Arch Dis Child 96:675-680 doi:10.1136/adc.2010.202010
  • Global child health

Childhood drowning and traditional rescue measures: case study from Matlab, Bangladesh

  1. D Bishai4
  1. 1Health Systems Program, Department of International Health, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland, USA
  2. 2Public Health Sciences Division, The International Centre for Diarrhoeal Disease Research, Dhaka, Bangladesh
  3. 3International Injury Research Unit, Department of International Health, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland, USA
  4. 4Department of Population, Family and Reproductive Health, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland, USA
  1. Correspondence to Dr Nagesh N Borse, CDC—Center for Global Health, 1600 Clifton Road, MS E-41, Atlanta, GA 30329, USA; nborse{at}
  • Accepted 19 January 2011
  • Published Online First 11 March 2011


Recent mortality data indicate that approximately half a million people drown each year worldwide, with more than 97% of such deaths occurring in low-income and middle-income countries. The purpose of this study was to examine verbal autopsy data on the circumstances of childhood drowning in Matlab, Bangladesh. The study analysed 10 years (1996–2005) of data which reported 489 deaths in children under 5 years and recorded preimmersion, immersion and postimmersion events. The data summarised household characteristics, age, gender and time of drowning event. The study also examined traditional rescue methods performed on children who were removed from the water OR found drowning. Of 489 deaths, 57% were aged 1–2 years and had a drowning mortality rate of 521 per 100 000 children. Most drowning events occurred during the morning (68%), in ponds (69%), and while the mother was busy doing household chores (70%). Traditional rescue methods were attempted in 55% of children and the most frequently reported measure was to spin the child over head (35%). Only 3% of families tried to perform resuscitation. Verbal autopsy data for Matlab is a useful resource for childhood injury research in a low-income country. The study is one of the first to publish data on traditional rescue practices performed on drowning children in rural Bangladesh. The findings suggest that interventions should be designed using locally identified risk factors to reduce childhood drowning incidents. Community-based resuscitation techniques and emergency medical systems are needed to improve postimmersion recovery of the child.


Childhood drowning is a neglected topic in low-income and middle-income countries, particularly in Bangladesh.1,,8 Drowning has gained little attention from policy-makers, or even public health professionals, due to lack of research, understanding and funds. The drowning rate in low-income and middle-income countries, including Bangladesh, is six times higher than in high-income countries (with rates of 7.8 and 1.2 per 100 000, respectively).1

There have been impressive declines in child mortality in Bangladesh, largely due to decreases in infectious disease.8,,10 This is also true in Matlab, a surveillance site located about 55 km south-east of the capital Dhaka. Child mortality for all causes of death in Matlab decreased from 115 per 1000 live births in 1984 to 13 per 1000 live births in 2005. However, mortality due to drowning was relatively constant from 1983 to 2005. Therefore, the proportional mortality rate due to drowning in children aged 1–4 years increased from 8% in 1984 to 54% in 2005.7,,11 As a consequence, drowning accounted for more than half of deaths in children aged 1–4 years in Matlab.6 7 10,,13

What is already known on this topic

  • Drowning is the most common cause of death in children aged 1–4 years in Matlab, Bangladesh.

  • Causal explanations for childhood drowning are primarily associated with ‘evil spirits’.

  • Communities performed traditional rescue practices that have no known benefit.

What this study adds

  • Verbal autopsy data from Matlab, Bangladesh is a vital resource for childhood drowning research; it is a model injury surveillance system for low-income countries.

  • Children aged 1–2 years are at higher risk for drowning; postdrowning medical care is more likely to be sought for boys than girls.

  • Traditional rescue methods are commonly performed (82%) (eg, spinning the child to remove water from the stomach, oil massage); the positive or negative effects of these techniques have not been measured.

In high-income countries, common interventions for drowning prevention include proper pool fencing, supervision, lifeguards and water safety training at a young age.14 15 However, differences in geographical, environmental, parental, social, cultural and behavioural factors associated with drowning in high-income countries compared with low-income and middle-income countries often make these interventions unsuitable for low-income and middle-income countries and such interventions are generally not found in these countries.5 6 16 The systematic collection of high-quality mortality data is a prerequisite when designing relevant drowning prevention programs.17 For this reason, there is an urgent need to identify relevant data sources to highlight the importance of investment in developing childhood drowning prevention strategies in low-income and middle-income countries.

The lack of reliable data on the levels and causes of mortality in low-income and middle-income countries continues to limit efforts to develop evidence-based healthcare interventions. Verbal autopsy data are the primary source of information for causes of death in populations lacking vital registration and medical certification.18 The International Centre for Diarrhoeal Disease Research, Bangladesh (ICDDR,B), operating in Matlab since 1966, gave the authors the opportunity to study childhood drowning because it maintains a comprehensive Health and Demographic Surveillance System (HDSS) in a rural area in Bangladesh.19 Under this surveillance system, community health workers use the death registration form to record the deceased person's name, gender, current surveillance identification number, village code, date of birth, date of death, a detailed description of events in the local Bangla language, the type of treatment and healthcare provider consulted (figure 1).

Figure 1

An example of childhood drowning verbal autopsy data form from Matlab Health and Demographic Surveillance System (HDSS), Bangladesh.

The primary purpose of this study was to examine the circumstances of childhood drowning using preimmersion, immersion and postimmersion events and various demographic and socioeconomic variables. The secondary purpose was to demonstrate the utility of verbal autopsy data for child injury research in low-income countries.


Data construction and setting

The Matlab HDSS covers a population of 225 000, 12% of whom are under 5 years old. Every resident of Matlab is assigned a unique identification number in the HDSS that connects them to a village, a bari (compound) and a household. Twice a month, each bari is visited by a trained ICDDR,B community health worker and information on demographic events such as births, deaths and migration is recorded.20 Data were extracted from the HDSS death certificate forms (figure 1) using three variables: the International Classification of Diseases code (ICD-9 code ‘E830.0’ and ICD-10 code ‘W-70’: Drowning and submersion following fall into natural water) for death by drowning21; the age at the time of death; and the year of death between January 1996 and December 2005. The data were sorted by year, village and personal identification number and translated from Bangla to English. More than 45 death registration record books were searched to identify verbal autopsies for all childhood drowning deaths during the study period. Later, a two-staged process was used for data extraction and double-data entry. A medical officer fluent in both Bangla and English checked the accuracy of the translation of 5% of randomly selected forms. Later, two datasets were checked for inconsistencies using Microsoft Visual Basic commands.

Preimmersion variables included the mother's and child's activity prior to the child's death. The child's location, time of death, type of water body and whether a person accompanied the child were also noted. Postimmersion variables included the time taken to find the body, any rescue measures taken and the type of healthcare professionals consulted. If there was no written description of a specific variable on the death form the variable was recorded as missing. Finally, the HDSS Registration of Health and Demographic Event Dataset, and the HDSS Socio-Economic Status Census Dataset (2005 and 1996) were merged with death data to obtain information about the socioeconomic status of the child's household.13 22 23

A descriptive analysis was conducted using variables such as household characteristics, age, gender and time of event. The mid-year population for the year 2000 was used to calculate rates and odds ratios (ORs).12 The study also recorded local traditional rescue methods performed on the drowning children according to age, gender and socioeconomic status. The data analysis was carried out in SAS version 9.1.

This study received ethical approval from the institutional review board of Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland and from ICDDR,B's Office of Research Administration located in Dhaka, Bangladesh.


HDSS data for 10 years (1996–2005) reported 489 drowning deaths in children under 5 years of age. Detailed verbal autopsy information was recorded for 78% (385) of these deaths (figure 2).

Figure 2

Flow chart of Matlab childhood drowning verbal autopsy data creation using Matlab Health and Demographic surveillance (HDSS), Bangladesh.

Descriptive epidemiology

Of all drowning deaths, 55% were boys and 57% were in children aged 1–2 years (table 1). The mortality rate due to drowning was 521 per 100 000 in 1-year-old children. The annual drowning mortality rate was nearly two times higher in Muslim children (202 per 100 000 children) than in children of other religions (114 per 100 000 children). Fifty-six per cent of deaths occurred in the monsoon season. The mother and father of the child were illiterate in 48% and 50% of cases, respectively, which is higher than the proportion of men and women in the population of Matlab aged between 25 and 49 years who are illiterate (29% and 37%, respectively).23 Seventy per cent of the children were from relatively poor families (ie, families with the lowest percentile asset index score) (table 1). This is in comparison to 48% of the Matlab population who are in the lowest socioeconomic status dataset.24

Table 1

Descriptive epidemiology of drowning deaths in children in Matlab, Bangladesh (1996–2005)*

Information on the child's activity prior to drowning was given for 252 deaths (52%), with 26% of children playing in or near water. Information on the mother's activity prior to the child drowning could be analysed for 197 cases (40%), with mothers busy doing household chores in 70% of cases (table 2). Mothers were responsible for supervising the child in 42% of cases; 17% of drowning deaths happened when no one was supervising the child. Older siblings and other children were responsible for supervising the child in 15% of drowning deaths.

Table 2

Comparison of study results with previously published studies on preimmersion, immersion and postimmersion events in Bangladesh

Information on the type of water body was given for 287 of the deaths, with 80% occurring in ponds or ditches near the household (table 2). Sixty-eight per cent of drowning deaths were reported in the morning (06:01 to 12:00) and in 84% of cases the mother was reported to be in the house or nearby at the time of the event (table 2). The mother was also the most likely person to find the drowning child (45%).

Rescue measures

Of 489 deaths, no information was given on rescue attempts for 31%. Of all postimmersion rescue measures reported, only 3% of children were given some form of resuscitation by family members or neighbours, 55% only received traditional rescue measures and 41% were found dead or were declared dead by family members, neighbours or members of the community without any medical consultation. Figure 3 shows the sequence of rescue events for 200 drowning deaths. Local rescue measures were attempted in 55% (108 children), 62% received at least one traditional rescue technique, 28% received two, and 11% received three or more (figure 3).

Figure 3

Post-immersion child rescue measures: analysis of verbal autopsy data on childhood drowning deaths in Matlab, Bangladesh.

Of all traditional rescue measures combined, the most frequent rescue technique was to spin the child to remove water from their stomach (35%), followed by treatment from the village practitioner (26%) and applying pressure to the child's stomach to remove water (22%) (data not shown). Other rescue measures documented included massaging with ash and putting the child in direct sunlight.

Data on the type of care provided and socioeconomic associations were also analysed (data not shown). For boys receiving medical care, the OR was 2.1 (95% CI 1.19 to 3.69, p=0.01) compared with girls, while girls were more likely to be treated using traditional methods. Children with literate mothers had a higher chance of receiving medical care (OR 1.66; CI 0.94 to 2.93, p=0.08) than children with illiterate mothers, however this difference was not significant. A comparison of the literacy levels of the fathers showed no significant differences.

The results also showed a potential gender preference among mothers, regardless of the mother's literacy level. Literate and illiterate mothers were more likely to seek medical care for a drowned boy compared with a drowned girl (literate mother OR 2.08, 95% CI 0.99 to 4.39, p=0.05; illiterate mother OR 2.10, 95% CI 0.86 to 5.10, p=0.1). Among poor households, the OR for a drowned boy to be taken for medical treatment compared with a drowned girl was 2.48 (95% CI 1.38 to 4.45, p=0.01). No significant results were found when the type of treatment was studied in relation to the child's age group (p=0.78), the number of children in the household (p=0.84), the father's literacy level (p=0.74) and the household poverty level (p=0.76).


Most countries need better injury data on mortality and morbidity to develop effective injury-prevention strategies.25,,27 The Matlab HDSS is an established vital resource for childhood disease research, and as demonstrated in this paper, is a good source of verbal autopsy data in a low-income country. The Matlab surveillance site in rural Bangladesh provides an excellent data source for understanding circumstances around childhood mortality, particularly death by drowning. Unlike death certificates, which have limited information about the drowning event, the verbal autopsy data provide in-depth information about a child's death. This study also shows the value of integrating injury data and risk factor surveillance into a well functioning HDSS setting.28

This study documented risk factors associated with drowning, that is, children aged 1–2 years, the monsoon season, morning time while the mother is busy with chores, and the presence of a pond or ditch around the household. These risk factors were found to be comparable with previous studies by Ahmed et al7 and from other areas of Bangladesh (table 2). Also similar to other studies in China and Vietnam, this study shows that the child's gender, the child's age, the mother's literacy level, seasonality and the household poverty level have a significant impact on the risk of childhood drowning.7 9 11 29,,34

For 42% of child deaths, the mother was reported to be responsible for child supervision, and 70% of these mothers reported being busy during mornings and afternoons with household chores, such as bathing, cleaning and washing clothes. These necessary daily activities cause distractions from direct child supervision leading to an increased risk of the child drowning. In developed countries, childhood drowning events often occur during recreation time when parents are momentarily distracted by eating, reading or talking on the phone.35 36 Developing locally relevant interventions to help with child supervision during these busy times is important. Educating the rural community on childcare and supervision is critical, emphasising that not only the mother but all adults in the family should help with childcare.

This study shows that a child's gender has an influence on the type of care sought after a drowning event. There is a clear social preference towards boys because boys tend to have greater school attendance (proportion of boys admitted to school vs. girls), a higher likelihood of vaccination, slightly longer breast-feeding periods and are more likely to receive oral rehydration therapy.37,,39 These preferences are noticeable irrespective of the mother's literacy level. In fact, when a mother's education was changed from no schooling to 6 or more years of schooling better health was reported in 70% of boys, while for girls it was only 32%.40 However, since most drowning events are fatal and boys are allowed to move freely outside the household it actually puts them at higher risk compared with young girls.41 In addition, girls are culturally protected against drowning because they are not allowed to bath or swim in public after a certain age.38 These gender differences are important and need to be taken into account when developing drowning intervention strategies.

The mortality rate for drowning was twice as high for Muslim children compared with children of other religions. Although this study did not explore the specific reasons for this, it might be because of what Ahmed et al7 referred to as ‘dwelling space’ being a risk factor for childhood drowning. Muslim families tend to have more dwelling space, which increases the risk of childhood drowning because households with more dwelling space tend to have bigger man-made ponds or ditches.42 Also having a pond near the house is a status symbol in rural areas of Bangladesh.

This study was one of the first to document care-seeking behaviours after a drowning event in a rural community in Bangladesh. Data from Matlab show a specific care-seeking pattern in which the community tries to rescue the drowning child. Either the child is immediately taken to the hospital or the child is treated with traditional rescue techniques, such as spinning the child to remove water from the stomach, massaging or applying pressure to the stomach. As a last measure, children treated using the traditional methods were also taken to traditional healers. In 29% of cases, the child was reported to be dead when found, however this was declared without any medical consultation. It is important to explore when and how members of the community determine a child is dead, especially without any medical consultation. More research is needed to review such practices.

The Matlab surveillance data do not specifically study or separate out direct flood-related, boating or water transport related deaths by drowning; these events might increase the rates presented here. This study does not provide any information on the disability burden due to drowning, which cannot be quantified for Matlab. Despite the good surveillance system in Matlab not all preimmersion and postimmersion data could be obtained and so there is some missing information. Furthermore, like all verbal autopsy data systems, there are limitations, including possible reporting and recall bias because information is collected by community health workers who ask mothers and others present in the household questions during their monthly visits. Data such as maternal age have not been analysed. Moreover, the data included ‘unknown’ variables for some drowning deaths and these have been added to reflect the quality of the data and the lack of information in such cases (table 2).


The tri-modal distribution of trauma fatalities for childhood injury shows that drowning causes the largest number of prehospital deaths compared with other injuries.43 Immediate rescue and treatment are therefore crucial in minimising disability and improving survival. Previous studies concluded that all drowning children should receive vigorous and aggressive treatment.44 45 However, this is not feasible in rural communities of Bangladesh where access to trauma care and cost of care are major concerns. Therefore, primary methods for prevention of drowning, such as adult supervision at all times and parental supervision aids, could be important interventions to improve child survival.

Countries like Bangladesh have an urgent need for childhood drowning prevention strategies that are appropriate, cost-efficient and effective.6 43 46 47 For rural communities in Bangladesh, this paper recommends further studies should examine the effectiveness of interventions that can be used to prevent drowning in children under 5 years. Interventions should be designed using locally identified risk factors to reduce childhood drowning events. In addition, community-based resuscitation techniques and emergency medical systems are needed to improve postimmersion recovery of the child.


The authors would like to thank the Matlab community for providing this valuable data. The authors would also like to thank the ICDDR,B staff – Ms Nazma Begum, Mr Sajib Rahman, Mr S.A.K.M. Mansur and many others who provided the necessary support during data collection and analysis.


  • At the time of the work Dr Borse was a doctoral candidate at Johns Hopkins and an International Fellow at the ICDDR,B.

  • Funding This research was funded by the United States Agency for International Development (USAID). This project was also partly supported by USAID Family Health and Child Survival Cooperative Agreement through Global Research Activity to Johns Hopkins University.

  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.


Free sample

This recent issue is free to all users to allow everyone the opportunity to see the full scope and typical content of ADC.
View free sample issue >>

Don't forget to sign up for content alerts so you keep up to date with all the articles as they are published.