**Factors Associated with Health-Related Quality of Life in Community-Dwelling Older Adults: A Multinomial Logistic Analysis**

**Encarnación Blanco-Reina 1,\*, Jenifer Valdellós 2, Ricardo Ocaña-Riola 3,4, María Rosa García-Merino 5, Lorena Aguilar-Cano 6, Gabriel Ariza-Zafra <sup>7</sup> and Inmaculada Bellido-Estévez <sup>1</sup>**


Received: 19 September 2019; Accepted: 25 October 2019; Published: 1 November 2019

**Abstract:** The main aim of this study was to determine the association of various clinical, functional and pharmacological factors with the physical (PCS) and mental (MCS) summary components of the health-related quality of life (HRQoL) of community-dwelling older adults. Design: Cross-sectional study. Patients and setting: Sample of 573 persons aged over 65 years, recruited at 12 primary healthcare centres in Málaga, Spain. Sociodemographic, clinical, functional, and comprehensive drug therapy data were collected. The main outcome was HRQoL assessed on the basis of the SF-12 questionnaire. A multinomial logistic regression model was constructed to study the relationship between independent variables and the HRQoL variable, divided into intervals. The average self-perceived HRQoL score was 43.2 (± 11.02) for the PCS and 48.5 (± 11.04) for the MCS. The factors associated with a poorer PCS were dependence for the instrumental activities of daily living (IADL), higher body mass index (BMI), number of medications, and presence of osteoarticular pathology. Female gender and the presence of a psychopathological disorder were associated with worse scores for the MCS. The condition that was most strongly associated with a poorer HRQoL (in both components, PCS and MCS) was that of frailty (odds ratio (OR) = 37.42, 95% confidence interval (CI) = 8.96–156.22, and OR = 20.95, 95% CI = 7.55–58.17, respectively). It is important to identify the determinant factors of a diminished HRQoL, especially if they are preventable or modifiable.

**Keywords:** health-related quality of life; older adults; frailty; medication; primary care

#### **1. Introduction**

The aging of the population is a global phenomenon that is producing dramatic sociodemographic transformations. In this respect, a recent study modelled life expectancy, all-cause mortality and cause of death forecasts for 250 causes of death from 2016 to 2040 in 195 countries and territories. For 2040, Japan, Singapore, Spain and Switzerland were forecast to have an average life expectancy exceeding 85 years for both genders, and another 59 countries, including China, were projected to surpass a life expectancy of 80 years. According to these forecasts, Spain will then have the greatest life expectancy in the world (85.8 years) [1]. This pattern of aging poses a major challenge to health and

social assistance services, as older people present more chronic conditions and generate higher per capita healthcare costs. Multimorbidity is strongly associated with adverse health outcomes, such as disability, dependence, mortality, increased need for health and social services and polymedication, and diminished health-related quality of life (HRQoL) [2].

Increased longevity should not be achieved at the expense of quality of life. A poor HRQoL has been associated with reduced activities of daily living, a higher frequency of hospitalisation and increased mortality [3,4]. Therefore, enhancing HRQoL should be a major consideration in the design and implementation of health care for older persons [5]. In addition, assigning a greater importance to the quality of life, in preference to disease-based outcomes, is consistent with the opinions expressed by such persons themselves [6,7]. As a consequence of the interest being generated by this topic, patient-perceived quality of life is now widely used as a measure of health care in clinical research and in health economics assessments [8].

HRQoL has been defined as "the subjective perception influenced by the current health status of the ability to perform activities important for the person" [9,10]. In response to heightened interest in assessing HRQoL and in determining the effectiveness of health care interventions in older patients, a range of instruments—both generic and specific—have been developed for these purposes.

Many factors can influence the HRQoL of older adults, including health status, social engagement and cognitive function [11]. In this area, studies have been undertaken to analyse the association between factors (such as female sex, age, functional impairment and comorbidities) and HRQoL, and in recent years, the role of medication as a determinant of overall health outcomes has emerged as an area of great interest. Various aspects of treatment regimens call for special attention, because they may have a negative influence on the HRQoL of older persons, such as polymedication, and potentially inappropriate or harmful medication [12–17]. It is important to recall that the benefits of medication should always outweigh the potential harm, and therefore individual circumstances should be taken into account. In recent studies, frailty, too, has been highlighted as a possible determinant of worsened quality of life [18–22]. Frailty is defined as an abnormal health state characterised by the loss of biological reserves, related to the aging process [23]. Therefore, in providing care for persons with frailty, special emphasis should be placed on quality of life considerations.

Evaluation of HRQoL can help clinicians to determine the needs of older patients and thus optimise their decision making. Therefore, we believe it is important to explore factors influencing HRQoL in this population in order to identify suitable intervention strategies. In this respect, we note that many previous studies have not addressed all the factors that might have a significant impact on HRQoL, and in some cases the results presented are contradictory. For all of these reasons, we consider it interesting to investigate HRQoL in older adults, resident in one of the countries where life expectancy is highest, to assess clinical, functional and pharmacological aspects of the question. In this study, we hypothesise that there are multiple predictive factors of a poor HRQoL and that they may affect its physical and mental components in different ways and to different degrees. Accordingly, our aim was to determine levels of HRQoL among community-dwelling older people and to analyse the factors associated with a poor HRQoL.

#### **2. Patients and Methods**

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

In this cross-sectional investigation, the study population was composed of 89,615 community-dwelling residents aged 65 years or more, living in Málaga, Spain. Assuming a standard deviation for HRQoL of 11.0 [24], an absolute precision of δ = 0.9 and a level of confidence of 1−α = 0.95, we calculated that the minimum sample size needed to estimate the average physical and mental HRQoL was 570. Finally, the total sample was composed of 573 persons. These patients were recruited from twelve primary care centres, by stratified random sampling designed to obtain a representative sample of the population, allocating the population in proportion to the size of each healthcare centre. Participants were selected randomly within each healthcare centre from a general list of healthcare cards issued by the Spanish National Health Service. The inclusion criteria were people 65 years of age or older, included in the database of healthcare cards, belonging to the outpatient setting (not institutionalized), and giving their informed consent to participate in the study (people who did not give consent were excluded).

#### *2.2. Data Collection and Measures*

To obtain the study data for analysis, patients were interviewed using a structured questionnaire. Further data were obtained from medication packaging and digital medical records. The questionnaire was used to obtain detailed information on the patients' regular drug use, together with clinical, functional and sociodemographic data. Clinical diagnoses were examined, and the number of chronic conditions presented by each participant was determined. Patients' independence in performing instrumental activities of daily living (IADL) was assessed using the Lawton scale [25]. In addition, the body mass index was determined for all patients, and frailty was assessed according to Fried's criteria (as robust, pre-frail or frail) [26].

Medication assessment. Data were obtained for the medication prescribed (indication, dosage and duration of treatments during the last three months or more). The presence of polymedication was considered, defining this as the regular use of five or more medications, as was that of potentially inappropriate medication (PIM), according to the STOPP v2 criteria (Screening Tool of Older Person's Potentially Inappropriate Prescriptions, version 2) [27]. The latter variable was operationalised as the percentage of patients receiving at least one PIM.

Quality of life assessment. The main study outcome was HRQoL assessed by the SF-12 questionnaire, a widely used generic instrument. The SF-12 is an abbreviated version of the Short Form-36 Health Survey (SF-36), in which a subset of 12 items/questions are used to derive summary scores for physical health (PCS score) and mental health (MCS score) [28]. The response options form Likert-type scales that assess the intensity and/or frequency of people's health status. The final score obtained can be between 0 and 100, where lower scores indicate worse, and higher scores better HRQoL. Using only one-third of the SF-36 items, the SF-12 reproduces the two summary scores originally developed for the SF-36 with remarkable accuracy, but more quickly and requiring less effort from the respondent [29]. The SF-12 has been validated for use in the USA, the UK, Spain and many other European countries [24].

#### *2.3. Statistical Analysis*

Exploratory data analysis and frequency tables were used to describe the study variables. Using the standardised scores of the SF-12 questionnaire for the Spanish population [24], the HRQoL score of each participant was assigned to one of the following intervals, taking into account the population reference group according to age and sex:


A multinomial logistic regression model was used to study the relationship between the independent variables and the HRQoL variable, grouped into the corresponding intervals [30]. A 5% significance level was assumed to indicate statistical significance. Statistical data analysis was performed using SPSS version 23.0 (IBM SPSS Statistics, Armonk, NY, USA).

#### *2.4. Ethical Considerations*

This study was conducted in accordance with the provisions of the 1975 Declaration of Helsinki, revised in 2013. The Málaga Clinical Research Ethics Committee approved the study (PI-0234-14), and informed consent was obtained from all patients prior to their inclusion.

#### **3. Results**

#### *3.1. Characteristics of the Study Population*

The study sample was composed of 573 patients, with a mean age of 73.1 years (standard deviation 5.5, range 65–104) and of whom 57.2% were female. Most lived with their partner (62.4%) or family (16.5%), but 21.1% lived alone. On average, each patient presented 7.8 chronic conditions (standard deviation 3.3, range 0–20). The most prevalent diagnoses were bone and joint disorders (mainly osteoarthritis of the knee, hip, hand and shoulder) (75.2%), hypertension (70.5%) and dyslipidaemia (51.6%). Most of the patients presented with overweight (40.9%) or obesity (45.7%) and their mean body mass index was 30.2 (standard deviation 5.1, range 17–54.5). Half were independently capable of performing IADL, and the mean score on the Lawton scale was 6.6 (standard deviation 1.8, range 0–8). Frailty was present in 137 patients (23.9%; 95% confidence interval (CI) = 20.5–27.5), according to Fried's criteria. The main characteristics of the study population are detailed in Table 1.

The prevalence of polymedication was 68% (95% CI = 64.1–71.7), and on average each patient consumed 6.8 drugs (standard deviation 4.0; range 0–23). The most widely prescribed drugs were omeprazole and acetaminophen, followed by aspirin, simvastatin, metformin, metamizole, enalapril and bromazepam. The use of potentially inappropriate medication, according to the STOPP v2 criteria, was identified in 66.8% of patients (95% CI = 62.9–70.6). The number of PIMs per patient ranged from 0–10 (mean 2.1, standard deviation 2.2). The most frequent PIMs detected were benzodiazepines (61% of all PIMs).

#### *3.2. Assessment of HRQoL and Analysis of Related Factors*

The patients' perceptions of their HRQoL produced an average score of 43.2 (standard deviation 11.02, range 16.2–65.4) for the physical component summary (PCS) and a somewhat higher one, 48.5 (standard deviation 11.04, range 14.1–66.6), for the mental component (MCS). Males obtained higher scores than females in both the PCS and the MCS, with average values of 45.91 vs. 41.27 and 51.95 vs. 45.91, respectively. Table 2 shows the distribution within the global sample among the categories of perceived HRQoL (very low, low, high and very high). A notable feature of this distribution is that a higher proportion of patients perceived their HRQoL as very high in the mental component than in the physical one.

To further examine the impact of the independent variables on the HRQoL categories, a multinomial logistic regression analysis was performed (Tables 3–5). The factors related to having high vs. very high HRQoL (P50–P80 and ≥ P80, respectively) were the level of dependence for IADL in the PCS, and BMI, respiratory disease and frailty in the MCS (Table 3). The presence of a low HRQoL (P20–P50) with respect to a very high score for the PCS was associated with the level of dependence for IADL, with accompanied living, with the presence of osteoarticular pathology and with frailty (Table 4). The odds of these older persons having a low HRQoL decrease by 30% for each additional point of independence on the Lawton scale (odds ratio (OR) = 0.70, 95% CI = 0.55–0.88). However, they double for those who do not live alone (OR = 2.13, 95% CI = 1.07–4.27), are 2.5 times greater for those with osteoarticular pathology compared to those without it (OR = 2.57, 95% CI = 1.35–4.85) and are seven times greater for those who are frail (OR = 7.43, 95% CI = 2.13–25.82) compared to those who are robust. In the case of the MCS, the presence of frailty was associated with a three times greater perception of low HRQoL (OR = 3.2, 95% CI = 1.21–8.46).


**Table 1.** Characteristics of study population (*n* = 573).

IADL: Instrumental Activities of Daily Living; BMI: Body Mass Index; PIM: Potentially Inappropriate Medication (according to STOPP v2 criteria).


**Table 2.** HRQoL assessment according to SF-12 questionnaire.

PCS: Physical Component Summary; MCS: Mental Component Summary; P: Percentile.

A very low perception of HRQoL (≤P20) in the PCS was also related to the level of dependence for IADL (OR = 0.62, 95% CI = 0.48–0.79), with the presence of osteoarticular pathology (OR = 4.38, 95% CI = 1.98–9.70) and with the states of prefrailty (OR = 4.19, 95% CI = 1.61–10.86) and frailty (OR = 37.42, 95% CI = 8.96–156.22). In addition, the odds of a very low HRQoL increase by 15% for each additional drug in the treatment regimen (OR = 1.15, 95% CI = 1.02–1.30) and by 8% for each unit increase in BMI (OR = 1.08, 95% CI = 1.01–1.15). However, age was inversely related to HRQoL. Thus, all other variables being equal, for each additional year of life, the odds of the patient perceiving a very low HRQoL decreased by 8% (Table 5). A similar pattern for the age effect was observed with respect to the mental component of HRQoL; thus, each additional year of life decreased the odds of a very low HRQoL by 7%. In the MCS, too, the states of prefrailty and frailty were associated with a very low HRQoL. In this mental component, and with all other variables being equal, the female patients were 88% more likely than the males to have a very low HRQoL. Finally, the presence of a psychopathology (usually anxiety and/or depression) was related to a poorer HRQoL; the presence of any such disorder quadrupled the odds of the patient perceiving a very low HRQoL (OR = 4.69, 95% CI = 2.60–8.47).

**Table 3.** Factors related to HRQoL. Multinomial logistic regression for High HRQoL (with respect to Very High HRQoL) for physical and mental health components (PCS-MCS).


PIMs: Potentially Inappropriate Medications (according to STOPP v2 criteria). \* *p* < 0.05; \*\* *p* < 0.01.

**Table 4.** Factors related to HRQoL. Multinomial Logistic Regression for Low HRQoL (with respect to Very High HRQoL) for physical and mental health components (PCS-MCS).


#### \* *p* < 0.05; \*\* *p* < 0.01.


**Table 5.** Factors related to HRQoL. Multinomial Logistic Regression for Very Low HRQoL (with respect to Very High HRQoL) for physical and mental health components (PCS-MCS).

\* *p* < 0.05; \*\* *p* < 0.01; \*\*\* *p* < 0.001.

#### **4. Discussion**

Overall, our results for perceived HRQoL among community-dwelling older patients in Malaga (southern Spain) are consistent with those reported in similar studies conducted in other regions of Spain [31,32] and elsewhere in Europe [33]. With respect to the components of physical and mental health, the average PCS score (43.2) is slightly higher than that obtained in previous research [31–34], while the MCS (48.5) is somewhat lower than the average value for six European countries (54.3), also obtained using the SF-12 [33]. This lower score in the MCS could be due to cultural differences, and/or the non-negligible prevalence of psychopathological disorders (mainly anxiety and/or depression) in our region (these pathologies affect 36% of the older population). Regarding the influence of age on HRQoL, previous studies have produced conflicting results [32–35]. Our investigation revealed an inverse relationship between advanced age and the odds of a very low HRQoL. Therefore, other questions such as comorbidities, frailty, dependence and other possible confounders being equal, as the patient ages there is a lower probability of him/her perceiving a very low HRQoL, and this is true for both the physical and the mental components. These findings might be explained in terms of a better psychological adaptation to aging and perhaps to the fact that with age comes not only wisdom, but also the attribution of greater meaning or value to life. As concerns the influence of gender, previous studies have reported a poorer HRQoL among women than among men [31–34,36,37], and our own results corroborate this difference. Thus, women obtained poorer scores in both components of HRQoL, although the association was only significant for the mental component of a very low quality of life. In the physical component, female gender is probably not a determinant factor of a poorer HRQoL due to the adjustment made for confounding variables such as frailty or osteoarticular pathology, both of which are more prevalent among women.

One of the most consistent predictors of a poor quality of life is the level of dependence for IADL. According to the multinomial logistic regression model, with a higher score on the Lawton scale, i.e., with greater independence, the possibility of a poorer HRQoL decreased significantly (in all the categories considered: high, low and very low, with respect to very high) always with respect to the physical component. It seems logical that the higher the degree of independence among older persons, in activities such as handling economic affairs, using the telephone or managing different forms of transport, the better the quality of life they will perceive. According to a recent study, functional dependence, together with the presence of depressive symptoms, would be an important factor mediating the well-known association between multimorbidity and a poor quality of life [38]. In this line, too, it has been shown that disability is one of the most significant conditions worsening the HRQoL [39]. Older adults who live in cohabitation are twice as likely to have a low physical HRQoL than are those living alone. This finding may be related to the level of functional dependence presented, but further study is needed to confirm this possibility.

Regarding medication, the number of drugs prescribed was positively associated with a very low HRQoL in the PCS. This finding is consistent with previous research, in which polymedication has been identified as a determinant factor of a poorer quality of life [12–14,40]. In our opinion, this association may be related to the side effects produced by the joint presence of several drugs within the patient. On the other hand, we found no evidence of any association between having at least one PIM and the perception of a poorer HRQoL. We speculate that such a relationship might not have been demonstrated because the STOPP v2 criteria contain a large number of items, of varying clinical significance, and therefore the impact produced on HRQoL by a single PIM might be slight. In other words, the sensitivity of this means of measuring the risk might be insufficient. Other authors, too, have failed to observe any significant association between the presence of a PIM and HRQoL, whether using the Beers [16] or the STOPP v2 criteria [41]. On the other hand, other previous studies do suggest that the prescription of drugs with a high anticholinergic load may be associated with a diminished quality of life [15–17,42].

Among this population, the most prevalent pathologies observed were osteoarthritis and hypertension. This is a common profile and similar to that reported in previous research in this field [39]. Our assessment of the impact on HRQoL of different conditions shows that physical disorders mostly affected the "physical" HRQoL while mental disorders mainly affected the "mental" aspect—as is only logical. The presence of bone and joint disease was significantly associated with a low or very low HRQoL, which corroborates previous findings in which this diagnosis was associated both with disability and with a diminished HRQoL [39]. We believe this relationship may be explained by the chronic pain which often accompanies these diseases, as well as a degree of physical limitation that is usually provoked. This association contrasts with its absence for other diagnoses, perhaps with a worse prognosis but of a more 'silent' nature, such as hypertension and diabetes mellitus. On the other hand, the presence of a psychopathology in an older person raises the possibility of his/her perceiving a very low HRQoL, in the mental component, by 400%, as has also been indicated in previous studies [12,31,32,43,44]. It would seem that the distress generated by these diseases has a negative impact on emotional regulation, motivation and other components of subjective perceptions of health and well-being. According to other researchers, and taking into account the significant impact on the patient's quality of life, we believe more screening for depression and anxiety among older populations should be performed, because these conditions tend to be under-diagnosed and also because a more active approach to this question would promote healthy aging [31]. In our sample, the BMI values observed presented two interesting aspects: on the one hand, at the mental level, a higher BMI was associated with a better HRQoL, but for the physical component it was significantly associated with a very low quality of life. It seems clear that overweight and obese patients perceive a poorer physical health [40,45–47]. This is corroborated by the fact that interventions aimed at achieving weight loss have been shown to improve the physical quality of life [48].

It is interesting to note that the factor most strongly associated with a diminished HRQoL was that of frailty, which severely affected both the MCS and the PCS, but especially the latter. The odds of a very low physical quality of life were 37 times greater among frail older persons than among those who were robust, while the MCS reflected 20 times greater odds of their presenting a poor mental quality of life. Previous studies, too, have reported this association [18–21,49], which was reinforced in a recent systematic review and meta-analysis that described it as clear and often substantial [22]. Although the growing numbers of frail older people pose a real challenge to health systems around the world, this state of pre-discapacity can in fact be prevented and treated. Furthermore, in our study sample, the pre-frail persons also perceived a poorer quality of life than those who were relatively robust. Accordingly, we believe it necessary to design and incorporate care programmes specifically adapted to this emerging population of frail and pre-frail patients, to help them age with a better quality of life.

In our opinion, the type of study described in this paper is useful for identifying the characteristics and clinical conditions that are associated with a poorer HRQoL, with particular attention to those factors that may be preventable or treatable. As improving the quality of life and well-being of older persons is a priority objective, it is of major importance to extend our knowledge of these questions.

The strengths of our study lie in the analysis made of a representative sample of healthcare centres, the global approach taken, and the great variety of clinical, functional and treatment data compiled. We acknowledge that selecting a sample population from a single region or country may result in a certain lack of external validity. Nevertheless, the sample examined in this study may be representative of the population of older adults in the ambulatory setting, which is where the largest number of such patients are to be found. Another limitation to our study is its cross-sectional design, which does not allow causal relationships to be established, although it can detect factors related to HRQoL.

With regard to the analysis performed, since HRQoL is a quantitative variable, a multivariate linear regression model might, in principle, be considered appropriate to identify the factors related to it. However, in the present case the linear model considered did not meet the conditions of homoscedasticity and normality of the residuals necessary for the correct estimation of the study parameters. Previous statistical research has reached similar conclusions, i.e., that HRQoL cannot be analysed using linear regression models [50]. Neither nonlinear modelling, nor generalised least squares nor other more complex statistical techniques overcame the problem of achieving an adequate fit, and therefore we adopted the solution of treating QoL as a qualitative variable and using a multinomial logistic regression model. This approach provided a good fit to the data considered.

#### **5. Conclusions**

In conclusion, many factors may be predictive of a poor HRQoL and they affect its physical and mental components in different ways. For the PCS, the associated factors were dependence for IADL, a higher BMI, the number of medications and the presence of osteoarticular pathology. The main factors associated with a lower MCS score were female gender and the presence of a psychopathological disorder. Some factors may be preventable or modifiable, and so recognising them and optimising the response made are crucial to the priority objective of enhancing the quality of life among older people. Clearly and consistently, the factor that was most strongly associated with a poorer overall HRQoL was the state of frailty (and also, albeit to a lesser extent, that of pre-frailty). Frailty is a dynamic syndrome, in which transitions are possible between the states of normality, pre-frailty and frailty. Accordingly, detecting frailty and addressing it in a suitable way are questions of major importance.

**Author Contributions:** Conceptualization, E.B.-R., R.O.-R., G.A.-Z.; Methodology, E.B.-R., R.O.-R.; Formal analysis, E.B.-R., R.O.-R.; Investigation, E.B.-R., R.O.-R., G.A.-Z., J.V., L.A.-C., M.R.G.-M., I.B.-E.; Resources, E.B.-R., I.B.-E.; Data curation: J.V., L.A.-C., M.R.G.-M.; Writing—original draft preparation, E.B.-R., R.O.-R., J.V.; Writing—review & editing, E.B.-R., R.O.-R., J.V., I.B.-E.; Visualization, E.B.-R., R.O.-R., I.B.-E.; Supervision, E.B.-R., I.B.-E.; Project administration, E.B.-R.; Funding acquisition: E.B.-R.

**Funding:** This research has been supported by grant funding provided by the Fundación Pública Andaluza Progreso y Salud, Consejería de Salud, Junta de Andalucía, through the Programme Proyectos de Investigación Biomédica (Grant number PI 0234/14). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

**Acknowledgments:** The authors wish to thank the Primary Care Management Team (Health District of Málaga) for providing access to the health centres and patient lists.

**Conflicts of Interest:** All authors declare that they have no conflict of interest.

#### **References**


© 2019 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 (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Multidiscipline Stroke Post-Acute Care Transfer System: Propensity-Score-Based Comparison of Functional Status**

**Chung-Yuan Wang 1, Hong-Hsi Hsien 2, Kuo-Wei Hung 3, Hsiu-Fen Lin 4, Hung-Yi Chiou 5,6, Shu-Chuan Jennifer Yeh 7,8, Yu-Jo Yeh <sup>7</sup> and Hon-Yi Shi 7,8,9,10,\***


Received: 2 July 2019; Accepted: 15 August 2019; Published: 16 August 2019

**Abstract:** Few studies have investigated the characteristics of stroke inpatients after post-acute care (PAC) rehabilitation, and few studies have applied propensity score matching (PSM) in a natural experimental design to examine the longitudinal impacts of a medical referral system on functional status. This study coupled a natural experimental design with PSM to assess the impact of a medical referral system in stroke patients and to examine the longitudinal effects of the system on functional status. The intervention was a hospital-based, function oriented, 12-week to 1-year rehabilitative PAC intervention for patients with cerebrovascular diseases. The average duration of PAC in the intra-hospital transfer group (31.52 days) was significantly shorter than that in the inter-hospital transfer group (37.1 days) (*p* < 0.001). The intra-hospital transfer group also had better functional outcomes. The training effect was larger in patients with moderate disability (Modified Rankin Scale, MRS = 3) and moderately severe disability (MRS = 4) compared to patients with slight disability (MRS = 2). Intensive post-stroke rehabilitative care delivered by per-diem payment is effective in terms of improving functional status. To construct a vertically integrated medical system, strengthening the qualified local hospitals with PAC wards, accelerating the inter-hospital transfer, and offering sufficient intensive rehabilitative PAC days are the most essential requirements.

**Keywords:** stroke; post-acute care; medical referral system; propensity score matching

#### **1. Introduction**

Acute stroke is a major cause of mortality and disability [1,2]. Stroke patients can incur considerable costs for medical care, including nursing, rehabilitative, and long term care. Therefore, many countries are attempting to establish comprehensive and integrated healthcare systems for stroke patients [3,4]. Post-acute care (PAC), which refers to medical care services that support the individual patient in recovery from illness or management of chronic disability, is aimed at enhancing the functional status of patients discharged from acute hospitalization [2–5]. Discharges to PAC facilities have increased

nearly 50% during the past 15 years, and PAC is a major contributor to hospitalization costs in the United States [5]. To control medical expenses, reform the medical referral system, and improve continuity of care, the Taiwan National Health Insurance Administration (NHIA) focused on stroke for its first national PAC project in 2014—post-acute care for cerebrovascular disease (PAC-CVD).

In Taiwan, beneficiaries are free to visit their preferred physicians and are not required to follow strict referral rules [6]. An efficient PAC system for stroke patients is needed to reduce unnecessary utilization of hospital resources and to ensure seamless care for these patients [7]. Stroke patients and their families expect that local hospitals can deliver PAC at a lower cost and with greater efficiency, effectiveness, and convenience. Our review of studies published in international journals, however, shows that most studies of PAC have analyzed a limited number of patients in a single hospital [8,9]. Additionally, few studies have used longitudinal data exceeding 1 year, and few studies have applied propensity score matching (PSM) in a natural experimental design to examine the longitudinal impacts of a medical referral system on functional status. Therefore, this study coupled a natural experimental design with PSM to assess the impact of the medical referral system in stroke patients and to examine the longitudinal effects of the medical referral system on functional status.

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

#### *2.1. The PAC Program*

In Taiwan, the multidisciplinary PAC stroke team consisted of neurologists, physiatrists, physical therapists, occupational therapists, speech therapists, and nurses. The PAC rehabilitation program was prescribed by the physiatrist, and it consisted of a complex program of universal activities that were performed at least three times per day. One hour of physical therapy, occupational therapy, or speech and swallowing therapy was carried out at each time. Notably, the fiscal incentive for a medical center to transfer a patient to a regional or a district hospital is mitigated by several factors, including the PAC-CVD transfer policy, the willingness of the stroke patient (or family) to accept further post-acute care in local hospitals, and whether the physician agrees to the transfer. These factors should cause health providers to reconsider the manner in which patient stays are controlled and to be mindful that the shortest lengths of stay (LOS) may not obtain the best outcome for the patient [10,11].

#### *2.2. Study Design and Sample*

The study population included all stroke patients admitted to PAC wards in four Taiwan hospitals between March 2014 and March 2018 (defined as ICD-9-CM codes 433.x, 434.x, and 436.x for ischemic stroke, and codes 430 and 431 for hemorrhagic stroke). The inclusion criteria were acute stroke and admission to PAC ward within 40 days after day of stroke onset. Another inclusion criterion was Modified Rankin Scale (MRS) level 2 to 4. Instead of focusing on patients who had received intensive in-patient rehabilitation, this national PAC project focused on the prevention of complications (e.g., pressure sores) in stroke patients who were bed-ridden (MRS 5). Patients who did not have major disability (MRS 0–1) were assumed to have undergone out-patient rehabilitation or were assumed to have resumed their pre-morbidity activities of daily living.

In observational studies, non-comparability between the intervention group and the comparison group can distort the estimation of the treatment effect [12,13]. The propensity score is a balancing score that can be used to compare groups that do not systematically differ. This study used PSM at the patient level to compare the baseline characteristics of the two groups, which increased the robustness of the analysis. A generalized estimating equation (GEE) model was used to cluster stroke patients treated by the same physician and to generate propensity scores for predicting the probability of the medical referral system. The covariates included patient demographics (age and gender), clinical attributes (stroke type, hypertension, hyperlipidemia, diabetes mellitus, atrial fibrillation, and previous stroke), quality of medical care (acute care LOS and PAC LOS), and pre-rehabilitation functional status. The caliper matching method ("greedy algorithm") was used for 1:1 PSM between the inter-hospital

transfer group and the intra-hospital transfer group. Thus, 483 patients in the inter-hospital transfer group were compared with an "all participants matched set" of 483 patients in the intra-hospital transfer group (Figure 1). These PAC stroke patients completed the pre-rehabilitation and the 12th week and first year post-rehabilitation assessments.

**Figure 1.** Flow chart of recruitment and study procedure.

#### *2.3. Functional Status Instruments*

The MRS scores of 0, 1, 2, 3, 4, and 5 are interpreted as no symptoms, no significant disability, slight disability, moderate disability, moderately severe disability, and severe disability, respectively [14]. The Barthel Index (BI) score was used to measure functional disability in daily life activities (e.g., eating, grooming, bathing, dressing, walking, transferring, staring, and controlling bladder and bowel) [15]. The score is for the 10-item BI ranges from 0 (totally dependent) to 100 (independent). The Functional Oral Intake Scale (FOIS) was used to assess functional oral intake in stroke patients with dysphagia [16]. The FOIS classifies swallowing function from level 1 (nothing by mouth) to level 7 (total oral diet with no restrictions). The Lawton-Brody Instrumental Activities of Daily Living Scale (IADL) is used to evaluate performance in daily life activities, including making telephone calls, shopping, preparing food, housekeeping, laundering, taking medicine, using transportation, and performing financial activities [17]. In the conventional use of the scale, women are scored in all eight domains, while men are not scored in the domains of preparing food, housekeeping, and laundering. The rationale for excluding these three domains in males is that performance of these tasks is subject to cultural differences in gender roles, which could compromise comparisons of the incidence of disability between men and women [18]. The EuroQoL five-dimensional (EQ-5D) measure is a self-assessment of mobility, self-care, usual activities, pain or discomfort, and anxiety or depression as part of a total health state [19]. The subject is required to score each item from 1 to 3 (no problem, some problem, or extreme problem, respectively). The Berg Balance Scale (BBS) is a scale of functional balance, including static and dynamic balance [20]. Each item on this 14-item scale is rated from 0 (poor balance) to 4 (good balance). The maximum score is 56. The Mini-Mental State Examination (MMSE) is the best-known short screening tool for cognitive impairment [21]. The MMSE tests orientation, attention, memory,

language, and visual–spatial skills. The maximum score is 30 points. An MMSE score below an education-adjusted cut-off score indicates cognitive impairment. The Taiwan version of these measures has been validated as a reliable and valid tool for measuring functional status in both clinical practice and research [22].

#### *2.4. Statistical Analysis*

The unit of analysis in this study was the individual stroke patient. Descriptive statistics were tabulated to depict the stroke patient demographics. For clarification, the values predicted by the regression models were used to illustrate the results, starting from before initiation of PAC until completion of 12 weeks to 1 year of follow up in the two matched study groups. Thus, the GEE models were used to estimate difference-in-differences models used to examine the effectiveness of the medical referral system. For the predicted values, standard errors in differences and standard errors in difference-in-differences were estimated using a bootstrap technique involving 1000 replications, with sample sizes equivalent to that of the original sample [23].

Hierarchical linear regression models were used to examine the roles of the MRS after accounting for demographic characteristics, clinical attributes, quality of care, and pre-rehabilitative function status. Five-step hierarchical linear regression models were used to analyze differences between explanatory factors. In Model 1, explanatory factors included age, gender, stroke type, hypertension, hyperlipidemia, diabetes mellitus, atrial fibrillation, previous stroke, length of stay in acute care, length of stay in post-acute care, and pre-rehabilitation functional status. Model 2 included Model 1 and MRS = 2; Model 3 included Model 1 and MRS = 3; Model 4 included Model 1 and MRS = 4; Model 5 included Model 1, MRS, and medical referral system. In each model, the adjusted *R*-square was estimated while adjusting for covariates. For each model, the variance inflation factor (VIF) was used to assess multicollinearity. No models showed multicollinearity.

Statistical analyses were performed using Stata Statistical Package, version 13.0 (Stata Corp, College Station, TX, USA). All tests were two-sided, and p values less than 0.05 were considered statistically significant.

#### **3. Results**

Table 1 compares the inter-hospital transfer group and the intra-hospital transfer group before and after PSM. Before PSM, all assessed characteristics significantly differed between the two groups (*p* < 0.05). After PSM, no variables significantly differed between the two groups.

All PAC stroke patients had significantly improved scores for functional measures at the 1-year follow-up survey (*p* < 0.001) (Table 2). When the 12th week post-rehabilitation scores were used as the baseline, functional status showed significant improvements at the first year post-rehabilitation (*p* < 0.001). All subscale scores continued to improve throughout the follow-up period. Additionally, in both groups, functional status scores at the first year post-rehabilitation were significantly higher than the functional status scores at the 12th week post-rehabilitation and the functional status scores at pre-rehabilitation (*p* < 0.001). Throughout the follow-up period, the intra-hospital transfer group also had significantly higher scores for functional status measures compared to the inter-hospital transfer group (*p* < 0.001).

Table 3 shows that Model 1 revealed a significant association between patient demographics and scores for functional status measures at the first year post-rehabilitation during the study period (*p* < 0.05). In Models 2–4, patients with MRS = 2 showed very little functional status improvement after adjustment for patient demographics; however, patients with MRS = 3 and MRS = 4 showed improvements. That is, rehabilitative PAC improved quality of life in patients with moderate disability (MRS = 3) or moderately severe disability (MRS = 4) but not in patients with slight disability (MRS = 2). After adjustment for all relevant influential factors, the intra-hospital transfer group had greater improvements in functional status scores compared to the inter-hospital transfer group.


 \*.


Activities of Daily Living Scale; BBS, Berg Balance Scale; MMSE, Mini-Mental State Examination.

 \* Values are expressed as mean ± standard deviation or *n* (%).


Mini-mental State Examination. † Difference indicates mean score for functional status at the 12th week after rehabilitation, mean score for functional status before rehabilitation, or mean score for functional status at the first year after rehabilitation. ‡ Difference indicates mean score for functional status in intra-hospital transfer group, or mean score for functional status in inter-hospital transfer group at each time point.

#### *J. Clin. Med.* **2019**, *8*, 1233



 \*.

Model 4 included Model 1 and MRS = 4; Model 5 included Model 1, MRS, and medical referral system.

#### **4. Discussion**

Our data for the percentage of stroke patients with vascular risk factors were consistent with previous reports. After PSM, the percentages of stroke patients with hypertension, hyperlipidemia, diabetes mellitus, atrial fibrillation in our study were 77%, 42%, 37%, and 10%, respectively. Previous studies of stroke patients have reported hypertension in 63–80%, hyperlipidemia in 40–49%, diabetes mellitus in 34–42%, and atrial fibrillation in 7.3–11% [24,25]. A review of studies performed in Asia found that Taiwan has a higher prevalence of hypertension, diabetes mellitus, and hyperlipidemia compared to Japan, Korea and Singapore [26]. A previous Taiwan study of stroke incidence and recurrence during 2000–2011 also reported that, although the prevalence of diabetes mellitus, hyperlipidemia and atrial fibrillation increased during this period, the rates of primary ischemic stroke and 1-year recurrence of stroke decreased by 9% and 18% respectively [27]. Factors that can have important effects on stroke incidence and recurrence include medication control, early detection of diseases, diet control, body weight control, life style adjustment and exercise.

To achieve a vertically integrated medical system, post-stroke care should be patient-centered, and inter-hospital transfer of stroke patients must be seamless. In Australia, stroke patients treated at hospitals with stroke coordinators had lower LOS and higher quality of evidence-based care compared to those treated at hospitals without stroke coordinators [28]. In the Taiwan national PAC project reported here, the case manager assigned to each stroke patient ensured efficient transfer of the patient to a local hospital. Therefore, the mean LOS for patients in acute stage was much lower in the intra-hospital transfer group compared to the inter-hospital transfer group (13.01 and 24.45 days, respectively). Since minimizing the time from admission to inter-hospital transfer decreases total LOS and total cost, implementation of a case manager and an efficient inter-hospital transfer system are essential for an integrated medical system. Other possible reasons in the inter-hospital transfer group include geographical variations in the distribution of physicians and medical resources, and differences in care quality and expertise among hospitals and individual providers. For stroke patients in acute stage, those treated at teaching hospitals and certified primary stroke centers have a lower mortality rate, greater availability of rehabilitative care, and lower ADL dependence status [29]. A possible reason for the superior outcomes obtained by teaching hospitals and certified primary care centers for stroke is that they tend to have a high volume of stroke patients, and clinicians who treat a high volume of patients tend to achieve high skill levels.

Before PSM, compared to the intra-hospital transfer group, the inter-hospital transfer group in this study had a lower functional status before PAC and more comorbidity (hypertension, hyperlipidemia, atrial fibrillation). Most of the patients in inter-hospital transfer group had been treated at a medical center. Compared to stroke patients treated at medical centers tend to have a higher severity of disability, are more likely to require intubation (e.g., tracheostomy tube, nasogastric tube, and urinary catheter tube), and tend to require more time to stabilize. These differences might explain why the inter-hospital transfer group in our study had a longer mean LOS in acute stage compared to the intra-hospital transfer group. However, further studies are needed to compare the service path and treatment costs in patients with varying severity of stroke and in stroke patients treated at different hospital levels.

After the acute stage, local low-volume rehabilitation facilities can usually provide adequate care [30]. Several studies have reported that, compared to skilled nursing facilities, intensive inpatient rehabilitation facilities achieve higher functional outcomes in PAC for stroke [30,31]. Additionally, duration of hospital stay and in-hospital mortality are related to socioeconomic status [32]. Therefore, increasing socioeconomic inequality over time has markedly increased inequality in stroke survival. Direct non-healthcare costs (including informal care costs, paid care costs and transportation costs) and rehabilitation costs are the largest post-stroke care costs [33,34]. Stroke patients who receive PAC and rehabilitation at local hospitals can reduce the physical, mental and economic burdens on their families. Our study analyzed data obtained from four local southern Taiwan hospitals that had the

largest volumes of stroke inpatients in PAC. Therefore, the data obtained in this work are highly representative of hospitals throughout Taiwan and have immediate applications.

In Japan, the rehabilitation is 3 hours, and the average LOS in rehabilitation facilities after discharge from tertiary hospitals is approximately 74.7 days [35]. The Taiwan PAC-CVD program provides the maximum of 12 weeks of services. Reimbursements are similar regardless of the hospital accreditation level (regional or district) and the equipment costs of the hospital. The PAC-CVD project enables stroke patients to access rehabilitation programs that are more intensive (in terms of frequency and duration of treatment) compared to those available under current NHIA provisions. The PAC-CVD project was expected to offer a continuous care model to restore function and reduce disability in stroke patients [8,9]. Most PAC-CVD studies published thus far have been studies of a limited case number in a single hospital [8,9]. In the post-acute stage, the mean LOS in the intra-hospital transfer group and the inter-hospital transfer group was 31.52 and 37.1 days, respectively. Another study of a large population of stroke patients reported a mean PAC stay of only 15.1 days [36]. Notably, the LOS of stroke patients in different countries is related to differences in post-stroke policy. For example, the average LOS for inpatient rehabilitation after stroke is approximately 1 month in Ireland, Switzerland, and Thailand [37]. In contrast, the average LOS inpatient rehabilitation after stroke is only 15.8 days in the United States, which is much shorter than that in western countries (e.g., 32.7 days in Germany, 35.3 days in Canada) [38–40]. The main reasons for the short LOS in the United States are the high economic burden of inpatient care for stroke patients and the availability of various well-established PAC rehabilitation facilities (e.g., inpatient rehabilitation hospitals, skilled nursing facilities, home health agency services, long term care hospitals) [4]. According to our multi-center data, the LOS for stroke patients who undergo PAC in Taiwan is similar to that in other countries. Notably, our data indicated that the effectiveness of rehabilitation training during the 12th week to the first year is better than in the first 12 weeks. These data confirm that a continuous rehabilitative PAC program is essential.

A study by Dewilde concluded that MRS level is a major determinant of medical resource use [41]. For this national PAC-CVD project, the MRS level was the main criterion for participation. Initially, only patients with MRS levels of 2 to 4 were eligible for transfer to hospitals with PAC wards. However, some people have proposed excluding patients in MRS level 2 or including patients in MRS level 5. In 2019, the Taiwan NHIA excluded MRS level 2 patients from this national PAC-CVD project. This study found that after adjusting for all relevant influential factors, the quality of life improvement was smaller in patients at MRS level 2 compared to patients at other MRS levels. Therefore, these data support the decision made by the Taiwan NHIA. In terms of maximizing efficiency in the use of limited healthcare resources, the decision to limit inpatient rehabilitation to patients with MRS 3 to 4 was not only justifiable, but necessary.

Although all research questions were adequately and satisfactorily addressed, two limitations are noted. This study only collected data for acute stroke patients for 40 days after stroke onset. Furthermore, this study only analyzed patients treated at four hospitals in south Taiwan. However, the numbers of patients treated in the PAC-programs at these hospitals were among the four highest of all district hospitals in south Taiwan. Further studies are needed to compare a PAC group and a control group in other regions of Taiwan and under current NHI regulations. Additionally, the two groups in this study were matched for demographic characteristics, clinical attributes, quality of care, and pre-rehabilitative functional status. Future studies could consider the use of inverse probability weighting rather than PSM.

#### **5. Conclusions**

In conclusion, this study showed that rehabilitative PAC improved outcomes of stroke rehabilitation. To achieve a vertically integrated medical system for stroke rehabilitation, the key requirements are improving the PAC ward qualifications of local hospitals, accelerating inter-hospital transfer, and ensuring a sufficient duration of intensive rehabilitative PAC. Early rehabilitation is important for successful restoration of health, confidence, and self-care ability in these patients.

**Author Contributions:** Study concept and design, C.-Y.W. and H.-Y.S.; acquisition of data, H.-H.H., K.-W.H., H.-F.L., H.-Y.C., S.-C.J.Y., and Y.-J.Y.; analysis and interpretation of data, C.-Y.W. and H.-Y.S.; drafting of the manuscript, C.-Y.W. and H.-Y.S.; revision of the manuscript, H.-H.H., K.-W.H., H.-F.L., H.-Y.C., S.-C.J.Y., and Y.-J.Y.; approval of the manuscript, all authors.

**Funding:** This study was supported by funding from the Pingtung Christian Hospital (PS104001) and Ministry of Science and Technology (MOST 104-2410-H-037-006-SS2 and MOST 106-2410-B-037-076) in Taiwan.

**Acknowledgments:** We thank all the individuals who participated in this study, without which this study would not have been possible.

**Conflicts of Interest:** Authors declare that they have no competing interest.

#### **References**


© 2019 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 (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Relationships between Multimorbidity and Suicidal Thoughts and Plans among Korean Adults**

**Youn Huh 1, Ga Eun Nam 2,\*, Yang-Hyun Kim <sup>2</sup> and Jun Hyung Lee 1,\***


Received: 30 June 2019; Accepted: 22 July 2019; Published: 24 July 2019

**Abstract:** Multimorbidity and suicide rates are rising simultaneously among Korean adults. To address this issue, we assessed the association between multimorbidity and suicidal behavior among adults aged ≥19 years in Korea. We analyzed the data from the 2017 Korea National Health and Nutrition Examination Survey. Multimorbidity was defined as experiencing two or more chronic diseases. We compared the presence of suicidal thoughts and plans according to multimorbidity using chi-square test, and examined the associations between multimorbidity and suicidal thoughts and plans using multiple binary logistic regression analyses. Multimorbidity was found in 30.8% of total participants. As the number of chronic diseases increased, the percentage of thoughts and plans tended to increase (*p* < 0.001 and *p* = 0.002). Among participants with multimorbidities, 8.5% had suicidal thoughts, whereas only 3.4% without multimorbidity had such thoughts (*p* < 0.001). Participants with multimorbidity had significantly higher odds of suicidal thoughts (OR = 2.14; 95% CI = 1.54–2.97) and suicidal plans (OR = 2.01; 95% CI = 1.08–3.73) compared to those without multimorbidity after adjusting confounding variables. **Conclusion:** People with multimorbidity had a higher prevalence of suicidal thoughts and plans. Early detection of and intervention for suicidal thoughts and plans are critical for suicide prevention among people with multimorbidity.

**Keywords:** chronic disease; multimorbidity; suicidal thoughts; suicidal plans

#### **1. Introduction**

The National Commission on Chronic Illness in the United States defines chronic illness as permanent illnesses accompanied by disability due to complication or injury, and those requiring special training for rehabilitation, long-term protection, monitoring, and treatment [1]. These individuals require continuous and comprehensive medical intervention, and communication and self-care is necessary [2]. Multimorbidity refers to the state of having multiple chronic illnesses simultaneously, potentially mixed with acute illnesses as well [3]. Unlike the concept of co-morbidity, which refers to how the impact of indicator diseases are influenced by co-morbid illnesses, the concept of multimorbidity is firmly centered on individuals with multimorbidity [4].

There has been an increase of chronic illnesses following the extension of longevity in society. As expected, in South Korea the prevalence of chronic illnesses and multimorbidity has been shown to increase with age. An analysis of older adults with one or more chronic illnesses revealed that 70.9% of older adults with chronic illnesses have multimorbidity, reaching an average of 4.1 chronic illnesses [5]. Another study found that older patients with multimorbidity had an average of 5.1 chronic illnesses, whereas patients without multimorbidity tended to have an average of 1.6 [5]. There have been several studies on the definition of multimorbidity [3,6]. Because people with two or more chronic medical

conditions are more common than those with three or more chronic medical conditions, former status is more frequently accepted as the definition of multimorbidity [6]. Studies adopting this definition reported that multimorbidity better reflected poor quality of life, impaired functioning, and increased use of medical facilities such as emergency admissions, particularly with higher numbers of coexisting conditions [7–11].

Suicide is one of the leading causes of death worldwide [12]. Among the countries in the Organization for Economic Co-operation and Development (OECD) as of 2012, the suicide rate (per the OECD standard population of 100,000 people) in South Korea was 29.1, which is the highest out of all OECD countries (average is 12.5). The specific risk factors associated with suicide among adults include depression and other mental disorders, illnesses, loss of function, the death of a partner, and other traumatic incidents [13]. By comparison, according to a 2006 report by the World Health Organization (WHO), the general risk factors of suicide include low socioeconomic status, education level, loss of a job, social stress, mental disorders, illnesses, and chronic pain [14,15]. One study found that 34% of people with suicidal thoughts have plans for suicide, 72% of people who plan for suicide actually attempt suicide, and 26% of people who have suicidal thoughts attempt suicide impulsively without having elaborated a prior plan to do so [16]. Therefore, suicidal thoughts and suicidal plans are important when discussing suicide.

Multimorbidity is associated with suicide-related risk, and primary care and mental health clinics need to evaluate for suicide ideation for patients with multimorbidity [17]. To address the impact of these issues, we aimed to investigate the increase in suicide-related variables in multimorbidity among Korean adults. Specifically, we examined the prevalence of suicidal thoughts and plans among adults who participated in the Korea National Health and Nutrition Examination Survey (KNHANES) in 2017, depending on whether they had multimorbidity.

#### **2. Methods**

#### *2.1. Participants and Data*

Drawing on raw data from the second year of the 7th KNHANES of 2017, we initially selected a total of 8127 individuals. Among them, we excluded people aged ≤18 years (*n* = 1938) and those who had any missing variables (*n* = 395), because people under the age of 18 did not perform suicide-related questionnaire in the 7th KNHANES of 2017. Finally, 5794 adults were considered in our analyses.

The KNHANES is a cross-sectional health surveillance system that assesses the status of and trends in national health and nutrition of South Korea to identify vulnerable groups to be prioritized in health policies, as well as evaluate whether existing health policies and projects are effective. It provides data on smoking habits, drinking habits, physical activities, and obesity, as well as various other statistics required by the WHO and OECD. The KNHANES comprises a health interview survey, a nutrition survey, a health examination survey, and data on demographic characteristics, diet, and health collected through personal interviews. Physical examinations along with blood and urine sampling were carried out at a mobile examination center. A stratified, multistage probability sampling design was used to select the household units that participated in the survey [18].

Our study adhered to the principles of the Declaration of Helsinki and was approved by the Institutional Review Board of Ilsan Paik Hospital (No. ISPAIK IRB 12 February 2017). Requirement of informed consent was waived because anonymous and de-identified information was used.

#### *2.2. Research Variables*

The chronic illnesses surveyed in the 7th KNHANES included hypertension, dyslipidemia, cerebral infarction (stroke), myocardial infarction, angina, osteoarthritis, rheumatoid arthritis, osteoporosis, tuberculosis, asthma, allergic rhinitis, depression, renal failure, atopic dermatitis, diabetes mellitus, thyroid disease, stomach cancer, liver cancer, colon cancer, breast cancer, cervical cancer, lung cancer, thyroid cancer, other cancers, hepatitis B, hepatitis C, and cirrhosis. Participants who self-reported

not currently having a disease or a doctor's diagnosis of a disease were regarded as not having it; by contrast, participants who self-reported having a disease or a doctor's diagnosis of one were defined as having a chronic illness. Of these, stomach cancer, liver cancer, colon cancer, breast cancer, cervical cancer, lung cancer, thyroid cancer, and other cancers were combined into an overarching "cancer" category.

In the mental health section of surveys, participants were asked to answer the following questions (employing answer options of yes, no, or I do not know/no answer): "Have you seriously considered suicide in the last year?" and "Have you seriously planed suicide in the last year?"

People who had smoked at least 100 cigarettes in their lifetime and continued smoking at the time of survey were defined as current smokers. People who consumed at least one alcoholic beverage per month in a year prior to the survey were considered alcohol drinkers and those who did not were considered non-drinkers. Education level was categorized into two groups depending on whether participants had achieved a higher qualification than middle school graduation or not. Monthly personal income level was divided into two groups: the lowest quartile group and the second-lowest to highest quartile group.

#### *2.3. Analysis Method*

We combined the raw data from the 2017 KNHANES according to the KNHANES raw data analysis guidelines. Based also on a complex sample design, we conducted all analyses by assigning a dispersed stratification estimation, stratification variables, and weighted sample values. We divided participants who had 2 or more of the above-stated chronic illnesses as multimorbidity based on the definition used in the previous study [6]. Subsequently, a chi-square test was performed to evaluate proportions of suicidal thoughts and plan according to the number of chronic illness or the presence of multimorbidity. Then, to determine the association of suicidal thoughts and plans with multimorbidity, we conducted a multiple binary logistic regression analysis and calculated the odd ratio (OR) and 95% confidence intervals (CIs). Model 1 did not adjust. Model 2 adjusted for sex and age. Model 3 adjusted for sex, age, education, personal income, smoking status, and alcohol consumption. Since there are studies showing that income and education level are related the risk of suicide, we adjusted these variables [14,15]. Stratified analyses according to sex and age were also performed. All analyses were performed using SPSS Statistics 21.0 (IBM Corp., Armonk, NY, USA).

#### **3. Results**

#### *3.1. Participants' General Characteristics*

Participants' demographics are shown in Table 1. A total of 5794 adults were included. Of the entire sample, 30.8% had multimorbidity. The proportion of patients with multimorbidity increased with age. The proportion of women with multimorbidity was higher than that of men. In addition, the ratio of people with multimorbidity was higher when education level or income was low. Regarding smoking, the percentage of patients with multimorbidity was the highest among ex-smokers. The proportion of patients with multimorbidity was higher in non-drinkers than in drinkers.

**Table 1.** General characteristics of participants. Data are presented as % (standard error (SE)).



**Table 1.** *Cont.*

#### *3.2. Suicidal Thoughts and Plans According to the Number of Chronic Diseases*

Regarding suicidal ideation, 5.0% of participants had suicidal thoughts and 1.3% had suicidal plans. In the absence of chronic diseases, suicidal thoughts were present in 2.7% of the cases, and suicidal plans were present in 0.6%. If the patients had one chronic disease, suicidal thoughts were present in 4.5% of the cases, and suicidal plans were present in 1.6%. Suicidal thoughts and plans were present in 9.7%, and 2.2% of participants who had more than 3 chronic diseases, respectively. As the number of chronic diseases increased, the proportions of suicide thoughts and plans tended to increase (*p* < 0.001 and *p* = 0.002) (Table 2).

**Table 2.** Suicidal thoughts and plans according to the number of chronic diseases. Data are presented as % (standard error (SE)). \* Obtained by using chi-square test.


#### *3.3. Suicidal Thoughts and Plans According to the Presence of Multimorbidity*

Table 3 shows the relationship between multimorbidity and suicidal thoughts and plans. While 3.4% of adults without multimorbidity were found to have suicidal thoughts, 8.5% of adults with multimorbidity had such thoughts (*p* < 0.001). As for suicide plans, 2.1% of adults with multimorbidity and 1.0% without multimorbidity were found to have them, respectively (*p* = 0.003).


**Table 3.** Suicidal thoughts and plans among participants with and without multimorbidities. Values are presented as weighted % (standard error (SE)). \* Obtained by using chi-square test.

#### *3.4. Multivariable Logistic Regression Analysis Between Multimorbidity and Suicidal Thoughts and Plans*

Table 4 presents multivariable logistic regression analysis results regarding the association of multimorbidity with suicidal thoughts and plans. Compared to participants without multimorbidity, participants with multimorbidity had significantly higher odds of suicidal thoughts in all adjusted models (OR, 95% CI: 2.65, 1.95–3.60 in model 1; 2.49, 1.82–3.40 in model 2; 2.14, 1.54–2.97 in model 3). Stratified analysis with sex and age showed similar findings. In men, the odds of suicide thoughts increased by 3.1 times among people with multimorbidity compared to those without multimorbidity after adjusting for all potential variables (OR, 95% CI: 3.09, 1.84–5.20 in model 3). The odds of suicide thoughts increased by 3.2 times in younger people (19–40 years) with multimorbidity compared with those without it (OR, 95% CI: 3.15, 1.43–6.97 in model 3). For suicidal plans, compared to participants without multimorbidity, participants with it had higher chances of suicidal plans (OR, 95% CI: 2.14, 1.30–3.54 in model 1; 2.26, 1.22–4.20 in model 2; 2.01, 1.08–3.73 in model 3). Suicidal plans were 2.3 times more prevalent in women with multimorbidity than in those without multimorbidity (OR, 95% CI: 2.30, 1.07–4.97 in model 2); however, this association was attenuated after further adjustment in model 3. Suicidal plans were 2.5 times more prevalent in middle aged people (41–64 years) with multimorbidity than in those without multimorbidity (OR, 95% CI: 2.54, 1.12–5.77 in model 3). On the other hand, there was no statistically significant association between multimorbidity and suicidal plans among younger or older men.

**Table 4.** Associations of multimorbidity and suicidal thoughts and plans. OR, odds ratio; CI, confidence interval. Model 1 was not adjusted. Model 2 was adjusted for sex and age. Model 3 was adjusted for sex, age, education, personal income, smoking status, and alcohol consumption. Values were calculated by multivariable logistic regression analysis.



**Table 4.** *Cont.*

#### **4. Discussion**

In this study, it was shown that 30.8% of the total population of the 7th KNHANES has multimorbidity with more than two chronic diseases. As other studies have shown, the prevalence of multimorbidity increases with age, [19] with 60.4% of people aged 65 and older having multimorbidity. The proportion of women found with multimorbidity is overwhelmingly higher than that of men, which is the same as in previous studies [20]. Further, this study found evidence that level of education and smoking and drinking habits are related to multimorbidity.

We verified differences in suicidal thoughts and plans according to whether participants had multimorbidity. A comparative analysis was performed on these two groups drawing on data from the 2017 waves of the 7th KNHANES. Participants who had multimorbidity had a significantly greater prevalence of suicidal thoughts compared to those without such group of diseases. It was also found that the higher the number of chronic diseases, the higher the percentage of suicidal thoughts and suicidal plans.

As life extension continues to increase in South Korea, it has become necessary to change the perception of diseases among adults. This new understanding is essential for properly managing patients with multimorbidity [3]. Multimorbidity is highly likely to occur from polypharmacy, adverse drug side effects, or competing medical recommendations [21–23]. Moreover, if a person has multimorbidity, their risk of suicide appears to increase. The concurrence of chronic illnesses can lead to a decline in functional state and increase in mortality; therefore, more emphasis must be placed on the comprehensive impact of multimorbidity on patients' overall health and quality of life [24].

When a stratified analysis was conducted by age, younger people with multimorbidity were more at risk of suicide than those without multimorbidity. In this study, the associations of multimorbidity with suicidal thoughts in older adults aged over 65 were lower than in other age groups and did not affect suicidal plans. However, the suicide rate among older adults is increasing as the population of this group increases. According to South Korea's statistics for 2014, the suicide rate is approximately 37.5 per 100,000 people aged in their 60s, 57.6 per 100,000 people aged in their 70s, and 78.6 per 100,000 people aged in their 80s, which are all much higher than the rates of 17.8 and 27.9 of people in their 20s and 30s, respectively. These rates indicate that older adults have a higher risk of suicide compared to other age groups [25]. It is also relevant to mention that suicide attempts among older adults are more likely to lead to death compared to other age groups [25,26]. However, unlike suicide in other age groups, suicide among older adults is generally not impulsive, and diverse realistic factors tend to contribute to its occurrence [26,27].

When a stratified analysis was conducted by sex, there was a significant difference of suicidal thoughts according to multimorbidity in men and women. In this study, the associations of multimorbidity with suicidal thoughts in men are higher than in women. This shows that there is a difference in suicide between women and men, corroborating other studies [28–30]. However, suicidal plans according to multimorbidity in men and women did not significantly differ.

Although this study provides reliable basic data on multimorbidity and suicidal thoughts and plans in a representative sample of Koreans, it does not go without limitations. First, because of their nature, cross-sectional studies cannot explain causal relationships between disease status and the various suicide-related relevant variables among adults. Furthermore, we evaluated only the presence of suicidal thoughts in the last year, without considering the frequency of these thoughts. Second, because participants' disease status was recorded based on self-reports, it is possible that biases affected the data. Third, we could not perform a detailed survey of assessed suicidal thoughts and plans, as we relied on previously written questions. In addition, while we might have considered some of the factors that influence suicidal thoughts, we could not account for all confounding variables. Finally, because older adults only accounted for a small proportion of all participants in the KNHANES, there was a major difference in the number of participants between disease groups. Therefore, there is the chance of false negatives and underestimation for some age-related results.

Previous studies have analyzed suicidal thoughts in relation to chronic illnesses; however, this study is unique in verifying whether participants had suicidal thoughts and plans according to whether they had multimorbidity. The presence of suicidal thoughts is a key risk factor for death by suicide. Suicidal behavior often recur in people who have previously attempted suicide, and suicide attempts often occur in people who frequently think about suicide [31,32]. Therefore, to prevent suicide, people with suicidal thoughts must be categorized as a high-risk group for suicide attempts and be cared for accordingly. By extension, it is necessary to consider multimorbidity in suicide prevention measures, as these individuals appear at risk of suicide. Interventions by medical professionals should, therefore, consider these thoughts as reflective of a person's mental health status [33]. Four weeks before suicide death, about 50% visited medical instructions such as outpatients, inpatients, and emergency room visits, and 83% of them visited within one year [34]. Therefore, it is necessary to screen suicidal risk for high-risk patients.

In conclusion, early detection and intervention regarding suicidal thoughts is essential in suicide prevention strategies for older adults with multimorbidity. Further research is needed to determine the suicide-related risk by chronic diseases and any clusters of chronic diseases.

**Author Contributions:** G.E.N., and Y.H. were the principal investigators. They contributed substantially to the study design, literature search, collection and assembly of data, data analyses and data interpretation. G.E.N., Y.H., and J.H.L. wrote all drafts and the final version of the report. Y.H., and J.H.L. analysed data and created all the tables. G.E.N., Y.H., Y.-H.K., and J.H.L. contributed to the conception and design of the study, the collection and assembly of data, data analyses and data interpretation. All authors contributed to preparation of the report and approved the final version. G.E.N., and J.H.L. had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis and all authors had final responsibility for the decision to submit for publication.

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

#### **References**


© 2019 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 (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Two-Dimensional Laser-Align Device for Ultrasound-Guided Injection**

**Eric Chiwei Shiao <sup>1</sup> and Po-Ling Kuo 1,2,\***


Received: 30 May 2019; Accepted: 15 July 2019; Published: 18 July 2019

**Abstract:** Ultrasound-guided injection is a widely used technique, however, it takes substantial amounts of time for novices to master the skill. The most critical issue to improve the accuracy of the injection is to align the needle with the scan plane of the ultrasound beam and orient the needle angle after piercing skin to aim at the targeted tissue. In the present study, we developed a two-dimensional laser align device to assist physicians to accurately position the needle in the scan plane and advance it at an angle correctly pointing to the target. The device is inexpensive, light-weighted, and easy to fabricate and accommodate for any types of ultrasound probe. Statistical analysis revealed that the assistance with our device significantly reduced the successful targeting time and times of retargeting in comparison with the traditional freehand approach or only with in-plane assistance for inexperience subjects. Our results indicate that the device exhibits great potential in effectively reducing the learning time to master the skill and speeding up the procedure for ultrasound-guided injection.

**Keywords:** ultrasound-guided injection; laser assisted; long-axis injection

#### **1. Introduction**

As point-of-care has gained increasing attention, ultrasound-guided injection becomes an important and popular technique for physician. The real-time feature of ultrasonography allows clinicians to target tissue for treatment more accurately and efficiently. However, it takes substantial amounts of time for novices to master the skill. For example, the in-plane approach requires well-trained coordination between hand and eyes to accurately position the needle in the scan plane and advance it at an angle correctly pointing to the target. It is of great importance to develop tools that assist the novices in accessing the target more easily and hence to improve their learning curve.

Several devices have been developed to assist clinicians in aiming the target more quickly and precisely [1]. However, limitations still exist. For examples, the Infiniti PlusTM needle guidance system (CIVCO Medical Solutions, Kalona, IA, USA) lacks the flexibility for direct adjustment of needle angle once the needle is inserted into the tissue [2]; the electromagnetic tracking system [3], optical tracking system [4,5], and robotic assistance system [6] are too large and expensive to be used in regular clinics. In contrast, laser assisting approaches are relatively inexpensive and compact-sized. However, most of the recently developed laser-based systems for needle guidance [7–9] only focus on the assistance in positioning the needle in the scan plane, without aiding its insertion angle.

In this research, we developed a novel two-dimensional (2D) laser-align (LA) device to help clinicians position the needle in the scan plane and at the right angle to precisely access the targeted tissue in ultrasound-guided injection. The device is inexpensive, light-weighted, and easy to fabricate and accommodate for any types of transducer. The device is switchable among three modes: 2DLA assisting, one-dimensional (1D) in-plane LA assisting without guidance for the inserting angle of the needle, and no LA assisting, namely the traditional freehand approach. We further compared the changes in the performance of ultrasound-guided targeting for novices and experienced physicians when assisted by different modes.

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

#### *2.1. Device Fabrication*

The 2DLA device consisted of a 5 V power module containing two coin batteries, a laser module, a clip, and a transparent film printed with slanting lines corresponding to various angles for needle advancement. The laser module included two laser units and on-off and brightness were independently controlled by two sets of switch and variable resistor, respectively. This design allowed users to switch the device function among the 2D, 1D, or no LA assisting mode. The variable resistors let the user change the brightness of lasers in an environment of various darkness.

The 2DLA assistance was accomplished by composing the laser module with two miniature laser diodes and their associated lenses, such that a laser line and a laser point were projected along and on a line aligned with the scan plane of the ultrasound beam, respectively. The pointing laser unit was used to guide the needle angle when entering the tissue. This was done by matching the projected laser point with a particular mark on the surface of the syringe body, while the marks corresponding to various needle angle were determined using the method described below. The lining laser unit was converted from a pointing laser unit with a half cylinder shaped lens, which refracted a point light source into a line one. It emitted a sector plane which aligned with the scan plane of the ultrasound beam. The sector plane projected a line on the skin surface next to the transducer and the point where the line began indicated the point where the needle tip was expected to enter the skin.

The clip was custom-made by 3D printing to fit a 10 MHz ultrasonic probe (BenQ, Taipei, Taiwan). It firmly mounted the laser units onto the probe at a specific angle, such that the pointing laser always projected on the syringe surface whatever the needle angle was and provided sharp laser contrast for needle and syringe positioning. The angle was determined as follows. Given a syringe of body length *L* and its attached needle of length *l*, the syringe body should be confined within the region between the interior and exterior of two concentric circles of radius *L* and *l*, respectively, as the needle angle to the skin changes. It is assumed that the circles are centered at point *p*, where the needle tip was placed, as shown in Figure 1a. To ensure that the spot emitted from the pointing laser unit always projects on the syringe body as the needle angle changes, the unit could be positioned either at position *a* or *b* as shown in Figure 1a, both of which are close to the two ends of the syringe body when the needle angle was 90◦ to the skin, respectively, and the laser emitted from both positions is tangent to the inner circle. These setups allowed the projection to cover the whole syringe body and hence provided better spatial resolution for angle guidance.

Let the angle of the laser light with respect to the line vertical to the skin, and that of the needle to the skin be ϕ and θ, respectively, and the distance from the needle tip to the point where the laser light projected on the syringe body be *x* (Figure 1b), we have

$$
\varphi\_{\mathfrak{d}} = \sin^{-1}(l/\overline{p\mathfrak{a}}) = \sin^{-1}(l/L), \tag{1}
$$

$$
\varphi\_b = \cos^{-1}(l/L),
\tag{2}
$$

$$
\overline{p\overline{b}} = \mathcal{L} \times \cot \varphi = l \times \mathcal{L} / \sqrt{(\mathcal{L}^2 - l^2)},\tag{3}
$$

*x* = *h*(sin θ + tan(ϕ − θ) cos θ) = *h* sinϕsec(ϕ − θ), (4)

$$\mathbf{x}' \equiv d\mathbf{x}/d\theta = -\mathbf{h} \times \sin\varphi \tan(\varphi - \theta) \sec(\varphi - \theta),\tag{5}$$

where ϕ<sup>a</sup> and ϕ<sup>b</sup> represents the angle ϕ when the unit was set at position *a* and *b*, respectively, *h* denotes *pa* or *pb* in each condition, and *x*' is the slope of *x* against various θ.

**Figure 1.** Geometric representation of the proposed positioning of the laser units. (**a**) Proposed setups of the pointing laser unit. (**b**) Geometric characteristics of the setup. (**c**) Profiles of slope of *x* against various needle angles. (**d**) Geometric layout of the lining laser unit.

Figure 1c depicts the change in *x*' as the needle angle varied. Note that larger *x*' offered finer spatial resolution for the guided points projected on the syringe body. Thus, the laser emitted from position *a* were projected on a wider range at larger needle angle, while that of position *b* was opposite. Since the needle angle is usually smaller than 45◦ in regular clinical practice, we chose position *b* to setup the pointing laser unit. The projected range of the pointing laser calculated using Equation (2) and Equation (3) would be the widest as the needle angle varied from 0◦ to 30◦, if ϕ and *pb* were set to be 61.5◦ and 5.42 cm, respectively. An array of marks corresponding to various θ was determined using Equation (4) and labeled on the syringe. Thus, the needle was expected to enter the skin at a specific angle if the emitted laser spot was aligned with the mark corresponding to the angle.

The lining laser unit was set close to the pointing one to save the device size. Figure 1d demonstrates the geometric layout of the setup. The emitting angle of the laser sector plane is denoted by β, while the plane was aligned with the scan plane of the ultrasound beam and projected as a line on the surface of the skin or the syringe. The vertical distance between the unit and the skin surface is *H*, and *d* represents the distance between the end of the probe and the beginning of the projected line, which corresponds to point *p* in Figure 1a. Note that *d* is usually about 1 cm in clinical setting. Hence, the unit had to be tilted and γ represents the angle between the vertical line through the emitting point of the laser and the boundary of the sector plane, and we have

$$\gamma = \tan^{-1}(d/H). \tag{6}$$

Note that the unit was tilted by an angle of γ + 0.5β. We set β and γ as 52.7◦ and 9.46◦, respectively, and *H* = 6 cm, such that the two units were separated by 0.58 cm to reduce the device size.

A transparent film printed with slanting lines was attached to the ultrasound monitor. Each line represented a particular angle for needle entering and had the corresponding mark labeled on the syringe. The labels included marks for various angles and a line aligned with the long axis of the syringe. One was expected to enter the needle in the scan plane at correct angle to the target; once the syringe was aligned with the projected laser line and the pointing laser was pointed to the right mark that matching the angle determined by the slanting lines. Figure 2 illustrates the device prototype and the typical procedure of needle entering with the 2DLA assistance was demonstrated in Figure 3.

**Figure 2.** Schematic and photographs of the 2DLA device. (**a**) Diagram simulating the device in work. (**b**) A device assembled with an ultrasound probe. (**c**) The transparent film printed with the slanting lines for various needle entering angle and attached to an ultrasound screen.

**Figure 3.** Procedure of needle entering guided with the 2DLA mode. (**a**) Position the probe to clearly visualize the target (highlighted by the yellow circle) on the screen. (**b**) Turn on the lasers and place the needle tip right at the beginning point of the projected laser line (the end of the laser line close to the probe). (**c**) Reorient the needle to match the projected laser line with the straight line attached on the syringe surface, which indicates the long axis of the needle. (**d**) Determine the insertion angle (highlighted by the yellow circle) using the slanting lines attached on the screen. (**e**) Adjust the entering angle to match the laser spot projected by the pointing laser unit with the mark labeled on the syringe surface corresponding to the insertion angle determined in (**d**). (**f**) Enter the needle into the phantom with persistent needle orientation until the needle tip touches the target on the screen.

#### *2.2. Device Performance*

To evaluate the effectiveness of our device, twenty subjects were enrolled to perform ultrasound-guided targeting in phantoms. Six of the subjects were physicians experienced in ultrasound-guided injection and the other fourteen subjects had little experience in ultrasound imaging before this study.

The phantoms were mainly composed of liquid paraffin and thermoplastic styrene-butadiene rubber which allowed the phantoms being formed and reshaped easily. The thermal-cured phantoms were transparent, excellent for ultrasound transmission, and possessed elasticity similar to that of soft tissues. The transparency of the phantoms allowed direct visualization of the needle advancement during practice and the performance test. Blocks made of poly (dimethylsiloxane) and iron powders with high echogenicity were embedded in the phantoms to simulate targets. There were two types of phantoms used in this study (Figure 4a,b); one was for practicing and had targets embedded at the same distance to the phantom surface; and the other was made for the performance test and the distance of the simulated targets to the phantom surface varied from 1 to 2 cm.

**Figure 4.** Photographs of the two types of phantom and its use in the performance test. Phantoms for (**a**) practicing, (**b**) performance test for ultrasound-guided targeting and (**c**) A subject performed ultrasound-guided targeting with a masked phantom.

The performance test was conducted by asking all participants to needle three targets of different depths and each needling was guided with one of the three assisting modes in randomized order. The order of the target depths was randomized as well and the variation of target depth was to prevent subjects from being familiar with the target depth from the preceding needling. In the 2DLA mode, both the lining and pointing lasers were turned on; the subject was instructed to locate the target for injection on the B-mode image, place the needle tip at the beginning end of the projected laser line, align the needle and overlap the long-axis line labeled on the syringe body with the laser line, estimate the needle angle to the target using the slanting lines on the transparent film, match the mark on the syringe body corresponding to the angle with the laser point, and advance the needle until the tip touched the target and complete the task. The procedure in the 1DLA condition was similar except that the pointing laser was turned off and the subjects had to estimate the needle angle by themselves, while the laser units were completely turned off in the no LA assisting, freehand (FH) condition. Before the performance test, the participants had unlimited time to practice targeting in the three modes. In the performance test, the phantom was masked on the side facing the subject to prevent direct visualization of the needle advancement by the subject, while the opposite side of the phantom remained transparent for video recording (Figure 4c). For each assigned target, the subject was allowed to withdraw the needle unlimited times to adjust the needle angle until successful targeting, which was considered as the needle tip penetrated the target with the presence of the entire inserted needle on the screen. Performance was evaluated by the total duration spent from entering the needle to the phantom to the moment when a successful targeting was achieved; the number of needle withdrawals until successful targeting; and the last duration, defined as the time spent from the last withdrawal to the successful targeting, which was equal to the total time if there was no needle withdrawal.

#### *2.3. Statistics*

Statistical analysis was performed using IBM SPSS Statistics 25. The differences of the three performance variables grouped with respect to different subjects, task order, and target depths were analyzed with Kruskal-Wallis one-way ANOVA test. The differences of the performance conducted in the three modes were examined using Mann-Whitney U test. Significance was set as *p* < 0.05.

#### **3. Results**

We first asked whether the performance of individuals of the same experience level was significantly different, and whether the performance was affected by the task order and the target depths. Table 1 summarized the statistical results of the performance data (Table A1) grouped against different subjects, task order, and target depths, while the data were categorized based on the level of experience. Given the subjects of the same experience level, it appears that there was no significant difference between the

averaged total durations they spent to finish the three tasks, the number of needle withdrawals, and the last durations to achieve successful targeting. Likewise, the performance was not significantly different whether the task was the first or last conducted, or the target was positioned more superficially or deeper. Detailed data were listed in the Appendix A.

**Table 1.** Significances of differences between the means of the performance data categorized with respect to different subjects, task orders, and target depths. \* *p* < 0.05.


We next examined the performance differences in various assisting conditions. Figure 5 depicts the performance of the inexperienced and experienced subjects and the data were detailed in Table 2. As for the inexperienced subjects, they less frequently withdrew the needle for retargeting as more assistance was provided, while the last duration for successful targeting followed an opposite trend. As expected, they also spent significantly longer total duration in the FH condition than in the other two conditions. In contrast, the experienced subjects spent the shortest total duration and last duration in the FH condition than in the other two conditions. Detail *p*-values of comparisons between conditions were summarized in Table 3. As one may expect, the inexperienced subjects took significantly longer total duration and more frequent needle withdrawals than the experienced subjects in the FH condition. However, their performances were not significantly different when the LA assistances were employed, except that the experienced subjects spent significantly longer time in the last duration than the inexperienced ones in the 2DLA condition. Detail data of the two subject groups and the *p*-values of comparisons between them were summarized in Table 2.

**Table 2.** Means and standard deviations of inexperienced group and experienced group in FH, 1DLA, and 2DLA conditions with significance level of relation between two groups. \* *p* < 0.05.


**Table 3.** Significance of difference in the performance of the inexperienced and experienced subjects compared between various assistance conditions. \* *p* < 0.05, \*\* *p* < 0.01.


**Figure 5.** Comparisons of the performance of the subjects assisted by the three different modes. Data of (**a**) total duration, (**b**) withdrawal number, and (**c**) last duration of the inexperienced and experienced subjects performing in the freehand (FH), assisted with the lining laser (1DLA), and assisted with lining plus pointing laser (2DLA) condition. \* *p* < 0.05, \*\* *p* < 0.01.

#### **4. Discussion**

In the present work, we showed that for people new to ultrasound-guided injection, our novel device did help their performance in reducing the spent time and number of needle withdrawals for successful targeting. These results agree well with those reported previously [7–9], and we further demonstrated that our novel 2DLA design significantly helped the inexperienced operators to perform better than in the traditional freehand condition or simply assisted by a lining laser. By following the guidance provided by the pointing laser and the marks on the syringe, most of the inexperienced participants could pierce the target in one shot. This is reasonable since the 2DLA mode provided a vast amount of information for needle operation such that the subject's expertise in ultrasound-guided injection had a minor role. This may explain the fact that most of the performance of the inexperienced and experienced subjects was not significantly distinguishable when assisted in the two LA modes. Indeed, the experienced subjects even spent much longer time to achieve the successful targeting in the two LA modes than that in the FH condition, since they already had good eye-hand coordination and individual tempo for injection and may have been confused with the laser assistance. Furthermore, it may be cumbersome to check many parameters before needle advancement in the 2DLA mode, which may explain why it took longer time for the last successful targeting in both inexperienced and experienced groups. However, we believe this drawback can be overcome as the operators become familiar with the device. Nevertheless, this mode still improved the efficacy of the inexperienced operators as the total time was less, owing to the much fewer withdrawals. The results shown in Table 1 indicate that the randomly assigned targets of different depths were successful to prevent the participants from predicting the needle angle using the experience learned in preceding tests. Furthermore, these results suggest that the task difficulty perceived by the participants was not dependent on the target depth, nor the task order. Collectively, our results indicate that the 2DLA device has a high potential to speed up the learning curve of novices in ultrasound-guided injection.

With 5.5 × 5.0 × 2.8 cm in size, 32 g in weight, and a price of about \$20, our 2DLA device was compact, lightweight, inexpensive, easy to mount onto the probe, and provided immediate guidance if the operators need to change the target suddenly. Unlike the device designed by Daehee et al. [10], there is no direct contact between our device and patients' skin or needle, which reduced the risk of infection. These features render our device competitive in clinical use or medial training when compared with commercially available products. For example, although the Infiniti PlusTM needle guidance system provided by the CIVCO company has very affordable price (~\$10 for each set), it prohibits repeated adjustment of the needle orientation for injection of multiple targets, or immediate replacement of syringe of various size in conditions such as the combined operation of ultrasound-guided aspiration and injection [11]. The eTRAX needle tip tracking guidance system, which is also from the CIVCO, embeds an electromagnetic sensor in the needle tip to provide needle tracking, but it costs more than 2 thousand dollars and is roughly forty times larger than ours in volume [12]. The SCENERGY system from the Clear Guide Medical attaches stereocameras to the ultrasound probe and integrates the ultrasound image with the CT/MRI image to improve needle navigation, but the assistive device is much larger than ours, and it requires attachment of labels on the subject's skin for recognition, which may increase the risk of infection [13].

However, there are limitations in the present study. Although nonparametric tests were applied to make our results statistically stricter, the enrollment size was small. Furthermore, the marks labeled on the syringe for guidance of injection angle were calculated based on a flat skin surface, which may well approximate conditions when the radius of curvature of the skin is much larger than the probe size, such as scanning along the long axis of an arm. However, the guidance will be erroneous when entering the needle through skin surface with smaller radius of curvature, holding the ultrasound probe askew or using a curvilinear array probe. Additional calculation is required to fix the scale distortion arising in these conditions. For example, when our device is applied to a curvilinear probe, the slanting lines printed on the transparent film should be curved to match the sector image generated by the probe. Likewise, since the needle entering point is usually out of the image view when the probe is applied on a curved surface with a large curvature, such as scanning along the transverse plane of an arm, it is hard to predict the needle orientation unless the geometric condition between the probe and the entering point is given.

#### **5. Conclusions**

In summary, we developed a novel 2DLA device that provided guidance for needle orientation both for the image plane and the angle for entering. Statistical analysis revealed that the 2DLA mode significantly reduced the time for successful targeting and frequency of retargeting in comparison with the FH and 1DLA mode. Injecting quickly and accurately is one of the most challenging tasks during the training for ultrasound-guided injection. Our device exhibits great potential in effectively reducing the learning time to master the skill and speeding up the procedure.

**Author Contributions:** Conceptualization, E.C.S. and P.L.K.; methodology, E.C.S.; validation, P.L.K.; formal analysis, E.C.S.; investigation, E.C.S.; data curation, E.C.S.; writing—original draft preparation, E.C.S.; writing—review and editing, P.L.K.; visualization, E.C.S. and P.L.K.; supervision, P.L.K.; project administration, P.L.K.; funding acquisition, P.L.K.

**Funding:** This research was funded by the Ministry of Science and Technology, Taiwan, grant number 106-3114-B-038-001, 107-2321-B-038-002, 108-2321-B-038-003, and 107-2221-E-002-040; and theMinistry of Education, Taiwan, grant number 108L893902.

**Acknowledgments:** The authors are grateful for the valuable input about the tips of ultrasound-guided injection in clinical practices by Chueh-Hung Wu, Ming-Yen Hsiao, Shaw-Gang Shyu, and Yi-Chian Wang in National Taiwan University Hospital, Taipei, Taiwan.

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

#### **Appendix A**

**Table A1.** Data of each subject in inexperienced group and experienced group.



**Table A1.** *Cont.*

#### **References**


© 2019 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 (http://creativecommons.org/licenses/by/4.0/).

#### *Article*
