Introduction Globally, rivers are a common drowning location. In Australia, rivers are the leading location for fatal drowning. Limited information exists on exposure and impact on river drowning risk.
Methods Australian unintentional fatal river drowning data (sourced from coronial records) and nationally representative survey data on river visitation were used to estimate river drowning risk based on exposure for adults (18 years and older). Differences in river drowning rates per 100 000 (population and exposed population) were examined by sex, age group, activity prior to drowning, alcohol presence and watercraft usage.
Results Between 1 January 2014 and 31 December 2016, 151 people drowned in Australian rivers; 86% male and 40% aged 18–34 years. Of survey respondents, 73% had visited a river within the last 12 months. After adjusting for exposure: males were 7.6 times more likely to drown at rivers; female drowning rate increased by 50% (0.06–0.09 per 100 000); males aged 75+ years and females aged 55–74 years were at highest risk of river drowning; and swimming and recreating pose a high risk to both males and females. After adjusting for exposure, males were more likely to drown with alcohol present (RR=8.5; 95% CI 2.6 to 27.4) and in a watercraft-related incident (RR=25.5; 95% CI 3.5 to 186.9).
Conclusions Calculating exposure for river drowning is challenging due to diverse usage, time spent and number of visits. While males were more likely to drown, the differences between males and females narrow after adjusting for exposure. This is an important factor to consider when designing and implementing drowning prevention strategies to effectively target those at risk.
- descriptive epidemiology
- risk factor research
Statistics from Altmetric.com
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.
Accurately calculating exposure is an important, yet challenging aspect of injury prevention.1 Using exposure to estimate risk would allow for the development and monitoring of more effective prevention strategies.2 Determining suitable and accurate methods for calculating exposure is difficult.3 In road traffic research,4 a range of exposure methods have been used including distance travelled, type of vehicle, number of vehicles on the road, number of trips taken and time spent travelling.5 6 Such methods can broaden our understanding of different means of calculating and considering exposure. These challenges have, in part, been addressed (for drowning and other injuries) via the use of ad hoc survey data7–9 and direct observation.10
The WHO Global Report on Drowning estimates drowning claims 372 000 lives annually11; however, the methodology used is likely to underestimate the actual number of deaths.12 While the epidemiology and risk factors for drowning are increasingly being described, exposure to risk and how that varies based on location is an area where research is lacking.2 13 Previous studies have attempted to calculate exposure and drowning risk at beaches,8 10 14 for boaters,15 rivers in King County (USA)16 and as a result of various activities including fishing, rock fishing and swimming.1 Methods used include qualitative telephone surveys,1 14 16–18 observations,10 16 aerial surveys,16 remote camera observations16 and one-on-one interviews.15 16 19
Surveys have been used to explore alcohol consumption patterns around water,9 the role of different activities in non-fatal drownings of young adults19 and to explore visitation at different categories of aquatic location for residents of rural and remote areas in the Australian state of New South Wales.20 A recently proposed alternative method of calculating exposure has been to compare drowning rates per 100 000 population with rates per 100 000 km2 of water.21
Despite such studies at Australian beaches, there have been no specific exposure studies into drowning at Australian rivers, regardless of the fact that rivers claim over one-quarter of all lives lost to drowning, more than any other location.13 22 While the risk profile for unintentional fatal drowning in rivers is becoming increasingly understood,23 24 there is a need for exposure studies13 to better inform much-needed strategies for prevention. This study aims to examine how river drowning death rates vary when adjusted for exposure.
Data collection and analysis
Drowning deaths in rivers
Data on unintentional fatal drowning in Australian rivers, creeks and streams (henceforth referred to as rivers) among people aged 18 years and over between 1 January 2014 and 31 December 2016 were sourced through privileged access to the Australian National Coronial Information System (NCIS). The process for identifying cases of unintentional fatal river drowning in Australia has been previously outlined.22–24 Drowning deaths as a result of flooding were excluded as were not related to recreational activities undertaken in, on or beside rivers for the purposes of calculating exposure-based rates. Drowning deaths of international tourists (ie, no residential address in Australia) and children (aged 0–17 years) were removed as they were not surveyed.
Drowning data on sex, age group, activity prior to drowning, alcohol involvement and watercraft usage were used to calculate drowning rates. Alcohol involvement was coded as a ‘yes’ if the victim’s blood alcohol content (BAC) was ≥0.05%. As BAC readings of river drowning victims may be impacted by decomposition,25 this marker (≥0.05% the upper legal limit for operating a motor vehicle in all Australian states and territories and powered watercraft in most states and territories) was used to account for this. Where there was no BAC available (no body found, toxicology test not undertaken or toxicology report not attached to case file on the NCIS), this was deemed unknown.
River exposure data
Data on people’s exposure to rivers were sourced through a national computer-assisted telephone instrument (CATI) survey administered by Central Queensland (CQ) University.26 Based on an omnibus model, this survey uses cost-sharing participation, allowing researchers to embed a set of questions based on their specific research interests into a survey which collects data from a sizeable national sample. Researchers receive responses to their questions, as well as the core dataset of demographic and health variables.
The survey represents the geographical dispersion of the population of Australia, based on state and territory of respondent’s residence. The final sample contained a reasonably equal number of males and females (54.3% female). The sampling error for the total sample (n=1318) at a 95% CI was 0.027. The survey was conducted between 6 July 2015 and 14 August 2015.
The river usage questionnaire was developed specifically for this study by authors AEP and RCF. Respondents were asked if they had visited a river, creek or stream location within the last 12 months. If yes, the respondent continued on to the remaining questions about frequency of visitation, time spent at rivers, main activities being undertaken, alcohol consumption and watercraft usage. The question pertaining to watercraft usage categorises watercraft as any of the following: powered boat, canoe, jet ski, kayak, paddleboard, water skiing, tubing, houseboat, etc. The questionnaire was pilot tested by trained interviewers on a total of 40 randomly selected households prior to broader data collection. Interviewer comments (such as confusing wording, inadequate response categories and question order effect) and pretest frequency distributions were reviewed before modifications were made. There were no modifications required for the river usage question set. Face validity was not confirmed.
Deidentified responses were provided to the authors in an SPSS database. The main activities undertaken at rivers were provided via an open response question. These were initially coded into 11 categories or recorded as other/text (n=214) by CATI interviewers. This information was then recoded (by consensus by authors AEP and RCF) into five categories to match those used to describe drowning deaths. These activities prior to fatal drowning were non-aquatic activities (ie, those activities which did not involve interaction with the water such as walking beside the water, having a picnic, etc); fishing; swimming and recreating; watercraft and other. There remained 38 (17.8%) responses from the initial ‘other’ category that could not be recoded to the four groups. The activity prior to drowning categories and corresponding CATI survey activity categories can be seen in table 1.
This study used descriptive, retrospective, population-based analysis. River drowning rates per 100 000 adult (aged 18 years and older) population were calculated using the total number of drowning deaths between 2014 and 2016 as the numerator and the total population during 2014–2016 as the denominator. This 3-year period was chosen as being the year the CATI survey was conducted (2015) as well as a year either side of this to account for yearly variations in fatalities.
For river drowning rates based on exposure, the numerator remained the same and the proportional basis of the population (as determined by the CATI survey results) were used as the denominator. These rates were also stratified by sex, sex and age group, activity undertaken and activity by sex, alcohol presence by sex and age group and watercraft usage by sex and males by age group only. Population data were sourced using the Australian Bureau of Statistics population data cubes for the month of June for 2014, 2015 and 2016.27
Data coding and analysis was conducted in SPSS V.20.28 Univariate and Χ2 analysis (with 95% CI) were used. A modified Bonferroni test suggested by Keppel29 has been applied (p<0.01). Non-parametric Χ2 analysis was also conducted using the proportional basis of the population as the assumed outcome numbers. RR was calculated using the MedCalc statistical software.30 Variable analysis which produced less than four cases will be concealed using ’not presented' in adherence with ethical requirements.
River drowning deaths
There were 195 drowning deaths in rivers in Australia during the study period. Of these, 16 (8.2%) were children (0–17 years of age) and 28 (14.4%) were international tourists. Once removed, 151 drowning deaths of adults aged 18 years and older with an Australian residential postcode remained (86.4% male). Male-to-female ratio remains constant by age group, activity and blood alcohol presence. Twenty-nine per cent of all river drowning deaths involved alcohol (table 2). In a further 23.2% of cases (n=35), alcohol involvement was unknown.
Of the 1318 survey responses, 73.3% of respondents had visited a river within the last 12 months. Males (74.7%) and females (72.2%) visited a river in similar proportions, with people aged 35–54 years reporting the highest visitation numbers. Activities that were non-aquatic in nature were the most commonly undertaken activities (71.4%). A higher proportion of females participated in non-aquatic activities (χ2=19.0; p<0.001), whereas males were more likely to fish (χ2= 11.5; p=0.001) (table 3).
Forty-five per cent of respondents visited a river once in the last 12 months, or once every 3 months. When visiting a river, 51.3% spent between 1 and 5 hours at the river. Nine per cent (9.5%) of visitors stayed 24 hours or more (overnight). Those who reported visiting the river everyday most commonly spent <1 hour (χ2= 83.1; p<0.001). Those aged 75+ years were more likely to stay at a river for <1 hour (χ2= 21.5; p=0.001). Approximately one-quarter (25.6%) of those attending rivers used watercraft.
Males were significantly more likely to consume alcohol at rivers (χ2= 9.2; p=0.002) (table 3). Those who reported consuming alcohol were more likely to use watercraft (χ2= 60.1; p<0.001) and stay 24 hours or more at rivers (ie, overnight) (χ2= 85.7; p<0.001).
Adjusting for exposure
Males drown at a rate that is 7.6 times that of females, however females reported slightly lower exposure to rivers (72.1% compared with 74.6% for males) meaning the exposure adjusted rate of female drowning increased from 0.06 to 0.09). When adjusting for exposure, the fatal drowning rate for males aged 75+ years moves from 0.55 to 0.95, representing an RR that is slightly higher (RR=1.1; 95% CI 0.6 to 2.1) than that of males aged 18–34 years. Females aged 55–74 years recorded the highest rate of female river drowning (0.10), increasing to a rate of 0.13 per 100 000 exposed. This represents an RR one and a half times (RR=1.5; 95% CI 0.44 to 5.2) that of females aged 18–34 years (table 4).
After adjusting for exposure, the RR for swimming and recreating increased from 1.2 to 12.4 when compared with the risk of a fall from a non-aquatic activity (RR=12.4; 95% CI 7.7 to 20.0). The RR for swimming and recreating increased by a factor of 9 for males (RR=1.10–10.59) and 11 times for females after adjusting for exposure (RR=2.50–28.00) (table 5).
After adjusting for exposure, the male rate of alcohol-related river drowning increased from 0.15 to 1.00/100 000. Males were eight times (RR=8.5; 95% CI 2.6 to 27.4) more likely to drown with alcohol present in their bloodstream than females. For males aged 18–35 years, after adjusting for exposure, the rate of alcohol-related river drowning increases from 0.20 to 0.97 and from 0.11 to 0.76 per 100 000 exposed for males aged 55–74 years. Females aged 55–74 years had the largest change in alcohol-related river drowning rates after adjusting for exposure, 0.01–0.14 per 100 000 (table 6).
The RR of drowning as a result of a watercraft-related incident when comparing males with females, decreased from 30 times (RR=30.0; 95% CI 4.1 to 220.0) to 25 times (RR=25.5; 95% CI 3.5 to 186.9) after adjusting for exposure (table 7).
Exposure is complex and challenging to calculate,5 7 13 15 particularly in dynamic environments such as rivers15 where people move in and out of the water and multiple activities are undertaken. This study explored river exposure by activity to establish improved estimates of drowning risk. In Australia, 73% of the adult population visits a river at least once in a 12-month period, most commonly for non-aquatic activities. Even after adjusting for exposure, males still have a higher rate of drowning in rivers than females, posing challenges for prevention.
There are a multiplicity of factors which impact exposure at rivers including number and duration of visit and type and duration of activities undertaken. People spend different lengths of time undertaking activities in, out, near and away from the water, as well as a varying numbers of visits. CATI surveys are a popular option among researchers to gather population-level data,31 however they are not without their limitations.32 In using a CATI survey to calculate river-exposed drowning risk, the authors found it was challenging for respondents to recall length of time spent performing an activity, especially when undertaking multiple activities. Respondents were asked to nominate the main activity they undertook, however as people often undertook multiple activities, results present the minimum number of people who undertook that activity. This therefore means that activity-based rates presented represent the maximum possible due to only one activity being recorded. Similar issues were faced by other exposure-based research where self-reported visitation and activity data may suffer from recall bias.1 33 Observational research, including novel strategies such as aerial surveys and remote camera observations,16 may add greater detail and overcome some of the methodological constraints, but is not without its own limitations.34
Rivers present the challenge of having three types of exposure; namely beside the river (eg, walking, picnicking and fishing from the river bank), in the river (eg, swimming and recreating) and on the river (eg, using watercraft). Questions remain regarding how activity and distance of activity from river (eg, walking beside the river vs picnicking) impact risk. It is not known what a safe distance from the river is and how this differs based on type of river and geography of the location. Similarly, watercraft usage changes risk, so too does boat size,35 powered or unpowered,36 seat position in the boat,37 lifejacket wear36 38 and alcohol consumption.15 23 A constraint of this study is that watercraft usage was combined in the CATI survey and used to calculate rates. Further research, potentially in the form of observational studies, exploring watercraft usage at rivers, is warranted.
Exposure and sex
Males make up 80% of drowning statistics worldwide,11 in Australia39 and in Australian rivers.22 There was no difference in visitation between males and females at rivers, however the female drowning rate increased by 50% when adjusting for exposure. While exposure explains a small amount of the variation in drowning rates between males and females, visitation alone based on any time in the last 12 months is not a predictor of drowning risk and further work needs to be undertaken in developing a more sensitive measure of self-reported exposure to rivers.
A rationale for the variation between males and females may be the activity undertaken and the location of that activity, males have been identified as undertaking risky behaviours40 41 and visiting unsafe locations.22 42 Even at a young age, males are identified as having lower levels of water safety knowledge43 and poorer swimming skills44 than their female peers. Further research is required to examine behaviours and attitudes towards safety of female river-goers and how this differs from males.
Alcohol is widely consumed across Australia (9.70 L of pure alcohol consumed per capita in 2016)45 and is a known risk factor for drowning. Twenty-nine per cent of the deaths in this study had a BAC≥0.05%,23 whereas the proportion reporting drinking at rivers was 16%. Based on these findings, drinking at rivers appears to increase drowning risk by 1.8 times (RR=1.8; 95% CI 1.05 to 3.12). It should be noted that alcohol was present in 314 (41%) river drowning deaths,23 however this may also include a build-up of alcohol due to decomposition.25
Alcohol consumption impacts the risk of injury or death due to the way people interact with the water and the physiological effects on the body.46 The location of the river, as well as activity being conducted may also impact on alcohol consumption and risk. It is unclear if needing to drive a motor vehicle to the river (or operate watercraft once at the river) impacts on alcohol consumption, however even if the driver does not drink, passengers may consume alcohol, potentially at higher levels, although this assumption needs further testing in the field.
This study identified challenges associated with quantifying alcohol-related drowning risk; in particular understanding location-specific alcohol use, activities undertaken while or immediately after consuming alcohol and the self-reported nature of alcohol consumption. This may impact the accuracy of the exposure-adjusted alcohol-related river drowning rates presented in this study. Future research with different methodological approaches (such as population-based case-control studies15) will be required to help fully understand the role and impact of alcohol on drowning risk at rivers.
Alcohol has also been identified as a significant risk factor for drowning (both in rivers and other aquatic locations) in other high-income countries, such as Canada,47 Sweden,48 Finland,49 the USA9 and New Zealand.40 Exposure studies in such countries should also be conducted in order to explore whether drinking alcohol at rivers increases drowning risk.
Future research: validating findings in the field
There are thousands of kilometres of rivers where people recreate and more work needs to be done in understanding how people are using these spaces. A river user may visit a particular river or spot on the river for boating or water skiing and a different location for swimming, for example, all of which impact risk. This study provides exposure information at a population level, noting the challenges associated with self-reporting.31
Further community-based research in the form of prospective studies is required, such as using diaries or mobile applications to collect visitation and participation data direct from the community or validating this study’s findings in the field at known river drowning locations.13 Other technological solutions for conducting observational studies, such as aerial surveys and remote camera observations may also be explored.16 Developing a set of common variables for river exposure studies will be important both in Australia and internationally, to allow for comparison between and across studies, to enhance our knowledge of river usage and safety.
Strengths and limitations
This study is the first to examine river exposure on a national basis. The survey tool was nationally representative allowing for broader extrapolation to the Australian public. By combining survey data with a total population survey of fatalities in the year immediately before, during and after the survey was conducted, this study provides more accurate river drowning rates, both overall and by subpopulation, to better inform river drowning prevention efforts.
Given the high burden of river drowning internationally, and the lack of published data on river exposure, this study has global relevance, especially in high-income and middle-income countries where there is significant recreational aquatic activity.
There are, however, limitations associated with this study. The CATI survey collects self-reported responses and has the limitations of self-reported data. The survey is cross-sectional in nature and therefore, different studies are needed to support causality. The survey also asks respondents to reflect on their behaviour within the last 12 months, which may introduce recall bias. The CATI survey data were collected between 6 July 2015 and 7 August 2015, being the winter season in Australia which may have impacted on the answers respondents gave. Information race and ethnicity was not collected in the survey and is poorly reported in the coronial data placing further limitations on revising river drowning risk based on exposure for different ethnicities. This is an area that warrants further attention. Information on type of watercraft was not collected by the survey tool and therefore further refined fatal drowning rates based on different types of watercraft were not able to be calculated. As risk may differ based on watercraft-type, this is also an area worthy of further research.
Analysis of activity by age group beyond watercraft incidents for males was not undertaken. Given the high exposure-adjusted rates of fatal river drowning for females aged 55–74 years and males 75 years and older, this warrants further research to better inform prevention efforts among these age groups.
Available data on unintentional drowning fatalities for the 3 years used to calculate risk ratios may not offer all details on the circumstances of the drowning. As 15.2% of cases remained open (ie, under investigation), information may change pending the outcome of coronial investigations.
The findings of this study present, for the first time, nationally representative data on river exposure in Australia. When combined with data on unintentional fatal drowning in Australia, a clearer understanding of those at increased risk appears. This study has identified those at highest risk are males, in particular those aged 75 years and older and females aged 55–74 years. Such findings should be considered when developing prevention strategies and in undertaking advocacy for prevention among target groups.
What is already known on the subject
Globally, rivers are a leading location for drowning, with risk factors including being male and alcohol.
No previous study has examined river drowning risk based on exposure.
What this study adds
Females were more likely to report participating in activities putting them at risk of a fall (χ2=19.0; p<0.001), and males fishing (χ2=11.5; p=0.001).
The 55–74 years age group was identified as being at an increased risk of alcohol-related river drowning (males 0.79 and females 1.68/100 000 exposed).
Compared with females, males were more likely to drown due to a watercraft-related incident (RR=25.5; 95% CI 3.5 to 186.9).
Contributors AEP and RCF conceptualised the study, designed the survey questionnaire, conducted the analysis, drafted and revised the manuscript. AEP collated and analysed the fatal drowning data. PAL provided oversight and advice on study design, analysis and revised the manuscript. All authors approve the submitted manuscript.
Funding This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests None declared.
Patient consent Not required.
Ethics approval Ethics approval for the fatal drowning data was received from the Victorian Department of Justice (JHREC-CF/16/19581) and James Cook University (HREC-H6282). The CATI survey received approval by the Human Ethics Research Review Panel at Central Queensland University before administration to the general public (H14/09-203).
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement Researchers can apply for approval to access the coronial data presented in this study. For those interested, please contact the Australian National Coronial Information System for more information: email@example.com.