IJE Advance Access originally published online on February 28, 2005
International Journal of Epidemiology 2005 34(4):765-771; doi:10.1093/ije/dyi025
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Special Theme: Mental Health |
Temporal variation in deaths related to alcohol intoxication and drinking
1 National Research and Development Centre for Welfare and Health (STAKES), Helsinki, Finland
2 Population Research Unit, Department of Sociology, University of Helsinki, Helsinki, Finland
3 International Centre for Health and Society, Department of Epidemiology and Public Health, University College London Medical School, London, UK
* Corresponding author. STAKES, PO Box 220, FIN-00531 Helsinki, Finland. E-mail: pia.makela{at}stakes.fi
 |
Abstract |
|---|
 Top Abstract DataMethodsResultsDiscussionConclusionsReferences |
|---|
Background Temporal variation in deaths related to alcohol intoxication is examined using two approaches. First, we examine the risk of these deaths during festivals, on the day preceding them, and on the three days that immediately follow them. Second, we assess the weekday variation in deaths, and compare this with survey-based data on weekday variations in drinking. Previously no data existed on the temporal association between intoxication-related deaths and drinking occasions according to the severity of intoxication.
Methods We used population registration data on 15–69-year-old men and women, linked with the national cause of death register for the years 1987–2001, and the Finnish Drinking Habits Survey carried out in 2000. Intoxication-related deaths were defined on the basis of underlying and contributory causes of death.
Results The largest increased risk of intoxication-related deaths was observed for Midsummer Day [Observed deaths/Expected deaths (O/E) = 2.88 (95% confidence interval 2.48–3.31) for men and O/E = 2.21 (1.43–3.27) for women respectively], Midsummer Eve [O/E = 2.70 (2.32–3.12) and 3.18 (2.23–4.41)], May Day [O/E = 1.80 (1.50–2.16) and 2.65 (1.79–3.79)], Christmas Eve [O/E = 1.58 (1.29–1.91) and 2.21 (1.43–3.27)], and New Year's Day [O/E = 1.48 (1.20–1.80) and 1.77 (1.08–2.74)]. Among men, the increased risk at Midsummer lasted for three consecutive days. The weekday distribution of different levels of intoxication and of intoxication-related deaths was similar, with a clear increase observed on Friday, Sunday, and, particularly, Saturday.
Conclusions Intoxication-related deaths peak during weekends and around festival days when alcohol is widely consumed in excess. Public awareness of the risks attached to binge drinking should be increased.
Keywords Temporal variation, alcohol related deaths, intoxication
Accepted 20 December 2004
Alcohol intoxication strongly increases the risk of accidental deaths and growing evidence indicates that it is also causally associated with the risk of dying from cardiovascular diseases.1 In most countries, intoxication-related deaths can only be identified in case studies where the number of such deaths is small and hence, the range of research questions that can be studied is limited. In Finland, the routine nationwide registration of alcohol-related deaths, particularly accidental deaths, makes it possible to obtain extensive data on intoxication-related deaths (fatal alcohol poisonings and deaths for which alcohol intoxication is a contributory cause) that can be used flexibly for research purposes.2 Here we examine the temporal variation of deaths related to alcohol intoxication.
Previously, Poikolainen, Leppänen, and Vuori3 used time series analysis to examine the temporal variation of fatal alcohol poisonings (that make up altogether about one-quarter of all intoxication-related deaths1) in Finland. They found that these deaths were 2.2-fold on May Day compared with the average outside public holidays, 1.9-fold on Midsummer Day, and 1.6-fold on Christmas Day.
Arfken examined the temporal pattern (7 weekdays x 7 time periods within each day) of drinking and related this to the temporal pattern of different types of traffic accidents.4 Information on the involvement of alcohol in the accidents was available only occasionally, and the data did not allow the researchers to examine the temporal variation in different quantities drunk. According to the results, both drinking and accidents peaked on Friday and Saturday evenings.
Cherpitel et al. reported alcohol involvement in an emergency-room study carried out in Contra Costa, California and in Trieste, Italy.5 In Contra Costa, but not in Trieste, those sampled on weekend evenings were more likely to have positive breathalyser readings and to report drinking prior to the event than those sampled on weekdays.
A temporal pattern of alcohol-associated cardiac rhythm disorders, with a peak after weekends and holidays, was reported in 1978 in New Jersey.6 Additionally, a series of studies reported the weekday variation in deaths due to cardiovascular diseases, alcohol poisoning, accidents, and violence in Russia and Scotland—countries with a heavy weekend binge-drinking pattern. Alcohol poisoning, accidents, and violence have been reported to peak particularly on Saturdays and Sundays, cardiovascular diseases additionally on Mondays.7,8,9
In this paper, we consider two different viewpoints to study the temporal pattern of intoxication-related deaths. First, we examine how big an increase is observed in such deaths on festival days and in the days that immediately follow. On these days people are more engaged with drinking than during ordinary weekends, and very heavy drinking occurs more often. In the week preceding the festivals, alcohol sales have been reported to double for May Day, triple for Midsummer Day, and triple for Christmas Day compared with an average week (internal report by Alko, the alcohol monopoly of Finland, in 1991).
Second, we assess the weekday variation (outside festival seasons) in deaths caused by intoxication, and compare this with survey-based data on the weekday variation in drinking. Previously, information on drinking or not drinking on different weekdays has mainly been used. Here, we can differentiate between drinking occasions with differing degrees of intoxication.
We will also look at the temporal pattern of non-alcohol-related deaths in order to ensure that the pattern is specific to intoxication-related (or more generally alcohol-related) deaths. Additionally, we examine whether the temporal pattern of intoxication-related deaths varies with age.
|
Data |
|---|
|
TopAbstractDataMethodsResultsDiscussionConclusionsReferences |
|---|
Register data on intoxication-related deaths
This study linked death records in the period 1987–2001 to population registration data on Finnish men and women aged 15–69 years at baseline at the end of 1986. During this period, intoxication-related mortality among men was stable while it increased somewhat among women. Deaths, were classified according to the Finnish edition of the 9th version of the International Classification of Diseases and Causes of Death (FICD9). Less than 0.3% of deaths could not be linked to population registration records. Altogether, men in our study population lived about 24.75 million person years, and experienced about 233 000 deaths, of which 24 509 were related to alcohol intoxication. The corresponding figures among women were 25.93, 140 000 and 4113. Statistics Finland carried out the linkage of datasets by means of personal identity codes.
Data on deaths related to alcohol intoxication were derived from the national cause of death register maintained at Statistics Finland. These consist of two main categories of causes of death: fatal alcohol poisonings and deaths where the certifying doctor has diagnosed alcohol intoxication (code 3050A in FICD9) as a contributory cause of death. The temporal pattern of these two causes was roughly similar and hence they were combined in these analyses. When the contributory cause of death is alcohol intoxication, the underlying cause of death has been reported to be a somatic disease, most often a disease of the circulatory system, in one-fifth of the cases, and either accidental (
40%; with drownings and water traffic accidents, motor vehicle accidents, fire/flames/heat/cold, and falls as the most common categories) or violent (30% suicide, 6% homicide) in four-fifths of the cases.2 In 1995, the last year for which data is available, 97.9% of all accidental and violent deaths in the 0–64 age group went through a medicolegal autopsy.10 This and many other factors in the procedure of certifying the cause of death makes the alcohol-related cause of death diagnoses more reliable in Finland than in most other countries (for more detail see ref. 2).
We defined non-alcohol-related deaths as all deaths excluding (i) deaths related to alcohol intoxication as defined above, and (ii) deaths related to alcohol in other, more chronic ways (e.g. cirrhosis of the liver or alcohol dependence as an underlying or contributory cause).
Finnish drinking habits survey
Data on drinking patterns came from the Finnish Drinking Habits Survey carried out from September to October 2000 by face-to-face interviews. The sample was a simple random sample (n = 2500, response rate 78%) of Finns aged 15–69 years, excluding the Åland Islands, the homeless, the institutionalized and people who were known not to reside in their official place of residence (the three latter groups make up 1.5% of the total population). Weights calculated by post-stratification for sex, age, and geographical region were used in order to restore the population representation of the respondent sample.
The data on drinking patterns used here are based on detailed questions on the drinking patterns of the respondents in the 7 days preceding the interview. We derived the volume of consumption (in centilitres of 100% alcohol) on the basis of the reported beverages consumed. By additionally using information on the length of the drinking occasion and the respondent's weight, the blood alcohol level (BAC) was derived.11 Of the respondents 1157 reported 2162 drinking occasions. We used data on 1922 drinking occasions for which the weekday and BAC estimate is available. The weekly rhythm of drinking has been relatively stable over the past few decades.12 Weekday was defined here as the day of the week when the drinking occasion ended, because the risk of fatality is expected to increase towards the end of the drinking occasion, which is often after midnight (the day following the initiation of the drinking occasion). Drinking occasions were divided into light (BAC < 0.05%), moderate (BAC 
0.05 but <0.15%), and heavy (BAC 0.15%). Originally, the two latter categories were further divided into two, but due to the similarity of the weekday variation in these drinking pattern categories, they were combined. In the heaviest category, 61% of occasions were in the BAC range 0.15–0.25, 20% were in the range 0.25–0.35, and 18% had an estimated BAC 0.35.
|
Methods |
|---|
|
TopAbstractDataMethodsResultsDiscussionConclusionsReferences |
|---|
We calculated mortality rate ratios (O/E) for the days preceding festivals (Christmas, New Year, May Day, and Midsummer) and the following four days by dividing the observed number of intoxication-related (non-alcohol-related) deaths (O) by the expected number of intoxication-related (non-alcohol-related) deaths (E) for men and women separately. 95% confidence intervals were based on the Poisson distribution13. The expected number of intoxication-related deaths on festival days and succeeding days were calculated as total observed intoxication-related deaths in 1987–2001 divided by 365.25.
Similar rate ratios were calculated for weekdays. For weekdays the expected number of intoxication-related deaths was calculated as total observed intoxication-related deaths in 1987–2001 divided by the number of specific weekdays in 1987–2001. In these calculations the summer holiday season (dates 1.6–31.8, including Midsummer) and other festivals [(Christmas and New Year (23.12–1.1), May Day (30.4–1.5), and Easter (from Maundy Thursday to Easter Monday)] were excluded from the numerator and denominator.
|
Results |
|---|
|
TopAbstractDataMethodsResultsDiscussionConclusionsReferences |
|---|
Festivals
Among men, Midsummer Day and Midsummer Eve showed the highest increase in intoxication-related deaths, followed by May Day and Christmas Eve (Table 1). Only at Christmas was the increase higher on the day preceding the festival than on the festival day itself. This may reflect the fact that in Finland, Christmas Eve is widely considered as a bigger festival day than Christmas Day. The increased risk at Midsummer could be observed on three consecutive days, from Midsummer Eve to the first day after Midsummer Day. Among women, Midsummer Eve showed the highest increased risk, followed by May Day, Christmas Eve, and Midsummer Day (Table 2). Unlike men, women did not show increased risk on the days following Midsummer Day.
|
|
The temporal pattern observed is specific to deaths related to alcohol intoxication (Tables 1 and 2). The rate ratios for non-alcohol-related deaths were only marginally above 1 (1.03–1.08). When festival days were analysed separately (results not shown in tables), only the non-alcohol-related rate ratios for Christmas Eve and the days following it were statistically significant (point estimates 1.13–1.15 for Christmas Eve, Christmas Day and the two days after Christmas Day).
Because the temporal pattern of intoxication-related deaths was similar among men and women, the potential age differences were studied with data for men and women combined (Table 3). The estimated increased risk on festival eves and festival days was highest in the youngest age group (15–34 years), but the differences between the age groups were not very big. The examination of individual festival days (results not shown in tables) did not change this result substantially.
|
Weekday variation
Among both men and women, deaths related to alcohol intoxication were at their lowest on Mondays, were clearly increased on Fridays and Sundays, and peaked on Saturdays (Figures 1 and 2). These three days covered 51% of all intoxication-related deaths among men and 53% among women. The increase in deaths compared with what was expected on the basis of a uniform distribution was statistically significant on Friday, Saturday, and Sunday for men and on Saturday and Sunday for women. On Saturday the observed number of cases was about 1.5–1.6-fold as compared with the expection. This temporal pattern was specific to deaths related to alcohol intoxication: the rate ratio for non-alcohol-related deaths varied randomly between 0.98 and 1.01.
|
|
The weekday distribution of drinking was very similar to that observed for deaths caused by intoxication (Figure 1), with most drinking occurring on Fridays, Sundays, and, particularly, Saturdays (when the weekday at the end of the drinking occasion was used). This was true for the weekly volume of drinking, as well as for light (BAC < 0.05%), moderate (0.05% < BAC < 0.15%), and heavy (BAC > 0.15%) drinking occasions. The heaviest drinking occasions were most concentrated on weekends: 55%, 75%, and 86% of the light, moderate, and heavy drinking occasions, respectively, took place on the three weekend days among men; among women the corresponding proportions were 68%, 78%, and 85%.
On average there were 7.2 intoxication-related deaths among men and women on an ordinary Saturday. This number of intoxication-related deaths was exceeded on Christmas Eve (8.7), New Year's Day (7.9), May Day (10.1), Midsummer Eve (14.5), and Midsummer Day (14.5). The weekday variation of deaths related to intoxication and drinking were rather similar in the 15–34, 35–49, and 50–69 age groups (Figure 3). Drinking and intoxication-related deaths appear to be marginally more evenly distributed in the oldest age group than in the other age groups. In contrast, drinking and deaths seem to be slightly more concentrated on Sundays (including Saturday night after midnight) in the youngest age group.
|
|
Discussion |
|---|
|
TopAbstractDataMethodsResultsDiscussionConclusionsReferences |
|---|
Summary of results
This study showed that there is a clear peak in intoxication-related deaths on those festival days and the days preceding them when alcohol is traditionally consumed in large quantities and also during weekends when different types of drinking occasions are more common than during the week. The elevated risk at the most risky festivals clearly exceeded that on an ordinary weekend day. The temporal pattern observed was similar across different age groups and it was specific to deaths related to alcohol intoxication—a temporal pattern was not observed for non-alcohol-related deaths.
Intoxication-related deaths during festivals
Previously, Poikolainen et al.3 reported that the increased risk of fatal alcohol poisoning was the highest on May Day, followed by Midsummer, and Christmas. We show that for all intoxication-related deaths Midsummer (Eve and Day) was the most risky time, followed by May Day and Christmas Eve. At Midsummer, there was an excess of men's intoxication-related deaths for three consecutive days. In Finland, summer holidays often start at Midsummer, and this combination of a big drinking festival with no obligation to return to work seems to be lethal for many Finnish men.
In most cases, (Christmas being the most notable exception) there were more deaths on the festival day itself and not the day preceding the festival, even though it is the eve when most drinking usually takes place in Finland. An apparent explanation for this is that the most severe state of intoxication is usually reached only after midnight and deaths that take place in the early hours of the morning are counted as occurring on the festival day.
Weekday variation in intoxication-related deaths
In this study, it was possible for the first time to examine the temporal covariation of intoxication-related deaths and drinking occasions taking into account the amount of alcohol drunk or, to be precise, taking into account the severity of intoxication. The results showed that a similar proportion of light drinking occasions as of intoxication-related deaths was concentrated at weekends, while this concentration at weekends was clearly higher for moderate and heavy drinking occasions.
It is, however, unlikely that this strong association between intoxication-related deaths and light drinking correctly reflects the true causal contribution of the different drinking levels on intoxication-related deaths. It is true that lighter drinking occasions are much more numerous than heavier drinking occasions (62%, 25%, and 13% among men for the light, moderate, and heavier drinking categories used above, and 78%, 20%, and 3%, respectively, among women), but the risk of fatalities attached to the higher intoxication levels reached in heavier drinking occasions would seem to more than compensate for the differences in prevalence. Evidence from other studies indicates that a BAC of 0.05% is associated with a relative risk of mortality that varies approximately between 1 and 3.5 compared with a risk of 1.0 at a BAC of 0%. Correspondingly, a BAC of 0.1% is associated with a relative risk that is between 2.5 and 13 and a BAC of 0.15 is associated with a relative risk between 11 and 4514 (as estimated from a graph).
One potential explanation for this discrepancy is that injury and the death resulting from it do not always occur on the same day, but the distribution of the day of death presumably has a long tail, with deaths occurring less and less frequently the more time elapses after the injury. This process would even out the temporal distribution of intoxication-related deaths compared with intoxication-related injuries. However, for this to be an important explanation, one would expect the mortality rate on Mondays to be much higher than we observed, because many deaths resulting from injuries that take place during the weekend would occur then.
The homeless and institutionalized could not be included in the survey data used, whereas their intoxication-related deaths are included in our material. According to a study of drinking habits among alcoholics,15 weekend days were by far the most common drinking days among alcoholics. However, long drinking bouts, which were more common among the subjects more prone to alcoholism, who are also more likely to be homeless or institutionalized, were observed to be less bound to weekly and working rhythms. This could be one reason for the fact that intoxication-related deaths were somewhat less concentrated at weekends than might have been expected on the basis of the distribution of heavy drinking occasions reported in the survey.
|
Conclusions |
|---|
|
TopAbstractDataMethodsResultsDiscussionConclusionsReferences |
|---|
This study confirmed the expectation that intoxication-related deaths peak during the times when alcohol is consumed the most: during weekends and festivals. These results should form the basis for information campaigns, particularly before festivals, to increase the awareness of the risks attached to unrestrained drinking.
KEY MESSAGES
|
|
Acknowledgments |
|---|
Academy of Finland has supported Pekka Martikainen (grant 41498, 70631 and 48600) and Pia Mäkelä (grant 200852). We are grateful to the Central Statistical Office of Finland for permission (TK-53-1591-02) to use the data.
|
References |
|---|
|
TopAbstractDataMethodsResultsDiscussionConclusionsReferences |
|---|
1 Poikolainen K. It can be bad for the heart, too—drinking patterns and coronary heart disease. Addiction 1998;93:1757–59.[CrossRef][ISI][Medline]
2 Mäkelä P. Alcohol-related mortality by age and sex and its impact on life expectancy: estimates based on the Finnish death register. Eur J Public Health 1998;8:43–51.
3 Poikolainen K, Leppänen K, Vuori E. Alcohol sales and fatal alcohol poisonings: a time-series analysis. Addiction 2002;97:1037–40.[CrossRef][ISI][Medline]
4 Arfken CL. Temporal pattern of alcohol consumption in the United States. Alcohol Clin Exp Res 1988;12:137–42.[CrossRef][ISI][Medline]
5 Cherpitel CJ, Flaminio D, Poldrugo F. Alcohol and casualties in the emergency room: a US–Italy comparison of weekdays and weekend evenings. Addict Res 1993;1:223–38.
6 Ettinger PO, Wu CF, De La Cruz C, Weisse AB, Ahmed SS, Regan TJ. Arrhythmias and the ‘Holiday Heart’: Alcohol-associated cardiac rhythm disorders. Am Heart J 1978;95:555–62.[CrossRef][ISI][Medline]
7 Chenet L, McKee M, Leon D, Shkolnikov V, Vassin S. Alcohol and cardiovascular mortality in Moscow; new evidence of a causal association. J Epidemiol Community Health 1998;52:772–74.[Abstract]
8 Chenet L, Britton A, Kalediene R, Petrauskiene J. Daily variations in deaths in Lithuania: the possible contribution of binge drinking. Int J Epidemiol 2001;30:743–48.
9 Evans C, Chalmers J, Capewell S et al. ‘I don't like Mondays’—day of the week of coronary heart disease deaths in Scotland: study of routinely collected data. BMJ 1993; 320:218–19.
10 Statistics Finland. Causes of Death 1995. Health. Helsinki: Statistics Finland, 1996:5.
11 Finnish drinking habits. Results from interview surveys held in 1968, 1976 and 1984. Simpura J (ed.) Helsinki: Finnish Foundation for Alcohol Studies, 1987.
12 Simpura J, Metso L. Juomisen tunnit [The hours of drinking, in Finnish]. Alkoholipolitiikka 1995;60:231–44.
13 Statistics with confidence: confidence intervals and statistical guidelines. Gardner MJ, Altman DG (eds). London: BMJ, 1989.
14 Romelsjö A. Alcohol consumption and unintentional injury, suicide, violence, work performance, and inter-generational effects. In: Holder HD, Edwards G (eds). Alcohol and Public Policy: Evidence and Issues. Oxford: Oxford University Press, 1995, pp. 114–42.
15 Ahlström-Laakso S. Drinking habits among alcoholics. Helsinki: The Finnish Foundation for Alcohol Studies, Vol. 21, 1975.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||