Alcohol consumption and traffic and non-traffic accidents in Australia, 1924-26 Michael Livingston 1, 2 & Heng Jiang 1 1 Centre for Alcohol Policy Research, Turning Point Alcohol and Drug Centre, Melbourne, Victoria, Australia; 2 Drug Policy Modelling Program, National Drug and Alcohol Research Centre, The University of New South Wales, New South Wales, Australia Abstract Background: There is a significant body of literature demonstrating that the rate of alcoholrelated harm in a society is strongly correlated with per-capita alcohol consumption. In particular, studies have shown that mortality from a range of causes changes along with population level consumption. However, most of these studies have focussed on chronic outcomes such as liver cirrhosis, with few exploring whether accident-related mortality is affected by per-capita alcohol consumption. Aims: This study aims to evaluate the relationship between changes in per capita alcohol consumption and traffic and non-traffic accident deaths in Australia from 1924 to 26. We will also examine the impact of the introduction of random breath testing (RBT) on both types of accident mortality. Methods: Gender and age specific traffic and non-traffic mortality rates (15 years and above) were analysed in relation to per capita alcohol consumption using Box-Jenkins techniques for time series analysis. The external effect of the introduction of RBT was measured by inserting a dummy variable in the time series models. Results: Statistically significant associations between per capita alcohol consumption and both types of accident mortality were found in both males and females. The results suggest that an increase in per capita alcohol consumption of 1 litre was accompanied by an increase in accident mortality 3.4 among males and.5 among females per 1 inhabitants. A 1- litre increase in per capita alcohol consumption corresponded to an increase in non-traffic mortality about 3. among males and.9 among females. The association between alcohol consumption and traffic accidents was stronger among younger people than older people. However, the relationship between alcohol consumption and non-traffic accidents was weaker in younger people than older people. The estimated effects of the introduction of RBT show significantly reductions in traffic mortality in Australia, particularly for males and young people. Discussion and conclusions: Changes in alcohol consumption have had substantial effect on both traffic and non-traffic accidents. Policies for controlling alcohol consumption would effectively help to reduce traffic and non-traffic accidents in Australia. Specific policies aimed at reducing drink-driving have demonstrably reduced traffic accident deaths. Introduction It is well-recognized that alcohol intoxication contributes to a large proportion of traffic and non-traffic accidents in many countries (Skog 21a). Previous studies show that about 1% to 6% of all fatal traffic crashes appear to be alcohol related (Norström and Rossow 213) 1
Furthermore, alcohol also plays a significant role in non-traffic accident, such as accidental falls and drowning, accident caused by fire, and other accidents (Rehm et al. 21). Alcohol contributes to over 3 deaths each year in Australia, and is implicated in, 5% of assaults, 3% of car accidents, 44% of fire injuries, 34% of falls and drowning and 12 % of suicides (NCETA 24). There is significant evidence that the level of alcohol consumption in a given society is related to the rate of alcohol-related problems (Norström and Ramstedt 25). Studies examining whether accident mortality is related to population drinking levels are less common, but the existing evidence suggests a link, particularly in countries with intoxicationfocussed drinking patterns. Two studies on data from Europe demonstrated significant associations between per-capita consumption and accident mortality, with the effect sizes varying across countries. On the whole, changes in population drinking affected accident mortality more in Northern Europe than Southern Europe, but the effect on traffic accident mortality was higher in the Southern European countries (Skog 21a; 21b). Other analyses have also shown a strong association between per-capita consumption and accident mortality in North America (e.g. (Skog 23)), but there have been no Australian studies in this area. The role of alcohol in accidents probably by varies in different gender, age groups and historical periods. Gender differences in the association between alcohol consumption and fatal accidents have been found in some previous studies, such as Skog (23) and Ramstedt (28), with effects on male mortality much higher than on female mortality. Similarly, previous studies have found varying effects across age group, with alcohol affecting traffic deaths among young people more than older people (Ramstedt 28), while the effects on total accident mortality were smallest in the middle age group (3-49 years) compared with the young age group (15-29 years) and old age group (5-69 years) (Skog 21b). Random Breath Testing is a widely used drink-driving countermeasure in many countries and is particularly common in Australia (Rowland et al. 212). RBT was introduced in Australia in 1976 and detailed analyses have demonstrated that its introduction significantly reduced alcohol-related crashes and fatalities (Henstridge et al. 1997; Peek-Asa 1999). The main aim of this study is to present the first estimates of the extent that changes in per capita alcohol consumption influence mortality rates of fatal traffic and non-traffic accidents in Australia. As part of this analysis we will explore the impact of WW2 and the introduction of RBT in Australia on accident deaths. Methods A proxy for per-capita alcohol consumption was constructed using data on the sale of alcohol sourced from the Australian Bureau of Statistics (ABS 211). Early data were converted from gallons or proof gallons to litres of pure alcohol. This was then converted to litres of pure alcohol per resident aged 15 and older, with population data provided by the Australian Institute of Health and Welfare (AIHW) (AIHW 28). Mortality data were provided by the AIHW on the basis of historical death certificate data. The AIHW have developed standardised historical mortality databases which track the relevant cause of death codes across the various classification schemes used in Australia. The death rates were agestandardised for men and women separately and ex-pressed per 1, population, using indirect standardisation to 21 data from the Australian Census. Age-group specific mortality rates are unstandardized. 2
The one-off event dummy was constructed to represent the on-going impact of RBT on accidents in Australia. As RBT was first implemented in Australian in 1976, the RBT dummy variable is coded as 1 after 1975 and between 1924 and 1975. This represents a simplification of the roll-out of RBT in Australia, which was introduced in different states at different times, but will provide a sufficient proxy for the purposes of this study. We also control for the effect of World War 2 on accident mortality, as mortality data during these years excluded service personnel and are thus artificially low for young males (as can be seen in the negative associations between WW2 and young male mortality presented below). 7 6 5 Traffic 15-29 yrs Traffic 3-49 yrs Traffic 5-69 yrs Traffic 7 yrs + 3 25 2 Non-traffic 15-29 yrs Non-traffic 3-49 yrs Non-traffic 5-69 yrs Non-traffic 7 yrs + 4 15 3 2 1 1 5 193 194 195 196 197 198 199 2 193 194 195 196 197 198 199 2 l7 l6 l5 Female traffic Male traffic Total traffic Alcohol 14 12 1 8 7 6 Fmale non-traffic Male non-traffic Total non-traffic Alcohol 16 14 12 l4 l3 8 6 5 4 3 1 8 6 l2 4 2 4 l1 2 1 2 l 193 194 195 196 197 198 199 2 193 194 195 196 197 198 199 2 Figure 1 Trend in per capita alcohol consumption (15 years and older, refer to right axis) and, gender and age specific traffic and non-traffic accidents mortality rates (per 1 inhabitants) in Australia 1924-26 Traffic and non-traffic accidents caused by intoxication will not be affected by historical alcohol consumption in the way that chronic mortality outcomes like cirrhosis are. Therefore, no lagged effects are considered in the current study. A autoregressive moving average model (ARIMA), known as Box-Jenkins (Box and Jenkins 197) approach was employed to estimate the associations between alcohol consumption and, traffic and non-traffic accidents mortality in Australia. One-off event dummy variables were included in initial models to assess the impact of World War 2 and introduction of RBT in Australia since 1976 on the fatal traffic and non-traffic accidents. More elaboration about ARIMA models have been presented elsewhere (Skog (21a), Livingston and Wilkinson (213). As the trends in mortality rates were typically non-stationary (as shown in Figure 1), all models were conducted based on differenced data. The model fit was evaluated with the aid of the Box- 3
Ljung portmanteau test of the first 1 autocorrelations, Q(1). The model structures used are reported below, alongside the output of the models. Results The results of the time series analyses between per capita alcohol consumption and traffic and non-traffic accidents mortality are reproduced in Table 1 for males, females and total mortality. The estimates in Table 1 suggest that an increase in 1-litre per capita alcohol consumption was associated with an increase in traffic mortality about 2 per 1 inhabitants. Increase in population drinking level would lead greater effect on male (3.4) traffic mortality than females (.5). The results also suggest that a 1-litre increase in per capita alcohol consumption corresponded to an increase in non-traffic mortality about 3 among males and.9 among females per 1 inhabitants. The estimates of event dummies indicate that both World War 2 and the introduction of RBT since 1976 had generated significant negative effect on the traffic accidents mortality in Australia. Table 1 Estimated effect of alcohol consumption on gender specific fatal traffic and nontraffic accident mortality Male Female Total Traffic accidents Coef. SE Coef. SE Coef. SE Alcohol consumption 3.372** 1.27.534*.353 1.961**.622 WW 2 (1939-45) -3.35** 1.12 -.869**.33-2.13**.613 RBT since 1976-1.171 ( * ).642 -.543**.197 -.83*.389 Constant.551.47.239**.124.387.246 Q (lag 1) 7.644, p=.664 7.915, p=.442 11.515, p=.319 Model specification (,1,) (,1,3) (,1,) Non-traffic accidents Alcohol consumption 3.1**.65.884*.421 1.962**.493 WW 2 (1939-45) -.572.539.351.378 -.166.442 RBT since 1976 1.37.312.33.231.851.265 Constant -1.167**.196 -.227.145 -.697**.166 Q (lag 1) 5.985, p=.741 8.794, p=.457 7.613, p=.574 Model specification (,1,1) (,1,1) (,1,1) *p<.5, **p<.1, ( * ) p<.1. Q-tests for uncorrelated residuals do not suggest unsystematic variation p<.1. In contrast, the impacts of RBT and WW2 on non-traffic mortality were statistically insignificant. The estimated effect of aggregate alcohol consumption on male and female traffic and non-traffic accidents are greatly different (males are nearly four times higher than females). There are two possible explanations for this difference (Skog 21a). First, due to the qualitative differences in drinking patterns, frequencies and drunken comportment, males could experience more accidents per litre of alcohol than females. Second, generally, males drink more than females on average. Hence, a 1-litre increase in per capita alcohol 4
consumption typically implies a much greater increase in male alcohol consumption than in female consumption. The results of time series analysis of fatal traffic and non-traffic accident mortality by age are presented in Table 2. The findings suggest that the effect of alcohol consumption on fatal traffic accident mortality is positive and statistically significant (p<.5) only in young and middle age groups. A 1-litre increase in per capita alcohol consumption led to a rise in traffic accident mortality rates about 4 among 15-29 years and nearly 2 among 3-49 years per 1 inhabitants. Introduction of RBT negatively and significantly affected fatal traffic accidents in nearly all age groups (except the oldest), with particularly large effects for young males (Tables 1 and 2). The estimates for non-traffic accidents show a different pattern, with the highest impact in the older age groups. No significant impact of RBT was found on fatal non-traffic accidents. Table 2 Estimated effect of alcohol consumption on age specific fatal traffic and nontraffic accident mortality 15-29 years 3-49 years 5-69 years 7 years and plus Traffic accidents Coef. SE Coef. SE Coef. SE Coef. SE Alcohol consumption 4.34* 1.317 1.57*.629.971.844 2.229 1.541 WW 2 (1939-45) -3.659** 1.319-1.622**.62-2.147**.744-3.657** 1.367 RBT since 1976-1.373*.846 -.599*.393-1.9*.474-1.422.868 Constant.683 ( * ).538.296 ( * ).249.498 ( * ).298.68.545 Q (lag 1) 4.34, p=.99 6.41, p=.75 7.899, p=.162 8.876, p=.449 Model specification (,1,1) (,1,) (,1,1) (,1,1) Non-traffic accidents Alcohol consumption 1.796**.345 2.281**.533 2.937**.75 8.21 ( * ) 4.233 WW 2 (1939-45) -.542**.189 -.924 ( * ).478 -.151.626 5.323 3.738 RBT since 1976.747.129.854.295 1.243.359 2.715 2.236 Constant -.631**.8 -.668**.185-1.235**.225-2.682 ( * ) 1.42 Q (lag 1) 1.462, p=.234 7.34, p=66 13.985, p=.123 11.582, p=.238 Model specification (1,1,2) (,1,1) (,1,1) (,1,1) *p<.5, **p<.1, ( * ) p<.1. Q-tests for uncorrelated residuals do not suggest unsystematic variation p<.1. Conclusions This study has verified that changes in alcohol consumption have had substantial effects on both traffic and non-traffic accidents in Australia. The findings suggest that the strongest impact on fatal traffic accident mortality rates of a change in population-level alcohol consumption is in relation to male accidents, particularly for 15-29 year olds, with nonsignificant effects in the older age groups. In contrast, the strongest association between alcohol consumption and non-traffic mortality was found in the oldest age groups. This reflects the underlying cause of death distribution, with older Australian more likely to die in non-traffic accidents and younger Australians in traffic accidents. 5
As expected, the impact of the introduction of RBT on the traffic accidents was negative and significant in Australia for both males and females, with particularly large impacts on young drivers. It is worth noting that the effect sizes for RBT produced here are likely to be underestimates, as they estimate the impact of RBT on traffic mortality over and above its impact via changes in total consumption. The findings of this research also suggest that reducing population drinking levels can lead a reduction in traffic and non-traffic accidents mortality rates. It highlights the importance of policies aimed at reducing the total amount of alcohol consumed to reduce alcohol-related harm. Policies that reduce total consumption (e.g. liquor licensing controls and alcohol taxation) are likely to produce reductions in accident mortality. These policies can be complemented by specific approaches designed to reduce accident mortality (e.g. RBT) which, as we have shown here, produce additional reductions in alcohol-related harm. References AIHW. (28). GRIM (General Record of Incidence of Mortality) Books. Australian Institute of Health and Welfare, Canberra. Australian Bureau of Statistics. (211). Apparent Consumption of Alcohol: Extended Time Series, 1944 1945 to 28 29. Australian Bureau of Statistics, Canberra. Box, G. E. P. & Jenkins, G. M. (197). Time Series Analysis, Forecasting and Control. Holden-Day, San Francisco. Henstridge, J., Homel, R. and Mackay, P. (1997). The Long-Term Effects of Random Breath Testing in Four Australian States: A Time Series Analysis. Australia: Federal Office of Road Safety, Canberra. Livingston, M. & Wilkinson, C. (213). Per-capita alcohol consumption and all-cause male mortality in Australia, 1911-26. Alcohol and Alcoholism, 48(2): 196-21. NCETA. (24). Alcohol and Other Drugs: A Handbook for Health Professionals. Australian Government Department of Health and Ageing, Canberra. Norström, T. & Ramstedt, M. (25). Mortality and population drinking: a review of the literature. Drug and Alcohol Review, 24(6): 537-547. Norström, T. & Rossow, I. (213). Population drinking and drink driving in Norway and Sweden: an analysis of historical data 1957 89. Addiction, (In ealy view). Peek-Asa, C. (1999). The effect of random alcohol screening in reducing motor vehicle crash injuries. American journal of preventive medicine, 16(1): 57-67. Ramstedt, M. (28). Alcohol and fatal accidents in the United States A time series analysis for 195 22. Accident Analysis & Prevention, 4(4): 1273-1281. Rehm, J., D. Baliunas, et al. (21). The relation between different dimensions of alcohol consumption and burden of disease: an overview. Addiction, 15(5): 817-843. Rowland, B., J. W. Toumbourou, et al. (212). Reducing alcohol-impaired driving in community sports clubs: evaluating the Good Sports program. Journal of Studies on Alcohol and Drugs, 73(2): 316-327. Skog, O.-J. (21a). "Alcohol consumption and mortality rates from traffic accidents, accidental falls, and other accidents in 14 European countries." Addiction, 96(s1): 49-58. Skog, O.-J. (21b). Alcohol consumption and overall accident mortality in 14 European countries. Addiction, 96(s1): 35-47. Skog, O.-J. (23). Alcohol consumption and fatal accidents in Canada, 195-1998. Addiction, 98(7): 883-893. 6