Social class, health behaviour, and mortality among men and women in eastern FinlandBMJ 1995; 311 doi: https://doi.org/10.1136/bmj.311.7005.589 (Published 02 September 1995) Cite this as: BMJ 1995;311:589
- Juha Pekkanen, head of departmenta,
- Jaakko Tuomilehto, professorb,
- Antti Uutela, head of laboratoryb,
- Erkki Vartiainen, head of laboratoryb,
- Aulikki Nissinen, professorc
- aDepartment of Environmental Epidemiology, National Public Health Institute, PO Box 95, FIN-70701 Kuopio, Finland
- bDepartment of Epidemiology and Health Promotion, National Public Health Institute, Mannerheimintie 166, FIN-00300 Helsinki, Finland
- cDepartment of Community Health and General Practice, University of Kuopio, PL 1627, FIN-70211 Kuopio, Finland
- Correspondence to: Dr Pekkanen.
- Accepted 26 June 1995
Objective: To evaluate the associations between social class as defined by occupation, health behaviour, and mortality from all causes and coronary heart disease among middle aged men and women in eastern Finland.
Design: Prospective observational study of two independent, random population samples examined in 1972 and 1977.
Setting: North Karelia and Kuopio, Finland.
Subjects: 8967 men and 9694 women aged 30-64 years at the beginning of the follow up study. The subjects were followed up for mortality up till 1987 by using the National Death Registry.
Measurements and main results: Altogether 1429 men and 620 women died during the follow up, 603 men and 164 women of coronary heart disease. Among both sexes, compared with white collar workers unskilled blue collar workers had more adverse risk factors and also higher mortality due to coronary heart disease, other cardiovascular diseases, cancer, violent causes, and all other causes. Among men the age adjusted relative risk for all cause mortality in unskilled blue collar workers v white collar workers was reduced from 1.86 (95% confidence interval 1.55 to 2.22) to 1.47 (1.23 to 1.77) when adjusted for smoking, serum cholesterol concentration, hypertension, body mass index, and physical activity in leisure time. Among women the corresponding reduction in hazard ratio was from 1.49 (1.15 to 1.92) to 1.39 (1.07 to 1.81). The respective hazard ratios for coronary heart disease were 1.54 (1.16 to 2.02) and 1.22 (0.92 to 1.61) among men and 1.74 (1.05 to 2.90) and 1.66 (0.99 to 2.79) among women.
Conclusions: Unfavourable cardiovascular risk factors and high mortality are concentrated among lower social classes in Finland. Among men about half of the excess coronary and all cause mortality among unskilled blue collar workers was associated with their unfavourable risk factor profile. The association was smaller in women.
No previous study has assessed what proportion of the excess coronary mortality among women in lower social classes is associated with risk factors
In this Finnish study about half of the excess mortality among men in lower social classes but a clearly smaller proportion among women was associated with their more adverse cardiovascular risk factor profile
It may be possible to narrow differences in health between social classes through improvements in health behaviour, especially among men
This by itself will not be sufficient, but the increasing health inequalities need to be dealt with at multiple levels, from general social policy to everyday practice
Differences in mortality between the social classes are clearly larger in Finland than in other Nordic countries and also possibly somewhat larger than in Great Britain.1 2 During the 1980s the relative differences in mortality between the social classes have increased in Finland.3 4 Social class measured in middle age and mortality from future coronary heart disease are interwoven in a complex network of influences, and more adverse cardiovascular risk factors among the lower social classes explain only part of their higher mortality.5 6 7 8 9 10 11 12 13 14 Other factors include current material conditions, social mobility, psychosocial factors, and environmental influences as well as health during pregnancy and childhood.15 According to most previous analyses7 8 9 10 11 12 14 about one quarter to one half of the excess coronary risk among men in lower social classes is associated with well known, preventable cardiovascular risk factors. None of these previous analyses has, however, studied women. In the present article we examine the associations between social class, as defined by occupation, cardiovascular risk factors, and mortality among 18661 middle aged men and women in eastern Finland.
Study population and methods
In connection with the North Karelia Project independent 6.6% random samples of men and women born in 1913-47 and living in the counties of North Karelia and Kuopio were invited for examination in 1972 (men and women aged 25-59) and in 1977 (men and women aged 30-64).16 The only larger urban area in North Karelia is Joensuu, a town of 50000 inhabitants. An additional 6.6% random sample of those living in Joensuu was also examined and added to the present analyses to increase sample size among non-farmers. Participation rates were 92.4% in 1972 and 89.6% in 1977.17 Smoking, physical activity in leisure time, and use of antihypertensive drugs were determined from a self administered questionnaire. Blood pressure was measured once with subjects seated, and serum cholesterol concentration was analysed with the Liebermann-Burchard method.
Occupational status was obtained through record linkage with the 1970 population census for those examined in 1972 and with the 1975 population census for those examined in 1977. Subjects were divided into four social classes on the basis of the 21 grade socioeconomic variable in the population census database by following the classification used by Valkonen et al.3 The four social classes used were (a) white collar workers--for example, those with university degrees, technicians, nurses, book keepers, cashiers, bank clerks, sales staff, telephone operators, noncommissioned officers; (b) skilled blue collar workers--for example, postal workers, janitors, railway staff, workers in mining and manufacturing work (except assistant workers), fire fighters, cooks, waiters, hairdressers; (c) unskilled blue collar workers--for example, forestry workers, assistant workers, kitchen hands, cleaners, nursery nurses, messengers, lift attendants; and (d) farmers--own account workers at farms and agricultural employers.
Pensioners were classified according to their past occupation. Employers and own account workers in occupations other than agriculture, students, pensioners with missing past occupation, and others with missing data on socioeconomic status were excluded from the analyses (n=1355). Housewives and others without present or past occupation were classified according to the head of the household. The final analysis included 8967 men and 9694 women aged 30 to 64 years. None of the subjects had missing data on smoking.
Follow up of mortality in the study population was achieved through record linkage with the National Death Registry. By the end of 1987 there had been 2049 deaths, 767 due to coronary heart disease (International Classification of Diseases (eighth revision) codes 410-414).
Hypertension was defined as systolic blood pressure >159 mm Hg or diastolic blood pressure >94 mm Hg or use of antihypertensive drugs during the past week. Smokers who had quit smoking during the past six months were classified as current smokers. Obesity was defined as body mass index exceeding 27 kg/m2. Physical activity in leisure time was determined by using a four level question describing the usual intensity of physical activity in leisure time. Those subjects who mostly read, watched television, or did other non-strenuous activities were classified as having little physical activity in leisure time. Those reporting as having been unemployed most of the year were classified as unemployed.
Means were age standardised by analysis of covariance and proportions and mortality rates by the direct method by using study population as the standard. Differences in risk factors attributable to social class were tested by using logistic regression models for categorical variables and analysis of covariance for continuous variables. Mortality was analysed with Cox's proportional hazards model.18 Statistical tests for differences in social class were global tests for heterogeneity with 3 df, except in tables IV and V, in which tests for trend are also presented. Farmers were excluded from these tests for trend.
As the associations for social class with mortality among both sexes were similar in the two cohorts and in the two areas, only pooled estimates are presented. Area and cohort were adjusted for by stratification in the Cox's models. Inspection of the residuals of the final Cox's models revealed no outliers and no departures from the proportionality assumption for social class.
In the mortality analysis serum cholesterol concentration, body mass index, and square of body mass index were used as continuous variables; smoking was divided into three categories (as in table VI), and other variables were dichotomised.
The largest social class among men was skilled blue collar workers, but among women it was white collar workers (table I). Differences in education were large between the social classes. Among men, unskilled blue collar workers were more often unmarried and unemployed than men in other classes. There were also large differences in car ownership. Among women these differences were smaller.
Levels of all risk factors, except obesity among men and smoking among women, changed to a less favourable direction from white collar workers to unskilled blue collar workers (table II). Among men, farmers in general occupied an intermediate position but had the most unfavourable mean concentrations of serum cholesterol and physical activity in leisure time. Among women, farmers had the lowest prevalence of smoking but had unfavourable levels of other risk factors.
All cause mortality was highest among unskilled blue collar workers of both sexes (table III). Relative differences in mortality due to cancer and violent causes of death between social classes were larger among men than among women, otherwise differences were similar. Mortality among farmers was intermediate to that in white collar and unskilled blue collar workers.
In age adjusted analyses unskilled blue collar workers had 54% higher coronary mortality than white collar workers (hazard ratio 1.54; table IV). About one third of this excess mortality was associated with smoking. When serum cholesterol concentration and prevalence of hypertension were further adjusted for, the excess mortality among unskilled blue collar workers was reduced to 20%. Further adjustment for obesity and physical activity in leisure time had little effect. Among women adjustment for smoking widened the social class differences, especially for the female farmers. Further adjustment for other risk factors reduced the differences slightly below the age adjusted figures. Further adjustment for height, in addition to all the other risk factors analysed, reduced the hazard ratio in unskilled blue collar workers v white collar workers to 1.18 (95% confidence interval 0.89 to 1.57) among men and to 1.57 (0.93 to 2.66) among women.
Among men differences in all cause mortality between social classes were larger than in mortality from coronary heart disease (table V). About one third of the excess mortality among the unskilled blue collar workers was associated with smoking. Only a minor proportion of the excess mortality among women in lower social classes was associated with the risk factors examined. Further adjustment for height, in addition to all the other risk factors analysed, reduced the hazard ratio in unskilled blue collar workers v white collar workers to 1.42 (1.18 to 1.71) among men and to 1.29 (0.99 to 1.69) among women.
The effects of smoking and social class on all cause mortality were also examined in a stratified analysis (table VI). Given the wide confidence intervals in several strata the associations of social class with all cause mortality were in general similar to the estimates in table V. The hazard ratios, however, tended to be smaller among men who were former smokers. Also, when we calculated risk ratios for smoking were calculated from the table, former smoking proved more harmful among white collar men (risk ratio 1.56) than among skilled (1.30) or unskilled (0.96) blue collar workers, which suggests residual confounding by socioeconomic status.
The higher mortality due to most causes in the lower social classes has been shown repeatedly both in Finland3 4 6 as in other industrialised Western countries.19 20 Based on our results about half of the excess mortality among male unskilled blue collar workers in eastern Finland was associated with their more adverse cardiovascular risk factor profile. Among women a clearly smaller proportion of the association between social class and mortality was associated with the risk factors analysed.
Several previous prospective studies have examined this issue among middle aged men but none among women. In the Whitehall study of British civil servants the excess risk of coronary death in the lowest compared with the highest grade of employment was 170% (relative risk 2.7) when adjusted for age and 110% (relative risk 2.1) when adjusted also for other cardiovascular risk factors. Therefore, about one third of the excess mortality due to coronary heart disease in the lowest grade was associated with their more adverse risk factor profile.7 10 Similarly, an estimate of about one half can be derived from the population based British Regional Heart8 and Oslo studies5 and a study of United States employees,11 an estimate of about one third from a study in the former Soviet Union12 and from a cross sectional study in Scotland,14 and an estimate of about one quarter from a study in Sweden.9 In contrast, a Danish study suggested a very small effect of smoking,13 which in other studies has been the dominant factor.
Taken together the results from most of these studies are remarkably consistent given the small number of deaths in several studies and the differences in the association of social class with risk factors over time21 and place. For example, in the older British studies average serum cholesterol concentrations were higher in upper social classes7 8 in contrast within the present and other Scandinavian studies5 9 22 23 and the more recent Scottish study.14 It should therefore be clearly acknowleged that the associations of social class with risk factors and mortality are not universal but may change in time and place. Given this, it is surprising that several articles do not even mention the date of their baseline examination.
Among women adjustment for smoking increased the observed differences in social class especially among the female farmers, who rarely smoked. Overall, among women the excess mortality in the lower social classes had clearly less association with the risk factors studied than among men. Today in Finland the risk factors studied may explain more of the social class differences among women as currently more Finnish women in the lower classes smoke compared with women in the higher social classes.21 23 We also did not have information on the women's high density lipoprotein cholesterol concentrations, which may be a more important risk factor than total cholesterol among women.24
The clear social class differences in height were also able to explain part of the differences in mortality. Height, however, reflects factors operating in early life that are no longer modifiable in adulthood.15
In the present analyses women's social class was primarily determined on their own present or past occupation not the husband's, as has been done in most studies.25 In the beginning of the 1970s about 60% of middle aged women were working in their own occupation in the present study areas.26 27 Also, previous analyses from Finland have shown that classification by either own or spouse's occupation has little effect on the occupational mortality differentials.28 Therefore, misclassification of the women's occupation is an unlikely explanation for the present findings. Farmers in the present classification are a very heterogenous group, and they cannot easily be placed in the more or less hierarchical classification of workers.2 Therefore farmers were excluded from the analyses with trend test but included in the overall comparisons.
Mortality was predicted by using factors which exert their influence at different levels of the hypothetical causal pathway that finally leads to death. When all these factors are entered in the analysis at the same time their relative importance is strongly influenced by the degree of measurement error in each variable.29 This makes it very hard to respond to the question of whether cardiovascular mortality is more strongly associated with social class than lifestyle. Such a distinction is not even very helpful as the socioeconomic status of a person and his or her lifestyle are so tightly linked.
The present estimate of the proportion of the association between social class and mortality that is mediated through risk factors may be either too low or too high. Including other risk factors, like haemostatic factors,30 in the analysis could have increased the power of the risk factors to “explain” mortality differences between social classes. Also having repeated measurements of the risk factors would have strengthened their association with mortality.31 On the other hand, we used only a single, broad measure of socioeconomic status. Therefore the social classes used are heterogeneous in respect to socioeconomic position and, for example, smoking can be related to socioeconmic position also within the social class categories.32 This suggestion is also supported by the present results. Use of several measures of socioeconomic position10 or a finer classification would probably have strengthened the association between socioeconomic status and mortality.
A sizeable part of the excess mortality in lower social classes seems to be mediated through the known cardiovascular risk factors. Therefore, it may be possible to narrow social class differences in health through improvements in health behaviour among the lower social classes. In accordance with the Health for All 2000 programme of the World Health Organisation one of the major aims of health policy in Finland is to narrow inequalities in health.33 Relative differences in mortality between social classes have, however, increased in Finland in the 1980s3 4 as in some other Western industrialised countries.19 34 At the same time educational differences in obesity in eastern Finland have increased among both sexes, and smoking among women has changed from a habit of the more educated in the 1970s to a habit of the less educated in the 1980s.21 In contrast, differences in serum cholesterol concentration and blood pressure have remained stable. This indicates that social class differences in mortality may not diminish in the future in Finland.
To decrease the health inequalities between socioeconomic groups present health promotion methods should be reformulated to take into account the needs and abilities of the lower social classes. This by itself will not be sufficient, but the increasing inequalities in health need to be dealt with at multiple levels, from general social policy to everyday practice.35 36
Conflict of interest None.