| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
Correspondence: Address correspondence to Peter A. Bath, Health Informatics Research Group, Centre for Health Information Management Research (CHIMR), Department of Information Studies, University of Sheffield, Western Bank, Sheffield S10 2TN, United Kingdom. E-mail: p.a.bath{at}shef.ac.uk
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionReferences |
|---|
Key Words: Self-rated health • Self-rated change in health • Mortality • Gender differences • Social engagement
Although the majority of studies reviewed by Idler and Benyamini (1997) showed that self-rated health remained predictive of mortality, several studies provided some evidence that self-rated health is not an independent predictor of mortality, but that it may be explained by health and other related factors, and that these factors may be different among men and women. The evidence is not consistent across the studies, which may be related to differences in study design, phrasing of questions (e.g., whether perceptions of current health or how health has changed are being assessed), length of follow-up period, and the factors that are included in the regression models. Therefore, to develop a better understanding of self-rated health, research is needed into the effect of such variables on the ability of self-rated health to predict mortality. The aim of the research described here was to determine within a single study whether the ability of self-rated health to predict mortality can be explained by health and related factors, the effect of different follow-up periods on the self-rated health mortality relationship, and the relative importance of perceptions of current health as opposed to perceptions of how health has changed. These issues will be investigated in men and women separately to determine any differences between the genders.
|
Methods |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
General physical health was assessed using a health index containing 14 previously validated items (Ebrahim et al., 1987). The health index scored from zero (no health problems) to 14 (multiple health problems) covering the presence or absence of (a) heart, stomach, eyesight, sleep, or foot problems; (b) dizziness, headaches, urinary incontinence, arthritis, and falls; (c) long-term disabilities and mobility status; and (d) use of health and social services and prescribed medications.
Survey assessments were composed of self-rated health (Bath, 1999; How would you rate your present health?) with five response categories (poor, fair, average, good, excellent) and self-rated change in health (Compared with 5 years ago, do you think you are...?) with five response categories (much more healthy, more healthy, about as healthy, less healthy, much less healthy). For the analyses described here, people who reported being much more healthy and more healthy were combined into a single group.
Levels of customary physical activity (CPA) likely to promote muscle strength, joint flexibility, or stamina were assessed using detailed activity inventories administered by trained interviewers. Operational criteria used for the selection of activities have been reported previously (Morgan, 1998). All activities were divided into seven mutually exclusive functional categories: (a) outdoor productive activities (e.g., gardening, house and car maintenance); (b) indoor productive activities (e.g., housework, decorating, indoor maintenance); (c) walking (purposeful walking outside the house or garden); (d) shopping (i.e., continuous ambulatory behavior associated with shopping); (e) leisure activities (e.g., cycling, swimming); (f) strength activities (e.g., climbing high steps, dragging heavy loads); and (g) joint flexibility activities (e.g., reaching for high shelves, bending for low shelves). Factor analysis (principal components with varimax rotation) of the 1985 activity inventories extracted first principal components loading on outdoor, strength, and flexibility activities among men, and shopping, outdoor, indoor, strength, and flexibility activities among women (Morgan et al., 1991). Respondents were divided into three activity groupings according to separate tertile ranges for men and women for the 1985 CPA1 factor scores (high, intermediate, and low activity groups; Bath & Morgan, 1998). Although there were differences in these factors for men and women, the issue of interest was the relative level of physical activity with respect to mortality within the gender groups, not the precise nature of those activities.
Depression was assessed using the 14-item Symptoms of Anxiety and Depression (SAD) scale, derived from the Delusions, Symptoms, and States Inventory (DSSI) (Bedford, Foulds, & Sheffield, 1976). The SAD scale focuses exclusively on recent symptoms, and comprises two seven-item subscales relating to anxiety and depression. In a clinical validation exercise conducted at baseline, total SAD scores of ≥ 6 (with depression subscale scores ≥ 4) showed high levels of concordance with clinical diagnostic ratings of depression made by experienced psychiatrists (kappa coefficient = 0.7, 
) (Morgan, Dallosso, Arie, Byrne, & Waite, 1987). Respondents were classified as depressed or not, according to these criteria.
Survey assessments of social activity were undertaken to act as an index of well-being and as a control variable for the social component of many physical activities. The Brief Assessment of Social Engagement (BASE) scale was formed from a 20-item scale, previously reported, with an overall reliability alpha of 0.7 (Morgan, Dallosso, & Ebrahim, 1985).
Statistical Analyses
In the 4-year period from the interview date to April 30, 1989, the study received notification of 242 deaths (106 men and 136 women). In the 12-year period from the interview date to January 31, 1998, the study received notification of 665 deaths (287 men and 378 women).
I assessed relationships between self-rated health and mortality in Cox proportional hazards regression models with survival time (measured in number of days from baseline assessment to death or censorship [for those people still alive] on April 30, 1989 [4-year models] or January 31, 1998 [12-year models]) as the dependent variable.
For all models, I explored the effect of self-rated health and self-rated change in health. I explored them on their own in the first instance for the 4-year and 12-year models for self-rated health and self-rated change in health (Models 1). Successively, I included as covariates (a) age group (aged 65–74; aged 75 and older; Models 2), (b) health index (Models 3), (c) level of customary physical activity (Models 4), (d) whether depressed or not (Models 5), and (e) social engagement score (Models 6). The models were as follows:
In each of the above models, I calculated the overall p value for self-rated health/self-rated change in health in relation to mortality. I calculated hazard ratios (HR) and 95% confidence intervals (CI) for each category of self-rated health relative to "Excellent" self-rated health and each category of self-rated change in health relative to "Much more/more healthy," by entering the categories as dummy (indicator) variables.
I developed separate unadjusted and age-adjusted models containing a gender-interaction term for self-rated health for 4-year and 12-year mortality to test formally the differences seen in the stratified models.
For the remaining variables, I calculated the HR for each category of variable relative to the reference category and for each increment in the health index score and social engagement score.
I used a forced entry approach to variable selection in all models. In all models, I analyzed men and women separately. I conducted all analyses by using the Statistical Package for the Social Sciences (Version 10).
|
Results |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
: 258 men (67%) rated their health as excellent or good compared with 357 women (58%), and 69 men (18%) rated their health as poor or fair compared with 144 women (24%). There was a significant association between gender and self-rated change in health, 
men (6%) responded that they were much less healthy than 5 years previously compared with 65 women (11%).
|
, but not among women. One hundred fifty men (71%) aged 65–74 years rated their health as excellent or very good compared with 107 men (63%) aged 75 and older. There was a significant association between self-rated change in health and age group among men, 
, and among women, 
. Sixty-five men (30%) aged 65–74 responded that they were either less healthy or much less healthy compared with 72 (42%) of men aged 75 and older. Eighty-nine women (32%) aged 65–74 responded that they were either less healthy or much less healthy compared with 139 (42%) of women aged 75 and older.
Self-Rated Health and 4-Year Mortality
Self-rated health predicted 4-year mortality among men and women in unadjusted models (Table 2). There was significantly increased mortality among men reporting poor (HR = 2.31; 95%CI = 1.0–5.28) and average (HR = 1.96; 95%CI = 1.0–3.80) self-rated health and among women reporting poor (HR = 4.96; 95%CI = 2.23–11.03) self-rated health (Models 1), compared with people reporting excellent self-rated health. When a separate model with a gender–self-rated health interaction term was developed to test the differences seen in the stratified models the difference in the -2 log likelihood ratios (
-2LLR) for the two models was 10.36 (
) suggesting that the difference between men and women in the magnitude of the HR for self-rated health was significant. However, in age-adjusted models these differences were no longer significant (
).
|
). When age and health were included as covariates (Model 3), self-rated health no longer predicted mortality among men. Among women, self-rated health was predictive of 4-year mortality when age (Model 2), health (Model 3), and customary physical activity (Model 4) were included as covariates but was no longer predictive when depression (Model 5) and social engagement were included (Model 6).
Self-Rated Health and 12-Year Mortality
Self-rated health predicted 12-year mortality among men and women in unadjusted models (Table 3). There was significantly increased mortality among men reporting poor (
) and average (
) self-rated health and among women reporting poor (
) and fair (
) self-rated health (Models 1), compared with people reporting excellent self-rated health. When I developed a separate model with a gender–self-rated health interaction term, the differences between the models were not significant (
), suggesting that the difference between men and women in the magnitude of the HRs for self-rated health was not significant.
|
) was predictive. When I included level of customary physical activity (Model 4), depression (Model 5), and social engagement score (Model 6) as covariates, self-rated health no longer predicted mortality among men, although the 95% CIs for poor self-rated health excluded 1.00 in Model 4.
Among women, self-rated health still predicted mortality when I included age and health as covariates (Models 2 and 3), although only poor self-rated health (
) was predictive in Model 3. Self-rated health was no longer a significant predictor of mortality when I added level of customary physical activity (Model 4), depression (Model 5), and social engagement (Model 6) as covariates.
Self-Rated Change in Health and 4-Year Mortality
Self-rated change in health predicted 4-year mortality among women (
) in unadjusted models, but not among men. There was significantly increased mortality among women (
) who reported being much less healthy (Model 1), compared with women who reported being more or much more healthy. However, when I included age as a covariate (Models 2), although self-rated change in health still predicted mortality in women (
), the 95% CI for women who reported being much less healthy included 1.00 (
).
Self-Rated Change in Health and 12-Year Mortality
Self-rated change in health predicted 12-year mortality among men and women in unadjusted models (Table 4). There was significantly increased mortality among men (
) and among women (
) who reported being much less healthy (Model 1), compared with people who reported being more or much more healthy. Self-rated change in health remained predictive of mortality when age was included as a covariate (Model 2), but was no longer predictive in either men or women when I included health in Model 3. However, the 95% CIs for men who reported being less healthy excluded 1.00 when I included age (Model 2), health (Model 3), physical activity (Model 4), and depression (Model 5) as covariates.
|
|
Discussion |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
Although, in unadjusted models, poor self-rated health was a significantly greater predictor of 4-year mortality in women compared with men, this was explained by age in this study. The magnitude of the difference in the HR for poor self-rated health for 12-year mortality was much less, and although in addition fair self-rated health was a significant predictor among women in age-adjusted models, the differences between the genders were not significant. Overall, these results suggest that the magnitude of the predictive ability of self-rated health does not differ between men and women.
Differences between men and women in the factors that explain the self-rated health–mortality relationship have emerged from this study and support the findings of previous studies in this respect (Franks, Gold, & Clancy, 1996; Hays et al., 1996; Simons, McCallum, Friedlander, & Simons, 1996; Van Doorn, 1998; Van Doorn & Kasl, 1998). In the study by Hays and colleagues (1996), lower self-rated health among women was explained by health, sociodemographics, social engagement, health practices, social support, religiousness, depression, and negative life events but not among men. Similarly, in the Franks and colleagues study (1996), lower self-rated health was explained by health, sociodemographics, social engagement, health practices, health insurance, and access to health care among women but remained predictive of mortality among men. Van Doorn and Kasl (1998) showed that age, health, health practices, self-rated life expectancy, and parental longevity explained the self-rated health–mortality relationship among women but not among men. In a study of 565 couples, Van Doorn (1998) showed that age, health, health practices, and spouse ratings of limited function and life expectancy accounted for this relationship among women whereas poor self-rated health remained predictive among men. In contrast, Simons and colleagues (1996) showed that sociodemographics, health, social networks, health practices, and depression accounted for the relationship between self-rated health and mortality among men but not among women. Although in the current study, the relationship between self-rated health and mortality was explained by age, physical and mental health, and physical and social activity in both men and women, there is some evidence that the relative importance of these factors is different for men and women. The order in which I added these covariates (demographic, physical then psychological) to successive models may have had an influence on the stage at which self-rated health no longer predicted mortality. Further research examining the separate effects of these covariates, may give a greater understanding of their relative importance in explaining the self-rated health–mortality relationship, but is beyond the scope of this study. However, the results presented here are useful in developing researchers' understanding of the factors mediating the self-rated health–mortality relationship in men and women and provide further justification for analyzing men and women separately when studying self-rated health. The results also indicate the importance of the length of follow-up period in the self-rated health–mortality relationship.
In the shorter term, age and physical health explained the self-rated health–mortality relationship among men, whereas among women the inclusion of age, physical health, and physical activity did not wholly account for it. It was only when I added mental health and social engagement that self-rated health no longer predicted mortality. Self-rated health appears a more robust predictor of mortality among women in the short-term, a finding consistent with the Simons and colleagues study (1996) which had a follow-up period of 6 years, and which, although included social networks as a covariate, did not include physical or social activity.
Among men, length of follow-up period appeared to have little effect on the predictive ability of self-rated health, given that self-rated health did not predict longer-term mortality when age and health were included and that the HRs for poor self-rated health were similar in Models 1 and 2 for 4- and 12-year mortality. However, among women, self-rated health appears a less robust predictor of mortality among women in the longer term, given that it no longer predicted mortality when I included age, health, and physical activity, and that the HRs for poor self-rated health were substantially less in Models 1 to 3 for 12-year mortality compared with 4-year mortality.
Although self-rated change in health predicted long- and short-term mortality among men and women, it was less robust than self-rated health. The ability of self-rated change in health to predict 4-year mortality was explained by the higher proportions of men and women aged 75 years and older who reported being less or much less healthy, compared with people aged 65–74 years. This suggests that people in the older group had noticed a deterioration in their health that increased their short-term mortality. In the longer term, age alone did not explain the relationship between self-rated change in health and mortality, but it was explained by age and health. This suggests that these people may be starting to perceive a decline in health that is not affecting their short-term survival, but as the years pass the health problems may develop and start to influence their mortality.
To summarize, the findings from this study provide some evidence of differences among men and women both in the factors that explain the self-rated health–mortality relationship and in the effects of length of follow-up period on the relationship. These differences indicate that gender, duration of follow-up period, and precise wording of questions need to be considered and are worthy of further examination in future research. Although further research investigating these factors could usefully extend our understanding of the self-rated health–mortality relationship, uncertainty will continue to remain about the importance of other factors, such as cross-cultural differences, that may explain the lack of consistency in some of these studies. Although this study was undertaken in the United Kingdom, the studies outlined above were undertaken in Denmark, the United States, and Australia. One possible way to determine the importance of the cross-cultural factors affecting the self-rated health–mortality relationship would be through combining data from a number of studies and undertaking meta-analyses, although this would not be without difficulties given the differences in variables used and wording of questions among the studies.
The results presented here emphasize the importance of social activity in enhancing short- and long-term survival among older people. The benefits of social engagement are clear from the study and appear independent of the capacity to undertake, and benefit from, physical activity. What is not clear from this study are the benefits of individual social and physical activities, and further research should evaluate the relative importance of general and specific activities in older men and women.
|
Footnotes |
|---|
1 Health Informatics Research Group, Centre for Health Information Management Research, Department of Information Studies, University of Sheffield, Western Bank, United Kingdom.
Decision Editor: Laurence G. Branch, PhD
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  J. B. Dowd and A. Zajacova Does the predictive power of self-rated health for subsequent mortality risk vary by socioeconomic status in the US? Int. J. Epidemiol., December 1, 2007; 36(6): 1214 - 1221. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 H. Litwin and S. Shiovitz-Ezra Network Type and Mortality Risk in Later Life Gerontologist, December 1, 2006; 46(6): 735 - 743. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
B. Han, C. Phillips, L. Ferrucci, K. Bandeen-Roche, M. Jylha, J. Kasper, and J. M. Guralnik Change in Self-Rated Health and Mortality Among Community-Dwelling Disabled Older Women Gerontologist, April 1, 2005; 45(2): 216 - 221. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
D. J. H. Deeg and P. A. Bath Self-Rated Health, Gender, and Mortality in Older Persons: Introduction to a Special Section Gerontologist, June 1, 2003; 43(3): 369 - 371. [Full Text] [PDF]
|
|
|
|
|
|
E. L. Idler Discussion: Gender Differences in Self-Rated Health, in Mortality, and in the Relationship Between the Two Gerontologist, June 1, 2003; 43(3): 372 - 375. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
