The effect of burn mechanism on pediatric mortality in Malawi: A propensity weighted analysis

Abstract

Introduction:

The burden of global trauma disproportionately affects low- and middle-income countries, with a high incidence in children. Thermal injury represents one of the most severe forms of trauma and is associated with remarkable morbidity and mortality. The predictors of burn mortality have been well described (age, % total body surface area burn [TBSA], and presence of inhalation injury). However, the contribution of the burn mechanism as a predictor of burn mortality is not well delineated.

Methods:

This is a retrospective analysis of prospectively collected data, utilizing the Kamuzu Central Hospital (KCH) Burn Surveillance Registry from May 2011 to August 2019. Pediatric patients (≤12 years) with flame and scald burns were included in the study. Basic demographic variables including sex, age, time to presentation, %TBSA, surgical intervention, burn mechanism, and in-hospital mortality outcome was collected. Bivariate analysis comparing demographic, burn characteristics, surgical intervention, and patient outcomes were performed. Standardized estimates were adjusted using inverse-probability of treatment weights (IPTW) to account for confounding. Following weighting, logistic regression modeling was performed to determine the odds of in-hospital mortality based on burn mechanism.

Results:

During the study period, 2364 patients presented to KCH for burns and included in the database with 1794 (75.9%) pediatric patients. Of these, 488 (27.6%) and 1280 (72.4%) were injured by flame and scald burns, respectively. Males were 47.2% (n = 230) and 59.2% (n = 755) of the flame and scald burn cohorts, respectively (p < 0.001.) Patients presenting with flame burns compared to scald burns were older (4. 7 ± 3.1 vs. 2.7 ± 2.3 years, p < 0.001) with greater %TBSA burns (17.8 [IQR 10–28] vs 12 [IQR 7–20], p < 0.001). Surgery was performed for 42.2% (n = 206) and 19.9% (n = 140) of the flame and scald burn cohorts, respectively (p < 0.001.) Flame burns had a 2.6x greater odds of in-hospital mortality compared to scald burns (p < 0.001) after controlling for sex, %TBSA, age, time to presentation, and surgical status.

Conclusion:

In this propensity-weighted analysis, we show that burn mechanism, specifically flame burns, resulted in a nearly 3-fold increase in odds of in-hospital mortality compared to scald burns. Our results emphasize flame and scald burns have major differences in the inflammatory response, metabolic profile over time, and outcomes. We may further utilize these differences to develop specialized treatments for each burn mechanism to potentially prevent metabolic dysfunction and improve clinical outcomes.

1. Introduction

Burn due to thermal injury is among the most devastatinginjuries globally and ranks fifth in terms of trauma mortality [1]. There is a disproportionate burden of burn injuries in low- and middle-income countries (LMIC). Over 90% of global burn deaths occur in LMIC, which generally lack the resources to decrease burn incidence and mortality [2–5]. Burn injury and the resulting unique metabolic and immunological dysfunction makes treatment expensive and challenging, particularly in resource-poor regions. Sub-Saharan African countries are no exception, as the per capita health care spending is low when compared to high-income countries (HICs).

Collectively, Africa bears 15% of overall global burn mortality, second only to Southeast Asia, and carry the highest incidence of pediatric hospital admission for burns [3,6,7]. Epidemiological data in sub-Saharan Africa indicate that burns are primarily a disease of children, claiming the lives of 16,000 children aged 5 years or younger in 2000 [3–5,8–10]. Burns are also well-documented as a disease of poverty, both at the country and individual levels [11–13]. Specifically, low socioeconomic status, lack of child supervision, and floor-level open fire cooking practices have been identified as risk factors for burns [3,13,14]. As such, the vulnerability of children in LMIC to burn morbidity and mortality cannot be understated.

The major predictors of burn mortality are advanced age, a greater percentage of total body surface area (%TBSA) affected, and smoke inhalation, although the incidence of smoke inhalation in LMIC is low compared to HICs [14–18]. It has been suggested that the burn injury mechanism is associated with outcomes [19]. Despite the distinct epidemiology of pediatric burns in LMIC, burn mechanism has not been considered as a predictor of mortality in well-utilized burn mortality prediction models such as the Baux or Revised Baux Score [20,21]. A study from South Africa shows in-hospital mortality for flame versus other burns was 27.8 and 2.8%, respectively [22]. The primary objective of this study was to determine the effect of burn mechanism on in-hospital mortality in a pediatric population in central Malawi.

2. Methods

We performed a retrospective analysis of the prospectively collected data utilizing the Kamuzu Central Hospital (KCH) Burn Surveillance Registry from January 2009 through July 2017. KCH is a 900-bed hospital and tertiary referral center for central Malawi, with a catchment area of about six million people.

2.1. Study setting

Malawi is a country in southern-eastern Africa with a population of 18 million people with a rural population of 84%, a life expectancy of 63.8 years, and a Human Development Index rank of 170 out of 187 countries [23]. It is the fourth poorest country in sub-Saharan Africa. The health care system is tiered with Primary Health Centers providing basic health care and closest to the majority of the population and staffed by clinical officers and no physicians. The next level of health care is the District General Hospital, which is minimally staffed by physicians and has the basic surgical capacity; one is located in every district in Malawi. Lastly, there are the Central Hospitals, which serve as the referral centers for specialty care. There are four central hospitals in Malawi and KCH serves the central region of Malawi. Burn units are located at two of the four central hospitals.

The KCH burn unit contains 31 beds and houses both pediatric and adult patients. During the study period, care was provided by a consultant plastic or general surgeon, two specialized burn clinical officers, and burn-trained nursing staff. All patients presenting to KCH with a burn injury are captured within the registry. Pediatric patients (≤12 years) who were injured by flame or scald burns were included in the study. All other burn mechanisms, including electrical or chemical, were excluded from the study.

The study population was analyzed using descriptive statistics in the overall sample and stratified by burn mechanism (flame and scald). Univariate analysis was utilized to determine data distribution and missing values. There were less than 5% missing data after the application of the inclusion and exclusion criteria. Bivariate analysis was performed over burn mechanism. To report measures of central tendency for univariate and bivariate analysis, we utilized means (±standard deviation) or medians (interquartile range), if the covariates were not normally distributed. To compare the distribution of exposure across demographic variables, χ² for categorical variables, Student’s T-Test for normally distributed continuous variables, and Kruskal-Wallis for not normally distributed continuous variables were used.

The propensity scores were calculated based on patient and injury factors determined to influence the in-hospital mortality outcome, including sex, age mechanism of burn, time to presentation, TBSA, surgery intervention, and in-hospital antibiotic use. The primary objective was to balance the two injury mechanism cohorts to reduce the bias of patient factors and injury patterns to better determine in-hospital mortality based on mechanism. The inverse probabilities of treatment weight (IPTW) were generated from the inverse of the propensity scores to balance the groups based on the burn mechanism. To check balance, a logistic regression analysis was performed with mechanism of injury as the dependent variable and previously stated variables as independent variables. Comparisons of covariates based on burn mechanism among groups found no significant differences as all p-values were greater than 0.05, confirming balance, Table 2.

Table 2 -.

Logistic regression predicting mechanism of burn prior to and after propensity score weighted matching.

	Unweighted logistic regression			weighted logistic regression
	Odds ratio	95% CI	p-value	Odds ratio	95% CI	p-value
Age	1.26	1.20–1.33	<0.001	1.00	0.95–1.05	1.0
Sex	1.50	1.16–1.95	0.002	0.97	0.72–1.29	0.8
Percent TBSA	1.04	1.03–1.05	<0.001	0.99	0.98–1.01	0.4
Time to Presentation
0–6 h	Ref	***	***	Ref	***	***
7–12 h	0.98	0.59–1.64	1.0	1.15	0.63–2.11	0.6
12–24h	1.34	0.84–2.13	0.2	0.97	0.56–1.66	0.9
24–48 h	1.60	0.75–3.39	0.2	1.66	0.70–3.95	0.3
>48 h	2.62	1.64–4.17	<0.001	1.17	0.69–1.99	0.6
Antibiotics Prescribed	1.20	0.92–1.57	0.2	1.01	0.75–1.37	0.9
Surgery	5.26	3.88–7.13	<0.001	0.97	0.70–1.33	0.8

In-hospital mortality is defined in this study as death occurring during the hospital stay. Preadmission and post-discharge mortality outcomes are not captured in our dataset. To determine the influence of the burn mechanism on in-hospital mortality, we performed an unweighted and propensity score-weighted logistic regression. A priori, the variables included in the model were sex, age, TBSA, and if a patient had surgical intervention. Other variables significant after bivariate analysis at p < 0.05 over burn mechanism were also included in the multivariate model. The fully adjusted model included sex, age, %TBSA, surgical intervention, in-hospital antibiotics, and time to presentation. A backward elimination approach was performed to reduce errors with the removal of variables based on p > 0.05. The use of in-hospital antibiotics was removed from the final model based on these criteria. Precision was maintained as the confidence interval narrowed. Less than 10% change was seen in the odds ratios, so a reduction of bias was obtained.

This analysis was performed using StataCorp v14.2, College Station, Texas. Confidence intervals are reported at 95% and alpha was set at 0.05 for this study. This study was approved by Malawi’s National Health Science Research Committee and the University of North Carolina Institutional Review Boards.

3. Results

From May 2011 to August 2019, 2364 patients were captured in the KCH Burn Surveillance Registry. Pediatric patients (≤12 years) composed 1794 (75.8%) of the entire burn cohort. Of these, 1768 (98.6%) had a flame or scald burn mechanism. Flame and scald burns accounted for 27.6% (n = 488), and 72.4% (n = 1280), respectively. Overall, the cohort had an average age of 3.2 ± 2.7 years and was predominately male (n = 985, 55.9%). The median TBSA was 14% (IRQ 8–21). Patients predominately presented to the hospital 12–24 h after injury (n = 622, 35.8%). There was a 16.2% (n = 280) in-hospital mortality in the entire pediatric burn cohort. Table 1.

Table 1 -.

Unadjusted demographics by mechanism of burn.

	Overall n = 1768	Scald Burns n = 1280	Flame Burns n = 488	p-value
Age: μ ±SD	3.2 ±2.7	2.7 ±2.3	4.7 ±3.1	<0.001
Male Sex: n (%)	985 (55.9)	755 (59.2)	230 (47.2)	<0.001
BMI: median (IQR)	16.8 (15.4–19.2)	16.9 (15.4–19.2)	16.5 (15.6–19.2)	0.7
Percent TBSA: median (IQR)	14 (8–21)	12 (7–20)	17.8 (10–28)	<0.001
Traditional Medicine Used: n (%)	184 (10.9)	138 (11.3)	46 (9.9)	0.4
Time to Presentation: n (%)
0–6 h	208 (12.0)	166 (13.2)	42 (8.8)	0.01
7–12 h	387 (22.3)	315 (25.1)	72 (15.0)	<0.001
12–24h	622 (35.8)	480 (38.2)	142 (29.6)	0.001
24–48 h	67 (3.9)	43 (3.4)	24 (5.0)	0.1
>48h	453 (26.1)	253 (20.1)	200 (41.7)	<0.001
Antibiotics Prescribed: n (%)	930 (53.3)	634 (50.4)	296 (60.9)	<0.001
Surgery: n (%)	346 (19.5)	140 (10.9)	206 (42.2)	<0.001
Hospital Length of Stay: median (IQR)	12 (6–25)	10 (6–18)	23 (9–44)	<0.001
Time from Injury to Disposition: median (IQR)	13 (7–26)	11 (7–19)	25 (11–47)	<0.001
Time from Injury to Death: median (IQR)	9 (4–17)	8 (4–14)	11 (4–21)	0.07
Disposition: n (%) Died	280 (16.2)	161 (12.8)	119 (25.2)	<0.001
Discharge	1332 (76.9)	1025 (81.4)	307 (64.9)
Abscond	121 (7.9)	74 (5.9)	47 (9.9)

After stratifying by mechanism of burn, patients presenting with flame compared to scald burns were older (4. 7 ± 3.1 vs. 2.7 ± 2.3 years, p < 0.001) with greater TBSA burns (17.8 [IQR 10–28] vs. 12 [IQR 7–20], p < 0.001). Males accounted for 47.2% (n = 230) and 59.2% (n = 755) of the flame and scald burn cohorts, respectively (p < 0.001.) Operative intervention was performed for 42.2% (n = 206) and 19.9% (n = 140) of the flame and scald burn cohorts, respectively (p < 0.001.) Patients with flame burns and scald burns were more likely to present after 48 h (n = 200, 41.7%) and between 12–24 h (n = 480, 38.2%) after injury, respectively (p < 0.001.) In-hospital mortality was 25.3% (n = 119) and 12.8% (n = 161) after flame and scald burns, p < 0.001, respectively, Table 1.

In the propensity score weighted logistic regression predicting in-hospital mortality, injury following flame burn had increased odds of in-hospital mortality compared to scald burns (OR 2.53, 95% CI 1.66–3.84, p < 0.001). Also, increasing TBSA increased odds of in-hospital mortality (OR 1.10, 95% CI 1.08–1.12, p < 0.001). Undergoing surgical intervention (OR 0.37, 95% CI 0.21 – 0.66, p = 0.001) and older age (OR 0.87, 95% CI 0.79 – 0.96, p = 0.004) had decreased odds of in-hospital mortality (Table 3).

Table 3 -.

Unadjusted and propensity score weighted logistic regression predicting mortality.

	Unadjusted logistic regression			Adjusted logistic regression
	Odds Ratio	95% CI	p-value	Odds Ratio	95% CI	p-value
Flame Burn	2.59	1.77–3.78	<0.001	2.53	1.66–3.84	<0.001
Age	0.88	0.82–0.95	<0.001	0.87	0.79–0.96	0.004
Female Sex	1.12	0.82–1.53	0.5	1.52	0.99–2.34	0.06
Percent TBSA	1.10	1.08–1.11	<0.001	1.10	1.08–1.12	<0.001
Time to Presentation
0–6 h	Ref	***	***	Ref	***	***
7–12 h	0.83	0.48–1.44	0.5	0.65	0.29–1.45	0.3
12–24h	1.03	0.63–1.69	0.9	1.19	0.58–2.42	0.6
24–48h	1.30	0.54–3.13	0.6	1.79	0.49–6.52	0.4
>48 h	0.86	0.50–1.45	0.6	0.71	0.34–1.48	0.4
Surgery	0.34	0.21–0.55	<0.001	0.37	0.21–0.66	0.001

4. Discussion

Burns is a global public health problem, accounting for an estimated 180,000 deaths annually. [24] In this study, we show burn mechanism is a predictor of burn mortality. In a cohort of or flame burn mechanisms, patients with a flame burn had almost three times increased odds of mortality in a propensity-weighted analysis after controlling for pertinent covariates.

Worldwide, children are at risk for burn injuries with a global rate of 3.9 deaths per 100,000 population due to a mixture of curiosity and awkwardness. [25] In children, the level of motor development does not match the cognitive and intellectual development and results in an inability to assess risk [26]. Scald burns are the most common cause of burn injuries in children [27]. Similar to other studies in sub-Saharan Africa, there is a preponderance of scald burns as compared to flame burns in this study with an incidence of 72%. Studies from Nigeria and South East Asia, show comparable scald burn incidences of 64.4–85% and 53–55%, respectively [28–30].

Most studies suggest that burns in children occur most frequently in the home, and in particular in the kitchen. It has been suggested that the location within the home of the heating equipment and the structure of the kitchen may present significant risks to children. [31]Unfortunately in many parts of sub-Saharan Africa, the open fires used for cooking, typically at ground level within easy reach of children, and the use of paraffin lamps for heating contribute to increased burn incidence [32–34]. In most countries in sub-Saharan Africa, burn injury over 40% of total body surface area is usually fatal [9].

The mechanism by which flame burns impart higher mortality is currently incompletely understood, but the association between flame burns and inhalation injury may be a factor. Several studies show a significant association between flame burns and inhalation injury. Studies from Nigeria by Olawoye et al. and others [35,36] show a high association between inhalation injury and increased risk of mortality. At our center in Malawi, pediatric patients with burn inhalation injury rarely present to our burn center. They may have succumbed to the inhalation injury before presentation or inhalation injury is indeed uncommon in this rural population, as relatively few fires take place in enclosed, multistory buildings where victims are exposed to prolonged smoke inhalation [9]. Kraft et al. investigated differences in clinical outcomes between scald and flame burn in children and they show that patients with flame burns had a higher likelihood of increased burn depth and increased mortality [19]. Furthermore, they show that flame burns led to significantly increased hypermetabolic and inflammatory responses as well as an increased risk of burn sepsis and multisystem organ failure when compared to scald burns [19]. There are few studies in the region that separate scald versus flame burn mortality in the pediatric population. A study by Lari et al. [37] from Kuwait showed a 1.4% versus 8.8% mortality for scald and flame burns, respectively. Iregbulam et al., in Enugu, Nigeria shows 9.8% mortality in a cohort of predominantly scald burns [38], which is similar to our findings. Mortality for flame burn in our resource poor setting is comparable to flame burn mortality in South Africa [22]).

There are limitations to this study inherent to any study with a retrospective methodology. Firstly, the burn registry is subject to selection bias as we only have information in those who presented to our burn unit and lack information on those that did not survive prior to presentation for either burn mechanism. This may have led to an underestimation of the lethality of flame burns. Secondly, the registry did not contain information regarding burn depth during the study period, and therefore we could not control for this in the propensity-weighted analysis. Lastly, given the rarity of inhalation injury in our environment, our findings may not be generalizable to other settings where inhalation injury is more common.

5. Conclusion

Flame burns confer higher mortality when compared to scalds in a pediatric burn cohort in a resource-poor setting in sub-Saharan Africa. The underlying mechanism may be related to increased burn depth and the resulting inflammatory, metabolic, and infectious sequelae. The recognition of this finding may help guide burn treatment to mitigate metabolic and infectious complications by striving for aggressive resuscitation, early excision and would coverage, and optimization of nutrition to improve clinical outcomes.