Exploring the burden of fatal drowning and data characteristics in three high income countries: Australia, Canada and New Zealand

Drowning is a leading and preventable cause of death that has suffered an attention deficit, due in part to a lack of quality data. The development of drowning prevention interventions requires data systems that provide detailed and nuanced information on the circumstances of drowning. The current estimate, acknowledged as an underreport of the true burden of drowning [1‐5], is 372,000 fatal drownings globally per year [6]. Improving drowning data in countries has been identified as a key strategy by the World Health Organization (WHO) to better understand the full extent and circumstances of drowning, to target interventions and evaluate their effectiveness [7].

The use of International Classification of Diseases (ICD) codes to explore drowning is a common approach but provides a limited understanding of causal factors, impacting the development and reporting on the effectiveness of prevention strategies [8]. ICD-10 coding provides limited information on drowning location and provides no information on activity. Drowning deaths due to water transportation incidents, and as a result of flooding are also commonly excluded from global estimates [1].

The circumstances leading to drowning are complex. In high-income countries people interact with water primarily for recreation [9‐13]. By contrast, drowning in low- and middle-income countries often occurs as a result of interactions with water due to daily life or occupational endeavours [14‐16]. Prevention strategies, must therefore differ to suit the environment and the unique causal factors contributing to drowning risk in different contexts.

The comparison of data collation and database management methodologies has been used in other areas of research to work towards improving health [17‐20]. Cross-country comparison of drowning data has also been conducted previously, with capture based only on ICD-10 codes [21]. Such comparisons are valuable to identify strengths, weaknesses and commonalities in data collation and coding, as well as enable sharing of successful interventions, across countries with similar drowning burden and risk factors. However the ability to compare relies on consistency of coding and/or transparency of definitions and frameworks used to collate and code data to allow for mapping. There currently exists no definitive coding framework for drowning data. The ILCOR Advisory Statement – 2015 Revised Utstein-Style Recommended Guidelines for Reporting of Data From Drowning-Related Resuscitation [22] (henceforth referred to as the ILCOR Drowning Data Guideline), through an established consensus process, proposes updated guidelines for reporting data from studies of resuscitation from drowning. These Guidelines can be considered a useful starting point for developing a minimum core dataset for drowning suitable for use in countries with and without death registry, although the Guidelines are used for improving clinical outcomes rather than prevention.

Death investigation is vital to understanding the epidemiology of drowning, due to the rich detail that can be derived from case documentation [23]. Australia, Canada and New Zealand, through the benefit of well-resourced death registry and coronial systems, have three of the best unintentional fatal drowning databases in the world. All are high-income, English speaking countries, with similar culture and population distribution. All have Indigenous populations at increased risk of drowning (Table 1).

This study aims to examine three of the most comprehensive fatal drowning databases in the world to: (1) describe data collection and coding; and (2) compare crude fatal drowning rates, demographics and risk factors.

For the purposes of this study, only deaths where unintentional drowning was indicated as a primary or contributing cause of death were included. Excluded were all other water-related fatalities (i.e. spinal injury, hypothermia) where drowning was not a contributing factor. The process for the collection and collation of data in each of the three countries is discussed below.

Of the 55 variables made available for this study, 19 were consistent across all three databases. Consistently collected variables included date of incident, age group, sex, residential status, category of aquatic location, activity prior to drowning and if the drowning occurred during a MFE, among others (Table 2). When mapped against the ILCOR Drowning Data Guideline [22], 6 variables were consistently collected by all countries, with five being able to be compared within this study. The five compared within the study were date of incident (referred to in ILCOR Data Guideline as incident date and time of day); age (age – birthdate); sex (sex); first nation’s peoples (race and ethnic categories); and medical condition (pre-existing illness). The variable of case ID (victim identifier) was consistently collected but not able to be compared (Table 2).

The authors recommend a minimum core dataset of 11 variables. The recommended variables are: Case ID; date of incident; primary cause of death; age in years (used to code to age group); sex; ethnicity; aquatic location category; activity prior to drowning; victim alcohol involvement; victim drug involvement; and resuscitation enacted – yes/no/unknown (Table 2).

There were 51 categories collected across the three countries for location of drowning (ranging from nine to 27). (Table 3) Activity being undertaken immediately prior to drowning was classified into 53 categories across the three countries. The number of activity categories ranged from six in Canada to 32 in New Zealand (Table 4).

A total of 8176 cases of unintentional fatal drowning occurred in the three countries (Australia 34.1%, Canada 55.9%, New Zealand 9.9%). Crude fatal drowning rates per 100,000 population have been as low as 1.06 in Australia in 2014 and as high as 2.21 in New Zealand in 2008. When comparing the relative change between the first and the last year’s drowning rates for each nation, all three have achieved reductions, with the highest reduction seen in New Zealand (− 24.7%) (Table 5).

The proportion of child drowning ranged from 12.1% in Canada to 17.6% in Australia (X² = 46.4;p < 0.001). There was no difference in the proportion of people aged over 50 years drowning by country (Table 6).

New Zealand (4.9%) and Australia (4.7%) recorded a significant proportion of fatal drownings among international tourists (X² = 28.9;p < 0.001). Twenty-four percent (23.9%) of drowning deaths in New Zealand were Indigenous peoples (X² = 195.6;p < 0.001) (Table 6).

There were differences in locations of drowning in each country. Rivers were the leading location for drowning in Australia (27.1%), lakes/ponds in Canada (36.4%) and ocean/harbour locations in New Zealand (31.3%). Australia was the only country to record a statistically significant drowning burden in swimming pools (X² = 66.3;p < 0.001) (Table 6).

Activity prior to drowning also varied, with aquatic activity common in Australia (28.4%) and New Zealand (43.7%) (X² = 72.8;p < 0.001) and bathing (9.6%; X² = 91.1;p < 0.001) and boating-related incidents (24.0%; X² = 37.1;p < 0.001) common in Canada (Table 6).

Alcohol was a common risk factor with known involvement in 36.0% of Canadian drownings and 25.8% of Australian drowning. Drugs were most commonly involved in Australia (27.0%; X² = 350.5;p < 0.001) and Canadian drowning fatalities (24.4%). Pre-existing medical conditions were present in a similar proportion of cases in Australia (37.2%) (X² = 304.5;p < 0.001) and Canada (37.1%) (Table 7).

Canada recorded the highest proportion of drowning deaths where lifejackets were not worn (67.4%). Sixteen percent (15.7%) of Canada’s drowning fatalities during the study period occurred as a result of a MFE (X² = 67.7;p < 0.001) (Table 7).

Drowning prevention is a “wicked problem” [32] with a range of risk factors requiring targeted prevention strategies [33, 34]. Robust high-quality data underpins evidence for effective policy, interventions and education [34]. Currently there is a move across high-income countries to develop sophisticated drowning data systems [1, 11]. This study compares the drowning data collection of three countries who have similar language, culture, aquatic engagement and drowning prevention. Better data collection systems should enable the creation of targeted and effective drowning prevention interventions. It is likely that, as systems grow, more variables will be collected. However, reliability and validity of these variables will need to be tested, and then likely reduced to a core set of important indicators, such as those suggested in the ILCOR Drowning Data Guideline [22], with local variation enabled to address specific challenges.

There currently exists no definitive guide to drowning data collection and coding variables (including sub-categories). The ILCOR Drowning Data Guideline [22], although focusing on reporting data associated with drowning-related resuscitation, represents a useful first step. Six variables collected by countries within this study directly map to the ILCOR recommendation (five of which were able to be compared), however for those focused on drowning prevention, key variables required for the development of drowning prevention strategies are not discussed. For those countries focused on collecting and collating drowning data with a view to prevention, information on drowning causal factors is vital, including detail on aquatic location and activity being undertaken immediately prior to drowning. As a result, the authors of this study propose a core minimum dataset of 11 variables including age, sex, location, activity, date of incident, primary cause of death and drug and alcohol involvement). Such variables are included based on a philosophy of data collection for drowning prevention and quality and availability of data. As countries become more sophisticated and drowning data collection matures, further variables can be added, however this core dataset should be maintained. Across the study period, 8176 cases of fatal, unintentional drowning were captured, with all three countries achieving reductions in crude drowning rates of at least 10%. Detailed comparison across the three countries was challenging due to differing definitions, coding and data collected. It appears that the risk factors for drowning are similar across the countries with males (80.3%), drugs (23.1%), alcohol (30.6%) and pre-existing medical conditions (35.3%) contributing to the drowning burden. There were however, local variations between the datasets. For example Australia and New Zealand saw significant (p < 0.001) numbers in ocean/harbour and beach locations, whereas in Canada, drownings were more likely (p < 0.001) to occur in lakes/ponds and bathtubs.

For effective cross-country comparative studies, availability of quality data and comparable coding frameworks will be required, including coding hierarchy especially for cases where interpretation and therefore coding, could vary. Aside from basic drowning descriptors such as location and activity, all three countries collected rich information on risk and causal factors [17, 35] including data related to Indigenous peoples and lifejacket use. This suggests that all three countries identify Indigenous peoples as being at increased risk and lifejackets as an effective prevention strategy. With increased understanding of drowning risk factors, there will be an international need for minimum datasets and consistent coding.

Data quality poses a challenge for all undertaking epidemiological research. To help improve drowning data quality, there will be a need for engagement with the medico-legal process, including medical examiner, police and others involved in investigating drowning deaths. At a population level, ensuring the routine testing for, and collection of, data on presence of alcohol, drugs and pre-existing medical conditions will be vital for evidence-based prevention efforts.

The comparison of drowning data between countries allows for identification of similarities in drowning risk and therefore effective prevention, as well as informing potential improvements in data collection and coding.

Calculation of country-specific drowning rates would exclude non-residents. However, there is increasing concern for, and a need to develop an understanding of, those who drown outside their country of residence. There are two broad groupings of people: (1) tourists; and (2) refugees, migrants and people without a fixed address. In Canada visitor status is not routinely collected and had to be derived for this study. Though often small numbers (international tourists account for 4% of drowning deaths in Australia [36]) they are a unique subset requiring different prevention strategies.

With increasing focus on drowning among refugees, migrants and stateless peoples [37‐41], there is a unique challenge to calculate drowning rates worldwide. In Canada, during the study period three refugees drowned during a migration attempt and two illegal immigrants drowned. Drowning among non-residents is a challenge for prevention.

Collating and comparing data on sex and age group is reasonably straightforward whereas activity and location is more challenging. Within the three databases the number of categories within a variable differed, thus fidelity was lost, when merged. Detailed activity and location coding (i.e. greater than currently available in ICD-10 and more like what is proposed for ICD-11) will be critical for future international comparative studies. Alternative coding frameworks, such as that proposed by the International Classification of External Causes of Injury (ICECI) Coordination and Maintenance Group [42], provides greater detail and therefore more detailed evidence to support prevention efforts.

A strength is the drowning data represents all cases of unintentional fatal drowning, beyond the narrow inclusion criteria based only on primary cause of death with an ICD code of W65–74, which has been found to underreport drowning [1, 3, 43]. This however, means that outside of the three countries included in this study, data is not comparable unless the same inclusion criteria is applied. The data is longitudinal, yet retrospective in nature and is drawn from the coronial system, allowing causal and risk factor analysis to be conducted. Data in this study is based on cause of death and not ICD-10 coding, and may therefore differ from official cause of death statistics.

Fatal drowning data from all three countries is coronial data and details for cases which are open (i.e. under investigation) may change. Only those variables where comparison was possible have been included in the study. Assumptions have been made when mapping location and activity variables. Those drowning victims with unknown residential status or those on work or student visas have been retained for the purposes of rate calculations.

Drowning is a global public health threat, impacting both high income and low and middle income countries. A total of 8176 drowning deaths were recorded across the study period, with all countries achieving at least a 10% reduction in country-level crude drowning rates across the 10 year period of the study. Among the data coding methodologies used by the three countries, there were 55 variables available for analysis of which, 19 were consistently collected and 13 were comparable. To compare the variables of location and activity, there was a need to merge categories, thus reducing the fidelity of data available. Future studies involving more countries will require work to enable comparisons to occur. The authors have identified 11 variables that would form a core minimum dataset, however there is a need for the development and validation of category definitions for location and activity variables that are consistently applied to allow for global comparison.