The main finding from the study is that the correlation between self-rated heath and mortality was equally strong in most occupational classes and income quartiles, among both men and women, but the results also indicate some important exceptions that are also to be regarded as main findings. We found that mortality among the unemployed and economically inactive is higher than for the other socioeconomic groups with corresponding self-rated health. This may indicate that in studies in which self-rated health in the unemployed/economically inactive is compared with the occupationally/economically active, the actual health inequalities are underestimated, and this should be given prominence in the conclusions from these studies. A possible interpretation of this may be that mortality in these groups is associated with more than merely poor health, such as accidents.
Data from HUNT is presumed to provide a good representation of the country as a whole, since socioeconomic inequalities in mortality in the region correspond to the figures for the rest of the country (22). We also know that trends in both cause-specific mortality and the granting of a disability pension in Nord-Trøndelag County follow national trends (Statistics Norway) (23). It is therefore unlikely that the results can be attributed to characteristics of the sample. The fact that 90 % of a total population were followed over a long period is also a strength of the study, and something which places HUNT in a unique position among population studies.
It may be that we have not adjusted our models with a sufficient number of chronic diseases (such as depression) or lifestyle factors, which can possibly explain some of the «residual effect» among economically inactive persons. We can also take into account that it is 30 years since the study was conducted. It is difficult to assess whether the correlation between the concept of health and occupation/income is the same today as it was then, which may have a bearing on the transferability of the study. However, this limitation affects all studies with a long follow-up time, and we can see no good arguments to suggest that these factors may have affected the results to any great extent.
Our findings correspond to European studies, which have investigated the extent to which social class (2, 24), income (25) and education (19, 25) – (26) moderate the effect of self-rated health on mortality. Our study joins the series of those that conclude that there is no risk of over- or underestimating actual health inequalities by using self-rated health as a health measurement. Nevertheless, other studies have drawn different conclusions. In a French study, the authors observed that the predictive power of self-rated health was weaker in higher socioeconomic groups than in the lower groups. They therefore concluded that using self-rated health entails a danger of underestimating health inequalities as a result of income and occupational status (27). Correspondingly, a study of elderly Spaniards found a moderating effect of education on the relationship between self-rated health and mortality among men (28). However, here the predictive ability of self-rated health was strongest among those with high education. This effect was not observed among women. The fact that the literature in this field is inconsistent and that the results from a number of other studies indicate a moderating effect for socioeconomic factors on the correlation between self-rated health and mortality means, however, that the correlation should be studied more closely in several different contexts. It has been argued that future studies should also include mental illness, because this correlates strongly with self-rated health and may contribute to making the correlation with mortality unclear (26).
Our study shows that subjective health is a predictor of mortality irrespective of socioeconomic status among economically active persons in Norway. Although among males, ministers of the church, university teachers, lecturers and doctors have a life expectancy that is ten years longer than that of chefs, farm workers and sailors, it appears that they all consider the same criteria when they are asked to report on their own health (17). Among women we can draw the same parallels between physiotherapists, lecturers and teachers, all of whom have a life expectancy that is five years longer than for chefs, waitresses and metallurgy workers. To put it simply, none of the groups complain more about their health than other groups (17). Of course, it must be mentioned here that experiencing poor health may be as unpleasant for the individual as having poor health. This is illustrated by one of the most widely used definitions of health from the World Health Organization: «a state of complete physical, mental and social wellbeing, and not merely the absence of disease or infirmity» (29). Nevertheless, one reason why the predictive ability of self-rated health for mortality is so similar between socioeconomic groups might be that socioeconomic factors tell us little about actual health knowledge and health experience in Norway, because access to higher education and health services is universal.
Self-rated health is, of course, an intuitive and simplistic definition of health, but its imprecise quality may also be viewed as a strength. When people are asked to assess their health, more dimensions are included in their assessment than it is normally possible for a survey instrument or clinical examination to capture (30). Thus there is no gold standard for what actually constitutes good health, and self-experienced health and mortality also reflect different aspects of health. The fact that these largely conform to socioeconomic strata, as we have found here, is an important observation because it strengthens the credibility of findings from previous and upcoming studies in which self-report surveys are used to measure differences as well as to identify the mechanisms that create them.