Declines in Drowning: Title and subTitle BreakExploring the Epidemiology of Favorable Trends

Over the past quarter century, fatality rates in many unintentional injury categories have declined dramatically, both in the United States and in other industrialized countries.¹ The largest decline in number of deaths has been from unintentional drownings, which declined from 5700 in 1986 to 3959 in 1996.^{1 - 2} In contrast to other injuries, the causes for this decline are largely unexplained.

In any consideration of drownings, it is important to recognize that there are 2 distinct high-risk groups in the United States: children (both boys and girls) younger than 5 years and males aged 15 to 34 years. Despite the availability of effective interventions such as pool fencing and supervision, drowning rates have declined little among children and actually have increased among children younger than 1 year. Drowning rates by year of age are lowest at 10 years and then increase rapidly in males, peaking at 18 years. The largest declines in drowning rates nationally have occurred in the 10- to 19-year-old age group (with decreases of 5.8% and 5.4% per year from 1971 to 1988 in the age groups 10-14 years and 15-19 years, respectively). These declines are of particular interest because they seem to have occurred without specific interventions directed at these age groups.^{3 - 4}

The study in this issue of THE JOURNAL by Cummings and Quan⁵ is a welcome exploration of the declining rates in drowning. The study provides convincing evidence that medical care played little role in the downward trend in drownings. The authors suggest that 51% of the decline from 1975 through 1995 was due to reduced alcohol involvement, although the remaining 49% remains unexplained. This article raises some interesting questions: Why would the contribution of alcohol to drowning decline over time? What other factors might account for the dramatic declines in drowning? What factors might explain the long-term trends in the reduction of injury fatalities in general?

Alcohol use is an important factor in drownings and plays just as great a role as it does in highway crashes. A recent meta-analysis⁶ found similar proportions of intoxicated victims (blood alcohol concentration [BAC], ≥100 mg/dL [21.7 mmol/L]) between drownings and drivers killed on the highway (34.2% vs 32.8%). The study by Cummings and Quan⁵ observed a similar alcohol intoxication rate for drownings (30%) and offers convincing evidence that declining alcohol use by individuals in and around water may be responsible in part for declining drowning rates. The authors estimate that the proportion of deaths attributable to alcohol declined from 50% in 1975 to only 22% in 1995, and that alcohol-related drownings declined more than did total drownings (81% vs 59%). However, such ecological findings are difficult to interpret. It is possible, for example, that alcohol use is more often associated with drownings occurring in natural settings (eg, open water, rivers, streams, and quarries), locales where the most dramatic declines in drowning occurred. The propensity of persons to swim in these natural settings may have shifted toward more supervised venues (eg, public pools or beaches with lifeguards) where drinking is often prohibited. Thus, when trends are examined for all drownings, the results may combine reductions in risks due to the mixed effects of both alcohol and setting. Although the data on the declining use of alcohol are fairly convincing, unfortunately there are no corroborating longitudinal data on exposure to risk.

It is also possible that the authors may have overstated the decline in alcohol use. A factor that may have biased the results toward finding an influence of alcohol is that 31% of cases had missing BAC data (20% not tested and 11% invalid). Imputation as a means of compensating for missing alcohol data also is used for highway fatality research, but uses more complex models.⁷ The method used by Cummings and Quan assumes that cases with missing BACs had the same characteristics as those in the same age group with known BACs. It is likely that alcohol was less involved in cases not tested because medical examiners are less likely to measure BACs when alcohol involvement is not suspected. BACs were measured less frequently in earlier years, with 38% of cases having missing BACs in the period 1975-1979 compared with only 27% in the period 1990-1995. If these untested cases were more likely to have a zero BAC (a reasonable assumption), the imputation method used by the authors would artificially inflate estimates of alcohol involvement in earlier years. However, this potential bias is unlikely to change the conclusions of the study, especially as the actual rate for drownings that involve alcohol also declined over the study period, but may reduce the estimated size of the influence of alcohol.

Dramatic declines have occurred in both alcohol-related motor vehicle fatalities and drinking and driving over the past 20 years, particularly in groups at the highest drowning risk.⁸ It is possible, although unproven,⁹ that policies designed to reduce drinking and driving may indirectly have reduced alcohol-related drowning. This hypothesis is difficult to test without adequate data on alcohol use by persons on and around bodies of water. A recent study¹⁰ documented alcohol use among boaters, but did not provide any data on trends over time. Ideally, data on person-time spent at risk (on or close to the water) with elevated BACs over time would be informative, but unfortunately, such data are not available.

Although it is likely that changes in patterns of drinking in aquatic settings played a role in the overall declines in drowning, it is apparent that a large part of the decline is still unexplained. Reduction in risk can be achieved by reducing risk associated with exposure or by reducing exposure itself. For most drowning situations in natural bodies of water, the nature of the activity and its risks have changed little except perhaps for boating, for which safety regulations and programs were first initiated in the 1970s.¹¹ Reductions in drownings could be attributable, at least in part, to reduced exposure to open bodies of water, where most drownings occur.

Any change in exposure can have a dramatic effect on injury incidence. For example, teenage driver education programs were found to actually increase injury risk to adolescents because such programs encouraged driving at an earlier age (thus increasing exposure), whereas the training provided had little or no effect at decreasing the risk per mile traveled.¹² Concerns also have been raised that swimming education may potentially increase exposure to hazardous bodies of water for similar reasons.¹³ In Australia, laws that require bicycle helmets to be worn reduced not only head injuries, but also other injuries, apparently because the law also reduced overall exposure by decreasing the amount teenagers traveled by bicycle.¹⁴ Most of the decline in pedestrian and bicycle fatalities among children and adolescents is attributed to reduced exposure (eg, miles walked or cycled) as a result of increased car travel to school because of increased distances between home and school or bus and fears of injury, abduction, or both.¹⁵

Although declines in children's activities (exposure) may explain declines in pedestrian and bicycle deaths, there is only indirect evidence that exposure to hazardous open bodies of water may have declined. However, there is evidence that the increase in obesity in the population at risk is associated with decreases in overall activity levels and lower levels of fitness in Army recruits.¹⁶ Potential reasons for reduced activity include increased sedentary activities (eg, television watching, computer use, video games) and declines in physical education participation in high schools.^{17 - 19}

Societal changes and fears may also have restricted swimming or play in unsupervised settings. Instead of adolescents entertaining themselves (including walking or cycling to hazardous waterways), many more organized and safer recreational activities are now available. In addition, increased awareness of water pollution, closures of public swimming facilities because of fears of litigation, and increased availability of patrolled swimming sites and pools also may have reduced exposure to high-risk bodies of water.

Although reduced exposure to high-risk situations may appear beneficial, it also may have the untoward effect of leading us to the false conclusion that the inherent risk of the hazard has been reduced. If exposure were to increase some time in the future without proven prevention strategies in place, this trend may be reversed. Declines in activity also may have more long-reaching consequences on future health. The decreased physical activity levels and corresponding increased incidence of obesity can lead to increased risks of chronic disease in later life.²⁰ For now, whether the declines in drowning and other unintentional injuries will come at the price of an increased incidence of chronic diseases later in life remains an unanswered question.