Based on three southern French population-based cohorts on aging, the present study shows the existence of sex differences in the time to be spent in poor SRH across ages, whilst no differences were found regarding expected lifetime in good health. While the proportion of UHLE increased with age in both sexes, women had longer TLE and UHLE but similar HLE as men at all ages. When comparing the life expectancies calculated using the subjective health and the disability indicators (considered as a more objective indicator), the study reveals specific results by indicator and different trends by sex across ages. Indeed, the UHLE was longer than the DLE for young older people (with a larger gap in men) and shorter for older people (with a larger gap in women). As for good SRH, lifetime lived free of IADL-disability was similar in both sexes. It should be noted that, in both sexes, the HLE was lower than the DFLE at younger ages but became higher at older ages, with a larger and earlier gap in women. In addition, the proportion of DLE in TLE increased substantially with age (especially up to age 75 for women), whereas this dynamic was less important for the proportion of UHLE in TLE, which increased less rapidly and similarly for both sexes.
Sex differences in the expected years to be spent in poor and good SRH
As mentioned in introduction, the well-documented sex health-survival paradox supports that men and women have different health profiles. Women are more likely to have higher levels of morbidities (e.g., hypertension, depressive symptoms), poorer socioeconomic conditions, and chronic conditions that are often non-fatal but disabling (e.g., dementia, arthritis, frailty)4,5,25,40,41,42. In contrast, men are more likely to suffer from fatal diseases (e.g., heart disease, stroke)6,43. Although we did not use multimorbidity as an indicator of health in this study, our results confirm this paradox by showing that compared to men, women can expect to live longer but with poorer subjective health and in disability.
Our findings are consistent with previous research which used SRH as health indicators to compute HLE and UHLE30,34,44,45. Despite differences in TLE between studies, all results showed that women could expect to live longer and in poorer SRH than men30,34,44,45. Like our study, other studies have also reported that, while the expected number of years spent in good SRH was relatively similar between men and women, the proportion of HLE in TLE remained higher in men than in women30,45. However, in contrast to our sex-specific trends in HLE and UHLE, Belon et al. found among Brazilians older adults that women would spend their additional TLE in good rather than in poor SRH (e.g., at age 65: in women, 17.1 HLE out of 19.6 years of TLE vs. 13.5 HLE out of 15.6 years of TLE in men). They also observed that women and men would spend similar time in poor SRH (difference = + 0.4 years at age 65 and + 0.1 at age ≥ 80)29. Such findings differences with our study may reflect contrasted living conditions and culture (e.g., how people perceive aging? what does “being in good or poor health” in old age mean to them? what are the sex-specific social norms?). Indeed, as mentioned earlier, SRH measure includes/reflects also cultural specificities.
Comparisons of life expectancy dynamics based on SRH and disability indicators
We found that for both sexes, even though the HEs were correlated and decreased with age, the estimates computed using the disability indicator were almost systematically different from those obtained using the SRH indicator. The findings reinforce the fact that, although moderately related (medium effect size), SRH and disability remain complementary health measures for understanding and addressing the biopsychosocial conditions of individuals as they age. The present study provides an original contribution to the literature on the HEs by exploring the sex-specific patterns in the age trends of life expectancy spent in good and poor health calculated using SRH and disability indicators.
Our study highlights important age differences in the trends of HE based on SRH and disability indicators. Indeed, we found a reverse pattern in the trends of the expected time to live in poor SRH and in disability, across ages. Regardless of sex, the expected lifetime in poor SRH was longer than that in disability at youngest ages and becomes shorter with increasing age. Some hypotheses could explain these findings. Firstly, in early old age (before age 70 or 75), individuals are more likely to report poor health conditions without necessarily needing help to perform the activities of daily living. For example, in our analysis sample, at baseline, of the 729 people aged ≤ 75 who reported poor SRH, only 109 (15%) had reported IADL difficulties. At older ages, we observed situations where more people have disability but did not necessarily consider their health as poor. We can assume that the ‘young old’ are probably more critical of their health than the ‘oldest old’, who consider it ‘normal’ to experience limitations and declines in old age and therefore report relatively good subjective health despite disability. Therefore, they would not consider themselves to be in such a poor health even if they have IADL disability, since they are almost like their counterparts or healthier than those who are worse off (“downward social comparisons”). This supports the fact that SRH also reflects the differences in the individual’s current conditions and past experiences given their age and compared to other people they may know, as proposed by Jylhä M.21. Secondly, the reversal trend observed with age indicates that the presence of disability with advancing age does not necessarily nourish the feeling of being in poor health. Indeed, on the one hand, it can be hypothesized that if these limitations do not prevent individuals from continuing the activities they value despite advanced age (e.g., receiving visits from family or friends, playing cards, cooking…), they may feel less unhealthy than at younger age. On the other hand, in the presence of limitations, older people may reduce their standards for good subjective health (recalibration response shift) and/or their priorities (reprioritization response shift)27. Therefore, from a public health perspective, SRH and disability measures in older adults are not interchangeable but appear to be complementary. In addition, depending on the research or health system question issue, it may be appropriate to use these measures in combination or independently. From the perspective of healthcare and long-term care planning, it may be advisable to use disability measures, whereas from the perspective of monitoring the consequences of social inequalities in health, it may be more relevant to understand the dynamics of poor subjective health measures. Identifying people who are not yet disabled but who perceive themselves to be in poor health could also help to address related needs to improve well-being, quality of life and reduce the risk of health deterioration.
The present study also highlights sex differences in the trends of DLE and UHLE, across ages. Indeed, DLE became higher than UHLE earlier in women than in men. This result can be explained by the fact that women seemed to have a higher standard of subjective health at a younger age and therefore had a higher proportion of UHLE (46% of TLE at age 65) than men (38% of TLE at age 65). With age, women tended to revise their standard of subjective health downward, reducing the proportion of their remaining life expectancy in poor subjective health, even though more than half of their total remaining life expectancy will be spent in disability from age 75 onwards. This change in the standard of subjective health appears to occur at an older age in men.
Strengths and limitations
To our best knowledge, this is the first study that investigated sex-specific pattern in HLE based on subjective health and DFLE across older ages and to compare the two. We thus showed the extent to which the indicator used to assess health status can affect estimates of HEs. Therefore, this study provides original additional contribution in the HEs research in older adults. Secondly, we used a multi-state modelling analysis suitable for longitudinal data. Since longitudinal data, such as ours, cover a relatively long period, and that health behaviors and care environments can change, the probabilities of dying and/or being in good or poor health can also change. Therefore, it is important to emphasize that unlike prevalence-based methods, the IMaCh program used the current incidence of all possible transitions between states (including recovery) and the dynamics of survival rates over time to provide TLE and more realistic HEs37,38,39. Thirdly, we did not exclude from our analyses the participants who entered in an institution in later survey waves, as they are still being followed.
Despite these strengths, some methodological choices may have impacted our results and need to be discussed. Firstly, to define the poor SRH group we combined individuals who self-rated their health as fair, poor, and very poor. This classification may have influenced our estimates by underestimating the UHLE because participants who self-rated their health as fair could be considered as being in relatively good health. However, our classification is similar to that used in several studies34,35,36. Secondly, as the sample is not representative of the general French population and as a recent study has shown, the southern department to which Bordeaux belongs is one of those with the highest LE among the 100 French departments (25th /100 for women’s LE at age 60 and 22nd/100 for men)46, we cannot extrapolate our results to the French population as a whole. Thirdly, we observed some specific patterns in the AMI cohorts (retired farmers). Contrary to the other cohorts, DFLE was longer in women than in men that could be explained by different selection criteria according to sex. Indeed, only farmers insured in their own name could be included from the farmers’ health insurance reimbursement database (men were more likely to be in this situation). Women insured in their own name (potentially included in the sample) may have been different from women insured under their husbands’ contracts (not enrolled in the study). They may be more likely to be farm owners who are therefore not representative of all women in agriculture. Although this result differs from that of the two other cohorts it did not alter our conclusions. However, it should be noted that by pooling the three cohorts, we have sought to consider, to some extent, the different profiles that can be found in the general population (e.g., people living in rural, urban areas, being retired farmers). Moreover, as the design of these cohorts and the definition of the two health indicators were similar, pooling these cohorts also allowed us to have a reasonable sample size for stratification by sex, and increased the statistical power of our analyses. Fourthly, as there is a gap of several years between the real starting points of our three cohorts, the participants do not belong to the same birth generation, it is important to note that beyond age, perceptions of aging may have evolved from one generation to the next, leading to changes in the importance people attach to the different factors taken into account when self-assessing their health. Therefore, we cannot rule out a possible cohort effect in the change observed in the dynamics of the DLE and of the UHLE. However, the results regarding the age of change in the pooled sample and in the individual cohorts were consistent (the tipping point is between 75 and 77 years of age depending on cohort). Finally, our measure of IADL disability was less objective than that which can be obtained from a performance-based test. It is also important to stress that although the (Lawton validated) scale used, included the assessment of three additional IADLs items in women, this sex-specific assessment did not increase the probability of disability in women compared to men. Indeed, using the same scale, it has been shown that individuals enter into IADL disability through limitations in shopping and transportation, two IADL items that are common to both sexes47. In women, limitations in the three additional IADLs items occur later. In this study, as we were not interested in the number of IADL limitations presented by participants, the effect of this sex-specific assessment of IADL disability on our results was practically null. Our results showing that women had a higher DLE than men are consistent with the extensive literature on sex differences in DLE/DFLE14,34,48,49.