As far as we were able to discover, this study, which involved follow-up over a 10-year period, is the first in Brazil to investigate the association between SRH and mortality in a population of young adults. Men and women with “Fair/Poor” SRH had 2.1 and 3.4 times greater hazard of mortality, respectively, than those with “Very good” SRH, independently of reporting diagnoses of chronic diseases and other covariates.
A number of studies [4, 10], including review articles [2, 8] and meta-analysis [9], point to SRH as an independent predictor of mortality. In Latin America the only three studies of this subject – all evaluating populations of older adults – come from Brazil. Two of them also encountered greater hazard of mortality for individuals with worse SRH [18, 19]. In the third study, excess hazard ceased to be significant after adjustment for cognitive function [17].
The prevalences of the worst category of SRH estimated in our population were similar to those found among industrial workers in Brazil [28] and lower than observed in Brazil's overall population [7, 29]. The differences between our findings and those of population-based studies can be explained by the fact that our population was younger, had permanent employment and more schooling, which characterize better conditions of life and health than those of the overall population. The incidence of mortality in the cohort (0.25%) was also lower than the mortality rate observed in the population from 30 to 59 years old in Rio de Janeiro State (0.56%) (mean for the period from 1999 to 2006) [30]. The Pró-Saúde study cohort and the population of Rio de Janeiro State have similar profiles in terms of causes of mortality, except for external causes, which had more influence on mortality in the overall population than in the cohort, and diseases of the respiratory system, which occurred more often in the Pró-Saúde study population [30].
The results of previous studies have varied as regards whether hazard of death, comparing the worst and best SRH categories, is greater among men or women. Some estimated greater relative hazard of mortality among women than among men associated with “Poor” SRH [1, 16, 31]; others found greater relative hazard among the men [32, 33]. In our study, relative hazard of mortality was higher for women than for men. The fact that external causes were the second cause of death among the men and did not figure among causes of death for women may have contributed to this result, given that these conditions are less associated with self-rated health than other causes of death [4]. However, this difference between hazard of death for men and women must be interpreted with caution in our study, because there are few observations in some categories, (e.g., only 3 women in the “Very good” SRH category had died) generating imprecise estimates (model 4, Table 2).
The presence of diseases is identified as the main potential confounder of the relationship between SRH and mortality [11]. In our study, although inclusion of this variable contributed to reducing the strength of the association between “Fair/Poor” SRH and mortality (reduction of age-adjusted HR by 32.4% and 27.3% in men and women, respectively), SRH continued to be an independent predictor of mortality. Other studies that adjusted for the presence of diseases found similar results. Mossey & Shapiro [12], in a pioneering study of the association between SRH and mortality, showed that the mortality hazard associated with worse SRH was stronger than the mortality hazard associated with objective measures of health. In the study by Idler et al. [11], SRH was a significant predictor of mortality, even when physical health status was taken into consideration. In the same way, Mackenbach et al. [13] showed that adjusting for a set of self-reported chronic diseases, for socio-demographic variables and for behavioral risk factors, attenuated the effect of SRH on mortality by about 44%, in comparison to the effect measure adjusted only for sex and age; nonetheless, the excess mortality risk associated with worse SRH continued about four times greater. Other authors, on the contrary, observed that the presence of diseases explains the absence of an association between SRH and mortality [10, 16, 32].
Idler & Benyamini [8] suggest some possible interpretations for the effect of SRH on mortality, independently of the presence of diseases and other risk factors. SRH is an accurate, inclusive measure able to reflect symptoms of existing diseases still at prodromal stages, or even the influence of family risk factors on health. In addition, it represents a dynamic assessment that considers health trajectories and not just the health status at the time of assessment. It is also related to behaviors that affect health status, such as lesser adhesion to preventive practices and to treatment. Moreover, it is a measure that can indicate the presence or absence of psychosocial resources capable of attenuating decline in health. Manderbacka [34] suggests that in addition to the medical model of health, adopting health promotion messages and "healthy" lifestyles are important factors contributing to health assessments.
Many studies have indicated that the ability of SRH to predict mortality diminishes with increasing cohort follow-up time [10, 31, 35]. This result may possibly stem in part from the use of SRH measured at the baseline alone, making it a good predictor of early mortality, but not of late mortality. The studies that have investigated the association between SRH as a time-dependent covariate and mortality using Cox regression [14–16, 36, 37] are not that frequent, but the results are consistent.
In the study by Strawbridge & Wallhagen [14], time-dependent SRH was a predictor of mortality among women and men from 21 to 94 years of age (relative hazard = 1.44; CI95% 1.25-1.65). Han et al. [15] investigated SRH among older women at baseline and every six months for three years. Change in SRH from “Excellent” to “Poor” entailed twice the hazard of death as compared to stable “Excellent” SRH.
Some limitations of this study deserve mention. Our results might have been biased due to lack of complete information on changes of SRH over time for the participants. About 14% of them had only the first baseline SRH measure, and other 10% had only two SRH measures (baseline plus SRH recorded on stages 2 or 3). The potential effect of such problem on the results is unknown, but one might suppose that those who drop out would probably have worse health as compared to their earliest SRH evaluation and higher probability of death. Including only the first SRH assessment for these participants would probably lead to underestimation of the strength of the association between SRH and mortality. In addition, we didn’t have the measure of chronic conditions on stage 3, thus possibly slightly overestimating the independent effect of SRH. Last, it was not possible to update the status of some covariates in our analyses. However, considering that the population is made up of staff at a single public institution, changes in income and schooling are uncommon.
Moreover, deaths occurring after 2006 could not be identified, as they were not available in the Mortality Information System (SIM). However, the high proportion (96%) of deaths recorded in the university human resource system that were also found in the SIM database between 1999 and 2006 warrants our belief that the university records for deaths occurring from 2007 to 2009 are valid. Besides, the university records system is extremely reliable, as the family must notify the institution of any death in order to secure their right to a regular pension and funeral costs. Lastly, it cannot be guaranteed that residual confounding is absent, given that only some medical diagnoses were included as self-reported chronic diseases, and objective measures of health (biochemical tests, electrocardiogram, etc.) were not used. Nonetheless, we believe that this potential residual confounding is not considerable, because a number of studies using objective measures have arrived at results similar to ours [11, 12, 38].
Analysis of SRH in three categories (instead of the “SRH positive”/“SRH negative” dichotomy found in most research) and also the use of both SRH and presence of diseases as time-dependent covariates, make the results of this study more robust and comparable to those of the few studies that use a similar strategy. We consider that analysis of SRH that changes over time is the most appropriate analytical method for investigating the relationship between this variable and mortality, given that these alterations are frequent. Failure to incorporate such information can result in misclassification, which in our study would affect about 40% of the participants.
As regards external validity, the results obtained in the Pró-Saúde study cohort may represent an approximation to what is occurring in the middle strata of the economically active population of Brazil's major metropolises. Subsequent studies could investigate the role of SRH in predicting specific causes of mortality, which was not possible in the Pró-Saúde study given the small number of deaths. It would also be interesting to ascertain whether specific causes of mortality can explain the differences observed between men and women in the SRH-mortality relationship. It is also suggested that studies investigate different SRH trajectories over time, and their association with mortality.