Contribution of deaths related to alcohol use to socioeconomic variation in mortality: register based follow up studyBMJ 1997; 315 doi: http://dx.doi.org/10.1136/bmj.315.7102.211 (Published 26 July 1997) Cite this as: BMJ 1997;315:211
- a Population Research Unit, Department of Sociology, PO Box 18, 00014 University of Helsinki, Finland
- Correspondence to: Ms Mäkelä
- Accepted 25 April 1997
Objective: To estimate the contribution of excessive alcohol use to socioeconomic variation in mortality among men and women in Finland.
Design: Register based follow up study.
Subjects: The population covered by the 1985 and 1990 censuses, aged ≥20 in the follow up period 1987-93.
Main outcome measures: Total mortality and alcohol related mortality from all causes, from diseases, and from accidents and violence according to socioeconomic position. The excess mortality among other classes compared with upper non-manual employees and differences in life expectancy between the classes were used to measure mortality differentials.
Results: Alcohol related mortality constituted 11% of all mortality among men aged ≥20 and 2% among women and was higher among manual workers than among other classes. It accounted for 14% of the excess all cause mortality among manual workers over upper non-manual employees among men and 4% among women and for 24% and 9% of the differences in life expectancy, respectively. Half of the excess mortality from accidents and violence among male manual workers and 38% among female manual workers was accounted for by alcohol related deaths, whereas in diseases the role of alcohol was modest. The contribution of alcohol related deaths to relative mortality differentials weakened with age.
Conclusions: Class differentials in alcohol related mortality are an important factor in the socioeconomic mortality differentials in Finland, especially among men, among younger age groups, and in mortality from accidents and violence.
Alcohol related deaths constituted 11% of all deaths in Finland among men aged 20 and above and 2% among women; the corresponding proportions were much larger for accidental and violent deaths and smaller for deaths from diseases
Relative socioeconomic differentials were much larger in alcohol related mortality than in overall mortality, the largest rates being among manual workers
Alcohol related mortality accounted for 14% of the mortality differentials between manual workers and upper non-manual employees among men, 4% among women, and 24% and 9% of the differentials in life expectancy, respectively
The role of alcohol in the socioeconomic differentials was modest in deaths from diseases but substantial in accidental and violent deaths—for example, one half of the difference between upper non-manual employees and manual workers in accidental and violent mortality could be attributed to alcohol related deaths
The impact of alcohol on relative socioeconomic mortality differentials increased with decreasing age
In all countries for which data exist, the lower socioeconomic classes have higher mortality than higher classes.1 2 The possible causes of this gradient have been widely discussed,3 4 5 but few quantitive estimates of the contribution of specific factors have been reported.6 7
Alcohol consumption is an important determinant of premature death, particularly among men.8 9 10 Its role has usually been considered only in passing in accounts of causes of socioeconomic differentials in mortality, partly because explanatory studies based on epidemiological data have focused mainly on mortality from coronary heart disease, for which alcohol is not a major risk factor.
Our aim was to give a quantitive estimate of the contribution of excessive alcohol use to socioeconomic variation in mortality among Finnish men and women by category of cause of death. Comprehensive data from the death register that have been linked with census data offer a unique opportunity to study this problem. The data do not include information on consumption, but inferences about the role of alcohol are made on the basis of individual level, as given in death certificates.
Data and methods
The data used in this study were extracted from two large data files compiled by Statistics Finland, which include census records of all people covered by the 1985 and 1990 censuses linked with all death records for the years 1987-90 and 1991-3, respectively, by means of personal identification codes.11 The analysis comprises men and women aged 20 and above in 1987-93. Social class could not be adequately measured for younger Finns in our data.
Deaths caused by excessive alcohol use were defined as those for which there was a reference to alcohol in the death certificate. They are here called alcohol related deaths and are described in detail in the Appendix 1. There were 20 835 such deaths in the data (see table 1). Deaths for which the underlying cause was explicitly attributed to alcohol use constituted 40% of these deaths. In the 60% remaining, at least one of the contributory causes was either alcoholic intoxication or a disease explicitly attributed to alcohol use.
Socioeconomic position was measured by social class obtained from census records. It is based on own occupation for economically active people. Economically inactive family members—for example, housewives—are classified according to the head of the household. Data on earlier socioeconomic position were used for pensioners, students, unemployed people, and men in military service. More details about the classification are given elsewhere.12
Two ways of measuring socioeconomic inequalities in mortality were applied. The first was based on mortality standardised for age and was calculated by using direct standardisation with the combined Finnish male and female populations in 1987-93 as the standard. The excess mortality among other classes compared with upper non-manual employees was defined as RR-1, where RR is the ratio of the rates concerned. Confidence intervals for these ratios were calculated by using a formula given by Rothman.13 The contribution of alcohol related mortality to inequality was obtained by comparing the excess mortality in all deaths and that in deaths not related to alcohol.
The second measure of inequality was the difference in life expectancy at age 20 between upper non-manual employees and other classes. Life tables for the classes were calculated in the conventional manner with five year age groups.14 The reduction in life expectancy due to alcohol related mortality was calculated by using cause elimination life tables.15 The contribution of alcohol to differences in life expectancy was then obtained by comparing the differences between upper non-manual employees and the other social classes in life expectancy and in the reduction in life expectancy due to alcohol.
In the population aged 20 and above, 11% of all deaths in men and 2% of all deaths in women were alcohol related (table 1). The proportion was much larger in deaths from accidents and violence than in deaths from diseases: 47% of the former were alcohol related among men and 17% among women. Among men, the proportion of alcohol related deaths was the largest in manual workers even though their overall mortality was the highest as well. The relative impact of alcohol on mortality was the highest at ages 20-50, whereas absolute mortality related to alcohol use peaked at a later age, closer to 60 years (fig 1).
Among men, alcohol related mortality was almost three times higher among manual workers than among upper non-manual employees, and among women it was more than twice as high (table 2). Lower non-manual employees fell between these two classes, while farmers had a relatively low level of alcohol related mortality, especially among women. Generally, the differentials in alcohol related mortality were substantially larger than the differentials in overall mortality.
When alcohol related deaths were excluded, excess mortality among male manual workers diminished by 14%, whereas among women the decrease was fairly small at 4% (table 2). Alcohol related deaths thus accounted for 14% and 4% of excess mortality among male and female manual workers, respectively, compared with upper non-manual employees. The corresponding attributable proportions were somewhat lower among lower non-manual employees and in the group “others,” whereas among farmers they were negligible or even reversed.
Table 3) shows the differences between upper non-manual employees and other classes in life expectancy at age 20 and in the estimated reduction in life expectancy due to alcohol. Among male manual workers 1.5 years (24%) of the total 6.2 year difference in life expectancy and among women 0.3 years (9%) of the 3.0 year difference were attributable to alcohol related deaths.
Alcohol related mortality from accidents and violence was as much as 3.5 times higher among male manual workers than among upper non-manual employees and 2.7 times higher among women, while the corresponding ratios for diseases were 2.3 and 1.7 (table 4). The exclusion of alcohol related deaths from accidental and violent deaths diminished the excess mortality among manual workers by 49% in men and 38% in women. The corresponding decreases in manual workers' excess mortality from diseases were only a fraction of these figures because of the small proportion of alcohol related deaths among deaths from disease. Among both women and men, two thirds of the excess accidental and violent deaths of lower non-manual employees was attributable to alcohol related mortality.
The impact of alcohol related deaths on socioeconomic differentials from all cause mortality was higher in the younger age groups because of the larger proportions of accidental and violent deaths and of alcohol related deaths (table 5). There was a similar age pattern for deaths from disease. In accidental and violent deaths the effect of alcohol was substantial in all age groups.
In the study period alcohol related deaths constituted over 10% of all deaths among men aged 20 and above and 2% among women. The proportions were considerably larger in accidental and violent deaths than in deaths from disease. Alcohol related mortality was substantially higher among manual workers than among upper non-manual employees. Our results show that alcohol consumption is an important cause of socioeconomic differentials in mortality in Finland, particularly among men and among the young and middle aged population. A large part of the differentials in accidental and violent deaths could be attributed to alcohol related deaths, whereas the influence of alcohol use on the differentials in mortality from diseases was modest. The impact of excessive alcohol use on the differences in life expectancy was even larger than its relative impact on the differences in mortality. This is because life expectancy as a measure of mortality gives greater weight to death at a younger age, and the proportion of alcohol related deaths is larger among younger than older adult age groups.
Generalisability of results
The findings concerning the Finnish population in 1987-93 cannot be generalised to other countries. The contribution of alcohol use to socioeconomic differentials in mortality is likely to be larger or smaller than reported here, depending on factors such as the nature of the differences in alcohol use between socioeconomic groups and the general level of mortality from alcohol related causes.
The only report known to us that estimated the contribution of alcohol to socioeconomic mortality differences concerned Sweden in 1981-6.16 It reported a somewhat smaller effect than here, but the measure used for alcohol related deaths in that study would have covered only about 30% of alcohol related deaths covered by our measure.
The contribution of alcohol use is likely to be relatively small in countries such as the United Kingdom, where the overall mortality from both liver cirrhosis and from accidents and violence is relatively low17 and class differences in mortality from liver cirrhosis seem to be small and irregular.18 On the other hand, the contribution of alcohol is likely to be even larger than in Finland in countries such as France, where the alcohol related mortality is high and differences between occupational classes are large.17 19
Reliability of results
The reliability and the coverage of the information on alcohol related deaths in the Finnish death register were discussed in an earlier report.10 It seemed that the gross underestimation of alcohol related causes of death in death certificates which was common in many other countries was not as severe a problem in Finland. Among the most important reasons was that the death certificate is not a public document in Finland, which practically eliminates social stigma related to alcohol related diagnoses—for example, the proportion of deaths from liver cirrhosis reported to be alcohol related was as high as 90% among men in Finland. It would be lower if strong social stigmatisation occurred. Also, the rate of necropsies is high in Finland, especially in accidental and violent deaths; medicolegal necropsies were carried out in more than 97% of all accidental and violent deaths among people aged under 75 in 1987-93.20 Some underestimation does occur, particularly in alcohol related deaths from such diseases as cancer of the upper aerodigestive tract (oropharynx, larynx, and oesophagus) and stroke. The earlier report suggested that the underestimation of alcohol related deaths from these diseases is about 10% of all alcohol related mortality or almost 30% of mortality from alcohol related disease.10 Thus, the contribution of alcohol to socioeconomic differentials in mortality from disease may be somewhat underestimated here. On the other hand, some deaths not related to alcohol use may have been misclassified as alcohol related—for example, because it is not always possible to assess correctly the causal effect of a raised blood alcohol concentration on the death. The effect of these misclassifications on our results depends strongly on whether they have an association with social status. No evidence about this is available from Finland.
Moderate alcohol consumption has been observed to prevent deaths from coronary heart disease.21 A large survey from the United States showed that “moderate” or “frequent light” drinkers were more common in the upper social classes than in the lower ones, while abstinence was more common and heavy drinking slightly more common in the lower classes.22 The evidence from Finland points in the same direction.23 These results suggest that alcohol consumption may increase socioeconomic variation in mortality not only through more deaths caused by excess alcohol consumption in lower social classes, but also through a greater protective effect of alcohol against coronary disease in the higher classes. The confirmation of this hypothesis, and the estimation of the importance of it in a population, would require a survey on alcohol consumption large enough to estimate the consumption distribution in the population by sex, age, and socioeconomic position.
Alcohol use and socioeconomic variation in mortality
The framework of the Black report has been widely used in studies aiming to explain socioeconomic variation in mortality.3 It divides the explanations into four types: artefact, social selection, cultural or behavioural, and materialist. In this classification, alcohol consumption and other health related behaviours belong to behavioural-cultural explanations. Alcohol consumption and the differentials in it, however, also have their causes, and some of these are “materialist.” Problems in, for example, work or personal finances, the lack of opportunities to pursue leisure time activities, and inadequate parental resources to offer a stimulating environment for children may all be more common in lower socioeconomic groups and may increase the inclination towards heavy drinking. The part of the variation in mortality accounted for by alcohol consumption does not exclude the materialist explanation; excessive alcohol use can be partly seen as a pathway along which mortality and differentials in mortality are affected by material living conditions.
We thank Statistics Finland for providing the data for the analyses (permission TK-53-133-95), and Jari Hellanto for his help with the construction of the dataset.
Funding: Finnish Foundation for Alcohol Studies; the Research Council for Culture and Society of the Academy of Finland.
Conflict of interest: None.
The Finnish Classification of Diseases 1987 (FCD; see table 6) has been used in coding the causes of death.24 It is based on ICD-9 (international classification of diseases, ninth revision), but in part the Finnish classification is more detailed because five digit codes are used, and some categories have codes different from ICD.