**1. Introduction**

Globally, injuries are a leading cause of mortality and morbidity for children and adolescents. In 2017, the Global Burden of Disease study estimated 4.48 million injury deaths globally, an increase of 5.3% since 1990 [1]. However, some progress is being made in reducing injury-related deaths, with both years of life lost and age-standardized mortality rates decreasing between 1990 and 2017 [1]. Injury-related morbidity is also a significant global concern. New cases of non-fatal injury are increasing, with 520 million cases recorded globally in 2017, while years lived with a disability also increased [1].

Understanding and preventing injuries is complex as they may be intentional or unintentional, be due to a range of mechanisms such as road transport, falls, drowning, burns, poisoning, interpersonal violence and suicide, impact all ages and require a range of strategies to prevent them from occurring [1–4]. As the

risk factors and prevention strategies needed to address unintentional and intentional injuries often di ffer, there is a need for studies which explore the causes of injury by age groups.

Injury risk is impacted by external factors including determinants of health. Determinants of health, referred to as the causes of the causes [5], are the conditions in which people are born, grow, live, work and play which impact health [6]. Determinants which impact health, including injury risk, include socio-economic conditions, daily living conditions, education levels and individual health-related factors [7,8]. Addressing the determinants of health helps to prevent events from occurring [5]. Strategies to prevent injuries need to be designed to address injury risk using a range of strategies from downstream (individual level) to upstream (system level), otherwise they are likely to be ine ffective, especially when used in isolation, and must address underlying determinants of health [9].

Geographical remoteness and socio-economic status are determinants which impact health, including injury [10,11]. Poorer health outcomes are seen in rural dwelling populations, with greater hospitalization rates and disease burden [12,13]. This increased injury risk in rural locations also goes hand in hand with socio-economic disadvantage [14], which has been identified as a factor impacting injury risk [15–18]. Effective injury prevention strategies must consider these (and other) determinants of health when identifying areas of need and designing interventions.

Children and adolescents experience significant fatal and non-fatal burden due to injuries [19] and are particularly vulnerable to harm due to drowning [2,20], falls [3,21] and road tra ffic injuries [4,22]. Reducing injury-related mortality and morbidity is vital in order for nations to meet child and adolescent health targets within the Sustainable Development Goals [23,24]. Reducing injury-related harm among children and adolescents represents the area where the greatest health gains can be made [25].

While fatal child injury rates are declining, there are clear variations based on socioeconomic inequalities [26]. Therefore, within an Australian context, this study aims to explore injury-related mortality among children and adolescents to identify the impact of determinants of health, specifically rurality and a composite measure of socio-economic advantage and disadvantage (Index of Relative Socio-economic Advantage and Disadvantage (IRSAD)) of residential location, with an aim to inform future prevention e fforts.

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

This study reports a total population analysis of injury-related mortality among children and adolescents aged 0–19 years (henceforth referred to as children) in Australia between 1 January 2007 and 30 June 2017 (a period of 10 years), with a particular focus on determinants of health—namely remoteness and IRSAD of the child's residential location.

### *2.1. Data Source*

Cause of Death Unit Record File data were sourced from the Australian Bureau of Statistics (ABS). Data are provided to approved applicants only, but similar publicly available data to the Cause of Death data release collated by the ABS annually are provided [27]. Variables made available for analysis were date of death, sex, age group, jurisdiction of death (Australian state or territory and statistical local area), International Classification of Disease (ICD)-10 cause of death code and statistical local area. A statistical local area is a geographical area as used by the ABS. This study specifically uses statistical area level 2 which represents a community that interacts together socially and economically [28].

### *2.2. Case Identification and Data Cleaning*

All deaths that had a primary cause of death injury ICD-10-AM [29] code were selected; only cases where the incident occurred during the study period were included, people who were aged less than 20

years were included and those who resided in Australia (i.e., visitors to Australia were excluded). Prior to commencing data coding and analysis, a total of 75 cases were removed; being 54 overseas residents and 21 with unknown residence.

This study examines injury-related deaths that were registered and who died between 1 January 2007 and 31 December 2017; noting that particularly for 2017, this would represent an approximate under-numeration of 6% (this proportion is based on the proportion of people who died within a given year but whose death was not registered until the following year). This particularly impacts those deaths which occur later in the year i.e., November and December. As such, trends over time are explored on Australian financial years 1 July to 30 June, from 1 July 2007 to 30 June 2017.

### *2.3. Coding of Injury Mechanisms*

Fifteen categories of injury mechanism were collated using the ICD-10 codes. Due to small numbers of cases, the mechanisms of 'overexertion, strenuous and repetitive movements' (X50) and 'contact with venomous animals and plants' (X20-29) were grouped into 'Other'. The coding structure for the categories is described in Table 1.

> **Table 1.** Injury categories used in study and associated International Classification of Diseases (ICD)-10 codes.


### *2.4. Coding of Determinants of Health*

The impact of determinants of health on child injury risk was explored by remoteness and IRSAD. The remoteness of the child's residential location was calculated by matching the nine digit statistical local area (SLA) code to the corresponding Australian Standard Geographical Classification (ASGC) category (i.e., major cities, inner regional, outer regional, remote and very remote) [30].

The nine-digit SLA was also used to code the victim's residential location to the corresponding decile on the index of socio-economic advantage and disadvantage (IRSAD). IRSAD aligns the statistical local area with a decile ranking (1–10), with areas ranked 1 being the most disadvantaged. Victims' residential postcode current IRSAD was used as a proxy for their familial socio-economic status [31]. IRSAD includes 17 measures around: income, education, employment, car ownership, internet connection, disability,

family structure and renting status [32], which are combined to produce 10 IRSAD deciles. The IRSAD deciles were coded to low (deciles 1–3), mid (deciles 4–7) and high (deciles 8–10) for ease of analysis. Data between 2006 and 2011 were coded to the socio-economic indexes for areas (SEIFA) classification in 2011 and data between 2012 and 2017 were coded to SEIFA 2016 [33].

### *2.5. Statistical Analysis*

Temporal trends over time in fatal injury mechanism were explored using the linear calculation in Microsoft Excel 365 (Build: 13426.20274). Crude rates and relative risk (RR) with a 95% confidence interval (CI) were used to calculate the impact of remoteness of residential location on injury-related fatalities. Crude rates per 100,000 population were calculated for all children 0–19 years, by sex and by age group (i.e., 0–4 years, 5–9 years, 10–14 years and 15–19 years), using the population from June of each year [34]. Population data by ASGC classification are currently only available for years in which the national census has been conducted. Therefore, a two-year average for the population was calculated using census years (2011 [35] and 2016 [36]) and this was used with a 10-year average of the deaths to calculate crude annualized injury-related fatality rates for children per 100,000 population by ASGC remoteness classification. Rates were used to calculate relative risk (using a MedCalc calculator [37]), with a 95% confidence interval using the lowest rate as the control group.

Univariate and chi-square analyses (calculated in International Business Machines [IBM] Corporation Statistical Package for the Social Sciences [SPSS] V25 [38]) were used to explore the impact of IRSAD on injury-related fatalities. Population data by grouped IRSAD decile (low, mid, and high) and age group are not publicly available in Australia. Therefore, the proportional of the all-age population in Australia as at 2016 by grouped IRSAD decile was calculated and assumed to hold true for children aged 0–19 years across the study period. These proportions were used to calculate non-parametric chi-square tests of significance. A modified Bonferonni correction, as suggested by Keppel [39], was applied at the 0.05 level, where multiple categories within a variable have been analyzed.
