Article Text

Download PDFPDF

Drowning in Uganda: examining data from administrative sources
  1. Tessa Clemens1,2,
  2. Frederick Oporia3,
  3. Erin M Parker2,
  4. Merissa A Yellman2,
  5. Michael F Ballesteros2,
  6. Olive Kobusingye3
  1. 1 CDC Foundation, Atlanta, Georgia, USA
  2. 2 National Center for Injury Prevention and Control, Centers for Disease Control and Prevention, Atlanta, Georgia, USA
  3. 3 Disease Control and Environmental Health, School of Public Health, Makerere University, Kampala, Uganda
  1. Correspondence to Dr Tessa Clemens, National Center for Injury Prevention and Control, Centers for Disease Control and Prevention, Atlanta, GA, USA; tclemens{at}cdc.gov

Abstract

Background Drowning death rates in the African region are estimated to be the highest in the world. Data collection and surveillance for drowning in African countries are limited. We aimed to establish the availability of drowning data in multiple existing administrative data sources in Uganda and to describe the characteristics of drowning based on available data.

Methods We conducted a retrospective descriptive study in 60 districts in Uganda using existing administrative records on drowning cases from January 2016 to June 2018 in district police offices, marine police detachments, fire/rescue brigade detachments, and the largest mortuary in those districts. Data were systematically deduplicated to determine and quantify unique drowning cases.

Results A total of 1435 fatal and non-fatal drowning cases were recorded; 1009 (70%) in lakeside districts and 426 (30%) in non-lakeside districts. Of 1292 fatal cases, 1041 (81%) were identified in only one source. After deduplication, 1283 (89% of recorded cases; 1160 fatal, 123 non-fatal) unique drowning cases remained. Data completeness varied by source and variable. When demographic characteristics were known, fatal victims were predominantly male (n=876, 85%), and the average age was 24 years. In lakeside districts, 81% of fatal cases with a known activity at the time of drowning involved boating.

Conclusion Drowning cases are recorded in administrative sources in Uganda; however, opportunities to improve data coverage and completeness exist. An improved understanding of circumstances of drowning in both lakeside and non-lakeside districts in Uganda is required to plan drowning prevention strategies.

  • drowning
  • low-middle income country
  • mortality
  • surveillance
  • epidemiology

Data availability statement

Data may be obtained from a third party and are not publicly available. Data may be available upon reasonable request from the CDC Foundation:info@cdcfoundation.org.

Statistics from Altmetric.com

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.

Background

Drowning is the third leading cause of unintentional injury death globally; annually over 300 000 people die from drowning, and approximately 90% of deaths occur in low-income and middle-income countries (LMICs).1–3 Although drowning death rates in the WHO African region are estimated to be the highest in the world (approximately 8/100 000 population),3 data collection and surveillance for drowning in African countries are limited. A 2019 systematic review of the epidemiology of drowning in Africa identified 42 articles from 15 countries, 24 of which were from South Africa. Only 12 of the studies investigated drowning exclusively; the remaining reported on drowning as part of a wider study of all causes of death or injury deaths.4 The authors concluded that there are limited data on drowning in Africa. Developing drowning surveillance databases could help to address this gap, but establishing and maintaining such databases might be challenging for some African countries, where drowning competes with other public health priorities.4 5 Understanding what drowning data are available is an important first step in determining the need for additional data collection efforts and identifying appropriate drowning prevention interventions.6

Common sources of drowning data include hospital records, medicolegal autopsies, household surveys and targeted surveys; however, each of these sources has limitations. Hospital data in African countries often have minimal records of drowning cases7–10 or more commonly capture non-fatal drowning.11 Mortuaries typically investigate deaths due to external causes (eg, drowning), but families sometimes bypass the mortuary and quickly bury drowning victims.3 Generating population-based estimates from household surveys requires substantial resources due to large sample size requirements for capturing a sufficient number of drowning cases. Targeted surveys in small high-risk communities might identify high numbers of drowning cases12 but give limited insight into the circumstances of drowning across a country.

Using existing records might be a more sustainable approach for assessing drowning in low-resource settings. Previous studies have used mortuary records,13–17 death registries18 19 and hospital records11 to describe drowning in African countries. However, most of these studies collected data from a single city/site,11–17 and all used a single data source. Other potential sources of drowning data such as police and search and rescue organisations have not been well explored. Sarrassat and colleagues identified potential sources of drowning data through literature review and in-country networking in Tanzania and contacted authors and institutions to request aggregate data on drowning mortality.5 However, the majority of acquired data came from population-based surveys and hospital-based data from a few sites. The study identified a paucity of drowning death data and limited information on drowning circumstances.

Data collected from multiple sources have the potential to provide critical information about the burden and circumstances of drowning while leveraging existing resources. This study aimed to establish the availability of drowning data in multiple existing administrative data sources and to describe the characteristics of drowning in 60 districts in Uganda.

Methods

Study setting and design

We conducted a retrospective descriptive study using existing administrative records on drowning in district police offices, marine police detachments, fire/rescue brigade detachments and the largest mortuary in 60 districts in Uganda (figure 1). Data were abstracted for all reported cases of drowning—fatal and non-fatal—that occurred during the two-and-a-half-year study period (January 2016–June 2018). The most recent census (2014) defined 112 districts in Uganda; therefore, the 60 study districts represented 54% of Uganda’s districts and approximately 23 million people (two-thirds of the country’s population).20 Districts were selected to include 28 ‘lakeside’ and 32 ‘non-lakeside’ districts as well as districts from each geographical region of Uganda (Central, Eastern, Northern and Western). ‘Lakeside districts’ were defined as bordering one of the five largest lakes in Uganda (Victoria, Albert, Edward, Kyoga and George). ‘Non-lakeside districts’ were defined as not bordering the aforementioned lakes; however, they include other water bodies/sources such as smaller lakes, rivers, ponds, swamps, ditches and pits. The goal was to include different types of environments to capture the breadth of drowning circumstances in Uganda. We did not sample districts based on the completeness/reliability of data in the districts because that information was not known prior to data collection. Instead, within the lakeside/non-lakeside and regional strata, districts were sampled based on logistical, safety and security considerations.

Figure 1

Map of Uganda identifying the 60 districts selected for data collection and the largest lakes.

Data collection

We developed a standardised data abstraction tool using literature on drowning risk factors in LMICs, tools from previous drowning studies, and local knowledge of potential circumstances of drowning. The study team trained research assistants (RAs) to use the data abstraction tool to collect data from station diaries and mortuary registers. RAs also abstracted additional details from case files and postmortem reports whenever available.

A data manager programmed an electronic version of the tool on tablets using Open Data Kit software. Skip patterns and prompts reduced data entry errors. Data collection locations were confirmed using the Global Positioning System. Data quality was further enhanced through regular supervisory field visits, electronic data monitoring and double entry for approximately 25% of cases.

Measures

RAs recorded demographic and incident location information; risk factors (eg, activity at time of drowning, alcohol use); lifejacket use; and rescue-related details. Drowning was defined according to the internationally accepted definition: ‘the process of experiencing respiratory impairment from submersion/immersion in liquid’.”21

Data analysis

Duplicate cases that arose from drownings being reported in more than one source, or in more than one district, were manually matched based on victim name, type of water body, name of water body (when applicable), incident date and incident narrative. Matched cases were reviewed for consistency and completeness across all variables. We created a deduplicated dataset by combining matched cases into single unique cases. Descriptive statistics were reported to summarise the availability of drowning data across the administrative data sources and to describe the number and circumstances of drowning incidents recorded in these sources.

Patient and public involvement

Stakeholders in drowning prevention in Uganda (eg, marine police, lifesaving organisations) and members of the public who are impacted by drowning (eg, fisherfolk, landing site committee members) provided input that shaped the research question and recommended potential sources of drowning data. Stakeholders and the public were also central to dissemination of the study findings.

Results

During the two-and-a-half-year study period (January 2016–June 2018), we found 1435 fatal and non-fatal drowning cases (including duplicates across sources) recorded in the district police offices, marine police detachments, fire/rescue brigade detachments and the largest mortuaries in the 60 study districts. Of these cases, 1009 (70%) were recorded in lakeside districts and 426 (30%) in non-lakeside districts (table 1). The majority of cases were fatalities (n=1292, 90%). A similar number of fatal cases were found in district police (n=405, 31%), mortuary (n=390, 30%) and marine police (n=332, 26%) records, with substantially fewer found in the fire/rescue brigade records (n=165, 13%).

Table 1

Total number of drowning cases* reported in 60 districts by administrative source, January 2016–June 2018, Uganda†

The highest proportion of the 127 non-fatal cases were documented in fire/rescue brigade records (n=63, 50%) and marine police records (n=56, 44%). District police records included few non-fatal cases (n=8, 6%). Almost half (44%) of the non-fatal cases were from multiple victim incidents when other individuals died from drowning. In 16 cases (1%), survival status (fatal or non-fatal) could not be determined.

Of 1292 total fatal cases (including duplicates), 1160 unique fatal cases were identified (figure 2); 1041 (81%) were identified in only one source, while 119 unique individuals were represented in at least two sources (reflected as 251 (19%) cases prior to deduplication). Of the 127 non-fatal cases, all but 4 were unique cases, which were duplicated across different marine police detachments (data not shown). No single data source had substantial coverage of the fatal or non-fatal drowning cases. Coverage of all recorded fatal drowning cases was 35% (404/1160) in district police records, 34% (390/1160) in mortuary records, 28% (323/1160) in marine police records and 14% (165/1160) in fire/rescue brigade records.

Figure 2

Overlap of 1160 unique fatal drowning cases found in district police, marine police, fire/rescue brigade and mortuary records in 60 districts, January 2016–June 2018, Uganda. Ten cases were duplicated within the same source type (one across district police offices in different districts, nine across different marine police detachments); duplication within source types is not reflected. Figure developed using tool available at: http://bioinformatics.psb.ugent.be/webtools/Venn/.

Availability of information on characteristics of individuals who drowned and circumstances of drowning incidents varied by source (table 2). Age (90%) and sex (97%) were almost always available from mortuary records; however, circumstances such as type of water body (44%) and activity at time of drowning (12%) were less commonly available. For fatal cases in the other three sources, type of water body was most frequently available from marine police records (93%), followed by fire/rescue brigade records (90%) and district police records (88%). Activity that took place immediately before the drowning event was not frequently reported in any sources. When reported, it was most often available from marine police records (47%) followed by district police records (32%). An exact or estimated incident date could usually be determined from each of the sources, ranging from 98% available in district police records to 81% in fire/rescue brigade records.

Table 2

Completeness of key variables for drowning cases by source in 60 districts, January 2016–June 2018, Uganda*

There was a high proportion of missing information for variables related to circumstances and/or drowning risk factors for fatal cases across all sources. District of incident was missing for 21% of cases, district of residence was missing for 51% of cases and if the activity was occupational was unknown in 73% of cases. Information about whether the drowning incident had been witnessed (90% missing), whether a rescue had been attempted (93% missing) and drowning risk factors (such as alcohol use, 96% missing; and lifejacket non-use, 96% missing for boating cases) was rarely available.

After deduplication, 1160 unique fatal drowning cases and 123 unique non-fatal drowning cases were identified. A total of 804 (69%) fatal drowning cases were reported in lakeside districts and 356 (31%) were reported in non-lakeside districts. The number of fatal cases per district ranged from 1 to 124 in lakeside districts and 0 to 62 in non-lakeside districts. The most drowning deaths (n=124) were documented in Kalangala, a small island district with a highly mobile population where fishing and water transport are common. Table 3 describes characteristics of individuals who fatally drowned and the circumstances of those incidents. Exact age was recorded in 663 cases (57%). The average age was 24 years. The highest number of recorded drowning deaths occurred among young adults aged 20–24 years (n=99, 15%), followed by children aged 0–4 years (n=97, 15%). In lakeside districts, the highest number of drowning deaths occurred among young adults aged 25–29 years (n=71, 17%) and 20–24 years (n=69, 16%). In non-lakeside districts, children aged 0–4 years had the highest number of drowning deaths of all age groups (n=43, 18%). Sex was recorded in 1031 cases (89%); most cases were male (n=876, 85%). For females, drowning deaths were highest among young children aged 0–4 years (n=24).

Table 3

Characteristics of fatal drowning based on data available from administrative records in 60 districts, January 2016–June 2018, Uganda*

The type of water body where the drowning occurred was known for 888 cases (77%); lakes (n=434, 49%), rivers (n=150, 17%), dams (n=88, 10%) and ditches/pits (n=65, 7%) predominated. In lakeside districts, the majority of drowning deaths occurred in lakes (n=430, 70%), followed by rivers (n=70, 11%) and ditches/pits (n=30, 5%). In non-lakeside districts, only four deaths in smaller lakes were recorded. Drowning in these districts occurred commonly in rivers (n=80, 29%), dams (n=77, 28%) and ditches/pits (n=35, 13%). Activity was recorded for 29% of fatal cases (n=332). In lakeside districts, boating was by far the most common activity (n=200, 81%), followed by swimming/wading (n=24, 10%). In non-lakeside districts, the most common activities were swimming/wading (n=25, 30%) and collecting water/watering cattle (n=18, 21%).

Discussion

No single administrative data source was sufficient for understanding the burden or characteristics of drowning in Uganda. Different cases were available from different sources, and limited overlap of cases between sources suggests that the drowning burden is substantially higher than cases in any one source. The completeness of variable information differed by source. Using multiple data sources can contribute to a fuller initial picture of the drowning situation in Uganda. Previous studies describing drowning in similar settings that used only one administrative data source,13–19 most commonly mortuary data, might have undercounted drowning cases or missed details about circumstances, which were uncommon in Uganda’s mortuary data.

Our study revealed that administrative data sources in Uganda included a substantial number of fatal drowning cases but few cases of non-fatal drowning. This might not be representative of the true burden of non-fatal drowning or the ratio of fatal to non-fatal drowning cases. Local partners indicated that the term ‘drowning’ implies death; therefore, non-fatal drownings might not be recorded. The highest proportion of non-fatal cases was documented by the fire/rescue brigade, followed by the marine police, which both focus on responding to emergencies and rescuing individuals. Data on non-fatal drowning cases recorded in these sources might have been incidental to reporting on the rescue activity. Almost half of the non-fatal cases were related to multiple victim incidents when other individuals died, suggesting that non-fatal drowning cases might be more frequently recorded if they occur in conjunction with a drowning fatality. Studies from high-income countries have identified substantial burden from non-fatal drowning using hospital records,22 23 yet previous studies from Uganda suggest hospital data are not a useful source for identifying non-fatal drowning cases in this setting.7 8 Based on current systems, describing non-fatal drowning in Uganda might only be possible through population-based surveys. Such approaches have been useful for identifying non-fatal drowning cases in other low-resource settings.24

There is currently no mechanism for collating data from different administrative sources in Uganda. When duplication across sources occurred, cases were rarely linked by a standard case number. Duplication likely occurred when a drowning case was referred directly by one source to another. For example, if marine police retrieved the body of a drowning victim and forwarded it to a mortuary, this case might be recorded in both sources. However, based on the perception of administrative data source gatekeepers, we expected this to happen more often than it did. They indicated most drowning cases recorded in mortuaries should be referred by the police, and therefore should be in both records. However, 83% of mortuary cases in this study were not in other data sources. This could be the result of poor recordkeeping or because drowning cases from subdistrict police detachments were never reported up to the district-level police office.

The fact that no single source was sufficient for describing drowning in Uganda, and that there is no current mechanism for collating the data from existing sources, underscores the need for a coordinated drowning (or injury) surveillance system in Uganda. This is consistent with findings from a systematic review of existing drowning data sources in Tanzania5 and two systematic reviews on drowning epidemiology in Africa4 and South Africa,25 which emphasised a lack of strong, ongoing surveillance systems for identifying the true burden and circumstances of drowning. Uganda does not have vital statistics for births and deaths. While the institution of a national or district-level vital statistics system with comprehensive cause of death information would provide much needed data to inform public health practices, this could be many years in the future. Until such a system is in place, efforts to prevent drowning would benefit from drowning specific surveillance, especially in high burden communities. An effective surveillance system is more than access to strong data sources; it should also include an ongoing process to translate these data into action.

Given the extremely high rate of fatal drowning (502/100 000) found in lakeside communities in Uganda in a previous study,12 we expected drowning would be prevalent in lakeside districts in our study. However, drowning in non-lakeside districts was previously unexplored. Nearly one-third of the recorded drowning cases in this study were collected from administrative sources in the 32 non-lakeside districts, even though lakeside districts had an additional data source (marine police detachments). The estimated population of included non-lakeside districts (10 843 366) was 12% smaller than the estimated population of included lakeside districts (12 309 144)20— suggesting that drowning is also a substantial problem in non-lakeside districts. This is consistent with findings from Tanzania, where high fatal drowning rates (8.4 to 14.1/100 000) were found in four inland districts located away from the coast and major lakes.5 The most common age groups, types of water bodies and activities associated with drowning differed in non-lakeside compared with lakeside districts, suggesting that different prevention strategies might be required for non-lakeside areas.

From administrative data sources, we could describe some characteristics of drowning in Uganda. In particular, age, sex and type of water body were fairly complete across all sources. The findings related to these characteristics broadly align with those reported in previous studies, including the high prevalence of males and younger ages5 and the frequency of cases in lakes and related to boating in lakeside districts.12 A strength of our study was the use of multiple existing administrative data sources and coverage of 60 districts (approximately half the country and two-thirds of the population). Our findings demonstrate that existing administrative sources in Uganda and potentially in other low-resource settings might be useful for providing a preliminary understanding of the demographic characteristics of the drowning situation. However, many key pieces of information critical for prevention, such as circumstances of drowning incidents, were seldom recorded.

Our study has several limitations. First, districts were not randomly selected, so results might not be representative of the entire country, and some characteristics might be under-represented or over-represented. Second, due to the degree of missing information in the sources, some duplicate cases might still be present in the deduplicated dataset; however, we believe such instances are few. Third, documents such as station diaries, police files and postmortem reports were handwritten and often damaged. Some variables might have been misclassified due to poor legibility. Fourth, only drowning cases that were reported to district-level administrative data sources were captured; therefore, data might not be representative of all drowning cases. Local stakeholders believed that under-reporting of drowning is prevalent in Uganda, for reasons such as fear of repercussion from authorities and/or fear of postmortem costs. This inhibits the calculation of population-based estimates of drowning. Future work that attempts to triangulate data reported to administrative sources with cases identified by community members (eg, through community surveys or other primary data collection methods) would be useful for describing the full burden of drowning in Uganda.

Conclusion

Drowning cases are recorded in administrative sources in Uganda; however, there are opportunities to improve data coverage and completeness. Unlinked cases reported in multiple sources pose challenges for reporting an accurate number of unique drowning cases, and no one source was sufficient for finding all recorded drowning cases. While this study identified key characteristics of drowning in Uganda related to age, sex and type of water body, planning drowning prevention strategies will require a better understanding of the circumstances of drowning incidents.

Almost one-third of cases were reported in non-lakeside districts, indicating that further investigation and drowning prevention efforts in non-lakeside areas are warranted. Although this study provides valuable information about the burden and circumstances of drowning in Uganda, only drowning cases that were reported to district-level administrative data sources were captured. Therefore, data might not be representative of all drowning cases in the districts, and the findings likely underestimate the true extent of the drowning problem in Uganda.

What is already known on the subject

  • Estimated burden of drowning in Africa is high, but data documenting the burden and characteristics of drowning in Africa are limited.

  • Existing records have been used to describe the burden and circumstances of drowning in smaller areas or regions in some African countries but have seldom been used for describing drowning across an entire country.

  • Previous research suggests high rates of drowning in lakeside fishing communities in Uganda. Drowning in non-lakeside districts has not been explored.

What this study adds

  • Through an extensive review of administrative records in district police offices, marine police detachments, fire/rescue brigade detachments and mortuaries in 60 districts in Uganda, we verified that some fatal drowning cases are recorded in administrative sources, and we identified some key characteristics of drowning related to demographics, drowning location and type of water body; however, limitations on data coverage and completeness existed, and the sources did not contain many cases of non-fatal drowning.

  • No single administrative data source was sufficient for understanding the burden or characteristics of drowning in Uganda. Limited overlap of drowning cases between sources suggests that the drowning burden is substantially higher than the number of cases in any one source. Describing drowning based on only one data source does not capture the true burden of drowning in Uganda and might obscure the burden of drowning in similar settings.

  • Common characteristics of drowning in Uganda include male sex and young adult age. Frequent drowning locations and activities differed in lakeside and non-lakeside districts. Approximately one-third of fatal drowning cases occurred in non-lakeside districts, warranting additional investigation of drowning in non-lakeside districts. Planning drowning prevention strategies will require further understanding of the circumstances of drowning incidents.

Data availability statement

Data may be obtained from a third party and are not publicly available. Data may be available upon reasonable request from the CDC Foundation:info@cdcfoundation.org.

Ethics statements

Patient consent for publication

Ethics approval

Ethical approval for this study was obtained from Makerere University School of Public Health, Higher Degrees Research and Ethics Committee (HDREC), protocol number 621; with clearance from the Uganda National Council of Science and Technology (UNCST), approval SS 4777.

Acknowledgments

We gratefully acknowledge support from the Uganda Police Force and District Health Offices, staff in administrative offices and mortuaries who facilitated our access to data, and all local stakeholders who participated in the initial stakeholder meeting and final dissemination workshops. We appreciate the partnership of the CDC Foundation and Bloomberg Philanthropies. We gratefully acknowledge the hard work of the research assistants from Makerere University School of Public Health who collected data, as well as Mr Albert Ningwa, the data manager. We are indebted to Dr Erin Sauber-Schatz, Dr Robin Lee, Dr Milton Mutto, Dr Anthony Mubiru, Mr George Aluzimbi and Dr Anthony Mugeere for their invaluable insights.

References

Footnotes

  • Contributors TC, FO, EMP, MAY, MFB and OK contributed to the conception and design of the study. FO managed data collection and provided field supervision. TC manually matched cases, deduplicated the dataset and analysed data. TC, FO, EMP, MAY, MFB and OK interpreted the data. TC led the drafting of the manuscript. All authors contributed to manuscript preparation and reviewed and approved the final manuscript.

  • Funding This work was supported by Bloomberg Philanthropies (51606) through the CDC Foundation.

  • Disclaimer The findings and conclusions in this paper are those of the authors and do not necessarily represent the official position of the Centres for Disease Control and Prevention.

  • Map disclaimer The depiction of boundaries on this map does not imply the expression of any opinion whatsoever on the part of BMJ (or any member of its group) concerning the legal status of any country, territory, jurisdiction or area or of its authorities. This map is provided without any warranty of any kind, either express or implied.

  • 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.