Burn Admissions Across Low- and Middle-income Countries: A Repeated Cross-sectional Survey

Abstract Burn injuries have decreased markedly in high-income countries while the incidence of burns remains high in Low- and Middle-Income Countries (LMICs) where more than 90% of burns are thought to occur. However, the cause of burns in LMIC is poorly documented. The aim was to document the causes of severe burns and the changes over time. A cross-sectional survey was completed for 2014 and 2019 in eight burn centers across Africa, Asia, and Latin America: Cairo, Nairobi, Ibadan, Johannesburg, Dhaka, Kathmandu, Sao Paulo, and Guadalajara. The information summarised included demographics of burn patients, location, cause, and outcomes of burns. In total, 15,344 patients were admitted across all centers, 37% of burns were women and 36% of burns were children. Burns occurred mostly in household settings (43–79%). In Dhaka and Kathmandu, occupational burns were also common (32 and 43%, respectively). Hot liquid and flame burns were most common while electric burns were also common in Dhaka and Sao Paulo. The type of flame burns varies by center and year, in Dhaka, 77% resulted from solid fuel in 2014 while 74% of burns resulted from Liquefied Petroleum Gas in 2019. In Nairobi, a large proportion (32%) of burns were intentional self-harm or assault. The average length of stay in hospitals decreased from 2014 to 2019. The percentage of deaths ranged from 5% to 24%. Our data provide important information on the causes of severe burns which can provide guidance in how to approach the development of burn injury prevention programs in LMIC.

A burn is an injury to the skin or other tissue primarily caused by exposure to heat and hot substances. Burns are an important cause of death and survivors may be left with lifelong disabilities. 1 The World Health Organisation (WHO) estimates that 11 million burn injuries occur each year, of which 180,000 lead to death. 2 In high-income countries, there has been a large reduction in burn injuries. Burn injuries remain a massive issue in Low-and Middle-Income Countries (LMICs) and more than 90% of burns are thought to occur in these countries. 2 The majority of burns are small and do not require hospital admissions. Information on burn epidemiology is inconsistent, with high-income countries having more available data compared to LMICs. Obtaining epidemiological data on burns is not easy. From a data collection perspective, burns can be divided into large burns requiring hospital specialising in burn injury and smaller burns that do not lead to hospital admission. Collecting data on the latter is difficult and must be approached either by collecting records from a huge array of health facilities or by population surveys.
Large burns cause the greatest mortality and loss of amenity. Such burns can be ascertained through the records of centers in hospitals catering for severe burn injuries. This is not an ideal epidemiological method, since the dominator population for a particular burns center is hard to define. Nevertheless, studies of hospital admissions enable the trends in the incidence of burns to be observed over time (assuming no radical change in referral patterns) and can track changes in causes of burn injuries over time and between countries.
This paper aims to summarise information on burn admissions from eight burn centers in LMICs in 2014 and 2019. This information includes the total number of burn admissions, demographics of burn patients, setting of burn occurrence, type and cause of burns, and outcomes of burn admissions.

Study Design and Setting
This study design was a repeated cross-sectional survey. The observation periods cover the calendar years 2014 and 2019. In 2021, the local investigator at each participating burn center was asked to complete a survey based on the WHO global burn registry form from routinely collected data (see Supplementary Appendix 1). 3 A sample of the centers was contacted regarding the availability of data in their burn databases before the data extraction forms were sent. Nevertheless, the information available in each center varied from center to center as some centers did not collect certain data points included in the survey.
Eight centers from different LMICs contributed data: Cairo (Egypt); Nairobi (Kenya); Ibadan (Nigeria); Johannesburg (South Africa); Dhaka (Bangladesh); Kathmandu (Nepal); Sao Paulo (Brazil), and Guadalajara (Mexico). The centers were chosen as they cover Africa, Asia, and Latin America, they experience high burn rates and we have contacts in each of the centers. Details on the geopolitical characteristics of the countries and information on the burn centers are given in Table 1.

Africa
Egypt has a population of around 100 million people, just more than half the population live in rural areas. Our participating center is the Kasr Al-Ainy burn unit in Cairo University hospital. It is one of the oldest and best-established burn center and serves Cairo and other governates. It has 21 beds with an intensive care facility and an Intensive Care Unit (ICU). 4 There is another burn unit in Cairo in the Ain Shams University. Beyond this, we have little information on burn facilities in Cairo.
Kenya has a population of about 50 million people, more than 70% of people live in rural areas and about 40% are under 15-years-old. In Kenya, the main center is in the Kenyatta National Hospital in Nairobi, the leading referral hospital in the country with 100 beds for burn patients. 5 Nigeria has a population of about 200 million people, around 50% live in rural areas and more than 40% are less than 15-yearsold. Our participating center is in the University College Hospital in Ibadan which opened in 2011 and has a 12 bed ward. 6 South Africa has a population of about 59 million, around 35% of people live in rural areas. Burns are a major problem due to the proportion of people living in poorly constructed, combustible accommodation in urban areas. 7 In South Africa, the centers are an adult and pediatric burn units in Baragwanath Hospital in Johannesburg which is the largest hospital in Africa. 8 The adult burn unit consists of 24 ward beds and six ICU beds 9 while the pediatric burn unit has 24 ward beds and 8 ICU beds. 10 Asia Bangladesh has a population of about 160 million, more than 60% live in rural areas. In Bangladesh, our participating center is the Sheikh Hasina National Institute of Burn and Plastic Surgery in Dhaka which is the largest burn and plastic surgical care center in the world with 500 beds, 50 ICU beds, and 12 operation theatres. 11 This center is the only tertiary burn care center in Bangladesh and therefore, receives a wide range of patients from across the country. 12 This new hospital replaced the previous burns unit but continued the same routine data collection system. Nepal has a population of about 29 million and the majority of people live in rural areas. Nepal is a mountainous country with transportation and road access challengers. There are few facilities that provide organised burn care in Nepal. The Nepal Cleft and Burn Center, out participating center in this study is the only dedicated center in the country. 13 The center is based in Kirtipur Hospital in Kathmandu which has 30 beds (7 ICU beds) and two dedicated burn operating rooms which run six days a week. 14 The center received patients from all districts in Nepal and is nearly entirely dependent on donations. 15 Latin America Brazil has a population of more than 200 million people, only 20% are under 15-years-old and about 14% live in rural areas. In Brazil, the participating center is in the Federal University of Sao Paulo or Sao Paulo hospital, one of six burns care units in public hospitals serving the municipality of Sao Paulo which has a population of almost 20 million 16 Mexico has a population of about 120 million people, of which about 20% live in rural areas. In Mexico, the participating center is in the Guadalajara Civil Hospital. Most burn patients are hospitalised from the emergency department to the 14 bed ICU as it is the only facility for treating critically ill burn patients. 17

Data Collection
Local investigators were asked to extract data and summarise burn admissions for the calendar years 2014 and 2019. The month that burn admissions occurred and the number of admissions with a TBSA of more than 20% was recorded. Age groups were categorised as adults and children using the closest available cutoff point to 12 years of age. The setting in which burns occurred was categorised as household, occupational, other, or unknown and the location where the burn occurred was ategorized as urban (including peri-urban) or rural. The types of burns were ategorized as flame, hot surface, hot liquid, direct electric, chemical, other, or not known with further detail given on flame burns given when available. The cause of burns was categorised as unintentional, intentional self-harm, assault, or unknown. The outcomes of burn admissions were average length of hospital stay measured in days and the number of deaths that occurred.

Statistical Methods
Demographics of burn admission patients, details of burns admissions such as setting and location, type of burns, and outcomes of burn admissions were summarised using counts and percentages for categorical data and means or medians for numeric data. All analysis was performed in Stata v16.1. No comparisons were conducted since this is a descriptive study, there is no welldefined denominator population and even incidence rates ratios could be affected by changes in denominator populations. The percentage of patients with a burn of more than 20% TBSA ranged from 19% to 81% across the centers. In Johannesburg, only 19% and 21% of burn admissions had a TBSA of more than 20% in 2014 and 2019, respectively, while in Sao Paulo 73% and 81% of burn admissions had a TBSA of more than 20% in 2014 and 2019, respectively. In most centers, the TBSA of more than 20% stayed similar between 2014 and 2019, apart from in Cairo where it dropped from 58% to 31%, and Ibadan where it dropped from 66% to 43%.

Demographics of Burn Admissions Patients
For all centers with the sex of burn admissions patients reported, the percentage of men admitted was around 20% higher than women's admissions (about 60% men and 40% women) ( Table  3). In all centers apart from Nairobi, the percentage of male burn  (Table 4). In all centers with available data on the setting of burn admissions, household burns were most common (36% to 79%). In Dhaka and Kathmandu, occupational burns were also common, 32% and 37%, respectively. In Dhaka, burns also took place in settings other than household and occupational settings (24% in 2014 and 27% in 2019).
The type of burn flames varied largely depending on the center and year of burn admissions (Table 4). In Nairobi, the type of flame burn was mainly Kerosene in 2014 (72%), reducing in 2019 (52%), while the proportion due to LPG increased from 10% in 2014 to 35% in 2019. In Dhaka, the majority of flame burns changed from solid fuel (77%) in 2014 to LPG (74%) in 2019. In Cairo, the majority of flame burns were LPG regardless of the year (80% in 2014, 79% in 2019) followed by Kerosene (15% in 2014, 7% in 2019). In Ibadan, the majority of flame burns were solid fuel (62% in 2014 and 60% in 2019). The next most common type of flame burn was Kerosene in 2014 and LPG in 2019. The type of fuel associated with flame burns is either not recorded or mostly unknown in the other counties. Nevertheless, the rise in the proportion of flame burns due to LPG is a consistent finding among countries where data on fuel was recorded aligning in the switch from either solid fuel or kerosene to LPG seen across LMIC.

Intentional or Nonintentional Burns
In all centers, the majority of burn admissions were due to unintentional burns (63-97%) ( Table 6). In Nairobi, a substantial proportion of burns were caused by assault (22% in 2014, 20% in 2019) and intentional self-harm (10% in 2014, 12% in 2019). Differences between 2014 and 2019 in the cause of burns were limited due to missing information (in Guadalajara, Johannesburg, and Ibadan).

Outcome of Burn Admissions
The outcomes of burn admissions by center and year were reported ( Table 7). The mean length of stay in hospitals reduced in all centers from 2014 to 2019. The range of mean values varied from 11 to 33 days (median from 6 days to 26 days). In Nairobi and Dhaka, the average length of hospital stay is higher compared to the other centers. The percentage of deaths across the centers ranged from 5% to 42%. In Guadalajara, Johannesburg, and Sao Paulo, less than 10% of burn admission patients died. In Cairo and Ibadan, the percentage of deaths from burns admissions had dropped from 2014 to 2019 (7 and 20 percentage points, respectively).

DISCUSSION
As stated in the introduction, our data do not have denominator data which imposes a limitation on the inferences that we can draw. In addition, referral patterns likely changed during the course of the study so incident rate ratios may change as a result of a different spectrum of patients being admitted. For example, the proportion of children changed over time in some studies and children carry different risks of suffering burns of different causes. Thus, our data needs to be interpreted carefully and it is for this reason that we have not conducted statistical hypothesis tests. To add to the limitations of our study, not all centers collected the full WHO recommended list of items, for example recording the type of fuel associated with flame burns. What then can we infer from the data? First, the data provides confirmation that burns remain a very serious problem in LMIC. The very observation that the Dhaka center treats more than 5000 severe burns in a year, of whom about 1000 die, is evidence that burn injuries are a very serious issue in LMIC.
Second, some changes are stark and are, arguably, unlikely to have resulted purely from selection effects. Thus, we find evidence in the data that a switch to LPG, from either kerosene or biomass, is associated with two outcomes. First, the total number of flame injuries decreases. Second, the proportion of     (21) 14 (1) 417 (47) 58 (6) 7 (1) 44 (5) 165 (19) 18 (10) 133 ( 67 (8) 17 (2) 7 (1) 105 (12) 28 (29) 12 ( flame injuries that are due to LPG increases. These findings suggest that the switch from one fuel to another may introduce new hazards even if the overall risk decreases. Ideally, the switch from one fuel to another should be accompanied by measures, both education and environmental modification, to mitigate the new risk. While biomass and kerosene fuels are inherently risky to use, LPG if used properly should essentially be risk-free for burns. If cases are occurring, then the fuel is being improperly used and prevention efforts are urgently needed.
Third, the spectrum of burn types across the different centers is very large and again would appear to exceed what might be attributed to selection effects alone. As we have seen, the proportion of burns associated with different fuels types is variable. The two African centers, Nairobi and Johannesburg, register the highest proportions of injury due to "hot liquids"; in essence hot/boiling water. The seasonal effect on burn injuries is most marked in Kathmandu and is likely caused by use of electrical heaters during the short but sharp winter. Perhaps the most provocative finding with respect to geographic variation is the very high proportion of intentional burn injuries in Nairobi.
Our data show there is no 'one size fits all' solution when it comes to intervention design. Just as different cities experience a different spectrum of pollutants, so to do cities have different predominant causes of burns. While the intervention must be tailored to the main cause of local burns, the cause specific solutions developed in one city should inform those in another. What is needed is evaluation of interventions that might reduce burn incidence so that cities can learn from each other. For example, local authorities in Nairobi can benefit from a study showing that the risk of immolation can be reduced by a public education campaign. 18 Perhaps the largest implication of our findings, relate to the importance of better data collection regarding burns, especially severe burns, which have a high mortality and lifelong consequences. We hope that our findings can prompt renewed efforts to establish comprehensive burn registries. The WHO global burn registry form would enable burns units to collect data in a standardised way. It is the result of several years of collaboration lead by the WHO. It has been piloted to ensure that it is useable across a wide variety of settings and collects data on the epidemiology of burns along with clinical data. Widespread use of the WHO form would have overcome problems associated with the lack of standardisation that we encountered in this study. This in turn would provide a much more solid basis for initiatives to reduce burn injury. We therefore urge individual centers to adopt this tool that has been widely endorsed and that could be implemented at little to no marginal cost. We advocate for burn societies, public health agencies, and organisation of the United Nations, such as WHO, to vigorously promote use of the form. A registry or study based on the global burn registry would be limited to severe burns. We acknowledge that certain smaller burns, which do not prompt admission to a burn's unit nevertheless have severe consequences. The number of facilities to which smaller burns may be reported is vast and record keeping in these disparate units is of variable quality. However, a number of studies have been conducted in which surveys have been successfully used to elicit population-based incidence data. [19][20][21] Surveys are expensive and including burns data within wider registries may be a cost-effective way to harvest epidemiological data on burns in a regular basis. Table 6 In conclusion, our data provide important information on the causes of severe burns in LMIC and on how they vary by location. The findings show that burns remain a pressing issue even if we cannot provide a good numerical answer to the question of burn incidence and the data provide strong guidance in how to approach the development of burn injury prevention programs in LMIC.

SUPPLEMENTARY DATA
Supplementary data is available at Journal of Burn Care & Research online.