Article Text

Download PDFPDF

Determining child drowning mortality in the Sundarbans, India: applying the community knowledge approach
  1. Medhavi Gupta1,
  2. Soumyadeep Bhaumik2,
  3. Sujoy Roy3,
  4. Ranjan Kanti Panda3,
  5. Margaret Peden4,
  6. Jagnoor Jagnoor2
  1. 1 George Institute for Global Health, Newtown, New South Wales, Australia
  2. 2 Injury Division, The George Institute for Global Health India, New Delhi, India
  3. 3 The Child In Need Institute, Pailan, West Bengal, India
  4. 4 The George Institute for Global Health UK, London, UK
  1. Correspondence to Dr Jagnoor Jagnoor, The George Institute for Global Health India, New Delhi 110025, India; jjagnoor1{at}


Background The Sundarbans in India is a rural, forested region where children are exposed to a high risk of drowning due to its waterlogged geography. Current data collection systems capture few drowning deaths in this region.

Methods A community-based survey was conducted in the Sundarbans to determine the drowning mortality rate for children aged 1 to 4 years and 5 to 9 years. A community knowledge approach was used. Meetings were held with community residents and key informants to identify drowning deaths in the population. Identified deaths were verified by the child’s household through a structured survey, inquiring on the circumstances around the drowning death.

Results The drowning mortality rate for children aged 1 to 4 years was 243.8 per 100 000 children and for 5 to 9 years was 38.8 per 100 000 children. 58.0% of deaths were among children aged 1 to 2 years. No differences in rates between boys and girls were found. Most children drowned in ponds within 50 metres of their homes. Children were usually unaccompanied with their primary caretaker engaged in household work. A minority of children were treated by formal health providers.

Conclusions Drowning is a major cause of death among children in the Sundarbans, particularly those aged 1 to 4 years. Interventions keeping children in safe spaces away from water are urgently required. The results illustrate how routine data collection systems grossly underestimate drowning deaths, emphasising the importance of community-based surveys in capturing these deaths in rural low- and middle-income country contexts. The community knowledge approach provides a low-resource, validated methodology for this purpose.

  • drowning
  • epidemiology
  • surveys
  • mortality
  • child
  • low-middle income country

Data availability statement

Data are available upon reasonable request. Data will be shared upon reasonable request by contacting the Corresponding Author (

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


In 2018, WHO reported 320 000 drowning deaths globally,1 with an estimated 90% of drownings occurring in low- and middle-income countries (LMICs).2 Drowning is emerging as an important global health issue as more accurate data is collected.2–4

In India, drowning is reported to cause 62 000 deaths every year.5 Drowning is estimated to be biggest cause of death by injury in children under 5 years of age.5 Some coastal regions of India such as the Sundarbans in the northern state of West Bengal share many of the risk characteristics that lend to higher drowning rates in LMICs, including poor infrastructure, rurality, presence of unregulated open water, lack of safety awareness and inadequate health systems.6 7

Despite these risks, there is limited data on child drowning in the Sundarbans. The National Crime Records Bureau data captured only 1067 drowning deaths in West Bengal in 2015. There is consensus that official data sources underestimate drownings by up to 50% owing to poor reach of health systems and variations in data capture mechanisms.8 In neighbouring Bangladesh—which has a similar demographic and geographical context to West Bengal—rates of 121 deaths per 100 000 children for 1 to 4 year olds and 22 deaths per 100 000 children for 5 to 9 year olds have been captured through surveys.9–11 To ascertain a more accurate drowning rate in the Sundarbans to quantify the problem, a community-level survey is required.

Given the risks, there may be a requirement to develop and implement an intervention in the Sundarbans to address drowning. The six quality steps in quality intervention development (6SQUiD) framework outlines steps required for the development of evidence-based health interventions. The first step in the framework is defining and understanding the problem and its causes.12 Hence this project aimed to calculate the rate of fatal drownings in the Sundarbans for children aged 1 to 4 and 5 to 9 years, as they are likely to be at the highest risk of drowning.2 This information will assist in highlighting drowning as an issue to governments so that they may take sustainable action. A survey will also provide information on the circumstances around drowning for intervention design.13


Study settings

The Sundarbans region is a deltaic area in the state of West Bengal composed of over 100 islands.11 14 Over 4 million people live in this region of whom approximately 15.9% are aged 1 to 9 years. Many settlements are remote and separated from health services due to water and mangroves. The region also experiences an annual monsoon season from July to October when the presence of open water increases.15 Climate change is further increasing dangerous flooding events.16

A drowning mortality survey was conducted in the 19 blocks of the Sundarbans region situated in two districts of West Bengal (six blocks in North 24 Parganas and 13 in South 24 Parganas). The survey determined the rate of fatal drowning in children aged 1 to 9 years using the community knowledge approach (CKA) as outlined in Paul et al 17 (figure 1).

Figure 1

Community knowledge approach steps.

Conducting household surveys to ascertain mortality can be a resource-intensive process. The CKA is a validated methodology suitable for LMIC rural settings. This approach uses the collective knowledge of the community to identify deaths, involving both lay individuals and key informants such as community leaders and health workers. Community members identify households where deaths have occurred, which are then validated by those households through an extensive survey asking for detail on the circumstances around the death. This ensures two to three points of validation for each death, removing the possibility of overcounting. Iterations of the CKA have been applied in a range of LMICs and have accurately identified mortality rates.18–20 A comparison of the CKA to a household survey in Bangladesh with maternal and jaundice-associated deaths found that the approach captured 100% of maternal deaths and stillbirths and 80% of neonatal deaths.17 Although the CKA has not been used for drowning, the nature of child drowning as a salient event in rural communities means that residents are likely to be aware of these deaths.21

Sample size

Bangladesh experiences a similar context to West Bengal in relation to its rurality, geography and culture.3 22 23 In Bangladesh, drowning accounts for a yearly rate of 121 deaths per 100 000 children aged 1 to 4 years and 22 deaths per 100 000 children aged 5 to 9 years.9 Since there are no similar estimates for West Bengal these rates and a design effect of 2 were used to calculate the sample size for the current study.

Based on a 10% missing data rate, the total minimum sample required was 14 322 for children aged 1 to 4 years and 76 062 for children aged 5 to 9 years. Two hundred and five villages were randomly selected from census data through computer-aided randomisation to cover this population. Latest census data from 2011 was adjusted for rural population growth rates of 3.5% for South 24 Parganas blocks and 3.8% for North 24 Parganas blocks and used to calculate the denominator population. The recall period was 3 years to account for any annual variation while minimising participant recall issues.9

Data collection

Fatal drowning was defined as impairment to breathing caused by submersion in water or other liquid, resulting in the child’s death.24 Deaths were originally identified through lay community members through group interviews, and through key informants such as health workers and community leaders through one-on-one interviews. These deaths were validated by the household in which the event occurred through a survey.

Data collection occurred over a 4-month period from July to October 2019 by a team of 30 trained data collectors. All data collection occurred in Bengali. Teams of two data collectors visited villages and engaged residents to draw a map. The map was used to locate a central area such as a tea shop or community centre, where a group of 8 to 15 lay community members was gathered. After seeking the group’s consent to participate, data collectors compiled a list of deaths of children aged 1 to 9 years from the past 3 years, specifying which were from drowning. Data collectors then visited key informants from the same community and updated the list of child deaths after seeking their consent to participate. For every population of 1000, one community group and four key informants were interviewed. This was a similar ratio to previous applications of CKA in rural LMIC contexts.19 All participants were required to be over the age of 18 and usual residents living in the village for at least the past year. Community member and key informant interviews usually lasted between 5 to 15 min, and no compensation was given for participation.

Once the list of drowning deaths was complete for a village, data collectors visited the household of each identified death. A household member over 18 years old who was living there at the time of the event was surveyed to gather information on the drowning death.

The household survey was conducted using a predesigned tablet-based questionnaire with REDCap software, adapted from a similar survey conducted in Bangladesh and translated into Bengali.9 The responses were guided by the international classifcations of diseases (ICD-10) drowning codes. The survey took 10 to 15 min and asked participants about the circumstances around the drowning death including date, time, location, water source and activities before and after the event.

Data quality was ensured through 5% back checks of surveys halfway through data collection, and frequent random cross checks of data collection activities.


The drowning mortality rate was calculated per 100 000 children aged 1 to 4 years and 5 to 9 years using the SPSS software package.25 The t-tests were performed for differences in mean mortality rates by gender and age. Common circumstantial factors across deaths were identified through descriptive proportions and Pearson χ2 tests to confirm differences between proportions. χ2 cross tabulations were performed to assess any proportional differences between drownings of children aged 1 to 4 years and 5 to 9 years. Factors analysed included location of the water body, supervision at the time of incident, month of drowning, time of drowning and gender.


Participant demographics

The survey covered an estimated population of 867 380 in 205 villages across all 19 blocks of the Sundarbans region. This included an estimated 55 512 children aged 1 to 4 years and 82 401 children aged 5 to 9 years based on 2011 census data age breakdowns. Less than 5% of household survey participants refused to participate, often due to lack of time or distress associated with the event.

A total of 961 lay community-member group interviews (encompassing 11 027 participants) and 3844 key informant interviews were conducted. Most (2718; 70.7%) of key informants interviewed were women. See online supplemental file 1 for a breakdown of key informant type.

Community interviews identified 690 drowning deaths. Of these, 652 were verified by household members. From verified deaths, 502 were involving children aged 1 to 9 years and occurred during the 3-year period from October 2016 to September 2019, and were eligible to be included in the analysis.

Of deaths listed by community members, 63.0% (n=435) were identified by group interview and by at least one key informant, 25.6% (n=177) by key informants only and 11.4% (n=79) by community members only.

Of 652 household survey respondents, 46.5% (n=303) were mothers of the deceased child, 21.5% (n=140) were the father, 18.3% (n=199) were the child’s grandparents and 12.3% (n=80) were other relatives.

Drowning mortality rates

Over a 3-year recall period, 502 cases of drowning were identified from 496 households.

A higher rate of drowning was observed for children aged 1 to 4 years, at 243.8/100 000 per year (95% CI 213.2 to 274.4). The rate of drowning of children aged 5 to 9 years was found to be 38.8/100 000 per year (95% CI 26.6 to 51.0) and children aged 1 to 2 years accounted for 58.0% of all deaths. No significant differences were found between males and females (table 1).

Table 1

Average annual fatal drowning rates by age and sex

Common factors to drowning deaths

Common circumstances around drownings were analysed and reported for all children aged 1 to 9 years. Less than 3% of all variable data was missing from across household survey responses. Overall, few significant differences were found between common factors for children aged 1 to 4 years and 5 to 9 years. Differences found between age groups are specifically discussed below.

Timing of drowning deaths

Across 3 years, drowning deaths peaked during the monsoon season between July and October (231; 46.0%) (figure 2). Most drownings occurred during the day, with 52.7% of deaths (n=265) occurring between 10 a.m. to 2 p.m. (figure 3). Few child deaths (12; 2.2%) occurred during extreme weather events such as flooding.

Figure 2

Month of fatal drownings.

Figure 3

Time of fatal drownings.

Activities at time of death

Children were largely unaccompanied (406; 81.0%). A parent was accompanying the child in only 10.6% (n=53) of cases, followed by another adult (26; 5.2%) or a sibling (23; 4.6%). Household members reported that most children (488, 97.2%) were unable to float on their own.

The usual caretaker of the child was usually engaged in household work at the time of the death (390; 77.7%). Some were also working outside of the home (45; 9.0%) or sleeping (19; 3.8%).

Before the death, children were most often playing or working (including regular household chores and activities, not paid occupational work) near the water source (382; 76.1%) or walking along a water body before slipping (49; 9.8%). Only 6.8% (n=34) of children were playing in the water source at the time.

Location of drowning deaths

Most deaths occurred outdoors (464; 92.4%). Of the outdoor deaths, 95.3% (n=442) occurred in a pond. The water bodies in which children died outdoors were the yard of home (240; 51.9%), beside a road (115; 24.9%) or near the home’s veranda (56; 12.1%).

Almost all deaths occurred within 50 metres of the child’s home (417; 89.9%). A larger proportion of children aged 1 to 4 years died within 20 metres of the home than children aged 5 to 9 years. However, for 20 to 50 metres from the home, the proportion of deaths of children aged 5 to 9 years was higher.

Of the 7.6% (n=38) of indoor deaths, 78.4% (n=29) occurred in household reservoirs for storing water.

Removal of child from water

In 90.6% (n=444) of cases, the child was removed from the water by someone entering the water, as opposed to being assisted from the land (42; 8.6%) or other. In most cases, the child was recovered by a child’s relative or other caregiver (223, 45.1%) or another adult by-stander (157, 31.8%). In 22.6% of cases (n=93), another child such as a friend or sibling was the recoverer.

Health-seeking behaviour after child’s removal from water

Most children (383; 76.6%) were dead at the time of removal from the water source, while 21.8% (n=109) were unconscious. Only 1.2% (n=6) of children were conscious.

The most common immediate response to the drowning was spinning the child above an adults’ head (404; 80.5%) (figure 4).

Figure 4

Immediate responses after drowning.

Most children died at the spot of drowning (400; 79.9%) or at home (59; 11.8%). Only 16.6% (n=83) were taken for professional health treatment, most commonly a local doctor (53; 63.9%) or a registered doctor (28; 33.7%).

In half (42; 50.6%) of the children taken for health consultation (n=83), treatment was administered near the spot of drowning. Treatment was also sought in the child’s home (10; 12.3%), at block-level primary health centres or subcentres (11; 13.5%) or tertiary hospitals (10; 12.3%).

Common forms of transport to the place of health treatment were shared motorised transport such as e-rickshaws (62; 76.5%), followed by ambulance (10; 12.3%). Travel time was usually between 0 to 9 min (30; 37.0%) or 10 to 19 min (32; 39.5%). Most participants did not pay for treatment (65; 82%).

Reporting of deaths

Although 28.3% (n=142) of deaths were reported to the Civil and Vital Registrations Office, only 0.9% (n=4) of total deaths were issued a death certificate.


Our drowning mortality survey reports high drowning rates for children aged 1 to 9 years in the Sundarbans, particularly in the 1 to 4 years old age category at 243.8/100 000 children. Current rural child mortality rates of children aged 1 to 5 years are estimated to be 122/100 000 live births in West Bengal as based on the National Family Health Survey-4 (NFHS-4).10 The NFHS-4 does not present specific data for the Sundarbans, and is a sample-based survey which covers less than 0.1% of households.26 Hence, certain high-risk areas may be missed. In 2018, the National Crime Records Bureau reported a total of 959 deaths by drowning in all of West Bengal, which is a rate of 1/100 000 total population.27 This covers all ages and the entire state of West Bengal, and only includes cases that have been reported to police. These rates are far below our findings for the drowning mortality rates of children aged 1 to 4 years. Our results may guide intervention priorities for reducing under-5 child mortality and attainment of Sustainable Development Goal 3.2 targets.

The high proportion of drowning deaths of children aged 1 to 2 years is higher than other similar contexts.9 28 The location of drownings is primarily outside and 89.9% of outdoor deaths occurring within 50 metres of the home. One hypothesis for higher rates in younger children may be that water bodies are closer to homes in this region.29 Children wandering away even short distances are at risk. As children begin becoming mobile at this age, parents’ risk perception lacks. The area also has poor connection to piped water, meaning people frequently access ponds and expose children to these water bodies. Reduced access to piped water has been associated with higher drowning rates in Bangladesh.30 The proximity of drowning deaths to the home speaks to the requirement for interventions that provide safe spaces for children from water.

Unlike previous findings in LMIC rural contexts,9 28 no differences were found in drowning rates between genders in the 5 to 9 age group. Gender-based factors that make older boys more likely to drown, such as fewer responsibilities at home and riskier behaviour, may not matter if water bodies are closer to homes. While boys may be accessing water to play, girls may be accessing water for household chores.31 32

Most caregivers were engaged with housework at the time of the drowning event. This suggests caregivers are overburdened and unable to engage in both effective supervision and household responsibilities, as has been found in previous qualitative work in the region and other similar contexts.3 22 33

The WHO has recommended the provision of supervised childcare for the effective protection of children. Many deaths occur between 10 a.m. and 2 p.m., so supervision services during this time may be feasibly provided in the Sundarbans. In addition, WHO recommended the installation of home-based barriers to prevent younger children from accessing water bodies, including as pond fencing or door barriers. For more mobile 5 to 9 years old children, swim and rescue training was recommended.2 Rescue training is important to incorporate as almost one-quarter of drowned children were recovered by another child.

The survey also provided evidence of post-drowning malpractices. Many children were subjected to ineffective treatments such as spinning over an adult’s head or stomach compressions, and only a minority were taken for health consultation. Communities were not aware of appropriate rescue and resuscitation responses to drownings and were following old practices they had observed from others such as untrained local ‘quack’ doctors. Many were also unable to or unwilling to access timely healthcare. The provision of community rescue and resuscitation training would capacity build adults to appropriately respond to drowning events.2

Data inconsistencies

One inconsistency in the data was the over-represented deaths in the last three months of 2016, accounting for 18%. There were no reports of natural disasters at this time. One possible explanation is that more Anganwadi centres were opened in the region after 2015.34 Anganwadi centres are government-run centres that provide nutritional services, and to some extent childcare services, to children below 5 years of age. Mothers take their children to the centre for meals and some educational activities, potentially protecting children from drowning during this time due to increased supervision. However, state government reports show a decline in the number of beneficiaries of the ICDS (Integrated Child Development Services) programme between 2016 and 2019.35 Another contributor to lower rates in later years of the survey may be the implementation of the Sundarbans Embankment Reconstruction Project. Through this project, 5000 kilometres of embankments have been built in the last 3 years along coastal islands, protecting many of the Sundarbans islands from flooding and possibly reducing the risk of drowning.36 37

Implications for research methodology

To our knowledge, this was the first application of the CKA to injury mortality. Low reporting to Civil and Vital Registration Systems or hospitals emphasised the need for community-based surveys such as these to identify death in rural LMIC regions.

This approach is plausible only in rural settings with stronger community cohesion. Only 63% of deaths were identified by both group interview and at least one key informant. This demonstrates the need to engage a range of stakeholders at different levels of influence to ensure a comprehensive count of deaths.

Other surveys using the CKA approach should seek to use any local data collection systems. Some community-based government workers in our survey, such as Anganwadi workers and auxiliary nurses in primary health centres, kept record of child deaths in their communities. These key informants were prioritised for interviews.


One limitation of this survey is that some deaths may be missed by community members and key informants. If this were the case, the findings are at worst an underestimate of drowning in this region. In addition, household members interviewed were not always present at the time of the drowning and may have learnt about the circumstances from other members of the household.


The CKA is a valid methodology for estimating injury burden in rural LMIC contexts and can be useful in low-resource settings. The findings of this survey highlight high drowning rates among children aged 1 to 9 years in Sundarbans. To address this high health burden, a multisectoral drowning reduction programme must be introduced. These interventions should seek to create safe spaces for children and capacity build communities to response appropriately to drowning events.

What is already known on the subject

  • Drowning is estimated to cause 62 000 deaths per year in India, and is the largest killer of children under 5 years of age by injury.

  • Rural, coastal areas like the forested Sundarbans region in the north of India have many characteristics that increase the risk of drowning.

  • Routine data sources in India may miss most drowning events occurring in rural and remote areas due to poor reach of health and law enforcement systems.

What this study adds

  • The study found high rates of drowning in children in the Sundarbans region, especially in the 1 to 4 years old age group. This suggests similar work in rural, coastal areas is required to identify the true burden of drowning in India.

  • Common circumstances around drowning events in the Sundarbans region were identified.

  • This study provides guidance to researchers conducting injury mortality surveys in low- and middle-income country contexts on the use of the community knowledge approach, a low-resource methodology.

Data availability statement

Data are available upon reasonable request. Data will be shared upon reasonable request by contacting the Corresponding Author (

Ethics statements

Patient consent for publication

Ethics approval

Ethical approval was granted by the University of New South Wales Human Research Ethics Committee (HC 190274) and The George Institute for Global Health India Ethical Review Committee (06/2019).


We would like to acknowledge and thank the Royal National Lifeboat Institution and the Centre for Injury Prevention and Research, Bangladesh, for their support. We are thankful to the field staff who undertook data collection under challenging conditions.



  • Twitter @MedGupta, @margiepeden, @jjagnoor

  • Contributors MG and JJ formulated the research questions and methodology with inputs from SB. SR and RKP conducted data collection and quality maintenance activities with support from MG, SB and JJ. MG drafted the manuscript with inputs from JJ, SB, MP, SR and RKP.

  • Funding This project was supported by the University of New South Wales through the Research Training Program Scholarship (awarded to MG, no award number) and National Health and Medical Research Council (Australia) Early Career Fellowship funding (Application ID: APP1104745) (awarded to JJ).

  • Competing interests None declared.

  • Patient and public involvement Patients and/or the public were involved in the design, or conduct, or reporting, or dissemination plans of this research. Refer to the Methods section for further details.

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