Social Determinants Associated with Pediatric Burn Injury: A Population-Based, Case–Control Study

Abstract

Social determinants of health (SDoH) influence risk of injury. We conducted a population-based, case–control study to identify which social determinants influence burn injury in children. Children (≤16 years of age) admitted to a Canadian regional burn center between January 1, 1999 and March 30, 2017 were matched based on age, sex, and geographic location 1:5 with an uninjured control cohort from the general population. Population-level administrative data describing the SDoH at the Manitoba Center for Health Policy (MCHP) were compared between the cohorts. Specific SDoH were chosen based on a published systematic review conducted by the research team. In the final multivariable model, children from a low-income household odds ratio (OR) (95% confidence interval) 1.97 (1.46, 2.65), in care 1.57 (1.11, 2.21), from a family that received income assistance 1.71 (1.33, 2.19) and born to a teen mother 1.43 (1.13, 1.81) were significantly associated with an increased risk of pediatric burn injury. This study identified SDoH that are associated with an increased risk of burn injury. This case–control study supports the finding that children from a low-income household, children in care, from a family that received income assistance, and children born to a teen mother are at an elevated risk of burn injury. Identifying children at increased potential risk allows targeting of burn risk reduction and home safety programs.

Compared to adults, children are at a greater risk of burn injury.^1,2 Thinner skin anatomy results in more rapid full-thickness burn injury, leading to significant scarring.³ Postburn injury sequelae manifests as life-long increased rates of mental and physical illness, substance abuse, and suicide.⁴

The current view among burn care providers is that the majority of burn injuries are preventable, especially in children.² Burn research is therefore expanding to include prevention strategies.⁵ Social Determinants of Health (SDoH) are a broad range of social and economic factors that influence individual and population health.⁶ SDoH identifying individuals at risk of burn injury is one way of targeting prevention programs for greatest effect. Adverse SDoH are known to increase risk of physical injury in children; ⁷ however, the extent and influence of SDoH specifically on burn injury are less clear.

We recently published a systematic review examining SDoH and the influence on pediatric burn risk.⁸ The majority of published literature focused on the effect of income, geographic location, and the mental health of the child. Minimal research examined broader factors such as recent immigration, children of a teen mother, high residential mobility, or maternal mental health. Furthermore, previous research focused on children in the Middle East, Europe, and Australia. It is uncertain whether SDoH influences pediatric burn risk the same way in children from different environments. In Canada, the province of Manitoba provides a unique opportunity to examine the association of SDoH and burn injury on a population level, as healthcare is provided by a single insurance payer (the government) and information on economic and social factors is available in a linkable deidentified structure at the Manitoba Centre for Health Policy (MCHP).⁹

The goal of this case–control study is to evaluate the effects of SDoH on burn risk in children. Using matched controls from the general population, the Paediatric Burn Registry at a regional childrens’ hospital was linked with MCHP population-level administrative datasets. Informed by recent literature, we incorporated lesser studied SDoH into the case–control, filling the defined literature gap.⁸ This information is essential in order to target burn safety and prevention programs to those at greatest risk.

METHODS

Study Population

The case cohort consisted of children ≤16 years of age admitted a regional childrens’ hospital for a burn injury between January 1, 1999 and March 30, 2017 (n = 483). The burn (case) cohort was matched 1:5 to an uninjured control cohort (n = 2415) based on age, sex, and geographic region. Geographic regions were defined based on 11 regional health authorities in the province of Manitoba.¹⁰ An index date was defined for both cohorts as the date of the burn injury. To be included in either cohort, children must have held a valid Personal Health Information Number (PHIN) and Manitoba address for at least 1 year prior to index date. As part of the single payer health system in Manitoba, each resident is allocated a unique PHIN for reimbursement purposes. PHINs allow individual healthcare utilization in the province to be tracked.

Data Sources

Approval for this study was granted from the University of Manitoba’s Health Research Ethics Board and Health Information Privacy Committee (HIPC#2017/2018–75) prior to the study. Each patient PHIN was scrambled to protect privacy. Using the scrambled PHIN, deidentified data from the Pediatric Burn Registry at the Children’s Hospital were linked with the Population Health Research Data Repository located at the MCHP. The Burn Registry includes demographic data such as age and sex, as well as injury related factors.

MCHP datasets at the University of Manitoba contain deidentified health and socioeconomic information on residents of Manitoba for the purpose of health-related research. Extensive data have been collected as Manitoba has grown to greater than 1.3 million residents as of 2019.¹¹ Examples of data captured by the MCHP include hospitalizations, emergency department use, physician visits, drug utilization, vital statistics, census data, income level, and income assistance. The methodology of linking clinical data with administrative population-level datasets to evaluate risk factors for injury is unique.

SDoH Definitions

These validated definitions of SDoH included in this study arise are from the MCHP^12,13. Where applicable, these definitions were determined for the time prior to index date (at-risk period).

1. Low-income child: A child from a family belonging to the lowest-income quintile, a surrogate of neighborhood socioeconomic status, based on census data.

2. Child in care: A child or sibling removed from his or her family of origin and placed in the care of another adult due to concerns of neglectful care.

3. Rural: A child residing in a postal code outside of a city with a population of 50,000 people, based on residential postal code at index date.

4. Child from a family that has received income assistance: A child from a family that has received financial income assistance from the Employment and Income Assistance Program (EIA) in Manitoba.

5. Child with parent(s) involved in the justice system: A child with one or both parents involved in the justice system as a witness, victim, or accused, based on the data acquired via Prosecution Information and Scheduling Management (PRISM) developed by Manitoba Justice Prosecution Service.

6. Child of parents who did not graduate high school: Based on graduation status data from Manitoba Education.

7. Child with parent(s) in social housing: A child who has lived in social housing managed by Manitoba Housing and Community Development.

8. Child of parent(s) who immigrated: A child of parent from a native country that is not Canada. This is based on the probability of being in a family that immigrated calculated by MCHP.

9. Child from a family with high residential mobility: A child of a family that has moved residences three or more times within 10 person-years.

10. Child from a teen mom: A child of a mother who gave birth at age 19 or younger.

11. Child with a major mental health diagnosis: A child with an axis II mental health disorder, based on International Classification of Diseases (ICD-9) diagnosis codes 295 to 299.¹⁴

12. Child of a mother with a major mental health diagnosis: A child born to a mother with an axis II mental health disorder, based on ICD-9 diagnosis codes 295 to 299.¹⁴

13. Child from a mother with an Axis I mental health disorder: The presence of an axis I mental health disorder in the mother, measured using ICD codes. ICD-9-Clinical modification (ICD-9-CM) and the ICD-10-CA for anxiety (ICD-9-CM 300.0, 300.2, 300.3; ICD-10-CA F40, F41.0, F41.1, F41.3, F41.8, F41.9, F42, F43.1), depression (ICD-9-CM 296.2, 296.3, 296.5, 300.4, 309, 311; ICD-10-CA F31.3-F31.5, F32, F33, F34.1, F38.0, F38.1, F43.2, F43.8, F53.0) substance use disorders (ICD-9-CM 291, 292, 304, 305, 303; ICD-10-CA F10-F19, F55). “Any Axis 1 Disorder” combined anxiety, depression, and substance use into one variable. One or more hospitalization preindex date and/or one or more outpatient visits (physician billing) were considered an outcome diagnosis.

Analysis

Data analysis was performed using SAS version 9·4 (SAS Institute Inc., Cary, NC). Descriptive statistics on the two cohorts were performed and contingency tables were generated. Group differences were considered statistically different from each other when P < .05 (two-tailed). Conditional univariate analyses were performed on individual SDoH as covariates using case or control as the binary response. A multivariable logistic model was created including all covariates, whereas the final model included covariates with a relative rate odds ratio (OR) 95% confidence interval [CI] > 1 between cases and controls.

Role of Funding Source

The funding source had no role in study design, collection, analysis, interpretation, or writing of the manuscript.

RESULTS

Demographics

Descriptive comparisons of burn and control cohorts are shown in Table 1. No significant differences existed between the burn and control cohorts in distribution of age and sex. Table 2 should the total body surface area involvement of the burn injury/Table 3 outlines the mechanism of burn injury in the burn cohort. Scalds were the most common mechanism of injury representing 42.0% of all injuries.

Table 1.

Descriptive characteristics of burn and control cohorts

Demographic		Burn Cohort (n = 483)	Control Cohort (n = 2415)
Age	Mean ± SD (range)	5.46 ± 5.23 (0–17)	5.45 ± 5.24 (0–17)
	Median; IQR	3; 8	3; 8
Sex	Male, N (%)	312 (64.60)	1560 (64.60)
	Female, N (%)	171 (35.40)	855 (35.40)

Table 2.

Total body surface area

	Mean ± SD
Total burn surface area, %*	10.5 ± 13.4 (range 1–94.8)

Table 3.

Mechanism of burn injury

	N (%)
Chemical	8 (1.66)
Contact	38 (7.87)
Electrical	11 (2.28)
Fire/flame	150 (31.06)
Scald	203 (42.03)
Unspecified	73 (15.11)
Total	483 (100)

Social Determinants Prior to Burn Injury

Differences in SDoH between burn and control cohorts based on univariate analysis are shown in Table 4. Children with burn injury were more likely to come from a family of low income, been placed in care, come from a family who ever received income assistance, lived in social housing, had high residential mobility, or been born to a teen mother.

Table 4.

Social determinants of burn and control cohorts: univariate analysis

	Burn Cohort (N = 483) N (%)	Control Cohort (N = 2415) N (%)	Test Statistic Chi-sq.	P	Odds Ratio (95% CI)
Low Income	203 (42.03)	795 (32.92)	14.79	<.0001	1.48 (1.21, 1.80)
Child in Care	62 (12.84)	173 (7.16)	18.91	<.0001	2.10 (1.50, 2.93)
Rural	275 (56.94)	1350 (55.90)	3.37	.07	4.18 (0.91, 19.3)
Income Assist	180 (37.27)	583 (24.14)	35.75	<.0001	1.87 (1.52, 2.30)
Parental Justice	44 (9.11)	207 (8.57)	0.15	.70	1.07 (0.76, 1.50)
Parental Non-Grad	32 (6.63)	196 (8.12)	1.23	.27	0.80 (0.55, 1.18)
Housing	60 (12.42)	219 (9.07)	5.20	.02	1.42 (1.05, 1.93)
Immigrant	25 (5.18)	106 (4.39)	0.58	.45	1.19 (0.76, 1.86)
Residential Mobility	113 (23.40)	411 (17.02)	11.0	.0009	1.49 (1.18, 1.89)
Teen Mom	221 (45.76)	829 (34.33)	22.75	<.0001	1.61 (1.32,1.97)
Child Axis II	7 (1.45)	19 (0.79)	1.97	.16	1.89 (0.78, 4.59)
Maternal Axis II	47 (9.73)	205 (8.49)	0.78	.29	1.20 (0.86, 1.68)
Maternal Axis I	131 (27.70)	565 (23.71)	3.40	.07	1.24 (0.99, 154)

Income Assist, Child from household with history of income assistance; Parental Justice, Child with a parent involved in the justice system; Parental Non-Grad, Child of a parent who did not graduate high school; Immigrant. Child of a parent who has immigrated, Residential mobility, Child of a parent with a high residential mobility.

Parental Mental Health

As a sensitivity analysis, Axis I mental health of fathers and “any parent” was compared to mental health of mothers between the cohorts (Table 5). In the Burn Cohort (N = 476), there were 473 mothers identified (99.3% of children in this cohort) and 187 fathers identified (39.3%), whereas in the Control Cohort (N = 2394), there were 2383 mothers (99.5%) and 1120 fathers identified (46.8%). No significant differences were noted between the relative rates of maternal, paternal or “any parent” Axis I disorders. In our previous systematic review,⁸ maternal mental health was the most commonly evaluated factor. Given the lack of identified difference in the sensitivity analysis, as well as nearly 100% of children have an identified mother compared to less than 50% of children with an identified father, maternal mental health was chosen in the current study.

Table 5.

Parental mental health between burn and control cohorts

	Burn Cohort (N = 473) N (%)	Control Cohort (N = 2383) N (%)	Test Statistic Chi-sq.	P Chi-squared	Odds Ratio (95% CI)
Maternal Axis I	131 (27.70)	565 (23.71)	3.40	.07	1.24 (0.99, 1.54)
	Burn Cohort (N = 187)	Control Cohort (N = 1120)
Paternal Axis I	28 (14.97)	130 (11.61)	1.71	.19	1.34 (0.86, 2.09)
	Burn Cohort (N = 476)	Control Cohort (N = 2394)
Any (maternal, paternal, or both) Axis I	149 (31.30)	652 (27.23)	3.27	.07	1.22 (0.98, 1.51)

Correlation Matrix

Table 6 displays a correlation matrix to evaluate overlap between SDoH categories. Degree of similarity guided appropriate selection for a model for analysis. Correlation was defined as “low” = 0.1 to 0.3, “medium” = 0.3 to 0.5, “high = 0.5 to 0.9,” “very high” ≥ 0.9). No variables were associated with a “high” degree of correlation. A “medium” degree of correlation was found between income assistance and housing, residential mobility and teen mother. Due to the absence of “very high” correlation, the logistic regression model was created using all variables.

Table 6.

Correlation matrix

	Low Income	Child in Care	Rural	Income Assist	Parental Justice	Parental Non-Grad	Housing	Immigrant	Residential Mobility	Teen Mom	Child Axis II	Maternal Axis II	Maternal Axis I
Low Income
Child in Care	0.22, P < .0001
Rural	0.03, P = .07	0.02, P = .38
Income Assist	0.22, P < .0001	0.20, P < .0001	−0.13, P < .0001
Parental Justice	0.02, P = .36	−0.002, P = .93	0.02, P = .26	−0.03, P = .10
Parental Non-Grad	−0.01, P = .42	−0.01, P = .53	0.03, P = .09	0.04, P = .04	−0.02, P = .36
Housing	0.13, P < .0001	0.17, P < .0001	−0.10, P < .0001	0.46, P < .0001	0.03, P = .13	0.02, P = .34
Immigrant	−0.01, P = .56	−0.06, P = .0005	−0.11, P < .0001	−0.08, P < .0001	−0.02, P = .3	−0.06, P = .0006	−0.05, P = .009
Residential mobility	0.15, P < .0001	0.18, P < .0001	−0.24, P < .0001	0.38, P < .0001	−0.01, P = .56	0.02, P = .4	0.19, P < .0001	−0.09, P < .0001
Teen mom	0.26, P < .0001	0.20, P < .0001	0.13, P < .0001	0.33, P < .0001	0.03, P = .13	0.07, P = .0001	0.18, P < .0001	−0.13, P < .0001	0.17, P < .0001
Child Axis II	−0.007, P = .69	−0.001, P = .94	−0.02, P = .31	−0.03, P = .09	0.01, P = .60	−0.01, P = .44	0.006, P = .74	−0.003, P = .87	−0.03, P = .17	−0.03, P = .07
Maternal A xis II	−0.005, P = .80	0.09, P < .0001	−0.04, P = .04	0.06, P = .002	0.06, P = .003	0.009, P = .62	0.07, P < .0001	−0.04, P = .02	0.07, P = .0005	0.03, P = .11	−0.02, P = .40
Maternal Axis I	0.02, P = .41	0.15, P < .0001	−0.05, P = .006	−0.007, P = .70	−0.007, P = .70	0.03, P = .16	0.14, P < .0001	−0.03, P = .15	0.14, P < .0001	0.07, P = .0003	−0.01, P = 0.61	0.22, P < .0001

Logistic Regression

Table 7 shows ORs from conditional multivariable logistic regression including all covariates. Burn-injured children are more likely be in come from low-income households, been placed in care, come from a household that has received income assistance and been the child of a teen mom. Table 8 displays the conditional multivariable logistic regression model including only covariates with an OR 95% CI > 1. In the final multivariable model, children from a low-income household OR (95% CI) 1.97 (1.46, 2.65), in care 1.57 (1.11, 2.21), from a family that received income assistance 1.71 (1.33, 2.19) and born to a teen mother 1.43 (1.13, 1.81) were significantly associated with an increased risk of pediatric burn injury.

Table 7.

Conditional multivariable logistic regression, all covariates

	OR (95% CI)	P
Low Income	1.90 (1.40, 2.58)	<.0001
Child in Care	1.68 (1.17, 2.41)	.005
Rural	4.11 (0.87, 19.46)	.07
Income Assist	1.78 (1.35, 2.37)	<.0001
Parental Justice	1.18 (0.83, 1.70)	.36
Parental Non-Grad	0.73 (0.47, 1.14)	.16
Housing	0·79 (0.55, 1.14)	.21
Immigrant	1.54 (0.94, 2.52)	.09
Residential Mobility	1.09 (0.82, 1.46)	.55
Teen Mom	1.53 (1.20, 1.96)	.0006
Child Axis II	2.16 (0.81, 1.33)	.12
Maternal Axis II	1.06 (0.75, 1.52)	.73
Maternal Axis I	1.04 (0.81, 1.33)	.78

Table 8.

Conditional multivariable logistic regression, significant covariates only

	OR (95% CI)	P
Low Income	1.97 (1.46, 2.65)	<.0001
Child in Care	1.57 (1.11, 2.21)	.01
Income Assist	1.71 (1.33, 2.19)	<.0001
Teen Mom	1.43 (1.13, 1.81)	.003

DISCUSSION

This study suggests that children from a low income, in care, from a family that has received income assistance and children of a teen mom are at elevated risk of burn injury in Manitoba, Canada.

Previous research has indicated a relationship between children from socioeconomically deprived families and an increased risk of burn injury.^15,16 Much of this research has relied on indexes of deprivation as a surrogate marker of income level. Randall et al found burns to be more prevalent in the lowest socioeconomic quintiles based on Socio-Economic Indexes For Areas (SEIFA) quintiles, which include household income, education, employment, occupation, and housing.¹⁵ Baker et al relied on the Index of Multiple Deprivation to find that children from the most deprived quintile had relative risk of burn injury of 2.74 (95% CI 2.35, 3.20) compared to the least deprived quintile.¹⁶ Our study is unique in that it did not use a composite index measure of deprivation, rather it is based on census-derived income data. This isolated the effect of family income on burn injury risk. This study agrees with previous research, finding that children of families belonging to the lowest-income quintile had an OR of 1.97 (95% CI 1.46, 2.97).

Only one previous study analyzed the relationship between children in care and burn injury.⁸ Kendall-Grove et al included involvement of Child Protective Services as part of a composite measure of family dysfunction and found that 36% of parents of burned children reported significant dysfunction.¹⁷ In the final multivariable model, we found a significant association between burn injury and child being in care, OR 1.57 (95% CI 1.11, 2.21), P = .01. This novel SDoH presents a unique point of potential intervention for burn prevention programs (ie, educating foster care givers on burn prevention).

Several SDoH included in the logistic regression model had not previously been studied.⁸ Some of the SDoH that were not previously studied were found to be significant in the current multivariable model (eg, families receiving income assistance). It is recognized that there may be overlap between factors such as children from low income families, children from families who have received income assistance and children from social housing. However, in the final model, children living in social housing were not associated with an increased OR of burn injury, OR 0.79 (95% CI 0.55, 1.14). This outlines the importance of isolating individual risk factors to determine the unique influence that each SDoH has on pediatric burn injury.

Although no prior research has studied the influence of teenage mothers on pediatric burns directly, a mixed picture has been described with maternal age.⁸ Shah et al found that risk of scald injury in children decreases with increasing maternal age, with the OR for scald injury for a mother >40 years old = 0.32.¹⁸ However, mixed evidence exists as Barcelos et al found no increase in incidence of burn injury for children of teen mothers at age 12 or 48 months.¹⁹ Our study strengthens evidence that children born to a teen mother are at an increased risk of burn injury, OR 1.43 (95% CI 1.13, 1.81). Based on this information, our team is proposing to develop education material directed toward this at-risk population.

There is a dearth of literature on the relationship between parental mental health and risk of pediatric burn injury.⁸ Enns et al examined mental and physical outcomes in parents of children with burn injuries 2 years pre- and postindex date compared to matched controls.¹³ High parental rates of mental illness postburn injury were partly accounted for by high rates prior to the injury. Orton et al found a slightly increased risk of pediatric burn injury in children of mothers with a previous diagnosis of depression, OR 1.16 (95% CI 1.02, 1.32).²⁰ However, our study did not find a significant relationship between parental mental health and risk of pediatric burn injury, suggesting that other factors are at play.

Many of the SDoH analyzed in our conditional logistic regression model were not found to significantly influence pediatric burn risk. Despite previous research finding children from newcomers¹⁵ and children from less-educated parents^19,21 at an increased risk of burn injury, our findings did not support the same conclusion. Previous research has largely focused on children in Australia,¹⁵ South America,¹⁹ and the Middle East.²¹ The structure and strength of Canadian social supports may contribute to the lack of identified risk associated with these SDoH in our current study. This supports the concept that burns are preventable.

Burn Prevention

Burn injuries are often associated with multiple surgeries, expensive and frequent dressing changes, high rates of complications, and prolonged hospital stay. The cost of a single hospitalization from burn injury averages $84,798 in Canada.²² A targeted injury prevention budget for burn injuries should be strongly considered. Prevention of a small number of childhood burn hospitalizations per year could justify the cost of the prevention budget itself. Using the results of this study to focus the strategy around the highest-risk children would maximize the efficiency of a prevention program.

An example of this is in a recent report by Mitchell et al who analyzed injury-related hospitalizations in Australian children.²³ Total hospital costs of preventable injuries were $2.1 billion AUD from 2002 to 2012, with burn injuries having the highest mean hospital costs of any injury mechanism. No decrease in incidence of preventable injuries was noticed over the 10 years period, prompting the Australian federal government to implement an injury prevention budget of $900,000 AUD over 3 years to reduce rates of childhood injury. Strategies to reduce injuries include targeted approaches to children of highest risk including Aboriginal, homeless, socioeconomically disadvantaged, and children in remote and rural locations.

Prevention Strategies

Strategies for prevention are either passive or active. Passive strategies do not require a change in the participants’ behavior but are related to changing the environment; active strategies require changes in the participants’ attitude or activity. Passive strategies can be superior in the long-term, especially for participants that may have low compliance, such as those individuals identified at an increased risk.⁵ Consistent with previous literature,²⁴ scald injuries were the most prevalent mechanism of injury representing 42.0% of pediatric burns. Despite this, most Canadian provinces have no regulation with respect to maximal home water heater temperatures. There is a delicate balance with water temperature being hot enough to prevent Legionella spp. growth, and cool enough prevent scald injuries. The Government of Canada suggests installation of automatic or anti-scald mixing valves to circumvent both issues.²⁵ These valves input cool water at the distal end of the water piping to prevent scald injuries at spout points. However, this requires an additional expense on behalf of the purchaser, which can be especially problematic for the high-risk groups identified in this study. Residents living in a rented domicile or multiunit building may not be able to place a valve. Active strategies would include subsidized cost or credits for installing these devices, whereas passive strategies such as legislating installation of valves in all new construction may be more effective for the at-risk populations identified in this study.

One active strategy that has been previously employed is public education on burn injury prevention. Educational programs have been shown to be effective in reducing burn injuries in children.²⁴ Targeting educational programs to highest-risk children can maximize efficiency and usefulness of funds. However, it is well known that patients with multiple SDoH are less likely to be actively involved in their own healthcare, seek appropriate screening, and manage chronic diseases effectively.²⁶ Expecting patients to follow through with active strategies may pose an additional demand to those with an already heavy burden.

Regardless of the strategy used, regular evaluation of prevention programs is critical in assessing progress. Following a change in baseline prevention knowledge and safety practices can be used for more immediate results, rather than solely following incidence of burn injuries in children.⁵

Haddon Matrix

The Haddon Matrix is a previously validated paradigm that can be used to understand the sources of preventable injuries and discover unique intervention points for injury prevention. Application of the Hadden Matrix to burn injury divides the timeline to preevent, event, and postevent interventions. The Haddon Matrix has been largely underutilized in literature²⁷ despite the success of this framework, especially in burn prevention.^28,29 Previous work by Ytterstad et al established a framework to organize strategies for preventative intervention based on Haddon’s matrix.²⁹ Adopting a similar strategy for burn injury, Tables 9 and 10 were created. Table 10 contains unique interventions tailored to the highest-risk burn populations identified in this study.

Table 9.

Haddon’s matrix for burn injury

	Human	Physical Environment	Socioeconomic Environment
Preburn	1	2	3
Burn	4	5	6
Postburn	7	8	9

Table 10.

Interventions for targeted burn prevention

Interventions	Targeted Population	Haddon’s Matrix	Active/Passive
Basic burn prevention and first aid education for thermal injuries as part of “Families First” home visits for well babies	Teen mom; Low Income; Income Assistance	1, 2, 4, 5, 7	Active
Anti-scald mixing valve subsidized cost for low income families	Low Income; Income Assistance	2, 3, 5	Passive
Education for foster parents in burn prevention, first aid and home safety	Child in care	1, 2, 3, 4, 5	Active
Positive Parenting Program for new parents	Teen mom	1, 2, 4, 5	Active
Expanded childcare supports	Teen mom	1, 4, 7	Active
Legislation surrounding sources of flame or fire to be out of reach of children	Low income; Income assistance; Child in Care; Teen mom	1, 2	Passive
Expanded rehabilitation, school re-integration and risk avoidance programs	Low income; Income assistance; Child in Care; Teen mom	7, 8, 9	Active

Limitations

There are a number of limitations to this research project. First, administrative data analysis is limited to the underlying information collected. For example, mental disorders may be based on healthcare billing, which is also dependent on treatment seeking. Second, subgroups of children are more likely to be admitted regardless of burn severity, due to social circumstances. Low-income children may be preferentially admitted due to lack of care supports at home.³⁰ This may over-represent the proportion of low-income children in the burn cohort, inflating the effect on pediatric burn risk. Newcomers may be hesitant to seek medical care for burn injuries due to mistrust of the healthcare system, lack of knowledge, or communication barriers.³¹ This may underrepresent burn injury in children with an immigrant status. Previous literature indicates children living in a rural environment are at an elevated risk of burn injury.⁸ However, rural children transported from remote locations for burn care may be more likely to be admitted.³² The expense of transportation and lack of resources in rural settings for subsequent care may result in the overrepresentation in proportions of the rural burn cohort. The proportion of rural children (57%) outnumbers urban children in the burn cohort (43%), despite the majority of children living in an urban environment. In comparison, the proportion of the provincial province residing in urban Manitoba is 65%.³³ This concern of the subtle undocumented difference between selection criteria for admission was addressed by matching for residence. As a consequence of matching on residence, it was not possible to assess the effect of a rural living environment as a risk factor in this study. An additional limitation results from the inability of physician billing data in Manitoba to capture more than one ICD diagnosis per outpatient visit. A patient arriving with an extensive problem list may have less severe issues go undocumented. A substance abuse disorder or mental health diagnosis may hold less priority than exacerbating medical conditions. Thus, the parent of a burned child may have no record of a mental health diagnosis, leading to an underrepresentation of the proportion of children belonging to these categories. Finally, the definition of being in the family of a new immigrant in this study is imprecise. An algorithm developed by MCHP estimates the probability of a parent being a newcomer. This method of newcomer evaluation has been validated in previous literature²⁶ but remains an estimate.

Strengths

Population-based, administrative datasets allow access to a repository of information dating back several decades. These data permit the researcher access to a wide breadth of unique information that may otherwise be unavailable. Another benefit is the ability to link this valuable information with data registries. Linking administrative health data with the pediatric burn registry is a technique unique to Manitoba. This linkage allows a glimpse into the environment of the burned child prior to injury. Lastly, matching each of the burned children 1:5 with an uninjured child of the same age, sex and geographical location may reduce bias. Selection bias is mitigated, as this information is collected for all Manitoban residents.

CONCLUSION

Social Determinants of Health influence risk of burn injury in children from a low income, children in care, children from families with a history of income assistance, and children from a teen mother. Burn prevention programs targeting these highest-risk children will maximize efficiency of prevention budgets and prevent the life-long sequalae that result from pediatric burn injuries.

Funding The Manitoba Firefighters’ Burn Fund and the Canadian Institutes of Health Research Operating Grant #151519 supported this study.

Conflict of interest statement. No authors have any financial or personal relations to people or organizations. Author Contributions S.L. and J.G. conceived of the presented idea, verified the analytical methods, and developed the protocol. A.P. performed a systematic review under the guidance of J.G. and S.L. to inform the current research. J.G. took lead on development of the methodology, statistical analysis and calculations required, under the supervision and guidance of S.L. and D.C. A.P. designed manuscript tables with input from J.G. and S.L. J.G. took lead on writing of the methodology with input from S.L., D.C., and A.P. A.P. took lead on the writing of the introduction, results, discussion, and conclusion with input of S.L., J.G., and J.S. S.L. was responsible for overseeing progress of the project as lab supervisor.