| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
ORIGINAL ARTICLES |
From INSERM U687-IFR69 (A.S.-M., A.G.), Saint-Maurice Cédex, France; Department of Epidemiology and Public Health (A.S.-M., P.M., J.F., M.M., M.S.), University College London, UK; Centre de Gérontologie (A.S.-M.), Hôpital Ste Perine, AP-HP; Population Research Unit (P.M.), Department of Sociology, University of Helsinki, Finland.
Address correspondence and reprint requests to Archana Singh-Manoux, INSERM, U687, HNSM, 14 rue du Val d’Osne, 94415 Saint-Maurice Cédex, France. E-mail: Archana.Singh-Manoux{at}st-maurice.inserm.fr
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
Methods: Data (6316 men and 3035 women) are drawn from the Whitehall II study. SRH and covariates were measured at baseline (1985–1988) when the average age of individuals was 44.5 years (SD = 6.1). The mortality follow-up was available for a mean of 17.5 years and was classified as having occurred in the first 10 years or the subsequent follow-up period (range 6 to 9 years). The association between SRH and mortality was assessed using a Cox regression model with relative index of inequality (RII) to summarize associations.
Results: There were no sex differences in the association between SRH and mortality in either the short (p = .39) or the long term (p = .16). Sex-adjusted short-term association (RII = 3.80; 95% confidence interval (CI) 2.28, 6.35) was significantly (p = .004) stronger than the long-term association (RII = 1.56; 95% CI 1.04, 2.34). Explanatory variables accounted for 80% of the SRH-mortality association in men and 29% in women.
Conclusions: SRH predicts mortality equally well in men and women. However, the covariates explained a much larger proportion of the SRH-mortality relationship in men compared with women. In this middle-aged cohort, SRH predicts mortality strongly in the short term but only weakly in the long term.
Key Words: gender • mortality • self-rated health • Whitehall II study
Abbreviations: SRH = self-rated health; ECG = electrocardiogram; ANOVA = analysis of variance; RII = relative index of inequality; CI = confidence interval; SD = standard deviation.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
It is possible that some of this inconsistency in the association between SRH and mortality stems from the age at which the association between SRH and mortality is examined as the association has been shown to be weaker at older ages (1–3,8,9,28). Another possible factor is the length of follow-up; results on the elderly suggest that it is an important modifier of the SRH-mortality association (14–16), with some evidence of declining predictive ability over time in women (16). The objectives of this paper are to examine the link between SRH and mortality in a middle-aged sample with an attempt to address the following questions:
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
Measures
SRH was assessed at Phase 1 (1985–1988) of the study with this question: "Over the last 12 months would you say your health has been very good, good, average, poor, or very poor?"
Mortality
A total of 10,301 respondents (99.9%) were traced for mortality from the baseline through the national mortality register kept by the National Health Services Central Registry, by using the National Health Service identification number assigned to each British citizen. Mortality follow-up was available until September 30, 2004; a mean of 17.5 years with all surviving respondents having a minimum of 16 years of follow-up. Deaths were classified as having occurred in the first 10 years (year 0–10) or the subsequent period, ranging from 6 to 9 years (year 10+).
Explanatory Variables Measured at Baseline (Phase 1, 1985–1988)
Age calculated from birthdate taken from the questionnaire at baseline. Early life factors/parental longevity was assessed using two measures: height and age of death of parents. Height, an indicator of early life environment, was measured at the Phase 1 screening examination. Parental longevity was assessed from the response to questions regarding whether either parent had died and, if so, their age at the time of death. These responses were combined and grouped according to whether both, one, or neither parent died at or before the age of 70 years.
Sociodemographic measures were occupational position and marital status. Occupational position was the British civil service grade of employment at Phase 1; i.e., a 3-level variable representing high (administrative grades), intermediate (professional or executive grades), and low (clerical or support grades) grades. People in different grades differ with respect to salary, social status, and level of responsibility. Marital status was assessed by questionnaire and consisted of the following categories: married or cohabiting, never married, separated, divorced, or widowed.
Health Behaviors
The measure of smoking was a 5-level variable derived from several questions on smoking such as nonsmoker, ex-smoker, light smoker, medium smoker, and heavy smoker. Alcohol consumption was assessed by questions on the number of alcoholic drinks ("measures" of spirits, "glasses" of wine, and "pints" of beer) consumed in the last seven days. This was converted to the number of units of alcohol consumed in the last week. Frequency of fruit and vegetable consumption was assessed on an 8-point scale going from "seldom or never" to "2 or more times a day." Physical activity was assessed using questionnaire data; participants were asked about the frequency and duration of their participation in "mildly energetic" (e.g., weeding, general housework, and bicycle repair), "moderately energetic" (e.g., dancing, cycling, leisurely swimming), and "vigorous physical activity" (e.g., running, hard swimming, and squash). The frequency and duration measures were combined to three levels of activity: low, medium, and high.
Health
Seven measures of health were used. The presence of respiratory illness was detected using the Medical Research Council chronic bronchitis questionnaire (30). The category diabetes included self-report of doctor diagnosis or being on medication for diabetes. Electrocardiogram (ECG) abnormalities were probable/possible ischemia identified on ECG during the medical examination at baseline. The category hypertension included all participants on antihypertensive medication or with a systolic or diastolic blood pressure 
160 or 95 mm Hg, respectively. Diagnosed heart trouble was assessed through self-report of doctor diagnosis of coronary heart disease. Mental health was measured using caseness criteria (score 5) on the general health questionnaire (31). Sickness absence was assessed through a question on the number of sick days taken in the past year.
Statistical Analysis
Sex difference in SRH was assessed using the 
2 analysis. Descriptive analyses to examine the association between SRH and explanatory variables were conducted and tested using 2 analysis for trend for categorical variables and fitting a linear trend across the SRH categories for continuous variables. The distributions for the units of alcohol consumed and the number of sick days were skewed and therefore logged values of these measurements (after the addition of one to all values to remove the zeros) were used in all analyses.
Mortality rates for each SRH category (very good, good, average, poor, or very poor) were calculated using person-years at risk, expressed as deaths per 1000 person years. The association between SRH and mortality was examined using Cox regression to model survival time subsequent to the assessment of SRH for each individual. A key assumption of Cox regression is the proportionality of hazards assumption, requiring the hazard ratio to be constant over the entire follow-up period (here 16–19 years). This assumption was tested using an interaction term between log-time and SRH in Cox regression. As there was evidence for nonproportionality of hazards over time, we conducted subsequent analysis separately for the short (follow-up year 0–10) and long-term (year 10+) follow-up (32). This analysis was carried out using two strategies. The first was to model SRH as a categorical variable, where the first category ("very good" SRH) was the reference. Thus, the hazards ratio for each subsequent category provides the relative likelihood of death compared with those with "very good" SRH, the reference category.
The second strategy was to model the association between SRH and mortality using a summary measure called the relative index of inequality (RII) (33). The RII is useful as one hazard ratio replaces the four obtained using the categories of SRH. Furthermore, the analyses for assessment of the importance of explanatory factors to the association between SRH and mortality become less cumbersome. The RII is a regression-based measure, calculated by creating a scale from 0 to 1 to indicate the two extremes of an underlying SRH distribution. A value of 0 represents the best SRH and 1 represents the worst SRH. Each SRH category (very good, good, average, poor, and very poor) covers a range on this scale that is proportional to the number of individuals who endorsed that SRH category and it is given a value on the scale corresponding to the cumulative midpoint of its range. This procedure transforms a hierarchical categorical variable into a continuous variable. For example, a measure having the frequency distribution of 30%, 40%, 20%, 8%, and 2% would be transformed into a single scale taking the values 0.15, 0.50, 0.80, 0.94, and 0.99 for the five categories. The resulting Cox regression using the transformed SRH variable as a predictor estimates the hazard ratio for the worst SRH compared with the best SRH; the estimation takes into account the data from all SRH categories and the index is weighted to reflect the size of these categories. Thus, an RII of 1.5 indicates that the mortality hazard between the extreme ends of the SRH distribution is 1.5 times higher for the worst compared with the best SRH; an RII of 1.00 would indicate equal mortality hazard. The RIIs were compared (men and women; long term and short term) using a z test.
The covariates were included in the regression model in successive models, using predefined categories of early life factors, sociodemographic variables, health behaviors, and measures of health. The explanatory power of the covariates was examined by using the percentage reduction in RII (RIIcontrolling for age – RIIcontrolling for age and explanatory factor)/(RIIcontrolling for age – 1) x 100) when these variables were added to the model containing age, SRH, and mortality. The percentage reduction in RII calculated in this way indicates the extent to which the covariates explain the SRH-mortality association and not the extent to which they predict mortality themselves. The original continuous (logged values of units of alcohol consumed and number of sick days) and fully categorized explanatory variables were used in these models.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
|
The association between SRH and the variables are presented in Table 2 for men and Table 3 for women. In men (Table 2), SRH was associated with all explanatory variables except age (p = .52), parental longevity (p = .15), and presence of ECG abnormalities (p = .66). In women (Table 3), SRH was associated with all variables except parental longevity (p = .53), being married (p = .58), diabetes (p = .29), and presence of ECG abnormalities (p = .37).
|
|
Analysis of the association between SRH and mortality was conducted separately in the short term (follow-up first 10 years) and the long term (follow-up >10 years, range 16 to 19 years) as the test of nonproportionality of hazards over the total follow-up period indicated declining hazards over time (2 = 6.43, p = .01). Table 4 shows the mortality rates and the hazards ratio associated with the five categories of SRH in men and in women and the summary measure using RII. In almost all cases, we observed a monotonically increasing mortality with declining SRH. However, apart from the few men who rated their health as being "very poor," the trend in SRH with long-term mortality was weaker and statistically nonsignificant. Also, the RII confirmed this weaker association in the long term in men (RII = 1.62; 95% CI 0.98, 2.68) and in women (RII = 1.46; 95% CI 0.73, 2.93). The SRH-mortality relationship was not significantly different in the two sexes either in the short term (z = 0.53, p = .30) or the long term (z = 0.23, p = .41). In analyses combining men and women, the short-term association (RII = 3.80; 95% CI 2.28, 6.35) was significantly stronger (z = 2.68, p = .004) than the long-term association (RII = 1.56; 95% CI 1.04, 2.34).
|
Table 5 shows the SRH-mortality relationship after adjustment for explanatory variables, carried out only for the short-term associations as the long-term associations (in analysis stratified by sex) between SRH and mortality were not statistically significant. Eighty percent of the association between SRH and mortality in men, compared with only 29% in women, was explained by the covariates examined in this study. SRH continued to be associated significantly with mortality (follow-up year 0–10) in women after adjustment for all explanatory variables (RII = 3.63; 95% CI 1.25, 10.53). Measures of health had the strongest explanatory role in men (66%), and health behaviors (23%) and health measures (22%) were equally important in women. Health behaviors were slightly more important in men (31%) compared with women (23%). Early life factors had little or no explanatory role in either men or women.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
It is important to consider the implications of the analytical strategy adopted in this paper. The associations between SRH and mortality are usually examined by comparing the worst category ("very poor" SRH) with the best category ("very good" SRH) despite evidence of a dose-response relationship (4). This approach has been popular because it provides a summary index of the association between SRH and mortality. However, the results can be misleading and attempts have been made to use information from all five SRH categories. One approach is to dichotomize the 5-point measure by grouping the first two or three categories as "good" SRH and the others as "poor" SRH and then comparing these two groups. We use the RII, a summary measure that has the advantage of comparing mortality risk at the extremes of the SRH distribution, but it is estimated using data from all SRH groups and is weighted to account for the size of these groups.
This study adds to the existing literature by examining the SRH-mortality relationship over time. The follow-up times in previous studies have ranged from a few months (16) to as much as 27 years in a study on young men (11). The few studies to have explicitly examined differences in the strength of the association as a function of the length of follow-up were conducted on the elderly (14–16). The results from these studies were not consistent. Benyamini and associates found SRH to predict short-term mortality (4-year follow-up) but not long-term mortality (9-year follow-up) in either men or women aged over 75 years (14). Deeg and Kriegsman found the SRH mortality association only in men (aged 55 to 85 years), with hazards being similar in the short term (3-year follow-up) and the long term (7.5 year follow-up) (15). Grant and colleagues reported declining hazards in women but not in men over a 3-year period in individuals >70 years old (16).
It is possible that the normative decline in health in the elderly (thus changing SRH) over time could, in itself, lead to a decrease in the long-term association between SRH and mortality. The same could also be true in a middle-aged population. A recent study examined the 59-year longitudinal trajectory of SRH and concluded that it was relatively stable until the age of 50 years, with men consistently rating their health as better than women. After age 50, a steep decline in SRH among men left no gender differences in SRH by late adulthood (34). Given the age of our respondents (average age at baseline is 44.5 years, SD = 6.1), it is possible that changes in SRH account for some of the weakening of its relationship with mortality over the longer term. Furthermore, the measure of SRH asks individuals to rate their health over the last 12 months rather than future or anticipated health; even if SRH were a very accurate and inclusive measure of health status, it is not likely it would continue to be accurate over a prolonged time period. SRH in our data predicts long-term mortality (in analyses combining men and women) but it was substantially weaker than the short-term association.
In our sample, the "proportionality of hazards" assumption, required for Cox regression, held over the 10-year period. One measure of SRH will predict mortality in a middle-aged population for the following 10 years, referred to here as the "short-term" period. Thus, the evidence from our and other studies (16) indicates that the delineation of the follow-up period into short and long periods is dependent on the age of the study participants. As the decline in SRH intensifies around age 50 (34), the ability of SRH to predict mortality is likely to be better in younger populations. The implication of this finding for research and health monitoring is that, in older populations, more frequent assessments of SRH may be required. Until middle age, SRH is an easy and reliable measure, at least during the first 10 years of follow-up.
The SRH-mortality association was similar in men and women in the short term and long term. The documentation and explanation of gender differences in the SRH-mortality relationship have been a major focus of research in this field. Although some studies have found SRH to be a weak predictor of mortality in women (2–4,15,17–21), others have either found no sex differences (8,14) or stronger relationships in women (16,22). Our data show no sex differences in the overall SRH-mortality gradient in middle-aged men and women despite differences in the univariate distribution of SRH; more women reported poor health or very poor health. This finding suggests that differences across the SRH scale are being interpreted in a similar fashion by men and women.
Examining the association between SRH and explanatory variables reveals minor sex differences; thus, men and women appear to use similar criteria to judge SRH. Even when there are no sex differences in the prevalence of chronic diseases, men experience more severe forms of these conditions (27,35). In men, measures of health explained around 66% of the SRH-mortality association whereas they explained only 22% in women. As a result, all variables together explained a much larger proportion of the SRH-mortality relationship in men compared with women. The SRH-mortality association remained statistically significant after adjustment for all explanatory variables in women. Given the similarity in the SRH-mortality association in men and women and the differential impact of the explanatory variables, further research is required to understand the mechanisms that explain the SRH-mortality association in women. A starting point would be to look at cause-specific mortality, something our study is still underpowered to do. A further avenue of research could be to explore the impact of other measures of health.
There are some caveats to the results reported here. As the respondents are middle-aged, the mortality rate is low; in women, the results are based on a smaller number of deaths. Also, the Whitehall II study is not a population sample and it is likely that the explanatory variables will have different associations with both SRH and mortality in different samples. In this occupational cohort, there were no sex differences in the association between SRH and mortality but it is possible that working women differ from women in the general population.
Although the SRH-mortality relationship has been widely examined, it is little understood as there have been few attempts to examine factors that moderate this relationship and the conditions under which it strengthens, weakens, or disappears. Our results show that the SRH relationship is stronger at shorter follow-up periods. Furthermore, there were no sex differences in the strength of this association in this middle-aged cohort. SRH appears to be a multidimensional phenomenon (4,36) and was related to sociodemographic variables, health behaviors, and objective measures of health similarly in both men and women. These variables explained more of the SRH-mortality association in men than in women. Our results show SRH to be a pertinent, global measure of health status, particularly in the short term.
We thank all of the participating civil service departments and their welfare, personnel, and establishment officers; British Occupational Health and Safety Agency; British Council of Civil Service Unions; all participating civil servants in the Whitehall II study; and all members of the Whitehall II study team.
|
NOTES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
Received for publication May 15, 2006; revision received September 8, 2006.
DOI:10.1097/PSY.0b013e318030483a
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
This article has been cited by other articles:
 |
|
 |
|
  L. Ayalon and K. E. Covinsky Spouse-Rated vs Self-rated Health as Predictors of Mortality Arch Intern Med, December 14, 2009; 169(22): 2156 - 2161. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C Delpierre, V Lauwers-Cances, G D Datta, T Lang, and L Berkman Using self-rated health for analysing social inequalities in health: a risk for underestimating the gap between socioeconomic groups? J Epidemiol Community Health, June 1, 2009; 63(6): 426 - 432. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
L Giatti, S M Barreto, and C C. Cesar Household context and self-rated health: the effect of unemployment and informal work J Epidemiol Community Health, December 1, 2008; 62(12): 1079 - 1085. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. Sone, N. Nakaya, K. Ohmori, T. Shimazu, M. Higashiguchi, M. Kakizaki, N. Kikuchi, S. Kuriyama, and I. Tsuji Sense of Life Worth Living (Ikigai) and Mortality in Japan: Ohsaki Study Psychosom Med, July 1, 2008; 70(6): 709 - 715. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. Singh-Manoux, A. Dugravot, M. J Shipley, J. E Ferrie, P. Martikainen, M. Goldberg, and M. Zins The association between self-rated health and mortality in different socioeconomic groups in the GAZEL cohort study Int. J. Epidemiol., December 1, 2007; 36(6): 1222 - 1228. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
