Article Text

Download PDFPDF

Estimates of drowning morbidity and mortality adjusted for exposure to risk
  1. Rebecca J Mitchell1,2,
  2. Ann M Williamson1,
  3. Jake Olivier2,3
  1. 1Department of Aviation, University of New South Wales, Sydney, Australia
  2. 2NSW Injury Risk Management Research Centre, University of New South Wales, Sydney, Australia
  3. 3School of Mathematics and Statistics, University of New South Wales, Sydney, Australia
  1. Correspondence to Rebecca J Mitchell, NSW Injury Risk Management Research Centre, University of New South Wales, Sydney, NSW 2052, Australia; r.mitchell{at}unsw.edu.au

Abstract

Objectives To estimate the rate of unintentional drowning mortality and hospitalised morbidity using population-based, population-risk and person-time denominator data and to compare the estimates obtained. To then compare exposure-based rates for drowning with road traffic death rates.

Method Retrospective analysis of unintentional drowning mortality and hospitalised morbidity of New South Wales (NSW, Australia) residents 16+ years of age during 1 January to 31 December 2005. Information on population-risk and person-time risk was obtained from the 2005 NSW Population Health Survey. Analysis of road traffic death data from NSW and population and person-time risk estimates from the Survey of Vehicle Use, Household Travel Surveys and Roads and Traffic Authority Speed Surveys in 2005.

Results Estimated drowning mortality and hospitalised morbidity rates for adults were higher using population-risk and person-time risk exposures compared to a population-based exposure. Population-based estimates of road traffic mortality were four times higher than drowning mortality rates. In contrast, exposure adjusted person-time estimates for drowning were 200 times higher than road traffic fatalities.

Conclusions Many injury risks are underestimated when the total age-specific population is used to calculate an injury rate instead of actual population-risk or person-time exposure. This can result in the identification of misleading priorities for injury prevention. Drowning risk is strikingly higher than previously thought based on population-based estimates. This information is important for decision-making and policy development as it provides a basis for comparing the inherent risk in exposure to hazards with potential to cause injury.

  • Drowning
  • surveillance

Statistics from Altmetric.com

Introduction

Calculation of injury incidence rates is often hampered by the lack of appropriate denominators that reflect exposure. Failure to adjust injury rates for exposure to a hazard necessarily results in poor estimates of risk. This has been particularly so for estimates of unintentional drowning risk in Australia. Using population-based estimates, drowning in Australia occurred at an incidence of around 1.3 in every 100 000 people in 2003–04.1 This estimate assumes equal opportunity for drowning across the population which is clearly not the case. While almost all population-based estimates for injury suffer the same problem, drowning estimates are especially vulnerable compared to other injuries, such as road traffic injury. Drowning will only occur under very specific conditions: when exposed to water. This is not a trivial point as action on injury prevention is based mostly on these flawed population estimates of risk. For example, calculations of the drowning rate for adults typically use the total adult population as the denominator. This underestimates the risk of drowning as it includes adults who are never or very rarely exposed to water and consequently the risk of drowning. Ideally, only the number of adults who are exposed to bodies of water that have a drowning risk should be used to calculate the rate of adult drowning.

Population-risk (proportion of population exposed to potential hazard) and person-time risk (amount of time a person is exposed to potential hazard) have been taken into account in the calculation of injury risk in some settings, such as estimates of per kilometres travelled2 for the risk of injury in a motor vehicle crash or per million hours worked3 for the risk of occupational injury. However, this sort of exposure information has not been available previously for drowning in Australia. The aims of this paper are (1) to estimate the rate of unintentional drowning mortality and hospitalised morbidity of individuals aged 16 years or older using population-risk and person-time risk estimates; (2) to compare population-based, population-risk and person-time rates of drowning mortality and hospitalised morbidity; and (3) to compare population-based and person-time risk estimates for drowning and road traffic fatalities.

Methods

Risk estimates for drowning

Information on population-risk and person-time risk was obtained from the 2005 New South Wales (NSW) Population Health Survey. Information on drowning mortality and the number of NSW residents in 2005 came from the Australian Bureau of Statistics (ABS). Data on hospitalised morbidity was obtained from the NSW Admitted Patients Data collection (APDC). Only individuals aged 16 years and over were included in this analysis as this was the lower age limit of individuals in the NSW Population Health Survey.

NSW Population Health Survey

A total of 13 701 respondents aged 16 years and over were sampled in the 2005 NSW Population Health Survey, reflecting an overall response rate of 57.7%.4 Specific details concerning the survey methodology are described elsewhere,4 but are summarised here. A two-stage sampling process was used, with the sample stratified by each of the eight NSW Area Health Services (AHS). Within each AHS, households were randomly selected using a computer-generated list of telephone numbers; a single respondent was randomly selected from each household for interview.4 Interviews were conducted from February to December 2005 by trained computer-assisted telephone interviewers. Up to seven call-backs were made to establish contact with a household and up to five call-backs were made to contact a selected respondent.4

For the 11 271 respondents who had a valid response (ie, yes or no) for their exposure to water, 3033 (26.9%) stated they had been in or on the water, at a swimming pool, beach, lake, river, creek, stream or dam in the last 4 weeks, including for fishing; 2378 (21.1%) reported either swimming, fishing or rock fishing in the last 4 weeks; and 2055 (18.2%) respondents reported they had been swimming in the last 4 weeks. These individuals were asked to indicate the number of times they swam in the past 4 weeks, the location of their last swim, their swimming ability, and approximately how many hours they had spent swimming in the last 4 weeks. Questions were field tested prior to inclusion in the survey.

ABS mortality data file

Mortality data from 1 January 2005 to 31 December 2005 were obtained from the ABS mortality data file. Mortality data were coded using the International Classification of Diseases, version 10.5 Drowning mortality of NSW residents was identified using the following criteria:

  • the deceased was 16 years of age or older, a resident of NSW and the death occurred in Australia; and

  • an external cause code was related to unintentional drowning and submersion (excluding bathtubs) (ie, W67-W74 or Y21).

NSW APDC

Hospitalisation data were obtained from the APDC from 1 January 2005 to 31 December 2005. The APDC includes information on inpatient separations from NSW public and private hospitals, private day procedures and public psychiatric hospitals. It includes information on episodes of care in hospital that end with the discharge, transfer or death of the patient, or when the service category for the admitted patient changes. Also included are hospitalisations of NSW residents that occurred in another state or territory in Australia. Information collected includes patient demographics, diagnoses and clinical procedures. Hospitalised morbidity data were coded using the International Classification of Disease, 10th Revision, Australian Modified (ICD-10-AM).6 Cases were identified using the following criteria:

  • the hospitalisation was for a patient 16 years of age or older who was a resident of NSW;

  • a principal diagnosis was in the ICD-10-AM range S00-T98; and

  • an external cause code was related to unintentional drowning and submersion (excluding bathtubs) (ie, W67-W74 or Y21).

Data management and analysis

To reduce ‘multiple counts’, which occur when an individual has more than one episode of care for an injury, cases were excluded if they involved transfers or statistical discharges such as changes in hospitals or service categories (eg, acute to rehabilitation).7

The analysis used SAS statistical software, V.8.02.8 The SURVEYMEANS procedure was used to analyse the survey data and estimate proportions of reported water exposure by age and gender. Data were stratified by AHS and a sampling weight was applied to adjust for differences in the probabilities of selection among respondents due to the varying number of people living in each household, the number of residential telephone connections for the household, and the varying sampling fraction in each AHS.

Population-based exposure was calculated using age-adjusted rates for drowning mortality and hospitalised morbidity per 100 000 population. Age standardised rates were calculated using the estimated Australian residential population at 30 June 2001 as the standard population. Population-risk exposure was calculated per 100 000 population reported to be exposed to water using three different levels of exposure: (i) being in or on water, at a swimming pool, beach, lake, river, creek or stream or dam in the last 4 weeks; (ii) swimming, fishing or rock fishing in the last 4 weeks; and (iii) swimming in the last 4 weeks. The proportion of individuals who reported being exposed to water in each of these three settings was then applied to the NSW population data to obtain population estimates of the number of individuals exposed to water by age group and gender. Person-time risk exposure was calculated per 1000 h using the total hours reported swimming in the last 4 weeks from the NSW 2005 Population Health Survey as the total person-time exposed; 95% CIs were calculated assuming a Poisson distribution.9

Drowning mortality and hospitalised morbidity rates between the population-based, the three population-risk and the person-time risk exposures were compared using regression analyses to assess interactions between the difference in rates for mortality and hospitalised morbidity, sex and age group. Two-way interaction terms between combinations of mortality or hospitalised morbidity, age group, and sex were considered for the model. No interactions were significant at the 0.25 level and all were dropped from the model.

Risk estimates for road traffic fatalities

For comparison, population-based and person-time risk-based mortality rates were calculated for road traffic fatalities. Total road traffic fatalities for NSW and the Sydney Greater Metropolitan Area (GMA) were obtained for 2005 from the NSW Roads and Traffic Authority (RTA), together with licence-holder data for 2005.10 11 As there was no survey data to estimate person-time exposure to road traffic hazards, person-time exposure was estimated using two methods:

  1. The Household Travel Survey of private occupied dwellings in the Sydney GMA in 2005, which estimated that the average hours per year for car driver and passenger trip durations for residents in this area was 301.7 h.12

  2. Estimated road travel time using the total distance travelled per year in NSW from the ABS Survey of Motor Vehicle Use13 and estimates of the average speed travelled on some major roads in the Sydney region.14

The ABS survey involved a stratified sample of 16 000 vehicles across Australia selected to report on vehicle use by odometer readings over a 3-month period between 1 November 2005 and 31 October 2006. This survey estimated that 63 717 million kilometres were travelled in vehicles in NSW in 2005. Travel speeds were obtained from twice-yearly speed surveys on major routes in Sydney. These estimates suggested that travel speeds in the Sydney region were around 30–40 km per hour (kph) during peak hours. Combining distance travelled and estimated average speed, the average hours travelled on roads for NSW in 2005 was 314 h. As average speed outside the Sydney region is likely to be higher, the average hours travelled was also calculated for average speeds of 50 kph which resulted in a person-time estimate of 188 h. These person-time estimates were used as denominators to calculate mortality rates for road traffic crashes in NSW in 2005.

Results

Drowning mortality and hospitalised morbidity rates were consistently higher using population-risk and person-time risk exposures compared to the population-based exposure (χ2=88.47, df=4, p<0.0001; χ2=79.62, df=4, p<0.0001, respectively) (table 1). This is because as the proportion of individuals in the population who reported being in or on the water, involved in swimming or fishing, or who swam only are taken into account, the population exposed declines (table 2).

Table 1

Rate of unintentional drowning mortality and hospitalised morbidity in NSW using population-based, population-risk and person-time risk estimates, 2005

Table 2

Denominator data used in the calculation of incidence rates and hours exposed

For men, the drowning mortality rate varied by age group and exposure (figure 1). The drowning mortality rates were highest using the swimming-only population-risk data, with particularly high rates for men aged 65 years and over. For all age groups, the population-based mortality rate doubled when taking into account the proportion of the population who reported being in or on the water, increasing to over 10 times the population-based rate when only the proportion of the population who reported swimming-only were used as the denominator. Estimating drowning mortality for women was prohibitive due to rare events.

Figure 1

Rate of unintentional drowning mortality and hospitalised morbidity in NSW using population-based, population-risk and person-time risk estimates, 2005. Source: NSW Admitted Patient Data Collection and NSW Population Health Survey (HOIST). Centre for Epidemiology and Research, NSW Department of Health.

Hospitalised morbidity drowning rates were similar. The hospitalised morbidity rate for drowning varied by age group and exposure for both men and women, with each having higher rates of hospitalisation using the swimming-only population-risk exposure data (figure 1). All hospitalised morbidity rates calculated using population-risk data were higher than the population-based estimate.

Population-based road traffic mortality rates were at least four times higher than drowning mortality rates (7.5 compared to 0.9 per 100 000 population), but estimated person-time mortality rates showed the reverse, with drowning mortality rates per 1000 h exposure more than 200 times higher than equivalent exposure-adjusted rates for road traffic fatalities (table 3).

Table 3

Mortality rates for vehicle occupants using population-based and person-time risk estimates for road fatalities in NSW in 2005

Discussion

Injury risks are often underestimated as the total population is used to calculate an injury rate, rather than actual population-at-risk or by time exposed.15 The question is whether this really matters. These results show the impact of underestimation of risk. The risk of an individual drowning is greatly underestimated, by up to 10 times, by using population-based exposure data. A similar comparison for vehicle-occupant mortality showed the opposite; population-based denominators overestimate risk. Clearly, drowning is less frequent than road traffic fatalities as fewer people are exposed to water hazards, but when water exposure occurs, the risk of death is much higher compared to exposure to the road environment.

This study highlights the importance of establishing an appropriate denominator for the population-at-risk,16–18 as the denominator selected can affect the perceived importance of a particular injury issue.19 In terms of policy development, under- or overestimation of the true risk of injury20 can lead to poor identification of priorities for developing injury prevention policies and interventions, and inadequate resource allocation. Road safety is recognised as one of the major issues in injury prevention in Australia, and consequently attracts the highest levels of attention and funding. In absolute terms, this is appropriate as more people are killed on roads than in any other setting. Yet our current approach of using only population-based estimates of risk means that other important problems can be overlooked. While the person-based exposure estimates in this study indicate that there will be 2–4 deaths for every 10 million hours on the road, they also show that there will be 90 000 drownings for every 10 million hours of swimming. Reframing injury problems in this way with appropriate estimates of risk presents another, more valid, perspective for public health decision-making. The need to actively prevent injuries when they occur to many people is widely accepted, but we must not overlook the need to prevent injuries when people undertake activities that have a high risk of injury.

There are several limitations of the current analysis. First, the numerator for injury deaths in 2005 may be inaccurate due to delays in processing death registrations in NSW,21 which meant that there were around 200 fewer injury deaths recorded by the ABS in 2005. Some of these deaths may have involved drowning. In addition, cause was not specified for 19.6% of injury-related deaths and 9.3% of hospitalisations involving over 15 year olds in NSW (ie, ICD-10 or ICD-10-AM code X59), so the numerators for drowning mortality and hospitalised morbidly could be higher. Second, there are potential problems with exposure estimates for both drowning and road hazards. Data from the NSW Population Health Survey and the Household Travel Survey were self-reported and could suffer from recall and information bias. In addition, the NSW Population Health survey did not collect data in January 2005, a peak swimming time in Australia, so population-risk and person-time risk of an individual drowning or being hospitalised may have been overestimated. For defining road exposure, the most direct estimates were from the Household Travel Survey, but as this only covered the Sydney Greater Metropolitan Areas, exposure was calculated using vehicle distance travelled and estimates of average speeds. It is notable that the calculated method produced very similar person-time exposures for the Sydney region as the direct survey method (314 h compared to 301.7 h) so supporting the use of the calculated method. Finally, the 95% CIs for some rate calculations are wide due to low case counts, and these rate estimates should be interpreted with caution.

This analysis showed that population-based rates of drowning underestimate the true risk of drowning in NSW. Furthermore, this analysis only looked at risk of drowning using particular ICD-10-AM unintentional drowning and submersion codes. If additional activities were taken into account, such as watercraft-related activities, the rate of drowning per population-at-risk or person-time exposed is likely to be much higher. Information on population-risk and person-time risk is required to estimate injury risk more accurately. The results of this study relate specifically to one state of Australia. Actual rates are likely to differ for other Australian states and countries where exposure differs from NSW, such as low to middle income countries. The study, however, demonstrates an important point: the real risk of drowning is markedly higher than estimated using only population-based rates. Taking into account the relative risk of drowning as in this study, highlights the need to reassess our efforts to prevent drowning.

What is already known on the subject

  • Population-based measures of exposure are used routinely to calculate an estimate of the risk of injury.

  • There has been limited work conducted examining different measures of calculating injury risk that more closely reflect actual exposure, particularly for drowning.

What this study adds

  • This study provides evidence that population-based estimates of injury risk underestimate the risk of drowning compared to population-risk and person-time risk exposure estimates.

  • It shows that access to appropriate denominator data is essential to provide an accurate estimate of injury risk to more accurately inform policy development and prevention priorities.

Acknowledgments

The authors wish to thank the Centre for Epidemiology and Research at the NSW Health Department for providing access to the Health Outcomes and Information Statistical Toolkit (HOIST) to obtain data analysed in this study. The HOIST system refers to a data access, analysis and reporting facility established and operated by the Centre for Epidemiology and Research, Population Health Division, NSW Department of Health.

References

Footnotes

  • Funding RM and JO are supported by the NSW Injury Risk Management Research Centre (NSW IRMRC), with core funding provided by the NSW Health Department, the NSW Roads and Traffic Authority and the Motor Accidents Authority of NSW. AW is supported by an NHMRC Senior Research Fellowship.

  • Competing interests None.

  • Ethics approval This study was conducted with the approval of the NSW Population and Health Services Research Ethics Committee.

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

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.