Social position and mortality from respiratory diseases in males and females

Social position is known to be associated with mortality from a number of diseases and studies indicate that the social inequalities in health are increasing 1. Although there are strong indicators that the social differences in respiratory diseases are considerable 2, 3, few studies have focused on this disease entity using mortality as an outcome. Social gradient in smoking and occupational exposure to airborne pollutants explains only some of the association, and further exploration is called for.

Studying the association between social position and morbidity is complex because social position includes several dimensions. The choice of which measures to use depends on theories about pathways between social conditions and disease outcome. These pathways are not mutually exclusive and the use of several indicators of social position in studies of social inequalities in health, have been recommended.

In a previous study, the authors have shown that income and education are associated with lung function independently of smoking in both females and males. Subsequent risk of admission to hospital for chronic obstructive pulmonary disease (COPD) was three-fold higher for the lowest socioeconomic group compared with the highest 4. The present study explores the association between various measures of social status and mortality from respiratory disease and from COPD in 26,392 randomly sampled males and females followed prospectively for a mean of 12 yrs.

Methods

Results

Information from Statistics Denmark was obtained for 26,928 participants. The remaining 1,203 subjects were lost due to death before 1980 (n=1080), emigration (n=22), or no registered address in Denmark in 1980 (n=101). A further 536 did not report their level of education, leaving 26,392 (13,992 females and 12,400 males) for analyses. During follow-up there were 512 deaths from respiratory disease (226 females, 286 males); 314 were from COPD. Baseline data by level of education is shown in table 1⇓. Subjects, with higher levels of education tended to be younger, male, nonsmokers, and from the CCHS.

Results from univariate analyses of each indicator of social position on risk of mortality from pulmonary disease are given in table 2⇓. In females, only education was significantly associated with mortality. Since females are more often supported economically by their spouse, individual income was stratified by whether persons were living alone, but this did not alter the results (not shown). In males, education, housing, employment grade, and income were all significantly associated with respiratory mortality. Education and central heating were the only socioeconomic measures with similar strength of association in males and females.

Associations were explored further with adjustment for smoking and cohabitation yielding the results given in table 3⇓. Among females, only education was significantly associated with respiratory mortality. Among males, only central heating did not retain statistical significance and was omitted from the model. Although associations were attenuated, they all retained statistical significance after adjustment for smoking. Relative risks (RRs) associated with smoking were higher in females than in males.

Education was the only socioeconomic measure with similar strength of association in males and females and was explored further. When testing model assumptions, the authors found a lack of proportionality for hazard associated with education in females. This could be modelled by age-specific risk estimates for education, but the best model found included an interaction term between education and birth cohort. Initially, the effect of birth cohort was explored in three categories using cut-off points in 1920 and 1940. Since no additional effect of birth cohort was found for the group born after 1940, results are given for subjects born before and after 1920. Associations were strongest in the younger birth cohorts where a graded association was seen (table 4⇓). The difference between educational levels was not significant among females in the older birth cohorts, perhaps because relatively few females in these birth cohorts had higher levels of education. In males, there was also a tendency for larger differences in the younger birth cohorts but the difference between birth cohorts was not significant (table 3⇑).

Risk estimates were similar in males and females in both birth cohorts. In order to increase study power further, multivariate analyses were conducted on pooled data after formally testing for interactions between sex and other covariates and finding no indications of this. As expected, smoking was strongly associated with mortality from respiratory disease. RRs (95% confidence intervals) associated with smoking were 1.77 (1.22–2.58) in noninhaling smokers, 2.36 (1.63–3.41) in exsmokers and 9.53 (6.21–14.62) in heavy smokers who inhaled with intermediary risks in inhaling light and medium smokers. Adjustment for smoking did not affect estimated risks associated with education. Analyses using COPD deaths as the end-point gave similar results; in the younger birth cohorts RRs for the intermediary and highest level of education were 0.42 (0.28–0.61) and 0.05 (0.01–0.39), respectively. In the older birth cohorts, they were 0.49 (0.34–0.71) and 0.61 (0.34–1.11), respectively.

Changes in smoking habits during follow-up could cause residual confounding by smoking if the more educated smokers have higher quit-rates. Analysis of never-smokers only was not feasible due to the limited number of end-points in this group. A separate analysis of never- and exsmokers, however, yielded results similar to the ones given in table 4⇑. In addition, analyses including only the first 10 yrs of follow-up, and thus, limiting possible changes in smoking habits gave results similar to the ones presented. After adjustment for smoking, RRs for the intermediary and highest level of education in the younger birth cohorts were 0.36 (0.18–0.73) and 0.23 (0.03–1.65), respectively, and 0.73 (0.46–1.16) and 0.74 (0.32–1.71) in the older birth cohorts, respectively.

Discussion

This study confirms that level of education, as an indicator of social position is a predictor of respiratory mortality independently of smoking. Subjects with the lowest level of education had more than twice the risk of subjects with the highest level of education in both males and females. In males, but not in females, there was an additional effect of other indicators of social position, namely housing conditions, employment grade, income and cohabitation.

The difference in impact of education with birth cohort in females may be related to the increase in the average length of schooling. Of the subjects, 60% born before 1920 had <8 yrs of formal schooling versus 35% of those born after 1920. The large group in the older cohorts with a low level of school education is likely to have more variation in adult education, and this could cause the underestimation of the health gradient associated with education in these birth cohorts. The difference with birth cohort may also reflect an increased social gradient in disease as reported in other studies 1. Nonresponse and mortality are both higher among the deprived. This may introduce a bias, which, although likely to diminish during the follow-up period, has been shown to lead to underestimation of social class differences 8. Finally, it is possible that selective loss of low education cohort members with high mortality risk prior to the survey dates has affected risk estimates.

Several studies have reported an association between various socioeconomic measures and symptoms of respiratory disease or lung function 9, 10. In a study of male nonsmokers the difference in forced expiratory volume in one second (FEV₁) between the highest and lowest social class based on income and education was 400 mL 11. Similar results were found by the authors in a previous study on subjects from the CCHS, which is part of the present study population. In that study, the smoking adjusted difference between the highest and lowest social group also based on income and education was 400 mL for males and 259 for females. This difference in lung function was reflected in a three-fold higher risk of admission to hospital for COPD 4. In a well-designed Norwegian study, the odds ratio of COPD in primary versus university educated was 2.9 (1.3–6.5) after adjustment for occupational and smoking exposure 12. The present study extends this association to mortality data.

The different measures of social position are closely correlated, but in theory each measure also has unique advantages. Employment grades reflect structural factors as well as exposure and job-related psychosocial factors. Income directly reflects material conditions, and housing conditions also reflect factors that could have direct effect on respiratory morbidity, i.e. humidity, crowding, and indoor pollutants. Education, more than other measures tends to precede disease. Educational attainment also reflects cultural aspects, such as tools for understanding and implementing health advice, and is probably more closely related to parental social placement and thus in uteri and childhood exposures.

Low household income has also been shown to be associated with COPD independent of smoking in both males and females in several studies 13, 14. In a Canadian study, household income was related to self-reported COPD but more strongly in males than in females 15. In the present study, income, both household and personal, was associated with respiratory mortality in males only and the association was seen across the whole spectre of income. Assuming that the material conditions associated with a given income are the same for males and females, the findings reported here question whether income in this context merely represents material conditions or also reflects other aspects associated with social position.

Some of the social gradient in respiratory diseases may be caused by occupational exposures depending on the spectre of occupational exposure in the population under study. Females generally do not hold jobs with major exposure to dusts and fumes that could potentially cause respiratory disease, yet social differences in lung diseases are found in both sexes 12. A recent paper on the Whitehall II study showed that without adjustment for smoking self-reported incidence of chronic bronchitis was strongly associated with employment grade in males but not significantly in females 16. In the present study, employment grade was associated with respiratory mortality in males only. Occupational exposure to dusts and fumes is limited in the area from which this study population originates, and is not likely to explain the association between employment grade and respiratory health.

Poor housing conditions, home dampness with increased house dust mites and gas-stove usage are associated with respiratory symptoms and lower social status. Household crowding has been found to cause increased incidence of respiratory infections, but whether this is associated with development of COPD in adulthood is unsettled. In a British study, people in areas of bad housing were found to report more respiratory symptoms, and flats were found to be worse than houses 17. A Danish study focusing on housing found that smoking-adjusted FEV₁ was reduced by 10 mL·yr of living⁻¹ without central heating 18. However, these results could be caused by the correlation between housing and other measures of social grade. In the present study, the number of persons per room was associated with respiratory mortality in males, but not in females, and the use of central heating was not associated with mortality from respiratory disease after adjustment for other socioeconomic measures. Again, since the findings were not similar in males and females, the authors find it most likely that crowding and central heating do not directly affect lung health, but represent an underlying social reality.

Interestingly, the protective effect of cohabitation, which is perceived as a personal resource, was only seen in males. Being single or married is an indirect measure of various factors strongly affecting an individual's well-being, quality of social support, and health-related habits. That cohabitation is so strongly associated with mortality from respiratory disease in males is surprising. The authors do not know of studies showing that emotional support protects against development of respiratory disease.

Comparison of health differentials among females and males may provide useful insights into their causes. The social gradient in a females's health is particularly sensitive to how the inequality is defined. Social variation in morbidity in females is often found to be less than that of males 19, 20. Some studies have indicated that when a females's social position is classified according to the occupation of male and partner, male and female health gradients are more similar 21, 22. However, as pointed out, the various indices of social position also carry specific information and thus cannot be interchanged freely. The question of whether apparently smaller health gradients in females are due to imprecise measurement or represent real findings remains unsettled. It is possible that if there was a measure that gave a better guide to living and working circumstances throughout life, no sex differences would emerge. Certainly, if income, housing, and employment grade were merely indicators of material conditions and physical exposure, a priori expectations would be that there should be little or no differences between males and females.

Potential explanations of the social gradient in respiratory mortality include indirect selection, i.e. that factors predictive of ill health also predict future social position, e.g. parental social position. It is possible that the major determinant of adult social differences in lung disease is early childhood exposures. In the authors' previous study based on the CCHS, it was found that the difference in FEV₁ between high and low social groups was not age-dependent, indicating that the exposures causing the social gradient have mostly already taken place early in life 4. In the 1946 birth cohort of the British Medical Research Council, peak expiratory flow was associated with poor environment at the age of 2 yrs 23. Barker and co-workers 24–28 have shown that lower birth weight was associated with childhood respiratory disease and both were associated with adult lung function. Intrauterine influences, which retard foetal weight gain, may irrecoverably constrain the growth of the airways, and airways disease in childhood known to be associated with social conditions is undoubtedly also of importance in determining the incidence of COPD in adults. If parental social position, and thus early childhood exposure, was related strongly only to educational attainment in females, and in males also to income, housing and employment grade, this could explain the sex differences in the present study. Unfortunately, information on childhood circumstances or parental social situation were unavailable in the present study.

A shortcoming of the study is the lack of adjustment for changes in risk factors during follow-up. Socioeconomic factors, in particular education, are likely to remain stable but smoking habits may have changed and more so in the better educated 29. Between 5–15% of smokers quit between examinations with a 5–10‐yr interval 30. Given the natural history of chronic obstructive pulmonary disease, smoking cessation in subjects with established chronic obstructive pulmonary disease will have limited impact on mortality. In addition, analyses of never- and exsmokers, and analyses with limited duration of follow-up gave similar results. Thus, changes in smoking habits after study inclusion are not likely to explain the results. In conclusion, this study shows that a heavy smoker with the lowest level of education will have >15 times the risk of dying from respiratory disease compared with a nonsmoker with a high level of education. In these years, an increasing social gradient in smoking habits is seen that will increase the social gap in morbidity. Social position is a potentially avoidable cause of respiratory disease and further research is needed to explain the excess risk in the socioeconomically disadvantaged.