coming through primary routes (category (ii) above), but this time compared dentist referral vs. all other non-emergency (non-dental) routes (i.e., categories (v) vs. (vi) above).

For each analysis, baseline descriptive statistics were reported for the analytic population along with chi-square tests of associations between socio-demographic and clinical variables with diagnosis route. Univariable and multivariable logistic regression models were then developed to assess the likelihood of diagnosis route by socio-demographic characteristics with and without adjustment for confounders. Any variables significant in univariate analyses (likelihood ratio tests (LRT) *p* ≤ 0.05) were included in multivariable models. Models were reduced to contain only statistically significant variables (LRT *p* ≤ 0.05). Model goodness-of-fit was assessed, and care taken to avoid multicollinerity. The Akaike Information Criterion (AIC) was used to differentiate between competing models. All final models had adequate fit. Odds ratios (ORs) and 95% confidence intervals (CIs) were reported. Stata V.15 [22] was used for all analyses.

#### **3. Results**

#### *3.1. Patient Characteristics*

In total, 66,411 patients were diagnosed with a first, invasive primary HNC between 2006 and 2014. HNC diagnoses increased over time. Almost two-thirds of patients were aged 55–79 at diagnosis (64.1%); almost 70% were male (69.6%); and more than 4 out of 5 were of white ethnicity (81.2%). Diagnosis was associated with deprivation: a higher proportion of patients were resident in the most deprived areas (24.5%) than in the least deprived areas (15.7%). Most patients resided in an urban area at the time of diagnosis (82.4%). The site distribution was as follows: oral cavity (34.1%), larynx (23.9%), oropharynx (22.8%) and other HNC (19.2%). At diagnosis, the most common summary stage was stage IV (19.6%) and grade 2 (37.8%); stage was not recorded for most cases. Referral via each route to diagnosis was as follows: emergency (8.5%); all primary care routes (91.5%); 2WW (39.4%); standard care routes (52.1%); dentist (9.9%); and all other routes (81.6%). Demographic and clinical characteristics of the population are shown in Table 1. A full breakdown by individual route to diagnosis can be viewed in Supplementary Table S1.

**Table 1.** Demographic and clinical characteristics of HNC diagnosed during 2006–2014.



**Table 1.** *Cont.*

<sup>1</sup> Includes palate; <sup>2</sup> Other cancer site refers to nasopharynx, hypopharynx, salivary glands, other sites and nonspecific sites; <sup>3</sup> Non-white refers to other ethnic groups; <sup>4</sup> Unknown ethnicity refers to missing and unknown ethnicity; <sup>5</sup> Other stage refers to missing and unstageable tumours; <sup>6</sup> Unknown grade refers to unknown and missing tumour grades; <sup>7</sup> Measured using the Charlson Comorbidity Index; <sup>8</sup> Reported as a percentage of the analytical population (n = 66,411); <sup>9</sup> 2WW refers to 2WW (dentist), 2WW (GP) and 2WW (other); <sup>10</sup> Dentist refers to outpatient (dentist) and 2WW (dentist); <sup>11</sup> Emergency refers to A&E, emergency GP referral, emergency transfer, emergency admission or attendance; <sup>12</sup> All other non-emergency routes refers to GP referral, inpatient referral, outpatient (other referral), 2WW (GP) and 2WW (other). <sup>13</sup> Standard primary care-initiated routes refers to GP referral, inpatient referral, outpatient (other referral) and outpatient (dentist); <sup>14</sup> All primary care routes refers to GP referral, inpatient referral, outpatient (other referral), outpatient (dentist), 2WW (dentist), 2WW (GP) and 2WW (other) Abbreviations: GP: General practitioner; IMD: Index of multiple deprivation; 2WW: Two week wait.

#### *3.2. Analysis 1: Emergency Presentation vs. All Primary Care Routes*

In total, 8.5% of patients (n = 5676) were diagnosed through emergency presentation, compared to 91.5% identified through primary care (n = 60,735). The percentage of emergency presentations declined slightly over time from 9.6% in 2006–2008 to 7.9% in 2012–2014. In univariate analyses, several variables were associated with diagnosis through the emergency route. These were older age, being male, living in a more deprived area, having two or more comorbidities, non-white ethnic group, stage IV disease and higher-grade cancer (Table 2). Compared with oral cancers, cancers of the larynx and other HNCs were more likely to present through emergency routes.



<sup>1</sup> Emergency refers to A&E, emergency GP referral, emergency transfer, emergency admission or attendance; <sup>2</sup> All primary care routes refers to GP referral, inpatient referral, outpatient (dentist), outpatient (other referral), 2WW (dentist), 2WW (GP) and 2WW (other); <sup>3</sup> *p* values in bold are from LRT of the contribution of the variable to the model. Unbolded *p* values are from a test of whether the OR is different from 1; <sup>4</sup> Includes palate; <sup>5</sup> Other cancer site refers to nasopharynx, hypopharynx, salivary glands, other sites and non-specific sites; <sup>6</sup> Non-white refers to other ethnic groups; <sup>7</sup> Unknown refers to missing and unknown ethnicity; <sup>8</sup> Other stage refers to missing and unstageable tumours; <sup>9</sup> Unknown grade refers to unknown and missing tumour grades; <sup>10</sup> Measured using the Charlson Comorbidity Index. Abbreviations: A&E: Accident and emergency; CI: Confidence interval; IMD: Index of multiple deprivation; GP: General practitioner; LRT: likelihood ratio test; OR: Odds ratio: 2WW; Two week wait. Model adjusted for age at diagnosis, cancer site, deprivation category, period of diagnosis, ethnicity, urban/rural categorisation, stage, grade, and comorbidities.

Socio-demographic associations (apart from with sex) persisted in multivariable analyses and were statistically significant. Those aged 80 and over were almost twice as likely

to be diagnosed through emergency presentation (80+ years old vs. 20–54 years old; multivariable odds ratio (mvOR) 2.00, 95% CI 1.82, 2.19). There was also a consistent trend of increased likelihood of emergency diagnosis as the level of deprivation increased. Those patients resident in the most deprived areas were 1.82 times more likely to come through an emergency route than those patients resident in the least deprived areas (IMD5 vs. IMD 1; mvOR 1.82, 95% CI 1.65, 2.00). Non-white patients were 1.28 times more likely to be diagnosed via emergency presentation than white patients (non-whites vs. white; mvOR 1.28, 95% CI 1.13, 1.45). Patients residing in rural areas were significantly less likely to be referred through an emergency route (rural vs. urban mvOR; 0.91, 95% CI 0.84, 0.99). In terms of clinical variables, patients diagnosed with a higher-grade cancer were 1.45 times more likely to present through emergency routes (high vs. low grade; mvOR 1.45, 95% CI 1.09, 1.93). Stage I cancers were 82% less likely than stage IV cancers to be diagnosed via emergency presentation (I vs. IV; mvOR 0.18, 95% CI 0.15, 0.23).

#### *3.3. Analysis 2: 2WW vs. Standard Primary Care-Initiated Routes*

Of HNC patients who were diagnosed through a route initiated in primary care, just over 40% came through the urgent 2WW pathway (n = 26,148; 43.1%). This proportion rose over time from 36.0% in 2006–2008 to 49.8% in 2012–2014. When comparing patients referred via 2WW rather than via other standard care routes, the variables associated with an increased likelihood of urgent referral in univariate analyses were as follows: being aged 55–64 years old male, and of white ethnicity; having a cancer of the oropharynx, stage III and IV disease, grade 3 tumours, no comorbidities and residing in an area of higher deprivation. There was no observed variation by urban/rural residence. In multivariable analysis, associations with stage and grade did not persist. Patients aged 55–64 years were more likely to be referred via the urgent 2WW pathway than younger patients (55–64 years vs. 20–54 years; mvOR 1.18, 95% CI 1.13, 01.24); more modest increased risks were seen for the two older age-groups. Compared to cancers of the oral cavity, cancers of the oropharynx were more likely to been referred via 2WW (mvOR 1.64, 95% CI 1.57, 1.71). Patients were 1.43 more than 40% more likely to be referred by 2WW pathways if they resided in the most deprived areas (IMD5 vs. IMD1; mvOR 1.43, 95% CI 1.35, 1.50). Being female was associated with a reduced likelihood of 2WW referral (mvOR 0.76, 95% CI 0.73, 0.79) as was being from a non-white ethnic group (non-white ethnic group vs. white; mvOR 0.57, 95%CI 0.52, 0.62) (Table 3).

#### *3.4. Analysis 3: Dentist vs. All Other Non-Emergency Routes*

Overall, 10.8% (n = 6572) of all HNC patients who followed a non-emergency route were referred via a dentist. This percentage rose slightly from 9.6% in 2006–2008 to 11.8% in 2012–2014. In the univariable analysis, when compared with referral via all other routes, dental referral was associated with older age, female gender, residence in a less deprived area and having an oral cancer. All variables apart from urban/rural category and comorbidities were statistically significant in the final model. In multivariable analyses, patients aged 65–79 years old were most likely to be referred via the dentist (65–79 vs. 20–54 years; mvOR 1.13, 95%CI 1.05, 1.22) as were female patients (mvOR 1.27, 95% CI 1.20, 1.34) and those from a non-white ethnic group (non-white vs. white; mvOR 1.26, 95%CI 1.12, 1.43). Residence in an area of increasing deprivation was associated with a reduced chance of dental referral when compared to all other routes (IMD 5 vs. IMD1; mvOR 0.71, 95%CI 0.65, 0.78). Patients with stage I cancer (stage I vs. stage IV; mvOR 1.19, 95%CI 1.07, 1.32) were more likely to be referred via dental routes. Diagnoses via dental referral when compared to all other routes also increased over time (2012–2014 vs. 2006–2008; mvOR 1.22, 95% CI 1.12, 1.32) (Table 4).


**Table 3.** Likelihood (OR, 95% CI and *p* values) from logistic regression of 2WW versus standard primary care-initiated routes by socio-demographic and clinical characteristics for Analysis 2 (n = 60,735).

<sup>1</sup> 2WW refers to 2WW (dentist), 2WW (GP) and 2WW (other); <sup>2</sup> Standard primary care-initiated routes refers to GP referral, inpatient referral, outpatient (other referral) and outpatient (dentist); <sup>3</sup> *p* values in bold are from LRT of the contribution of the variable to the model. Unbolded *p* values are from a test of whether the OR is different from 1; <sup>4</sup> Includes palate; <sup>5</sup> Other cancer site refers to nasopharynx, hypopharynx, salivary glands, other sites and non-specific sites; <sup>6</sup> Non-White refers to other ethnic groups; <sup>7</sup> Unknown ethnicity refers to missing and unknown ethnicity; <sup>8</sup> Other stage refers to missing and unstageable tumours; <sup>9</sup> Unknown grade refers to unknown and missing tumour grades; <sup>10</sup> Measured using the Charlson Comorbidity Index. Abbreviations: CI: Confidence interval; IMD: Index of multiple deprivation; GP: General practitioner; LRT: Likelihood ratio tests; OR: Odds ratio: 2WW; Two week wait. Model adjusted for age at diagnosis, sex, cancer site, deprivation category, period of diagnosis, ethnicity, and comorbidities.


**Table 4.** Likelihood (OR, 95% CI and *p* values) from logistic regression of dentist versus all other non-emergency routes by socio-demographic and clinical characteristics for Analysis 3 (n = 60,735).

<sup>1</sup> Dentist refers to outpatient (dentist) and 2WW (dentist); <sup>2</sup> All other non-emergency routes refers to GP referral, inpatient referral, outpatient (other referral), 2WW (GP) and 2WW (other); <sup>3</sup> *p* values in bold are from LRT of the contribution of the variable to the model. Unbolded *p* values are from a test of whether the OR is different from 1; <sup>4</sup> Includes palate; <sup>5</sup> Other cancer site refers to nasopharynx, hypopharynx, salivary glands, other sites, and non-specific sites; <sup>6</sup> Non-White refers to other ethnic groups; <sup>7</sup> Unknown ethnicity refers to missing and unknown ethnicity; <sup>8</sup> Other stage refers to missing and unstageable tumours; <sup>9</sup> Unknown grade refers to unknown and missing tumour grades; <sup>10</sup> Measured using the Charlson Comorbidity Index. Abbreviations: CI: Confidence interval; IMD: Index of multiple deprivation; GP: General practitioner; LRT: Likelihood ratio tests; OR: Odds ratio: Model adjusted for age at diagnosis, sex, cancer site, deprivation category, period of diagnosis, ethnicity, stage, and grade.

#### **4. Discussion**

To our knowledge, this is the first comprehensive analysis of routes to diagnosis for HNC in England. In this population-based study, significant socio-demographic inequalities were observed and were shown to vary across diagnosis routes.

There were some indications in the results of positive changes over time, most notably the increase in those picked up through the urgent cancer referral route (2WW). How-ever, there are several areas of concern. The analysis showed that there has been an in-crease over time in the number of HNCs diagnosed, although the distribution of HNC cancer subsites has changed with the predominant tumour site being the oropharynx in 2012–2014, rather than larynx which was most common in 2006–2008. This echoes trends re-ported elsewhere [23] and likely reflects changes in risk factors such as a reduction in smoking prevalence and an increase in HPV-related cancers [24]. Although overall the number of patients diagnosed through the emergency route is relatively small compared to some other cancers [25,26], there was a small increase in the number (albeit not the percentage) of emergency presentations over time. This is concerning as emergency cancer presentations may be considered, in some ways, as a "failure" of the system, and indicative of significant delays or barriers to presentation.

#### *4.1. Emergency Route*

Those that were diagnosed through the emergency route were more likely to present with advanced disease, which is consistent with patterns of other cancers in the UK and internationally [3]. In terms of socio-demographic characteristics, emergency presentations were more often patients from urban areas and areas of greater deprivation, from non-white ethnic groups, and over the age of 65.

The association between older age and emergency presentation is supported by previous research in all cancers in England where likelihood of emergency presentation rose significantly in those over 70 years [27]. Whilst it is known that advanced age is a risk factor for HNC, the vague or non-specific nature of some HNC symptoms may mean that symptoms are not recognised as being of concern or are perceived as "normal" aging. Previous research has shown that cancer awareness is lower in this age group than among younger people [28]. It has also shown that people, and in particular older adults, can be more reticent to seek help in primary care due to a fear of wasting clinicians' time, particularly when symptoms are vague [29,30]. This fear may then reduce the chances of a person seeking help from primary care, resulting in a delay to diagnosis and an increased likelihood of an emergency presentation.

There are a growing number of reports on healthcare experiences of ethnic minority groups in the UK and internationally; people from ethnic minorities more often experience significant barriers to accessing healthcare, and once within the system, more often report poor experiences (for example: [31–34]). Much of the recent work has focused on the experience with COVID-19; however, it seems plausible that barriers such as lack of trust, inappropriate services and discrimination impacted help-seeking prior to COVID-19 too [35]. The finding here that patients from non-white ethnic groups are more likely to be diagnosed after an emergency presentation adds further to this accumulating evidence base [36].

Older age, deprivation and being from an ethnic minority have all been associated with suboptimal health literacy [37,38]. Health literacy is the extent to which an individual has the capacity, knowledge, understanding and confidence to access, understand, evaluate, use and navigate health and social care information and services [39]. It includes the capacity to communicate, assert and enact health decisions [40]. It has been associated—in other clinical areas—with less use of preventive health services and greater use of emergency services [39]. Given the socio-demographic patterns observed here, future research exploring the role of health literacy in emergency cancer presentation (and, more generally, across the entire cancer diagnosis pathway) would be of value.

#### *4.2. Urgent Cancer Referral (2WW)*

Those patients diagnosed through the urgent cancer referral route, compared to other routes which commenced in primary care, were more likely to be white, male, aged 55 years and older, resident in areas of greater deprivation and to have a cancer of the oropharynx.

The 2WW pathway requires that a patient meets a list of referral criteria for urgent investigation of a suspected cancer. Compared to other HNC tumours, oropharyngeal cancer more often presents with a neck lump/swelling [41]. This may mean that it is more likely to be recognised as potentially concerning by patients and primary care clinicians than vague, less specific (and perhaps more benign-seeming) symptoms, thus triggering a 2WW referral far quicker. Previous research on multiple different cancers has shown that those with vague symptoms delay attending primary care take a median of 34 days longer to diagnosis than those with alarm symptoms [42]. Moreover, the stereotypical "traditional" HNC patient is an older deprived male (likely with tobacco and alcohol addiction problems) [43]. It is therefore possible that primary care staff may be more likely to have a higher index of suspicion of a potentially serious underlying condition around individuals who match this profile, and therefore refer them for urgent investigation.

Some research in Denmark suggests that GPs suspect cancer in more patients than they refer onto cancer specific pathways, and that those patients who reported vague symptoms are less likely to be referred [44]. This suggests the possibility that those who do not display what the GP considers to be clear symptoms of a potential HNC, despite a suspicion of cancer, may not be being referred through the 2WW pathway.

#### *4.3. Dental Referral Route*

Patients from ethnic minorities, women, and those from less deprived areas, were more likely to have been referred through a dentist than through other primary care routes. As might be expected, oral cavity cancers were more often diagnosed through this route, but it is noteworthy that dentists also referred patients who were diagnosed with cancers elsewhere in the head and neck.

The dental system in the UK involves payment at the point of treatment, in contrast to the rest of primary care which is free at point of treatment; moreover, not all dental costs are subsidised by the State. There are significant barriers to accessing NHS dental services, including financial difficulties, lack of availability of services (i.e., no appointments being available), or lack of services being offered in the local area [45]. Our finding that people from more deprived areas were less likely to be diagnosed through the dental route may be explained by the cost of accessing dental check-ups and treatment. While some of those on the lowest incomes are entitled to free dental care, this involves the completion of lengthy forms [46]. Research has shown that areas of deprivation have far less NHS dentists (so called "dental deserts") [47], suggesting that those who may be entitled to free dental care may not be able to access a dentist. This is concerning given that dentists provide a potential route for early diagnosis of some HNC.

The finding that women are more likely to have been diagnosed from a dentist referral is supported by previous research which has shown women are more likely to have made an NHS dental appointment [48]. The association between being from an ethnic minority and diagnosed through a dentist is more striking. It has been reported that people from all minority ethnicity groups have greater mistrust of dentists, are less likely to have visited a dentist and, of those who have visited, are more likely to have done so because of a specific issue rather than a routine checkup [46]. Often, research which focusses on ethnicity and health outcomes is confounded by SES, which may not be controlled for in the analysis. However, our finding was apparent after adjusting for the effects of deprivation. It is consistent with results from a small study in London, which found that once SES had been considered, Asian people were far more likely, than white people, to have visited the dentist [49]. This is an area which would benefit from further investigation.

#### *4.4. Limitations*

This study had several known limitations associated with analyses of routine cancer registry data. Routine data sometimes have a significant amount of missing information and in this dataset, levels of missingness for summary stage and ethnicity were high. For the latter, this meant we could not explore whether there were differences between different ethnic minority groups, and further research on this topic would be of value. For the former, care is needed in making inferences from our findings. Completeness of stage details has improved over time in registry data, so subsequent studies would be of value to confirm the findings here. We took the decision not to exclude patients with missing information as the data were unlikely to be missing completely at random, and exclusion may have introduced bias. In addition, information was not available on risk factors for HNC, such as HPV status (which was not routinely tested for during the study period), and tobacco and alcohol use; these could be associated with patient diagnostic route. The registry provided a proxy variable for inferred HPV status based on tumour site and morphology, but this was not used in the final analysis as it was not more informative than cancer site alone. It is also likely that analyses are subject to residual confounding from comorbidities; the Charlson Comorbidity Index is a crude measure of the number of comorbidities that a patient has and only includes particular conditions documented during hospital admissions in a specified time period [50], so likely underestimates true levels of comorbidity. However, as comorbidities increase with age, and age was also included in the models, any residual confounding is likely somewhat mitigated.

Another important factor is that this data are from 2006 to 2014. While cancer pathways in England have not changed in the intervening years, it is possible that the frequency with which different routes are followed, or variations between socio-demographic groups, may have changed in the intervening years. In particular, the impact of the COVID-19 pandemic had a significant impact on cancer services; in England, urgent referrals decreased dramatically, and it is estimated that there will be substantial increases in cancer deaths due to delays in diagnosis and treatment [51,52]. Research investigating whether the inequalities in route to cancer diagnosis reported here have persisted since 2014, leading up to, during, and following the pandemic should be a priority. The current analysis could usefully serve as a baseline for such future work. Finally, these results may not be generalisable to all healthcare systems outside of England which may differ in terms of processing of diagnosis routes.

#### **5. Conclusions**

In conclusion, this population-based analysis of English cancer registry data showed significant socio-demographic inequalities in HNC routes to diagnosis. In many instances, groups who are already experiencing higher risk of HNC are further disadvantaged by inequalities inherent within their route to diagnosis. Understanding the reasons for these inequalities is the first step to being able to improve and speed the pathways to HNC diagnosis; this, in turn, would reduce inequalities and optimise patients' clinical outcomes.

**Supplementary Materials:** The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/ijerph192416723/s1, Figure S1: Route to diagnosis categorisation by each analysis, Table S1: Demographic and clinical characteristics of all individual HNC diagnosis routes during 2006–2014.

**Author Contributions:** Conceptualisation, L.S. and J.P.; Methodology, J.D. and L.S.; Formal analysis, J.D., L.S. and J.P.; Writing—original draft preparation, J.D. and R.N.; Writing review and editing, J.D., R.N., J.O., J.P. and L.S.; Supervision, L.S. and J.P.; Project administration, J.D.; Funding acquisition, L.S. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was conducted as part of JD's PhD, funded by Newcastle University. RN was funded by a grant from Cancer Research UK (A25618).

**Institutional Review Board Statement:** The study was conducted in accordance with the Declaration of Helsinki, and approved by the Institutional Review Board (or Ethics Committee) of Yorkshire and the Humber—South Yorkshire Research Ethics Committee (protocol code 206040; 16th November 2017).

**Informed Consent Statement:** Patient consent was waived due to only receiving pseudonymised data.

**Data Availability Statement:** The data used in this study were released to the authors for the purpose of this analysis; the authors are not permitted to share it. Interested individuals may apply for dataset including registrations of head and neck cancer over the study period from the current data controllers, NHS Digital.

**Acknowledgments:** This study was possible as a result of routine data collection by the NHS as part of standard cancer care and support. The data are collated and maintained by the NCRAS, previously part of PHE. Special mention is also noted to colleagues at the Office for Data Re-lease for their help during the data application process.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results.

#### **Abbreviations**

AIC: Akaike information criterion; CI(s): Confidence intervals; DCO: Death certificate only; GP: General practitioner; HES: Hospital Episode Statistics; HNC: Head and neck cancer; HPV: Human papilloma virus; IMD: Index of Multiple Deprivation; LRT: Likelihood ratio test; NCRAS: National Cancer Registration and Analysis Service; NCRD: National Cancer Registration Database; NCWT: National cancer waiting time; NHS: National Health Service; ODR: Office for Data Release; OR(s): Odds ratio(s); SES: socio-economic status; STROBE: Strengthening the Reporting of Observational Studies in Epidemiology; UK: United Kingdom; 2WW: Two week wait.

#### **References**


### *Article* **A Qualitative Evaluation of a** *Health Access Card* **for Refugees and Asylum Seekers in a City in Northern England**

**Malcolm Moffat 1,\*, Suzanne Nicholson 2, Joanne Darke 3, Melissa Brown 1, Stephen Minto 4, Sarah Sowden <sup>1</sup> and Judith Rankin <sup>1</sup>**


**Abstract:** Refugees and asylum seekers residing in the UK face multiple barriers to accessing healthcare. A *Health Access Card* information resource was launched in Newcastle upon Tyne in 2019 by Newcastle City Council, intended to guide refugees and asylum seekers living in the city, and the professional organisations that support them, to appropriate healthcare services provided locally. The aim of this qualitative evaluation was to explore service user and professional experiences of healthcare access and utilisation in Newcastle and perspectives on the *Health Access Card*. Eleven semi-structured interviews took place between February 2020 and March 2021. Participants provided diverse and compelling accounts of healthcare experiences and described cultural, financial and institutional barriers to care. Opportunities to improve healthcare access for these population groups included offering more bespoke support, additional language support, delivering training and education to healthcare professionals and reviewing the local support landscape to maximise the impact of collaboration and cross-sector working. Opportunities to improve the *Health Access Card* were also described, and these included providing translated versions and exploring the possibility of developing an accompanying digital resource.

**Keywords:** refugee; asylum seeker; health access; health information; intervention

#### **1. Introduction**

In 2021, more than 89,000,000 people worldwide were forcibly dispersed from their homes as a result of persecution, conflict, violence, human rights violations or events seriously disturbing public order [1]. This included more than 27,000,000 refugees (people who have fled war, violence, conflict or persecution and have crossed an international border to find safety in another country) and more than 4,000,000 asylum seekers (a refugee whose request for sanctuary has yet to be processed) [2,3]. A total of 56,495 people applied for asylum in the UK in 2021, and of the 14,572 applications that were processed 4083 (28%) were refused [4]. A refused asylum seeker refers to a person whose asylum claim has not been granted following initial review. By the end of 2021, 100,564 asylum seekers in the UK were still awaiting an initial decision on their asylum application [4]. Mass migration on this scale has the potential to significantly impact on healthcare provision in host countries. Research suggests that displaced migrants face multiple barriers, both structural and political, to accessing healthcare in their host countries, potentially leading to unmet need and poor-quality care [5]. Although the 1951 Convention relating to the Status of Refugees and its 1967 Protocol does not explicitly define refugees' right to healthcare, there is a more modern appreciation that access to healthcare should be regarded as a fundamental right for people seeking asylum, as is reflected in the International Organization for Migration's

**Citation:** Moffat, M.; Nicholson, S.; Darke, J.; Brown, M.; Minto, S.; Sowden, S.; Rankin, J. A Qualitative Evaluation of a *Health Access Card* for Refugees and Asylum Seekers in a City in Northern England. *Int. J. Environ. Res. Public Health* **2023**, *20*, 1429. https://doi.org/10.3390/ ijerph20021429

Academic Editor: Paul B. Tchounwou

Received: 28 October 2022 Revised: 3 January 2023 Accepted: 10 January 2023 Published: 12 January 2023

**Copyright:** © 2023 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

2019 Migration Governance Indicators and in the United Nation's Sustainable Development Goals [6–8]. Indeed, the UN Global Compact on Refugees states that "in line with national health care laws, policies and plans, and in support of host countries, States and relevant stakeholders will contribute resources and expertise to expand and enhance the quality of national health systems to facilitate access by refugees and host communities" [9].

In the UK, the National Health Service (NHS) is a comprehensive healthcare system providing care that is free at the point of access to all UK residents. Asylum seekers awaiting review and people who have been granted refugee status in the UK are entitled to free access to all elements of NHS care. Although some restrictions are placed on the NHS care that refused asylum seekers are entitled to use, access to primary care, Accident and Emergency and 111 services (telephone triage and advice) is free and universal. People who have been refused asylum may also use services providing treatment for specific infectious diseases, sexually transmitted diseases and treatment for conditions caused by torture, female genital mutilation and/or domestic/sexual violence. Access to health visitors, school nursing and family planning services should also be freely available to this population, as should end of life care services [10].

Despite these entitlements, it is known that refugees, asylum seekers and those whose asylum application has been refused are often inappropriately denied free UK NHS care, while some individuals may not seek it due to a lack of awareness [11]. Dominant themes that emerge from qualitative studies describing barriers to care among these populations include language barriers and inadequate access to interpreter services; limited understanding of the structure and function of the NHS; difficulty meeting the costs of dental care, prescription fees, and transport to appointments; an absence of timely and culturally sensitive mental health services, properly equipped to deal with the esoteric needs of these populations; and perceived feelings of discrimination relating to ethnicity, religion and immigration status alongside concerning professional attitudes [12–14]. This tension, between the expectation that people seeking asylum in the UK should have access to comprehensive healthcare and the reality of the often inadequate and incomplete care that many of them actually receive, has been at the heart of efforts to improve healthcare access in these populations. However, although barriers to care are well-documented, research examining the impact of interventions intended to address and overcome them is limited: a 2021 study explored qualitative perspectives on a pre-departure medical health assessment (MHA) for refugees accepted for resettlement prior to arrival in the UK, and the Doctors of the World Safe Surgeries initiative has championed improved access to primary care services for socially excluded groups including refugees and asylum seekers [15,16]. A systematic review published in 2021 explored primary care interventions delivered primarily in North American settings and found a lack of evaluations of community-focused approaches that might be expected to be particularly beneficial in these communities [17]. Evaluations of health information resources for these populations are seldom reported. With potentially more discriminatory changes to refugee and asylum seeker policy in the post-Brexit era, and with the passing of the Nationality and Borders Bill in 2022 that effectively criminalises asylum seekers who arrive in the UK via unregistered routes, an understanding of the challenges faced by these populations in accessing healthcare and of the opportunities offered by approaches that look to address some of these barriers is more important than ever.

In response to reports that new entrants to the North East of England (including refugees and asylum seekers) were struggling to navigate the local healthcare system, Newcastle City Council's (NCC) Public Health team, in collaboration with the Health and Race and Equality Forum (HAREF), Newcastle Council for Voluntary Services (NCVS) and the Regional Refugee Forum, developed and launched a *Health Access Card* for refugees and asylum seekers in 2019. Professional representatives from these and other organisations as well as refugees and asylum seekers living in Newcastle co-produced the resource, and 5000 copies of a folded, pocket-sized card were distributed to a number of key providers in healthcare and third sector organisations. The cards were made available in facilities

commonly accessed by members of these communities (such as GP practice waiting areas and community centres), and cards were also shared with professional staff in a number of organisations to be given out in person. It was intended that the card would be used by refugees and asylum seekers themselves, and also by professionals as an adjunct to conversations concerning healthcare access. An online pdf of the card is available to view at https://cdn.cityofsanctuary.org/uploads/sites/35/2019/05/Newcastle-Health-Access-Card-2.pdf, (accessed on 4 October 2019).

The aim of this qualitative evaluation was to explore perspectives on the impact and usefulness of the *Health Access Card* through semi-structured interviews with service users and professionals based in Newcastle, and to propose improvements and changes that could be considered for future iterations of the card. A secondary objective was to use these conversations as an opportunity to explore service user and professional perspectives on refugee and asylum seeker healthcare access and on barriers to care in Newcastle more generally.

#### **2. Methods**

Service users with refugee or asylum/failed asylum status and professional staff who work with these groups in Newcastle upon Tyne were invited to take part in a 30–40 min semi-structured interview with a researcher from Newcastle University. There were no exclusion criteria for participation if these basic eligibility criteria were met. Previous familiarity with the *Health Access Card* was not a prerequisite to participation. Participants were recruited through partner organisations including HAREF, the Action Foundation, Refugee Voices and NCC. Researchers had intended to recruit service user participants in person, by attending community group sessions and approaching potential participants with information about the study: this approach was not feasible following the implementation of COVID-19 lockdowns in March 2020, and potential service user participants were approached instead by staff in partner organisations. Although it was originally intended that professional participants would take part in focus group discussions, semi-structured interviews were undertaken instead in view of in-person group meetings being prohibited. Participants were provided with a study information leaflet to read prior to undertaking the interview and were asked to sign a consent form to confirm their participation. The information leaflet was only available in English, but interpreters were offered in circumstances where a potential participant who did not speak English expressed an interest in taking part in the study. Ultimately, interpreting services were not used—interviews with service users were conducted in English with their agreement. Two topic guides were prepared (for interviews with service user and professional participants), exploring participant perspectives on how the *Health Access Card* had been used, on the design and content of the card, and on how the card might be improved. Participants were also asked to describe their or their clients'/patients'/friends' experiences of healthcare in Newcastle, and to consider barriers to good care and opportunities to improve the healthcare offer for refugees and asylum seekers living in the city. Topic guides are included in the Supplementary Materials.

The first two interviews were conducted face to face; subsequent interviews took place online on the Microsoft Teams platform or by phone in view of COVID-19 lockdown restrictions. All interviews were conducted by MM; interviews were audio-recorded with participant consent and anonymised transcripts were produced by MM, MB and SM. Interview data were coded by MM using NVivo 12 software, and a thematic analysis of emergent themes was carried out. Thematic analysis is a flexible and intuitive method of qualitative analysis that involves examining multiple data outputs for recurring patterns and motifs that can be organised into themes [18]. Codes and emerging themes and subthemes from a sample of transcripts were also discussed by the research steering group (involving all authors). Illustrative quotations are provided, and we assign a P (professional) and SU (Service User) number for each quotation.

Ethical approval for the study was granted by Newcastle University Faculty of Medical Sciences (ref: 1847/18699/2019).

#### **3. Results**

Eleven participants took part in this qualitative evaluation between February 2020 and March 2021. Participant characteristics are described in Table 1.


**Table 1.** Participant characteristics.

Three themes and associated subthemes emerged from the participant interview data (see Figure 1).

**Figure 1.** Themes and subthemes from interviews with service users and professionals around the *Health Access Card*.

#### **4. Barriers to Healthcare for Refugees and Asylum Seekers**

#### *4.1. Experiences of Healthcare*

Where participants described positive experiences of using healthcare services in Newcastle, either their own experiences or those of friends, relatives or clients, these often related to interactions with individual clinical staff who had delivered tailored and compassionate care. GPs in particular were commended for some of their work with asylum seekers and refugees, and one professional participant highlighted the role that health visitors and midwives had played in arranging comprehensive care for their patients/clients. Referral to appropriate mental health support was mentioned as an example of good care:

"*We've had a few clients who have moved from the East end of the city over to the West end, and they won't change their GP because the GP they've got is great, they love their GP and they're wonderful and they don't want to change, so, and that GP has been quite happy to still have that person registered with them* . . . " (P1)

"*Health visitors and midwives are absolutely fantastic with this client group, yeah they really go above and beyond to try to help them as much as they can, so yeah I would say that's been really good, and I've done lots of joint working with midwives and health visitors* . . . " (P7)

Participants also described the important work done by third sector organisations in supporting refugees and asylum seekers to understand their healthcare rights, to access appropriate services, and to have recourse to non-NHS lifestyle and wellbeing support:

"*The [support organisation], that's for LBGT support for refugees and asylum seekers, I think, yeah it is it's on the bottom of the card, I suppose it's not physical health but more mental health support but they've been really good to work with, both for the [clients] and we've found them really helpful as well* . . . " (P3)

However, largely positive experiences of healthcare for refugees and asylum seekers were not necessarily replicated in different settings or for friends/family/clients, representing inconsistent care for this population group. One service user also described frustrations when initially trying to navigate the NHS system due to unfamiliarity with UK healthcare, but these frustrations were allayed when they became more aware of NHS practices:

"*At the start I was confused, but once I got used to the system it seems alright* ... *first time when I went to the emergency for example* ... *I used to wait really long time [and would think] oh my goodness what's that, but then I realised how the system works, how the patients are looked after* ... *but yeah, it was alright* . . . " (SU2)

All participants described direct or indirect awareness of refugees and asylum seekers in Newcastle having poor healthcare experiences and/or struggling to access the healthcare services and support that they needed. Several participants described negative experiences of accessing or utilising healthcare services that were related to language challenges. In particular, access to interpreters and translated literature was inconsistent and frequently inadequate, and these problems were exacerbated during the COVID-19 pandemic:

"*[The client] tried to ring up to get another prescription for, I think it was sleeping pills to help her because she struggles with anxiety and they said they can't, they can't have a meeting with her, the GP can't see her because they're not doing face to face [due to COVID-19 restrictions] and they don't have someone to be an interpreter so she'll just have to wait until after this.*" (P3)

Negative experiences of mental health management, and of accessing appropriate mental health support, were commonly reported, and these, again, were likely exacerbated by the pandemic:

"*I've got this client whose daughter is having quite severe psychotic episodes, she was admitted to hospital, she was, I don't know if it's like a young person's mental health team or something, she was under their care for a few weeks, and now nothing's happened, they haven't kind of followed up with anything, and she's had another quite serious episode in school where she was quite dangerous to other people, and I think she's just been, well I think they feel like she's been left* . . . " (P7)

Several participants described how they or their clients/patients were treated differently by healthcare services, often in terms of presenting complaints not being appropriately investigated/explored or of being expected to wait longer for review/access to services because of their asylum seeker/refugee status:

"*They feel like that because they don't understand, so they think because they are aliens [healthcare professionals] are reacting to them like that* ... *"they don't help us because we are foreigners"*" (SU1)

Experiences of problems in accessing appropriate and timely dental care were also widely reported:

"*It's always been dental care that they've had most problems with, so a lot of the dentists that they go to register with tell them that they've got to pay, even if they've got an HC2, I think for some of the dental practices maybe there's a misunderstanding with some of them* . . . " (P1)

Poor experiences relating to the restrictions placed on free access to healthcare services for refugees and asylum seekers were also described:

"*That lady I mentioned before who had the miscarriage, she was destitute at the time that happened and she got charged for the work they did when she had to be taken in for the miscarriage, and the letters are quite brutal in that they say that if you don't pay this within three months we will inform the Home Office and it will affect your claim, so people really panic and she found that really upsetting, having just gone through what she went through, and I went through Doctors of the World and they said yes the charges would stand currently.*" (P1)

#### *4.2. Cultural Barriers*

All participants described inadequate language support as a significant barrier to care. Often, this related to difficulties in accessing services due to problems arranging appointments or understanding what services were available; in other circumstances it related to interpreters not being offered or provided, or the standard of interpreter being inconsistent or inadequate; and sometimes it related to service users being sent or provided with literature/information about their care in a language that they were unable to understand:

"*We've had one learner who's sent me pictures of a letter from a consultant at a hospital to his GP, which he's been copied into, and it's all in English, he doesn't speak any English, so he hasn't received any kind of translated version of it, and it's talking, I mean I know those letters do talk in third person, I saw so-and-so, but it's really important the stuff that it's talking about, and describing that there was a communication barrier and that she's booked him in for a MRI scan anyway but she doesn't know how much he understands* . . . " (P3)

Cultural variation in the way in which service users understood and interpreted health and illness was also reported as a barrier to accessing care in a UK context. This was considered particularly important in the context of service users' often-extensive mental health needs:

"*I'm the only one who's accessed those services [counselling], because my family doesn't believe in mental health that much, like now they do but back then they didn't* ... " (SU3)

"*A lot of people come from cultures where mental health wouldn't be something that was even recognised as an issue, so being able to describe that in the first place [is a barrier]* . . . " (P4)

For a number of participants, limited understanding of and familiarity with the UK healthcare system and its practices and procedures in terms of referrals, waiting times, etc., was seen as a barrier to refugees and asylum seekers accessing healthcare services. Where

service users were perceived to access services inappropriately as a result of this, there was a feeling that this fuelled stigmatising and divisive rhetoric directed at these communities:

"*Another barrier is the assumption that people leaving their home and living here are used to the system, they assume that it's the same everywhere but it isn't, they'd be really frustrated if they went to my country for example and tried to access, which is not stressful for me as I understand it* . . . " (SU2)

#### *4.3. Financial Barriers*

Financial barriers to healthcare experienced by refugees and asylum seekers in Newcastle were commonly described. In some cases, this related to access to specific services for which failed asylum seekers would not be eligible for free care; in other cases, it related to the costs associated with, for example, travelling to an appointment, or, during COVID-19 lockdowns, paying for a supermarket delivery to reduce the likelihood of exposure to the virus in a public space:

"*A lot of [clients] are saying "oh well we walked for two hours to get here, if we'd paid for the bus we wouldn't have enough food for the day" so I think in terms of weighing up often it would be prioritising is it important enough to go to the doctors, more important than buying food for the week* . . . " (P3)

#### *4.4. Institutional Barriers*

Participants described institutional barriers to care that were rooted in NHS structures and in healthcare professionals' attitudes and behaviours. These barriers arose not only in relation to healthcare professionals' limited awareness of the healthcare rights and entitlements of these population groups, but also in relation to the professional understanding of the particular healthcare backgrounds of refugees and asylum seekers, many of whom have considerable mental and physical health needs arising from histories of trauma and/or torture:

"*I've worked with lots of people over the years who've kind of been in and out of various medical appointments talking about physical symptoms when actually it's turned out, you know talking about headaches talking about stomach aches talking about different things, and again some if that is their way of describing it with the language and the cultural barriers, and some of that is I think the practitioners' awareness of the particular needs of this population and that actually, you know, lots of people have experienced trauma and persecution and different things and actually for you to be aware of that, if people are coming in with headaches or different things, just explore some of that stuff* . . . " (P4)

#### **5. Opportunities to Improve the Healthcare Experiences of Refugees and Asylum Seekers**

#### *5.1. Bespoke Support*

Several participants proposed offering more individualised and bespoke healthcare support to refugees and asylum seekers in Newcastle, particularly upon arrival in the city as they begin to navigate the local health and care system for the first time. Offering this population the opportunity to visit healthcare premises under the supervision of a supportive and knowledgeable guide (from the client's accommodation provider or from a third sector/local authority organisation) was particularly well-supported:

"*Actually the most useful part was when a visitor came to my home and he actually took me and my husband to the places and he was speaking English but he tried to show how the places looked like, I didn't understand everything he said but it gave me the idea how to get used to the system, so he actually took us to the places and, yeah, tried really hard to explain* . . . " (SU2)

Additional, targeted mental health support was identified as a particular priority:

"*I think like there's more work to be done around people's mental health, really. Like a lot more work. It's such a huge area for our clients and for a lot of people, it's a really difficult thing to manage and to deal with and talk about*." (P2)

#### *5.2. Language Support*

Developing services that embed appropriate language support in every part of the healthcare pathway was identified as an opportunity to improve the healthcare experiences of these populations:

"*it's [the client's] right to be able to understand that information, so just having translated documents, and then people who understand the barriers that people might be experiencing when they're trying to access a service, would be really useful* . . . " (P3)

#### *5.3. Training for Healthcare Professionals*

Supporting healthcare staff to be more aware of the healthcare needs and personal circumstances of refugees and asylum seekers was identified as a priority opportunity by a number of participants. Good experiences of care almost always involved compassionate and understanding professionals, recognising and addressing the sometimes-unique needs of these population groups. Facilitating a better understanding of refugee and asylum seeker healthcare entitlement, among healthcare administrators as well as clinicians, was seen as an opportunity to improve care pathways, and potentially to encourage more empathy towards service users who have arrived in the UK in challenging circumstances. There was also a feeling that unconscious biases among healthcare staff should be challenged:

"*I think it's always a training issue. The more the surgery buys into training staff appropriately, the more awareness of someone's situation and know what to pick up on, the work that these surgeries have been doing is great, because you know there's a commitment there isn't there and their staff are going to have that level of training and you know they're kind of signed up to being welcoming to people and so, the more of that stuff that gets done the better, really.*" (P2)

#### *5.4. Funding and Capacity*

Several participants discussed the context in which healthcare access work with refugees and asylum seekers in Newcastle was currently funded and delivered and praised the role of the third sector in trying to support these populations' needs in a difficult economic climate. Trying to integrate and co-ordinate some of this work across the city's various networks was proposed as a local albeit imperfect response to ongoing challenges, in the absence of more comprehensive support from central government:

"*Because of various factors in Newcastle including years of you know austerity and funding cuts that have meant that lots of services that were previously there for asylum seekers and refugees are kind of contracted because of that, they [the voluntary sector] have been picking up asylum seekers much earlier in their journey so they've been picking up people who are just arriving in Newcastle, they've been picking up people who have been here and they haven't had their decisions yet and actually they've got lots of work around integration at those earliest levels around school access and health access and different things* . . . " (P4)

### **6. The** *Health Access Card*

#### *6.1. Content and Design*

Several participants commented favourably on the content of the information provided on the *Health Access Card*, mentioning in particular the useful information on the range of services available in an emergency situation and on how to access other services such as dentists and opticians:

"*Having that information on there that we can like point to, and make sure that people are informed and know how to use the kind of UK system of healthcare is really useful as well* . . . " (P3)

"*I think the information about the services is crucial, what service is provided at what time, and how, is important* . . . " (SU3)

However, one professional participant observed that the section on maternity care was potentially misleading, and should be revised and expanded:

"*I think what I'd like to see in there is something about midwives and health visitors as well, because clients don't always realise that when you're pregnant it's a midwife you would see most rather than your GP, so I think that section could be improved* ... " (P1)

The bright colours and effective use of images on the card was welcomed by participants:

"*It gives, like, images in its own self telling people that this is the thing you need to read if you're looking for an optician or a dentist because there's a picture of the glasses or a tooth, or the mother with the baby for pregnancy* ... *so I think it's easier in a graphical way* . . . " (SU3)

However, participants also felt that a lot of text on a small card might be off-putting to some people, and that the size of the card might result in it being missed or dismissed:

"*There is a lot of information on there now, obviously it's quite dense, there is a lot of writing on there, so anybody, even if they do speak reasonably good English anybody where English isn't a first language, it's probably going to be quite daunting*." (P5)

"*It doesn't really look like something important, so* ... *when you make something small, it doesn't look big, the best way to say it* ... *they don't make [you] take it serious[ly]* . . . " (SU1)

#### *6.2. Functionality and Distribution*

Participants commented favourably on the functionality of the card, with professionals reflecting on the usefulness of the resource as a signposting tool when having discussions about healthcare with clients and service users describing the card as compact and userfriendly. For several participants, if a client or friend expressed a particular healthcare need, the *Health Access Card* was a useful physical adjunct to verbal descriptions of what was available:

"*Clients in [name of charity], yes, some of them ask us how can I access this kind of healthcare, so I just gave it to them and, you know, point them to try this* . . . " (SU1)

However, for some professional participants, the fact that the card was only available in English was a barrier to them and their clients/patients using it more widely, and for one service user participant, the card was a poor substitute for the more bespoke and personalised support that a support worker might give.

Professional participants also described challenges in making the resource available to its target population—if they showed a service user the card, it would often be the first time they had seen it, and there was a feeling, among professionals and service users, that the card should be made available immediately upon arrival in Newcastle to maximise its usefulness. Professional participants rarely used the card in their own practice.

"*Frankly, when I came into the country, I wanted something like this, the only thing was there was nothing at that time* ... *now I know these places where I have to go.*" (SU3)

"*Most of the time when I've used it it's been the first time that someone has seen it, I've never given it out and someone's already had one or known about it.*" (P3)

#### *6.3. To Improve: Language and Content*

Participants recommended making the card available in other languages. Some participants also suggested that, if translated versions were not feasible, including a sentence or

two in the most commonly used languages in this population that directed service users to translation and interpreting services would be beneficial:

"*Another thing is, yeah, the best option would be if it is translated into as many languages as possible* . . . " (SU2)

One service user also cautioned against using abbreviations and acronyms that might not be familiar to people new to the UK/NHS:

"*I don't know what a GP is for example, I would add an explanation of this abbreviation* ... *it's still complicated, it's still really hard for me, everywhere abbreviations are used and it assumes that you know about it* . . . " (SU2)

Participants also emphasised the importance of including information on the card that was directly relevant to the experiences of refugees and asylum seekers, in particular with regards to bespoke services (including services that promoted physical and mental wellbeing as well as healthcare services) and to accessing interpreters if and when required:

"*Maybe if there were specific projects that were aimed at refugees they might be more accessible than just the general link to the website.*" (P3)

"*Something about certain organisations that are working with people from a similar [refugee and asylum] background* ... *so that they can tell them about the Health Access Card more properly, so that they can take them to the GP services because I know they have, like, health care champions* . . . " (SU3)

Participants were also keen for the card to provide some additional guidance on the nature and structure of the UK healthcare system:

"*In our countries doctors do not work like this so we just go to doctor and you can meet them right away, it's like a drop-in, you just go there and it's not ten minutes, so you can share whatever the problem is for as long as you want, but there are some different systems and they are expecting that to happen here but I've seen a lot of people, you know, complaining about how doctors are working.*" (SU1)

#### *6.4. To Improve: Presentation and Format*

Including more pictures on the card, to aid comprehension among service users with limited understanding of English, was suggested by several participants:

"*I think we need to make more use of images in explaining something, because images, you know, they're international, and there's a lot more graphic designers out there who can do, who can explain something graphically with a picture that people get the concept of* . . . " (P1)

"*I would like a bit more pictures, something more visual, rather than lots of writing I would add a bit more of this kind of pictures* ... *maybe I would add a bit, for example a tiny map, for example there is not many A&E department in Newcastle* . . . " (SU2)

Several participants proposed making the card available online as a digital resource, which would make it easier to provide translated versions and could include expanded content and links to other online resources:

"*Maybe consider having an expanded version of the cards available online, so you could have less text on the actual card. You could say, actually you can find more information on this website* . . . " (P5)

However, although it was suggested that the hardware required to access online resources was often available in these communities, reliable and consistent Wi-Fi access/data provision could be challenging:

"*The majority of patients, do have [internet] access, but you, of course, there are patients that don't either have devices or don't have data or Wi-Fi, so whenever you are providing service online you do have make sure that is also available in other formats as well.*" (P5)

#### **7. Discussion**

Participants in this study described diverse accounts of the healthcare experiences of refugees and asylum seekers in Newcastle. Good experiences tended to occur on account of healthcare and other support professionals providing compassionate, personalised care, but these experiences were not consistent. Negative experiences often related to challenges in accessing care in a language that service users were able to understand, difficulties navigating an unfamiliar healthcare system, and frustrations around the availability of dental care in particular. Related to these experiences, barriers to healthcare among these populations included inadequate language support, cultural unfamiliarity with NHS structures and processes, financial barriers (including the costs of travel to healthcare premises as well as the costs of accessing health and wellbeing services themselves), and inconsistent and often-poor understanding of refugee and asylum seeker healthcare needs and entitlements among healthcare professionals. Opportunities to address some of these barriers included offering more bespoke healthcare support to refugees and asylum seekers (particularly to new arrivals to the city), embedding language support at every stage of NHS care pathways, enabling healthcare professionals to better-understand refugee and asylum seeker care needs through additional training, and working with the range of excellent third sector organisations in Newcastle to co-ordinate the important services that they provide.

The Newcastle *Health Access Card* was found to be an effective and user-friendly resource for refugees and asylum seekers in the Newcastle upon Tyne that presents helpful content in a well-designed format. The information describing emergency/urgent care services and dental provision was especially welcome, and study participants appreciated the engaging use of bright colours and graphics. Professional participants described the card as a useful *aide memoire* during conversations about healthcare with service users. However, there was a sense that the information on the card describing services that are targeted at the refugee and asylum seeker population could be expanded, and that not making the card available in non-English translations was a barrier to it being utilised more widely. It was also suggested that the density of text on the card might be off-putting to service users with limited English language skills, and that the card had not necessarily reached the service users who might benefit most from having sight of it. Recommendations for policy and practice for future versions of the *Health Access Card* include for it to be made available in other languages and to avoid abbreviations/acronyms where possible; to consider reducing the volume of text and replacing some of this with additional pictures and graphics; to expand the content describing third sector support available to these population groups; and to explore the possibility of launching a more interactive, digital resource as an alternative to the paper version.

These findings should represent a call to action for those responsible for healthcare policy and practice in the Newcastle upon Tyne to tackle the barriers to healthcare experienced by refugees and asylum seekers and to spearhead the development of inclusive, culturally sensitive and responsive services in light of the professional and service user experiences described above, but they come at a time of political turmoil and economic uncertainty. Although the war in Ukraine and the resurgence of the Taliban in Afghanistan has inspired a compassionate and generous response to refugees and asylum seekers among many parts of the UK public, media portrayals of these populations remain provocative, and it is likely that public and overt efforts to improve access to healthcare services for these groups in particular, at a time of waiting list and workforce crises in the NHS, would be highly contentious [19].

In this challenging context, a population approach cognisant of the importance of the wider determinants of health is more important than ever. A 2019 study examining refugee integration in Newcastle highlighted the social barriers to integration in a post-austerity (and pre-COVID) context, and in many instances these barriers to integration also represent barriers to effective access to healthcare resources [20]. The mental health experiences of asylum seekers in Newcastle last appeared in the research literature seventeen years ago, and many of the barriers and experiences described in 2005 are repeated in this study [21]. The author of that paper also highlights the important role that social and economic circumstances play in determining (mental) health and wellbeing. An ethnopsychiatric approach to identifying and treating mental health presentations among migrant populations, that positions mental health in its appropriate cultural context, offers a more nuanced and patient-centred response to the significant burden of mental health need described in this paper, but delivering services of this nature places demands on providers that may, in the current UK healthcare climate, be unachievable [22].

The opportunity described in this study, to map and integrate the range of services currently provided by a range of local authority, NHS and third sector organisations, so as to better understand the comprehensiveness of current support and to deliver a more joined-up and consistent offer to refugees and asylum seekers, is persuasive in this context—as in other populations, people who are economically secure and socially connected are more likely to have better health and better healthcare access. Healthcare providers should facilitate and participate in these interdisciplinary conversations and should review how the support that they offer to these populations can be improved. This may include relatively minor changes such as ensuring that clinic letters describing management plans and test results are offered in translated versions, and upskilling patientfacing and administrative staff to be more aware of these patients' complex medical histories and healthcare entitlements. For those involved in developing health information resources for refugees and asylum seekers, this study demonstrates that there is an appetite for digital resources among these population groups, and previous research has identified a positive role for, for example, social media in supporting refugee youth to navigate and understand health systems in host countries [23]. The findings of the study also suggest that a resource that simply describes available services is potentially of less value than one that guides service users to third sector and other organisations that are able to provide more personalized and bespoke healthcare access support.

The period during which this study was undertaken, coinciding with the onset of the COVID-19 pandemic in February/March 2020 and with qualitative data collection continuing during UK lockdowns in the months that followed, served to shine a light on refugee healthcare access during crisis situations. The move away from face to face/inperson care exacerbated the barriers to healthcare already experienced by these populations, and in many instances removed meaningful access to support networks that previously would have looked to enable links to healthcare services. This had a harmful impact on service users' physical and, in particular, mental health, and services were slow to respond to a rapidly evolving situation. These findings are in keeping with the research evidence presented elsewhere [24]. However, the shift to online provision of some services also served to act as impetus to providers and support groups to improve their clients' digital access capabilities, and this, if sustained, offers healthcare organisations an opportunity to reconsider the means by which they engage with refugees and asylum seekers, and to look to overcome some of the barriers to care described above. The more specific experiences of refugees and asylum seekers in relation to health protection policies implemented to manage the pandemic response—including testing, quarantining and care and support for people required to isolate—were not explored in this study, but it is known that the accommodation given to vulnerable populations during the COVID-19 pandemic often fell short of providing a safe environment that was conducive to good population health management [25]. It is known anecdotally that similar challenges were faced by asylum seekers housed in temporary accommodation in Newcastle, and these experiences should be explored and documented and should inform the work of local authority and health protection practitioners in the event of future public health emergencies.

The strengths of this study include the range of professional partners involved in the research design and recruitment, the diverse sample of professional participants working in various important roles, the in-depth exploration of stories and experiences using semistructured interviews, and the robust thematic analysis involving several members of the research team. The study has significant limitations—due to pandemic restrictions and the

impact of these on recruitment, only three service participants were able to participate in semi-structured interviews, and those that did participate spoke good English and were well-established in Newcastle with settled refugee status, potentially unrepresentative of those with the most acute and urgent healthcare access needs. Several important voices, such as those of asylum seekers awaiting a Home Office decision and those of children, were not explored in this study. It is known, for example, that unaccompanied refugee minors are more likely to present with PTSD and other mental health conditions than children arriving in a host country with parents/other adult carers: the healthcare access experiences of this population are unlikely to be adequately represented in the findings of the current study [26]. Participants also had very limited experience of using the *Health Access Card* themselves or, in the case of professional participants, in their own practice, and while this is perhaps indicative of some of the challenges associated with the effective dissemination of the card, it makes any discussion of how the resource was used and the impact that it may have had impossible in the context of the current study.

#### **8. Conclusions**

This study sheds light on the impact of a simple but potentially wide-reaching health information resource for population groups that experience multiple complex barriers to healthcare, with important recommendations as to how the resource might be improved and expanded. It is the first study to consider the physical and mental healthcare needs of refugees and asylum seekers in Newcastle and the first to evaluate a bespoke healthcare access resource targeted at these groups, and the findings described herein are likely to be generalizable to over settings. It explores service user perspectives on barriers to healthcare alongside professional voices with extensive experience of the local and regional health and social care system, and the emerging themes complement the existing literature and offer an expanded exploration of cultural and language barriers in an urban UK context.

By virtue of the period during which interviews were conducted, the study was also able to consider refugee and asylum seeker healthcare experiences in the context of the COVID-19 pandemic. The healthcare access needs and experiences of refugees and asylum seekers newly arrived in the city who do not speak English are likely to be more extensive and complex, and future research should prioritise hearing these voices, as well as exploring the experiences of children and, in particular, unaccompanied minors. Researchers should also explore the healthcare access experiences of Ukrainian refugees, a group that was welcomed into Britain in large numbers as part of a national "Homes for Ukraine" scheme following Russia's invasion of Ukraine in 2022, but which we anecdotally know has faced similar healthcare access challenges to those described above.

**Supplementary Materials:** The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/ijerph20021429/s1, Supplementary Materials: Service user and professional participant topic guides.

**Author Contributions:** Conceptualisation: M.M., S.N., J.D., M.B., S.M., S.S. and J.R.; Methodology: M.M., S.N., J.D., M.B., S.M., S.S. and J.R.; Formal Analysis, M.M.; Investigation: M.M.; Resources: M.M., S.N. and J.D.; Data Curation: M.M., M.B. and S.M.; Writing—Original Draft Preparation: M.M.; Writing—Review and Editing: M.M., S.N., J.D., M.B., S.M., S.S. and J.R.; Visualisation: M.M.; Supervision: S.S. and J.R.; Project Administration: M.M., S.N. and J.D. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding. J.R. and S.S. are part-funded by the National Institute of Health and Care Research (NIHR) Applied Research Collaboration (ARC) North-East and North Cumbria (NIHR200173). S.S. is supported by Health Education England (HEE) and the National Institute for Health Research (NIHR) through an Integrated Clinical Academic Lecturer Fellowship (Ref CA-CL-2018-04-ST2-010) and RCF funding, NHS North of England Care System Support (NECS).

**Institutional Review Board Statement:** The study was conducted according to the guidelines of the Declaration of Helsinki and approved by the Ethics Committee of Newcastle University Faculty of Medical Sciences (protocol code 1847/18699/2019; date of approval 14 February 2020).

**Informed Consent Statement:** Informed consent was obtained from all subjects involved in the study.

**Data Availability Statement:** The data presented in this study are available on request from the corresponding author. The data are not publicly available due to participant confidentiality.

**Acknowledgments:** The authors acknowledge the generous contribution of the research participants and of the organisations in Newcastle that supported with recruitment to this study.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Disclaimer/Publisher's Note:** The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

### *Protocol* **Understanding the Lives of Aboriginal and Torres Strait Islander Women with Traumatic Brain Injury from Family Violence in Australia: A Qualitative Study Protocol**

**Michelle S. Fitts 1,2,3,\*, Jennifer Cullen 4,5, Gail Kingston 6, Yasmin Johnson 1, Elaine Wills <sup>1</sup> and Karen Soldatic 1,7**


**Abstract:** Globally, there is growing recognition of the connection between violence and head injuries. At present, little qualitative research exists around how surviving this experience impacts everyday life for women, particularly Aboriginal and Torres Strait Islander women. This project aims to explore the nature and context of these women's lives including living with the injury and to identify their needs and priorities during recovery. This 3-year exploratory project is being conducted across three Australian jurisdictions (Queensland, Northern Territory, and New South Wales). Qualitative interviews and discussion groups will be conducted with four key groups: Aboriginal and Torres Strait Islander women (aged 18+) who have acquired a head injury through family violence; their family members and/or carers; and hospital staff as well as government and non-government service providers who work with women who have experienced family violence. Nominated staff within community-based service providers will support the promotion of the project to women who have acquired a head injury through family violence. Hospital staff and service providers will be recruited using purposive and snowball sampling. Transcripts and fieldnotes will be analysed using narrative and descriptive phenomenological approaches. Reflection and research knowledge exchange and translation will be undertaken through service provider workshops.

**Keywords:** women; traumatic brain injury; violence; Australia; Aboriginal and Torres Strait Islander; care systems

#### **1. Introduction**

Exposure to violence has serious health outcomes for women [1], with the elimination of violence against women and children a recognised national priority in Australia [2,3]. Due to the recurrent nature of family violence [4], women are vulnerable to sustaining injuries that impact upon the functioning of their brain. Traumatic brain injury (TBI) is defined as damage to, or alteration of, brain function due to a blow or force to the head [5]. A subset of acquired brain injury (ABI), the experience of TBI is unique and can consist of various short- and long-term cognitive impacts as well as psychological and physical consequences. These changes can include memory loss, difficulty with motivation, lack of insight, sensory and perceptual problems, posttraumatic epilepsy, fatigue and sleep difficulties, mood changes, and anxiety [6–9]. Even mild TBI is a risk factor for the development of early onset dementia and other chronic health conditions [10]. Although there has been growing recognition of the intersection between family violence and head injury both in Australia and worldwide [11,12], this has yet to translate into significant

**Citation:** Fitts, M.S.; Cullen, J.; Kingston, G.; Johnson, Y.; Wills, E.; Soldatic, K. Understanding the Lives of Aboriginal and Torres Strait Islander Women with Traumatic Brain Injury from Family Violence in Australia: A Qualitative Study Protocol. *Int. J. Environ. Res. Public Health* **2023**, *20*, 1607. https:// doi.org/10.3390/ijerph20021607

Academic Editors: Jessica Sheringham and Sarah Sowden

Received: 25 October 2022 Revised: 8 January 2023 Accepted: 11 January 2023 Published: 16 January 2023

**Copyright:** © 2023 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

research action to listen to the voices of Aboriginal and Torres Strait Islander women who have acquired a head injury or been diagnosed with a TBI connected to family violence. The development of new knowledge is critical for informing robust, evidence-based family violence, disability and health policy, and practice.

Indigenous women in Australia and other settler nations such as Aotearoa New Zealand and the United States of America experience higher levels of violence than their non-Indigenous counterparts. Compared to other Australian women, Aboriginal and Torres Strait Islander women are more likely to be hospitalised due to violence and die due to injuries sustained from violence [13–15]. Concerningly, Aboriginal and Torres Strait Islander women in remote and very remote areas are more likely to be hospitalised for family violence than their urban counterparts (26.5 versus 2.8 per 1000) [13]. Violence against Aboriginal and Torres Strait Islander women is perpetuated by men of all cultural backgrounds including non-Indigenous men. In response to these high rates of violence, Aboriginal and Torres Strait Islander community leaders have led actions to prevent and address family violence [16,17]. Such trends are similarly reflected in assault-related head injury presentations [5,18]. One study found Aboriginal and Torres Strait Islander women are hospitalised with a head injury due to assault (1999–2005) 69 times the rate of other Australian women [5]. Factors such as alcohol are associated with TBI experienced by Indigenous peoples [19]. However, further research to understand the potential contribution of alcohol with TBI occurrence is warranted [20]. Incomplete and inaccurate data collection, together with the underreporting of violence by survivors of violence, also suggests that the published rates are, at best, an indication of the extent of violence experienced by women [21–23].

The terms domestic violence, family violence, and intimate partner violence are often used to describe violence against women. Although these terms are used interchangeably, the preferred term for Aboriginal and Torres Strait Islander communities tends to be family violence, as it encapsulates the extended nature of Aboriginal and Torres Strait Islander families and the kinship relationship within which a range of forms of violence occur [24]. For Aboriginal and Torres Strait Islander communities, family violence is a complex issue and must be seen in the context of wide-scale colonisation involving oppression, dispossession, massacres, the removal of children, and the loss of linguistic and cultural authority combined with the direct consequences of these policies on Aboriginal and Torres Strait Islander communities. Impacts include intergenerational trauma, economic and housing stress, unemployment as well as alcohol and other drug misuse [25–27].

Despite Aboriginal and Torres Strait Islander women experiencing high rates of head injury connected to family violence [5], research to document and examine their lived experiences remains scarce. Two reviews of TBI research with Indigenous populations identified a lack a studies investigating the lived experiences of Indigenous women in Aotearoa New Zealand, Australia, Canada, and the United States of America with head injury connected to violence [19,28]. Since then, two large scale Australian studies, 'The transition from hospital to home' (2016–2018; Queensland and Northern Territory) and 'Healing Right Way' (2017–2021; Western Australia), focused only on the transition period of Aboriginal and Torres Strait Islander peoples, which consists of hospital admission through to discharge and return to community and country [29–31]. Culturally responsive community rehabilitation models and resources were developed and implemented from these works [32,33]. Despite these promising advancements, the research area has narrowly focused on those patients who accessed hospital care alone. Thus, the models of rehabilitation developed are general in nature and may not recognise the unique issues that emerge when TBI is a direct outcome of family violence. Women who live with an acquired head injury as a result of family violence may not access health care, with fear shaping their daily lives and ability to seek help and access resources [34]. Other barriers to accessing health care include marginalisation, shame and stigma, and worries about confidentiality in tight-knit communities [31,35,36].

#### *Study Aims*

Although there are distinct bodies of literature examining family violence and TBI that demonstrate Aboriginal and Torres Strait Islander women experience high rates of both phenomena, the intersectionality of TBI and family violence has been overlooked [34]. Appropriate, effective, and equitable access to service providers that support and address the unmet needs and priorities of these women and their families has the potential to reduce the high incidence of head injury rates experienced by women and improve their health and well-being across their life course. With the increased recognition of both family violence and TBI in national initiatives [2], it is now critical to document and understand the beliefs, perceptions, and experiences of Aboriginal and Torres Strait Islander women with acquired head injury in the context of family violence. The aims of this project are to:


#### **2. Methods**

#### *2.1. Setting*

This project will be undertaken at three sites in three Australian jurisdictions (Queensland, Northern Territory, and New South Wales). Family violence research demonstrates recognition of the heterogeneity in the experiences of survivors and service providers across regional and remote areas [37]. Therefore, the involvement of the three locations will help to better understand the effects of social and geographical isolation on the ability of women to disclose, report, and seek help about family violence. In addition, these sites were selected to understand the impact of different programs, policies, and geography for women who live with head injury and the services who support them. Together, this will allow for a comparative analysis and close examination of three different jurisdictions. The population of project location 1 is approximately 25,000, with an Aboriginal and Torres Strait population of 4361 (17.6%). The population of project location 2 is approximately 230,000 with 18,008 (7.9%) of residents identifying as Aboriginal and/or Torres Islander [38]. In project location 3, approximately 1.4% (*n* = 13,426) of the total population (*n* = 936,433) identify as Aboriginal and/or Torres Strait Islander.

#### *2.2. Conceptual and Theoretical Innovation*

This project will draw upon socially-embedded phenomenology to develop a richer and more sophisticated understanding of the way acquired head injury disrupts a person's embodied being in the world. Exploration of the social creation of impairment through inequality, deconstructing the cultural construction of impairment, and analysing the personal significance of impairment identities is required to build upon existing research knowledge [39]. While previous TBI studies have made important contributions to understanding disability through TBI, under the medical model, there is a strong focus on outcomes that are valued within a Western framework. Indeed, under the medical model, the emphasis is on 'fixing' the impairment, so that people can 'function' in society as active participants (such as returning to employment) [40]. A general assumption is made that illnesses and disabilities are universal and invariant to the cultural and social contexts in which they exist [41]. In turn, the research and guidance for workforces who support these women are limited, with the intersections of cultural identity potentially complicating what

is already a complex issue. Within this project, an understanding of how different intersectionalities [42] such as gender, cultural belonging, and geographical location contribute to the women's experiences of services and systems after acquiring a head injury through family violence will be explored.

#### *2.3. Research Governance*

Aligning with the national research guidelines [43], the project will ensure Aboriginal and Torres Strait Islander peoples are involved in all aspects of the project. An advisory group consisting of members that represent a selection of service providers, community groups and hospitals participating in the project as well as representatives from national disability advocacy groups will meet twice a year (by teleconference). A critically reflective process will be completed at each site, entailing ongoing adjustments to the research process and incorporating iterative feedback into the research approach undertaken at each site [44,45]. This will enable the process to be locally guided to ensure that processes are responsive to local community requirements and local cultural protocols across the project, and simultaneously draw out comparative and contrasting aspects across each of the sites to inform a comprehensive interpretative analysis of the findings [46]. Individual advisors including Aboriginal and Torres Strait Islander women with lived experience of violence-related TBI will also help guide the project.

#### *2.4. Definition of Traumatic Brain Injury*

There is not one consistent definition used to define a TBI [47]. The inclusion criteria for the project are broad to include women with different severity levels of TBI. The project will define mild TBI (which can also be referred to as a head injury) as trauma to the head that is severe enough to cause neurological symptoms (including sensitivity to light, headache, and nausea). A moderate to severe TBI can be identified by one of the following: (1) loss of conscious for any duration, OR (2) post-traumatic amnesia > 24 h, OR (3) injury verified on a computerized tomography (CT) scan or magnetic resonance imaging (MRI). All major hospitals in the three project locations have CT and MRI facilities. The broad criteria also account for the suite of factors that can reduce accessibility to health care and specialist services following a TBI for women living in regional and remote communities [48,49].

#### *2.5. Participants and Recruitment*

The participants are outlined in Table 1. Aligning with the qualitative nature of the project, a non-probability sampling approach will be adopted, and no sample size calculation will be carried out [50]. The sampling aims to include information-rich cases and achieve in-depth understanding of the phenomena being explored rather than striving to meet a specific (statistically determined) sample size [51]. Sampling aims to reach a diverse range of participants (including age and location) and achieve data saturation across themes. The sample size of each group has also been determined by practical considerations including timeframes of qualitative fieldwork and reasonable workload requests for service providers who will support the recruitment of women who have experienced a head injury through family violence.

#### 2.5.1. Women Who Have Experienced a Head Injury through Family Violence

To preserve the safety of women, the research team will work closely with frontline service providers and community groups within each project site to purposefully sample women who fit the criteria. The broad inclusion criteria for this study are women who: (1) identify as Aboriginal and Torres Strait Islander; (2) are aged 18+; and (3) have experienced a head injury (or been diagnosed with a TBI) as a direct consequence of family violence. The recruitment of women will continue until data saturation is reached [52]. Frontline service providers are well-equipped to identify potential participants, with some service providers (such as legal and health services) having access to medical discharge summaries that confirm whether their patient or client would meet the study criteria. Other

services have active risk management plans and communication strategies in place with their clients, which makes them well-equipped to have knowledge and awareness of the different factors that may place each client at greater or lesser risk to participate in the study. Nominated service staff (such as women's groups coordinators, medical practitioners, case workers, and lawyers) will identify patients or clients who meet the eligibility criteria. Once a clinician has identified a patient that meets the eligibility criteria, a member of the research team will be notified. To ensure consent is conducted in an appropriate manner, the research team member will approach the patient or client with the assistance of an Aboriginal Interpreter Translator (where necessary) to explain the study fully. Potential participants will be provided with written information about the study, a short video about the project, face-to-face discussion with a research team member, and given an opportunity to ask questions about the project. To support nominated staff from service providers to identify women who meet the criteria, TBI education sessions will be delivered to participating service providers by two Aboriginal and Torres Strait Islander educators from a national brain injury organisation [53]. Aboriginal Interpreter Translators will also be employed to assist with data collection.


**Table 1.** Summary of participants, sampling, recruitment, and data collection.

#### 2.5.2. Family Members and Carers

Women who have experienced a head injury through family violence will be asked to nominate a family member or carer (aged 18+) to take part in an interview. Family members and carers can nominate to take part in an interview or small discussion group. Women with acquired head injury related to family violence can participate in the study without nominating a family member or carer.

#### 2.5.3. Hospital Staff

Direct invitations will be made to hospital staff and a 'snowballing' approach will also be used for recruitment, where participants will be able to recommend other staff to approach. A variety of disciplines will be targeted with the aim to include individuals who have lived experience of working with women who present to the hospital with head injury connected to family violence, their families, and perpetrators of family violence, and those who provide medical treatment to women who have sustained a head injury. Potential staff that will participate in an interview for the project include Indigenous hospital liaison officers, Aboriginal health workers, specialists, nurses, occupational therapists, physiotherapists and social workers.

#### 2.5.4. Service Professionals, Community Groups, and National Advocacy Groups

Frontline workers, community-based groups, and advocacy groups will be invited to participate in an individual interview or small discussion group. Frontline workers will represent primary health care (both Aboriginal community-controlled and government), the legal and justice sector (including magistrates, lawyers, victims support), housing and crisis accommodation, disability support services, and family violence services. Examples of community groups include women's and Elder groups. Representatives from national and regional disability and brain injury advocacy groups will also be invited to participate in an individual interview. This previously used approach [54] involves the selection of experts based on their knowledge and experience of the core research issue [55]. Prioritising contact with Aboriginal and Torres Strait Islander community groups and organisations, a list of potential participants will be developed by the research team. Sampling will occur through the networks of the research team and a 'snowballing' approach, with participants asked to recommend other relevant individuals and agencies to participate in the research [51]. As presented in Table 2, the interview protocol for service professionals aims to understand the delivery and practice of service models as well as to understand the knowledge of service professionals related to TBI and family violence.

**Table 2.** Overview of the interview and discussion group topics.


#### *2.6. Data Collection*

Interviews and discussion groups will be conducted by Aboriginal and non-Indigenous research team members. A semi-structured interview guide covering the topics listed in Table 2 will be used, but questions will also be informed by observations and new topics raised by the participant. Participants will have the opportunity to 'tell their story'. These methods have been rigorously applied in previous research with Aboriginal and Torres Strait Islander women and are drawn upon here as they can: (a) capture grounded subjective experiences and practices occurring locally and (b) effectively support the participation of highly marginalised groups who have a diverse range of skills, knowledge, and educational attainment [56]. Yarning is also a feature of Aboriginal and Torres Strait Islander convention for passing on information through informal conversations, reflecting the oral traditions that support the transmission of knowledge among Aboriginal and Torres Strait Islander peoples [57]. The interview guide was developed through a multi-phase process involving - aligning the interview questions with research questions, receiving feedback on the interview schedules, and piloting of the interview schedules [58].

#### *2.7. Workshops*

Once data have been collected from across all studies, service providers will be invited to participate in one-day workshops. Based on participatory models of qualitative research methods, discussion questions will be designed to elicit in-depth information about the existing knowledge and gather information and recommendations for the next steps in research, practice, and knowledge dissemination. Breakout groups will include at least one representative from different types of service providers and community groups to achieve triangulation of the data. Session summaries of the workshops with services will be presented to the collective group at each location to identify key themes and prioritise next steps. Discussions will be recorded in written and audio formats.

#### *2.8. Data Analysis*

All audio recorded interviews, discussion groups, and workshops will be transcribed verbatim. Transcripts, fieldnotes, and observations will be managed with NVivo 12 [59]. The transcripts will be analysed using a combination of narrative and descriptive phenomenological analyses. The aim of descriptive phenomenology is to describe particular phenomena, or the appearance of things, as lived experience [60]. The process is inductive and descriptive and seeks to record experiences from the viewpoint of the individual who had them without imposing a specific theoretical or conceptual framework on the study prior to collecting data [61]. The narrative analysis will focus on sense-making and Aboriginal and Torres Strait Islander women's changing identity and role post TBI from family violence. The two methodological approaches complement each other in terms of gaining knowledge of 'breadth' (narrative identity) and 'depth' (lived experiences), giving some support for a philosophical position that shows a person as both an active and passive agent, constructively making sense of their narrative identity as well as being constructed by their lived experiences. A constant comparative technique will be employed to systematically organise, compare, and understand the similarities and differences across the different participating groups and field sites, critically enriching the analysis and providing a substantive basis for theoretical extrapolation and affording critical points of comparative analysis across each of the locations.

#### *2.9. Ethics*

Ethics approval for this research has been obtained from the Central Australian Human Research Ethics Committee (CA-21-4160), Western Sydney University Human Research Ethics Committee (H14646), Townsville Hospital and Health Service Human Research Ethics Committee (HREC/QTHS/85271 and HREC/QTHS/88044), and the Aboriginal Health and Medical Research Council of New South Wales Human Research Ethics Committee (1922/22).

#### *2.10. Consent*

Participants will provide voluntary written informed consent. For women who experience severe impairment or are under a guardianship order, consent from a proxy for research participation from a person responsible for the person (such as a carer or guardian) will be sought. Participants may withdraw from the study at any time before dissemination of the findings, except for the discussion group participants, as it will be difficult to identify individual voices in the recording and transcript. Discussion group participants will be informed of this before consenting to the study. Participants will not be deceived in any way about the study objectives. All information regarding the study will be provided verbally and in writing prior to the interview. To minimise the risk of stigmatisation of potential participants and to ensure information regarding the study is understood fully by the participants, a flipchart that uses images and plain, easy English to describe key aspects of the project (e.g., what is a brain injury, reasons for the study, and participant rights) as well as video resources developed in a previous TBI study will be used [33].

#### *2.11. Potential Benefits and Risks*

Self-awareness plays an essential role in TBI rehabilitation and can impact motivation, safety, and rehabilitation goals during recovery [62]. Through self-exploration of their lived experiences, some participants may be able to fully explore their experiences, more fully explore their circumstances, and may also gain a new perspective. However, the recall of traumatic experiences by women and their family members may also cause discomfort or distress. Drawing upon the guidelines for working with women who have experienced family violence, all decisions within the research process will be driven by an awareness of the safety and ethical considerations [63–65]. Several strategies will be implemented to minimise any potential risk, and to identify the different levels of risk, of the women being identified as participating in the study. Promotion and recruitment through only service providers recognises the importance of recruiting women that are already connected with one or more of the participating services with access to ongoing support. Service providers have an existing awareness of the current life circumstances of the women referred to the project (e.g., if women are living in a high-risk environment). Existing connections between services and women who have experienced family members will also enable the research team to complete immediate referrals (with permission of the participant) back to the service for support should the women disclose that they are at risk of violence or have been identified by other community members as participating in the study. Other safeguards implemented for the safety and well-being of women and carers/support persons who take part in the study include the organisation of an experienced counsellor when interviews and discussion groups are conducted, and follow-up contact with each participant shortly after data collection.

#### *2.12. Dissemination*

A dissemination plan implemented within the project will ensure that the research findings and dissemination activities are controlled by Aboriginal and Torres Strait Islander peoples and service providers and ensure that the findings are disseminated throughout the course of the project in appropriate formats for the stakeholders. Stakeholders include research participants, government and non-government service providers, the health and legal sectors, state/territory and federal policymakers, the academic community, and advocacy groups. Findings will be disseminated through conference presentations and peer-review publications. Further dissemination activities will be determined in partnership with advisory groups and other service providers through the recommendations made in the workshops. Some of the expected dissemination formats include:


While this study will directly inform policy and practice within Queensland, the Northern Territory, and New South Wales, the findings will be disseminated to other relevant states/territory service providers, government ministers, and advocacy groups beyond these jurisdictions to ensure that an applicable national-level strategy can be shared. Publications will adhere to the CONSolIDated critERia for strengthening the reporting of health research involving Indigenous Peoples (the CONSIDER statement) [66] as well as the Consolidated Criteria for Reporting Qualitative Research [67].

#### **3. Conclusions**

This qualitative project will comprehensively explore and document the strengths, challenges, and nuances in the day to day lives of Aboriginal and Torres Strait Islander women with an acquired head injury connected to family violence. Through partnerships with key services, the evidence generated will enable service providers that work with these

women to better develop and tailor their services, programs, and workforce to support Aboriginal and Torres Strait Islander women and their families. The evidence may also help to inform resource allocation and provide vital information for governments to support the planning and development of equitable, holistic, appropriate care, and support that reflects the needs and priorities of Aboriginal and Torres Strait Islander Australian women experiencing head injury in the context of family violence. The evidence generated from this project is a critical step in addressing the unacceptable rates of head injury as a result of family violence among Aboriginal and Torres Strait Islander women.

**Author Contributions:** Conceptualisation, M.S.F.; Methodology, M.S.F., K.S., J.C. and G.K. Writing, original draft, M.S.F.; Writing–review and editing, K.S., J.C., G.K., Y.J. and E.W.; Funding acquisition, M.S.F. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by the Australian Research Council via a Discovery Early Career Research Award for M Fitts (#210100639). The funders have no role in the study design, data collection and analysis, decision to publish or preparation of the manuscript. The views expressed in this publication are those of the authors and do not necessarily reflect the views of the funders.

**Institutional Review Board Statement:** The Central Australian Human Research Ethics Committee (CA-21-4160), Western Sydney University Human Research Ethics Committee (H14646), Townsville Hospital and Health Service Human Research Ethics Committee (HREC/QTHS/85271 and HREC/ QTHS/88044), and the Aboriginal Health and Medical Research Council of New South Wales Human Research Ethics Committee (1922/22) have approved the study. Approval has also been received from Aboriginal, legal and health services, research committees, and boards. Dissemination will occur through stakeholder reports, workshops, presentations, peer-reviewed journal articles, and conference papers. Further dissemination will be determined in partnership with the project advisory group.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Acknowledgments:** The authors thank all study sites for their participation in the project including Elders, community leaders, advocates and service providers who are supporting the project. The authors would also like to thank the members of the advisory group and individual advisors for their advice on the study.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Disclaimer/Publisher's Note:** The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

### *Article* **Race and Ethnic Differences in the Protective Effect of Parental Educational Attainment on Subsequent Perceived Tobacco Norms among US Youth**

**Edward Adinkrah 1,\*, Babak Najand <sup>2</sup> and Angela Young-Brinn 1,2**


**\*** Correspondence: edwardadinkrah@cdrewu.edu

**Abstract:** Background: Although parental educational attainment is known to be associated with a lower prevalence of behaviors such as tobacco use, these effects are shown to be weaker for Black than White youth. It is important to study whether this difference is due to higher perceived tobacco use norms for Black youth. Aim: To study the association between parental educational attainment and perceived tobacco use norms overall and by race/ethnicity among youth in the US. Methods: The current study used four years of follow-up data from the Population Assessment of Tobacco and Health (PATH-Youth) study conducted between 2013 and 2017. All participants were 12- to 17-yearold non-smokers at baseline and were successfully followed for four years (*n* = 4329). The outcome of interest was perceived tobacco use norms risk at year four. The predictor of interest was baseline parental educational attainment, the moderator was race/ethnicity, and the covariates were age, sex, and parental marital status at baseline. Results: Our linear regressions in the pooled sample showed that higher parental educational attainment at baseline was predictive of perceived disapproval of tobacco use at year four; however, this association was weaker for Latino than non-Latino youth. Our stratified models also showed that higher parental educational attainment was associated with perceived tobacco use norms for non-Latino but not for Latino youth. Conclusion: The effect of high parental educational attainment on anti-tobacco norms differs between Latino and non-Latino youth. Latino youth with highly educated parents remain at risk of tobacco use, while non-Latino youth with highly educated parents show low susceptibility to tobacco use.

**Keywords:** population groups; risk behavior; perceived tobacco use norms; ethnic groups; academic achievement

#### **1. Introduction**

Youth is associated with heightened risk behaviors, including tobacco use [1]. However, socioeconomic status (SES) indicators such as parental educational attainment may lower youth risk-taking behaviors such as tobacco use [2]. Some of the many mechanisms that may explain the lower behavioral and health risk of high SES youth are social norms and beliefs that are not favorable toward tobacco (also called perceived tobacco use norms) [3], which are under the influence of peers, families [4], and other factors such as availability of tobacco in the areas, tobacco ads, and prevalence of tobacco use in the community, neighborhood, school, and family and friends [5].

However, the protective effects of parental educational attainment on youth risk behaviors such as tobacco use may differ between diverse racial and ethnic groups of youth [6]. In addition, according to a phenomenon called marginalization-related diminished returns (MDRs) [7–16], due to racism and social stratification, resources and assets may be associated with lower levels of economic, behavioral, developmental, and health outcomes for marginalized and racialized groups than White individuals [17,18].

**Citation:** Adinkrah, E.; Najand, B.; Young-Brinn, A. Race and Ethnic Differences in the Protective Effect of Parental Educational Attainment on Subsequent Perceived Tobacco Norms among US Youth. *Int. J. Environ. Res. Public Health* **2023**, *20*, 2517. https://doi.org/10.3390/ ijerph20032517

Academic Editors: Paul B. Tchounwou and Neal Doran

Received: 23 October 2022 Revised: 10 December 2022 Accepted: 27 December 2022 Published: 31 January 2023

**Copyright:** © 2023 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

Research has indicated that race may alter how SES influences health and behavioral problems such as tobacco use [19–29]. The association between parental educational attainment and a wide range of health problems varies between racial/ethnic groups of youth [30–32]. Fuller Rowell showed that the association between youth educational attainment and health is racialized [30–32]. Under racism and discrimination, high educational attainment may be linked to more distress and discrimination for Black than White youth [30–32]. Education gains may be linked to worse mental health for Black youth who live in a social context that may impose a higher level of psychological tax for their educational success or chronic poverty from childhood [30–32]. At all SES levels, Black students are discriminated against [33,34], and high SES Black youth attend worse schools than White youth [35]. Similarly, high-SES Black youth have family members who are more likely to be substance users than high-SES White youth [36]. When high-SES Black youth move to high-SES neighborhoods and schools (that are predominantly White), they become even more exposed [37,38] and vulnerable [39] to discrimination. As the education system differently treats Black and White youth [40,41], health gain due to education is weaker for Blacks than Whites [30–32].

According to the marginalization-related diminished returns (MDRs), SES resources and even non-economic resources may generate fewer behavioral, developmental, and health outcomes for marginalized and racialized groups such as Blacks and Latinos than non-Latino Whites [17,18]. While most of this literature is generated on the effects of SES on health outcomes for adults [16,19,21,23,29,42–44], non-SES factors such as self-efficacy may also be associated with lower health gain for Black than White individuals [45]. Similarly, positive affect [46,47] and happiness [48–50] may generate less health for Blacks than Whites. We explain this phenomenon through racism and societal inequalities: Even when SES and other resources are available, societal and environmental conditions such as social stratification, segregation, racism, and discrimination make it more difficult for Black and Latino than non-Latino White families and individuals to secure outcomes. In this view, what makes a large change for Whites may generate smaller real-life changes for Black individuals [45,51].

As shown by systematic reviews, behaviors such as tobacco consumption are under influence of cognitive elements such as perceived tobacco norms [52]. According to theories such as Theory of Planned Behavior (TPB) [52] and Theory of Reasoned Action (TRA) [53], perceived norms predict behaviors such as tobacco use. Perceived norms are different than actual norms and can be defined as what individuals think are the norms of their group [54]. For example, even when actual norms can be low, perceived norms can be high. Thus perceived norms are what people think is the norm, while actual norm is the reality of the society [55]. Cognitive elements such as perceived tobacco norms can be used as a marker of tobacco susceptibility and vulnerability [56].

Built on the MDRs literature on tobacco use risk [57,58], we conducted this study with two aims: the first was to test the association between parental educational attainment and perceived tobacco use norms overall. The second aim was to test the variation of this association by race. Our first hypothesis was that overall, high parental educational attainment is associated with lower perceived tobacco use norms in youth. Our second hypothesis was that this inverse association would be weaker for Latinos and Blacks than non-Latinos and Whites.

#### **2. Methods**

For this study, we conducted a secondary analysis of the first four years of the Population Assessment of Tobacco and Health (PATH-Youth) study data. The PATH-Youth is the state-of-the-art study of tobacco use of US youth. Data collection was performed between 2013 (baseline) and 2017 (follow up). Youth PATH data are publicly available to all individuals. This data set is fully de-identified and can be accessed here: https: //www.icpsr.umich.edu/web/NAHDAP/studies/36231 (accessed on 12 October 2022).

In the PATH study, participants are selected randomly. Stratified and clustered random samples were selected from all US states. Eligibility for inclusion in the current analysis were non-institutionalized members of US households, aged between 12 and 17 at baseline, having follow-up data for years (baseline and follow-up data), and being Latino or non-Latino White or Black. Participants were all never smoker at baseline. A total number of 4596 youth were entered who had and follow-up data for four years.

Study variables in this analysis included race, ethnicity, parental educational attainment, parental marital status, age, sex/gender, and perceived tobacco use norms. Age was a dichotomous variable 0 for lower than 15 and 1 for 15 and above. Gender was 1 for males and 0 for females. Parental educational attainment was the independent variable with five levels, and perceived tobacco use norms were the outcome. Both parental educational attainment and perceived tobacco use norms were treated as continuous measures. Perceived tobacco use norms were self-reported and measured using the following binary indicators: (a) People who are important to you: Their views on tobacco use in general, (b) People who are important to you: Their views on smoking cigarettes, (c) People who are important to you: Their views on using e-cigarettes or other electronic nicotine products, (d) People who are important to you: Their views on smoking traditional cigars, cigarillos, or filtered cigars, (e) People who are important to you: Their views on smoking shisha or hookah tobacco, (f) People who are important to you: Their views on using snus, and (g) People who are important to you: Their views on other types of smokeless tobacco. Each item was on a 1 (very positive) to 5 (very negative) response scale. The range of total scores was between 1 and 5, with a higher score indicating higher perceived tobacco use norms.

*Parental educational attainment.* Parental educational attainment was a five-level variable as below: 1 = "Some high school," 2 = "Completed high school," 3 = "Some college," 4 = "Completed college," 5 = "Graduate or professional school after college." This variable was a continuous variable.

*Parental marital status.* Parental marital status was a dichotomous variable that reflected married parents and any other condition (divorced, not married, partnered, etc.).

*Race.* Race was self-identified, treated as a nominal variable, and the moderator variable (White and Black). Race was the effect modifier (moderator). In this study race was a social rather than a biological variable. White was defined as a person having origins in any of the original peoples of Europe, the Middle East, or North Africa. Black or African American was defined as a person having origins in any of the Black racial groups of Africa. We used race as an effect modifier because MDRs theory suggests that due to racism and social stratification, returns of SES indicators such as parental education tend to be weaker for racialized groups.

*Ethnicity.* Ethnicity was self-identified as non-Latino, or Latino. We defined Latino as "a person of Cuban, Mexican, Puerto Rican, South or Central American, or other Spanish culture or origin regardless of race".

#### *Data Analysis*

Data analysis was performed using SPSS 24. SPSS was used for univariate, bivariate, and multivariable analysis. Univariate was descriptive statistics such as mean (standard deviation [SD]) and frequency (%). Bivariate included the Spearman correlation test. With the outcome being perceived tobacco use norms score at age 4, the predictor variable was parental educational attainment, and the moderators (effect modifiers) were race and ethnicity, and age, sex, and parental marital status as the covariates, six linear regression models were applied for multivariable modeling. *Model 1* and *Model 2* were run in the pooled sample. *Model 3* and *Model 4* were performed on non-Latino and Latino youth. *Model 5* and *Model 6* were performed on White and Black youth. *Model 1* did not have, and *Model 2* had the interaction term between race/ethnicity and parental education, our predictor variable. *Model 5* and *Model 6* were not shown because there were no race differences in associations. *Model 7* to *Model 10* were performed in race × ethnic groups. *Model 11* and *12* were performed by sex/gender. B, SE, 95% CI, and *p* were reported from each model.

#### **3. Results**

#### *3.1. Descriptive Data*

A total number of 4815 youth were entered who had and follow-up data for four years. Descriptive data are reported in Table 1.


**Table 1.** Descriptive data overall and by race in youth (*n* = 4329).

#### *3.2. Pooled Sample Models*

Table 2 presents the summary of linear regressions for *Model 1* and *Model 2* that were fitted to the pooled sample. As this model shows, higher parental educational attainment was associated with lower perceived tobacco use norms; however, this association was stronger for non-Latino than Latino youth. White and Black youth did not show difference in the slope of the effect of parental educational attainment on outcome.

**Table 2.** Pooled Sample models in US youth.


Outcome: Perceived tobacco use norms Score; Data: Population Assessment of Tobacco and Health (PATH).

#### *3.3. Ethnic Stratified Models*

Table 3 presents the summary of linear regressions for *Model 3* and *Model 4* that were fitted to White and Black youth, respectively. As these models show, higher parental educational attainment was associated with a lower perceived tobacco use norms for non-Latino but not for Latino youth.


**Table 3.** Stratified models in non-Latino and Latino youth.

Outcome: Perceived tobacco use norms Score; Data: Population Assessment of Tobacco and Health (PATH).

*3.4. Race* × *Ethnic Interactional Stratified Models*

As shown by *Models 5 to 8* performed in race by ethnic intersectional groups, parental education was associated with higher perceived tobacco use norms score in non-Latino Whites and non-Latino Blacks. This association was not significant for Latino White and Latino Black individuals (Table 4).

**Table 4.** Models in race × ethnicity groups.


Outcome: Perceived tobacco use norms Score; Data: Population Assessment of Tobacco and Health (PATH).

#### *3.5. Sex/Gender Stratified Models*

Due to low sample size, interaction between race or ethnicity with parental education did not show significance in our male or female youth. Table 5 shows the summary of these findings.


#### **4. Discussion**

The current study was performed with two main aims: one to evaluate the overall association between parental educational attainment and perceived tobacco use norms in US youth, and two to test variation in this association by race and ethnicity. The first aim showed an inverse association between parental educational attainment and perceived tobacco use norms overall. The second aim showed moderation by ethnicity not race. This protective association was weaker for Latino than non-Latino youth. This association did not differ between Black and White youth.

The inverse association between parental educational attainment and perceived tobacco use norms is in line with theories of fundamental causes, social determinants, social status, status syndrome, and several other models that explain the lower risk of high SES populations and individuals. Due to Jim Crow, historical racism, the legacy of slavery, social stratification, and segregation, Black-White differences in living conditions sustain across all levels of socioeconomic inequalities [59–62]. According to ecological theories, individuals who live in proximity to low SES neighborhoods, peers, schools, families, and friends will have a higher risk, including tobacco use risk [63]. However, many mechanisms may explain why low SES is associated with race, peer risk, and poor neighborhoods.

There are multiple studies that show racial and ethnic variation in the association between SES, health, and behaviors, with weaker associations in racial and ethnic minorities than non-Latino White youth [64]. There are also studies showing weaker associations between SES and tobacco risk in Black and Latino than non-Latino White individuals [16,19–23,25–29,65]. However, we are unaware of any past studies on racial and ethnic differences in the association between parental educational attainment and perceived tobacco use norms.

There are several studies on racial and ethnic variation in health-behavior association [30–32]. One of their studies showed that Black and Native American adolescents pay greater social costs with academic success than Whites; however, this is seen in highachieving schools with a smaller percentage of Black students [32]. In another study, they showed that the effects of educational attainment were weaker for Black than for whites, and only 8% of this difference was due to covariates. Analyses yielded consistent results. They concluded that the effects of educational attainment on inflammation levels are stronger for whites than for racial and ethnic minorities [31].

Most past research is conducted on Black, not Latino individuals. Our observation of a weaker association between parental educational attainment and perceived tobacco use norms in Latino than non-Latino youth is also in line with many previous publications on the MDRs. According to marginalization-related diminished returns, resources and assets generate fewer economic, behavioral, developmental, and health outcomes for marginalized groups than for White individuals. While most of this literature is generated on SES effects among adults, there are some studies showing that a sense of mastery, agency, and self-efficacy may be associated with lower health for Black than White individuals [45]. Similarly, positive affect [46,47], happiness [48–50], and a sense of health [66–68] may generate more life expectancy for Whites than Blacks [45,51]. The positive association between SES and John Henryism is also suggestive of the health risks that may be the price of success for Black individuals [69–73]. Hudson has published on the high costs of success for Black youth and young adults [70,74,75].

This study expanded the MDRs literature, which is written on tobacco use [57,58]. Previous work has shown that SES –tobacco use is racialized [57,58]. A study showed that education–tobacco knowledge is also racialized in the US [76]. This finding may be because high-SES White youth attend better schools than high-SES Black youth [35]. In addition, there are many challenges in the daily lives of Black youth in US schools [33,34]. Racial differences in the returns of education may be because of anti-Black discrimination at schools [33,34] or neighborhoods [37,38].

Our study is not without methodological limitations. First, all variables were selfreport. Thus, our results may be affected by reporting bias and social desirability. Second, our variables were measured from youth. Norms could be measured from the social network of the youth. We did not measure many potential confounders, such as drug availability at home or neighborhood conditions, such as proximity to tobacco outlets. In addition, this was a study with an imbalanced sample size (larger n for non-Latino and White than Latino and Black youth). However, our main inference was based on pooled sample analysis with interaction rather than stratified models, which have differential power. Our study explored sex differences in the relationship between parental educational attainment and youth's perceived tobacco use norms, however, the sample size was inadequate for race by sex by parental education interaction term. Despite these limitations, the major contribution of this study is to document MDRs for perceived tobacco use norms for the first time. We are not aware of any previous studies that suggest perceived tobacco use norms may have a role in higher-than-expected tobacco use of Black and Latino youth with highly educated parents.

Future research is needed on the social and environmental causes of the observed MDRs. Future research should test the role of advertisement exposure, the prevalence of smokers, as well as other contextual factors at school and neighborhood that may weaken the effect of parental educational attainment for ethnic minority youth. The role of high-risk peers, family, friends, proximity to tobacco outlets, and other contextual conditions should be tested in future multi-level research.

#### **5. Conclusions**

To conclude, although overall, high parental educational attainment is associated with lower perceived tobacco use norms, this inverse association is weaker for Latino than non-Latino youth. The diminished return of parental educational attainment on perceived tobacco use norms may be due to environmental and structural inequalities at family, school, or neighborhood due to the segregation of ethnic minority communities. Future research should test why and how the same MDRs could not be found for Black youth.

**Author Contributions:** Conceptualization, E.A., B.N., and A.Y.-B.; Formal analysis, B.N.; Investigation, A.Y.-B.; Software, E.A.; Writing—original draft, E.A., and B.N.; Writing—review and editing, E.A., and B.N. All authors have read and agreed to the published version of the manuscript.

**Funding:** As a scholar of the Clinical Research Education and Career Development (CRECD) program at Charles R. Drew University of Medicine and Science (CDU), Dr. Adinkrah's research-related activities were supported by the NIMHD/NIH Award # R25 MD007610. A.Y.-B. is funded and supported by the Tobacco-Related Disease Research Program (TRDRP) grant R00RG2347.

**Institutional Review Board Statement:** This study used publicly available PATH data. All data are fully de-identified. Thus, the study was not human subject research and exempt from full IRB review.

**Informed Consent Statement:** All youth provided assent. All parents provided consent.

**Data Availability Statement:** PATH data are publicly available here: https://www.icpsr.umich.edu/ web/NAHDAP/series/606 (accessed 12 October 2022).

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Disclaimer/Publisher's Note:** The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

#### MDPI

St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*International Journal of Environmental Research and Public Health* Editorial Office E-mail: ijerph@mdpi.com www.mdpi.com/journal/ijerph

MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel: +41 61 683 77 34

www.mdpi.com ISBN 978-3-0365-6979-6