This Article Abstract Full Text (PDF) Services Download to citation manager Citing Articles Citing Articles via HighWire PubMed PubMed Citation

Self-Rated Health in a Longitudinal Perspective: A 9-Year Follow-Up Twin Study

¹ Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm, Sweden.
² Department of Psychology, University of Southern California, Los Angeles.

Address correspondence to Pia Svedberg, Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Box 281, SE-171 77 Stockholm, Sweden. E-mail: pia.svedberg{at}meb.ki.se

	Abstract

TOP Abstract Method Results Discussion Conclusions References

 
Objectives. This study first considers age and cohort explanations for age-related changes in mean values and variance in self-rated health. Second, it evaluates the contributions of genes and environments to self-rated health measured longitudinally.

Methods. Subjects were participants in the Swedish Adoption/Twin Study of Aging. Self-rated health assessments were collected in four waves over a 9-year follow-up period, from one or both members of 788 twin pairs. Linear mixed effect models were used to test for differences in means and variances. Structural equation modeling provided estimates of genetic and environmental components of variance and contributions to stability.

Results. Changes in means and variance within cohorts seem to reflect illness. Earlier-born cohorts are more variable and have lower self-rated health. These cohort differences were not explained by childhood socioeconomic status. Correlations between time points reflect both environmental and genetic factors.

Discussion. Both genes and environments contribute to self-rated health longitudinally, and both age and cohort effects are seen. Age-related changes in self-rated health can be attributed to illness. Cohort differences are most likely attributable to socially mediated and individual-specific environmental factors.

Interest in the self-evaluation of health has grown over the last 25 years. The most provocative finding is that subjective health assessments are often superior to clinical assessments for predicting outcomes such as mortality and morbidity (Idler & Benyamini, 1997; Mossey & Shapiro, 1982; Thomas, Kelman, Kennedy, Ahn, & Yang, 1992). Twin studies that focus on self-rated health as an outcome have demonstrated that cross-sectional increases in variability with age are primarily due to environmental factors unique to the individual, but also to genetic factors (Christensen, Holm, McGue, Corder, & Vaupel, 1999; Harris, Pedersen, McClearn, Plomin, & Nesselroade, 1992a; Romeis et al., 2000; Svedberg, Lichtenstein, & Pedersen, 2001). It is unclear, however, whether these findings reflect differences in cohorts where earlier-born cohorts had more variability or whether the findings indicate a true age-related increase in variance.

The extent to which change in self-rated health over time reflects genetic or environmental influences has, to our knowledge, not been investigated previously. If environmental factors explain the correlation between follow-up occasions, this would suggest that stability in self-rated health reflects individual-specific environmental factors. These could be accidents or chance events that happen during the life span and that have enduring effects. If shared familial environmental factors explain the time-to-time correlation, childhood socioeconomic status (SES) or other rearing influences may be of importance. One of the most prominent factors contributing to cohort differences in Sweden the last century is SES. A large body of literature also indicates that SES is predictive of morbidity, mortality, and decline in functional status at different ages (e.g., House, Kessler, & Herzog, 1990; Liang et al., 2002; Seeman & Crimmins, 2001). Thus, it is reasonable to predict that childhood SES contributes to both cohort differences and longitudinal changes in self-rated health. If, instead, genetic factors explain stability in self-rated health, this would suggest that the consequences of genes are transmitted longitudinally. If there are changes in genetic influences at later follow-up occasions or among older age groups, this might imply that genetic influences tied to late-onset, chronic diseases are of importance for individual differences in self-rated health as people grow older.

Cross-sectional studies have demonstrated that increasing age, increasing number of symptoms, presence of a longstanding illness, and increasing number of new illnesses have shown a marked effect on subjects' self-rated health status (e.g., Murray, Dunn, & Tarnopolsky, 1982). Undén and Elofsson (1998) further showed that functional health and life style have a greater explanatory value for level of self-rated health status in older age groups than younger ones (younger than 44 years). However, cross-sectional studies do not allow one to distinguish properly the effects of age from aging due to coexisting processes contributing to differences between age groups such as sampling, selection, or different life experiences related to cohort (Idler, 1993; Maddox & Douglass, 1974; Schaie & Baltes, 1996).

Of the few longitudinal studies focusing on self-rated health as an outcome, some have reported change, while others report stability in level with age. Differences between studies are mainly due to differences between the study samples, differences in questions asked, and length of follow-up times. The Los Angeles Health Survey (Goldstein, Siegel, & Boyer, 1984) found that self-rated health remained relatively stable over 1 year. Changes in self-rated health status were not associated with any indicators of objective health or health beliefs. In their 15-year longitudinal study, Maddox and Douglass (1973) reported a substantial stability of self-rated health and persistent high positive congruence between medical evaluations and self-rated health. A Swedish longitudinal study showed no overall change in self-rated health from 60 to 67 years of age (Tibblin, Cato, & Svardsudd, 1990; Tibblin, Tibblin, Peciva, Kullman, & Svardsudd, 1990), and a Finnish study found similar results (Leinonen, Heikkinen, & Jylha, 1998), although another study reported a slight decline of self-rated health over an 8-year period (Markides & Lee, 1990). Nonetheless, results from the National Health and Nutrition Examination Survey I Follow-Up Study suggest that not only is self-rated health sensitive to deterioration in physical health over a 20-year period, but declines over time in self-rated health are associated with mortality (Ferraro & Kelley-Moore, 2001). In her 6-year follow-up study, Idler (1993) tested whether the observed optimism in self-rated health was attributable to changes in age, cohort, or survivorship with a sample of elderly persons (65 years or older). She concluded that none of the explanations could be ruled out; selective survivorship, processes of aging, and cohort differences all seem to play important roles. Taken together, cross-sectional and longitudinal studies suggest that age-related change in self-rated health may not be great, whereas cohort differences may be present but generally overlooked.

An increase in individual differences across age groups has been demonstrated cross-sectionally for several health-related variables, including self-rated health, although variance increases are not universal (Birren & Schroots, 1996; Schaie & Baltes, 1996). A review of biologically oriented studies in the gerontological literature shows that 75% of the cross-sectional studies show patterns of increasing variability, that is, increasing individual differences (Nelson & Dannefer, 1992). Individual differences can be traced to genetic or environmental influences. Genetic and familial factors represent illnesses, functional capacity, personality, SES, and other influences that in turn affect self-rated health status.

Age group differences in variance reported in previous cross-sectional gerontological genetic studies (Harris et al., 1992a; Svedberg et al., 2001) could also reflect cohort differences in which earlier-born cohorts had a greater range of socioeconomic and cultural influences than later-born cohorts. Alternatively, terminal decline, a rapid change in social, physiological, and psychological functioning prior to death, has been suggested as a possible source of individual differences in late life (Berg, 1996). When studying elderly people, the population can be presumed to include both elite survivors and an unknown number of individuals that are experiencing a terminal decline phase in their physical functioning.

Longitudinal quantitative genetic data help to distinguish true aging effects from cohort differences and also to investigate genetic and environmental explanations for increasing individual differences. Generally, longitudinal changes in both means and variance have been smaller than cross-sectional differences (Finkel, Reynolds, McArdle, Gatz, & Pedersen, 2003; Pedersen & Reynolds, 1998; Viken, Rose, Kaprio, & Koskenvuo, 1994). Longitudinal analyses of number of organ systems affected by disease were investigated in a Swedish twin study (Pedersen, Steffensson, Berg, Johansson, & McClearn, 1999). For those twin pairs that survived to the age of 80 or older, there were longitudinal increases in variance across a 30-year period, entirely attributable to increases in environmental influences. Finkel, Pedersen, and colleagues (2003) also found support for increases in environmental variance with age for grip strength and well-being, while genetic variance remained stable. Self-rated health has not been studied from a longitudinal quantitative genetic perspective.

Our goals were first to investigate how mean levels change with time and whether the increases in total variance and variance components with age previously reported in cross-sectional studies of self-rated health are replicated in a longitudinal study. Four age groups were constructed based on birth year. We predicted that mean levels within age groups would remain relatively stable across time and that longitudinal changes would be smaller than cross-sectional. Additionally, we test whether an index of illness accounted for any age-related changes in self-rated health and whether childhood SES accounted for any cohort differences in self-rated health. Second, we sought to investigate the extent to which genetic and/or environmental factors contributed to phenotypic stability over time, that is, correlations between time points.

	METHOD

TOP Abstract Method Results Discussion Conclusions References

 
Participants and Data Collection
Participants in this study are from the ongoing longitudinal Swedish Adoption/Twin Study of Aging (SATSA). This study is based on a subsample from the Swedish Twin Registry (Lichtenstein et al., 2002) and includes twin pairs reared together (TRT) and twin pairs separated before 11 years of age and reared apart (TRA). The sampling of SATSA twins and method of zygosity determination have been described in detail elsewhere (Pedersen et al., 1991). The data included in the present analyses were collected by questionnaires in four waves starting in 1984 with new questionnaires sent out at 3-year follow-up intervals (1987, 1990, and 1993). The analyses in this study are based on data from the 788 twin pairs (1,576 individuals) of known zygosity in which one or both members responded to the self-rated health items at least once (pair-wise response frequency 65%). Ages at the first questionnaire (1984) range from 26 to 86 years (M = 58.6, SD = 13.6), and 40% of the sample were male. For the purpose of comparisons with earlier results by Harris and colleagues (1992a), we divided data into four age groups based on age at first questionnaire (1984): younger than 50 years, 50–59 years, 60–69 years, and older than 70 years. These groups represent different birth cohorts. The age at the last measurement occasion of each age group is then the same as the age of first occasion of the next age group. Response patterns across measurement occasions by age group are presented in Table 1. Fifty-four percent of the sample participated at all four questionnaire occasions, and only 13 individuals had partners with missing information on all four of the questionnaires.

Measures
Self-rated health was measured by a scale including four questions (listed in Table 2), one of which asks respondents for a global rating of their health and three items about their health in the context of their own aging experience. Following the same procedure as Harris and colleagues (1992a), the items were standardized at the first questionnaire occasion to a mean of 0 and a standard deviation of 1 before summing. In questionnaires 2–4 (1987, 1990, 1993), the items were standardized relative to means and variances at the first questionnaire (1984) items and then summed, ensuring that the definition of self-rated health remained invariant across testing occasions. A more favorable self-rating of health is indicated by a higher value on the scale. The items included were reasonably homogeneous (Cronbach coefficient =.76) in 1984.

We also looked at SES in childhood (rearing home) measured on the first questionnaire in 1984. The SES Scale includes three components: material resources within the household, highest education, and highest occupational status of the parents. This scale is based on factor analyses. Variables were standardized to a mean of 0 and a standard deviation of 1 before summing. A higher score on the scale reflects higher SES level. For a more detailed description, see Lichtenstein and colleagues (1992).

Analytical Procedure
Analysis of means and variances
Linear mixed effect models with repeated measures and unstructured covariance structures, different for different age groups, were fitted to the data by running the Proc Mixed procedure in SAS (SAS/STAT, 1999–2001). First, a model was fitted to data to test for differences in means in self-rated health over time, within and between age groups. Second, a model was fitted to the data in the same manner, including sum of illness and childhood SES as covariates. Two-sample Student t tests were calculated using the SAS Proc ttest procedure to test for differences in mean childhood SES between age groups. Then, similar models were fitted to test for differences in variances.

Quantitative genetic analysis
Differences in similarity between zygosity groups provide information about genetic and environmental effects that may be present. For example, if the monozygotic (MZ) twin pairs are more similar than dizygotic (DZ) twin pairs, then genetic effects are indicated. Shared environmental effects refer to nongenetic influences that contribute to similarity within pairs of twins regardless of zygosity, for example, family environment and contact throughout life. Nonshared environmental effects are individual-specific influences not shared within a twin pair, like accidents or occupations. To the extent that twin pairs are dissimilar, nonshared environmental effects are indicated. Genetic differences also contribute to dissimilarities within DZ twin pairs.

To investigate what factors contribute to the variance in self-rated health over a 9-year time span in the four different age groups, a series of models was fitted to the raw data. Fitting to raw data allows for inclusion of single responders and therefore increases power in the analyses and provides population based estimates of total variance. The series of models began with the fully parameterized Cholesky model in a temporal context illustrated in Figure 1 (Loehlin, 1996; Neale & Maes, 2002). The figure depicts a simplified version of the model including only one twin in a pair. T₁–T₄ represent measurement of self-rated health at four successive time points (1984–1993). Three sources of variation were considered at each time: genetic (A_1–4), shared environmental (C_1–4), and nonshared environmental (E_1–4) variance (including measurement error). A, C, and E give information about anonymous influences that are not actually measured. Thus, A tells us that there are genetic influences but not which gene. Using this multivariate model, we can separate genetic and environmental effects specific to each time point from effects that are in common to the previous time points. Within this framework with four measurements (time points), the first genetic factor (A) loads on all of the measures, a second genetic factor loads on all but the first measure, a third genetic factor loads on all but the first two of the measures, etc. Shared environmental (C) and nonshared environmental (E) factors load on the four measurements in patterns similar to that of the genetic factors. The relative importance of genetic and environmental effects on each of the measurement occasions is calculated by squaring and summing the parameter estimates for each measure and dividing the squared parameter estimates by the sum of squares. Total variance estimates from the model and the decomposition into genetic and environmental variance at each measurement occasion will be presented. Heritability estimates for each time occasion may be obtained by taking the overall genetic variance over total variance. First, we fitted a full model (ACE) for each age group. A series of submodels were then fitted when we dropped common a, c, and e parameters that contribute to transmission from the full model (i.e., only time-specific a₁₁, a₂₂, a₃₃, a₄₄, c₁₁, c₂₂, c₃₃, c₄₄, e₁₁, e₂₂, e₃₃, and e₄₄ were kept in the model; see Figure 1) to identify a parsimonious explanation of the data and evaluate the hypothesis concerning transmission of stability. Nested models were compared by likelihood-ratio chi-square. Degrees of freedom are equal to the number of parameters deleted from the full model. The reduced models where common a, c, and e were dropped from the model were compared with the fit of the full model.

View larger version (27K): [in this window] [in a new window]   Figure 1. Cholesky path model depicting common and unique factors for genetic and environmental sources of variance for self-rated health, measured at four consecutive points in time. The figure is simplified and contains only one of the twins in the pair. A = genetic factors; a = genetic loadings; E = nonshared environmental factors; e = nonshared environmental loadings; C = shared environmental factors; c = shared environmental loadings

The standardized path coefficients from the Cholesky model can be used to estimate how the correlations between self-rated health at different time points are mediated and thereby describe what factors contribute to stability over time. To evaluate longitudinal stability, the proportions of the phenotypic correlation attributable to a, c, and e between time points for each age group were calculated (Plomin & DeFries, 1981). The genetic component of the estimated phenotypic correlation (r_P) for self-rated health between times 1 and 2 is a₁₁*a₂₁, the shared environmental portion is c₁₁*c₂₁, and the nonshared environmental portion is e₁₁*e₂₁. Following the same procedure among the other follow-up occasions, the estimated genetic portion of the phenotypic correlation for self-rated health between times 2 and 3 is (a₂₁*a₃₁) + (a₂₂ * a₃₂), the shared environmental portion is (c₂₁*c₃₁) + (c₂₂*c₃₂), and the nonshared environmental portion is (e₂₁*e₃₁) + (e₂₂*e₃₂), and so on.

Because the SATSA sample consists of both TRT and TRA, a second series of analyses was undertaken where rearing status was taken into account. Inclusion of TRT and TRA enables us to estimate the importance of similarities in the rearing environment, as distinct from other shared environmental influences that might occur in adult life. In these analyses, a fourth parameter was included by a unity correlation between these latent factors for TRT and a zero correlation for TRA, representing influences that result in greater similarity of TRT than TRA (Lichtenstein et al., 1992; Plomin, DeFries, & McClearn, 1990).

	RESULTS

TOP Abstract Method Results Discussion Conclusions References

 
Analysis of Means and Variances
Mean values and standard deviations for the self-rated health scale at baseline and at follow-up questionnaire occasions are described in Table 3 by age group. Younger adults were more positive about their health status than older adults at baseline. A linear mixed effect model was fitted to the data controlling for sex, rearing, and zygosity. Figure 2 illustrates mean values and associated 95% confidence intervals of self-rated health at the four measurement occasions within the four age groups regardless of participation pattern. There were statistically significant mean effects for age group and time and a significant age group-by-time interaction (p <.05). Within each age group, change between successive pair-wise time points (1984 and 1987, 1987 and 1990, and 1990 and 1993) was not significant (p >.05), although statistically significant changes between time points more distant apart were found for the youngest and the two oldest age groups. For the age group younger than 50 years, mean self-rated health was significantly lower in 1993 than in 1984 and 1987 (p <.05). For the age group 60–69 years, the mean value in 1993 was significantly lower than the three earlier years (p <.003). For the oldest age group (70 years and older), the mean value in 1993 was statistically lower than in the years 1987 and 1984, and the mean value in 1990 was statistically lower than in the year 1984 (p <.05).

View larger version (13K): [in this window] [in a new window]   Figure 2. Estimated mean values of self-rated health and associated two-sided 95% confidence intervals by age group and four time points (1984–1993)

When we included childhood SES and sum of illness variables as linear covariates in the model, the main effects of age group, time, and sum of illness were significant. The age group-by-time interaction was no longer significant. Results are illustrated in Figure 3. There were no changes in mean values of self-rated health within age groups between 1987, 1990, and 1993. The time effect was due to each age group having significantly lower self-rated health at baseline in 1984 compared with all of the following measurement occasions (1987, 1990, and 1993; p <.01,.01, and.02 respectively). These results indicate that sum of illness accounted for a significant amount of change over time in self-rated health but did not explain cohort differences. Age group differences remained. The oldest age group showed statistically significant lower mean self-rated health than age groups 50–59 years and younger than 50 years, and the age group 60–69 years showed significantly lower mean self-rated health than age group younger than 50 years.

View larger version (14K): [in this window] [in a new window]   Figure 3. Estimated mean values of self-rated health and associated two-sided 95% confidence intervals by age and four time points controlling for sum of illness and socioeconomic status

View larger version (12K): [in this window] [in a new window]   Figure 4. Estimated variances of self-rated health and associated two-sided 95% confidence intervals by age group and four time points (1984–1993)

View larger version (13K): [in this window] [in a new window]   Figure 5. Estimated variances of self-rated health and associated two-sided 95% confidence intervals by age and four time points controlling for sum of illness and socioeconomic status

If the missing data can be considered missing at random, the estimates obtained from the maximum likelihood estimation are unbiased. If this assumption is not true, the missing data are not ignorable and the missing mechanisms should be modeled (Little & Rubin, 1987). With the advantage of technical and conceptual simplicity, we performed a series of sensitivity analyses, testing the effects of assumptions regarding the "true" responses by those who missed a measurement occasion. We computed multiple imputations assuming correlated multivariate normal distribution for the data within twin pairs and then recalculated our original analysis. Data missing because of death were not imputed, but the data available before death were used. In four separate analyses, the imputed values were then decreased (or increased) by –0.5, +0.5, –2, and +2 units, respectively, to test the consequences of assuming that nonresponse represented worse (or better) true values. The results were consistent with the original analyses of means and variance development (data not shown), with only minor differences from the original results. When imputed values were negatively adjusted, there were only slightly lower means and higher variances, whereas for a positive adjustment, the opposite was found.

Reasons for nonresponses were available for 44% of the missing subjects, with 28% having died during the follow-up period and 16% indicating that they were too sick to participate. On this basis, we believe the nonobserved self-rated health data represent poorer ratings rather than the contrary. This assumption is also supported by evaluating the means and variances for those who participated in the questionnaire follow-ups as a function of number of occasions of participation. There is a pattern resembling terminal decline with decreases in means at the last occasion of measurement (see Figure 6). We can see by visual inspection that mean values are lower for those who participated only at the first questionnaire occasion compared with those who participated at more than one occasion. Those participating in all four occasions have the greatest mean level stability, whereas the variance is higher for those who participated only the first time and lower for those who participated at several occasions. Fewer twins participated at all four occasions in the oldest age group; therefore, we also looked specifically at this group. The twins who participated at the fourth occasion had lower variability in 1984 than those who dropped out over the course of the study. This finding suggests that these individuals are farther away from death and hence less variable because they are all healthier.

View larger version (20K): [in this window] [in a new window]   Figure 6. Means and variances for self-rated health for all twins participating in the questionnaires in a sequential order

Quantitative Genetic Analysis
A fundamental assumption of quantitative genetic analysis based on twin data is that variances are equal in MZ and DZ twins. Analyses of variance in SAS and Mx yielded no differences in means and variances between MZ and DZ twin groups for the self-rated health scale except for the age group 50–59 years. Violations of this assumption typically result in a worsening of fit of the genetic models but no influence on parameter estimates.

There were no significant differences in intrapair similarity between the two rearing groups, and shared rearing environmental effects did not explain significant proportions of the variance. Thus, in the following, we will focus on the results where rearing is not specifically taken into account.

Figure 7 illustrates the total raw variances and the genetic and environmental variances based on the most parsimonious model evaluated from the four occasions 1984–1993 for each of the four age groups as well as average total variance over time for each age group. Table 4 shows the fit statistics of the model fitting procedures for the full and most parsimonious models based on Akaike information criteria. The most parsimonious model for all of the age groups suggests that shared environmental factors (C) in common to the four measurement occasions could be dropped from the model; that is, only shared environmental factors unique to each time point were kept in the model. Nonshared environmental variance was the greatest source of variance for all age groups at most of the measurement occasions, except for the last measurement occasion for the oldest age group. Most of the remaining variance was explained by genetic factors.

View larger version (22K): [in this window] [in a new window]   Figure 7. Total raw variances divided into genetic and environmental variance of self-rated health at four occasions (1984–1993), within four age groups. The best model is presented where we dropped shared environmental effects (C). Time-specific A, C, and E are still in the model as well as common A and common E. A = genetic variance; C = shared environmental variance; E = nonshared environmental variance. aTwins younger than 50 years were all born after 1934, age group 50–59 years was born between 1925 and 1934, age group 60–69 years was born between 1915 and 1924, and age group older than 70 years was born before 1915.

	DISCUSSION

TOP Abstract Method Results Discussion Conclusions References

Cohort Versus Longitudinal Effects
We found that previous cross-sectional reports of a decrease in mean level and an increase in total variance with age are not fully replicated in longitudinal approaches. We found mean levels to remain relatively stable over a period of 9-years within each age group after inclusion of morbidity data (sum of illness) in our model. However, we also found that younger age groups report a better self-rated health status than the older age groups. Age differences in mean level have been reported earlier, and in general, older people are more likely to rate their health as poor (e.g., Murray et al., 1982). Consistent with the role of health in accounting for self-rated health (Ferraro & Kelley-Moore, 2001; Undén & Elofsson, 1998), the effect of change over time was largely eliminated when disease status was taken into account in our analysis. However, cohort differences remained, suggesting that other factors play an important role for self-rated health across different age groups. Different birth cohorts might be affected by societal changes in different ways (Fritzell & Lundberg, 2000). There has been strong economic development in Sweden since 1940. Twins born 1914 or earlier grew up when Sweden was fairly poor and during a time of multiple epidemics. Industrialization increased the standard of living, and the age groups born after World War I might have gained more from that development than the prewar cohorts. In support of this hypothesis, our comparison of SES during childhood for the different age groups demonstrates that younger age groups had a higher childhood SES level. However, inclusion of childhood SES in our mixed effect model did not explain the cohort differences in self-rated health to any great extent. Societal changes not tapped by our SES measure are more likely to explain these differences.

Our results also show that the increase in total variance is greater between age groups than longitudinal contrasts within age groups. This finding is consistent with the literature contrasting cross-sectional with longitudinal findings within the same study for other gerontological variables (Birren & Birren, 1990; Nelson & Dannefer, 1992; Schaie & Baltes, 1996). The age at the last measurement occasion of each age group is the same as the age of first occasion of the next age group. The total variance estimates at the last occasion for the youngest age group matches the estimates of the first occasion for the second age group. However, there are clearer differences in the start and ending points for the last two age groups (60–69 years and 70 years and older), suggesting clearer cohort differences and little longitudinal change for these age groups. Neither childhood SES nor an index of morbidity explained these findings.

Sources of Variance
The contributions of genetic and environmental variance to the total variance for all four age groups replicate well what has been found in previous cross-sectional studies (Christensen et al., 1999; Harris et al., 1992a,b; Lichtenstein & Pedersen, 1995; Romeis et al., 2000; Svedberg et al., 2001). Both genetic and environmental variance appear to increase longitudinally, although for the youngest age group in our present study, there is not much variability. The increased variability is due to both genetic and environmental sources of influence. Increasing genetic variability might result from the fact that the twins are coming into the age when many chronic diseases that are influenced by genetic effects have their debut, for example, coronary heart disease, hypertension, diabetes, etc. Some people become affected, and others remain disease-free, leading to increased individual differences.

Stability in Self-Rated Health Over Time
Stability of self-rated health, that is, interoccasion phenotypic correlation, is lower for the youngest age group compared with the older age groups, which might reflect the restricted range in variation or the onset of health-related conditions. Primarily nonshared environmental factors contribute to stability for the youngest age group, and both genetic and environmental factors contribute fairly equally to stability in the other age groups. A somewhat higher environmental component of stability indicates that the same environmental factors had an impact at more than one occasion. Such nonshared environmental factors could, for example, include social relationships, education, and accidents that have an enduring effect on self-rated health. Equal mediation by genetic and environmental effects has also been found for personality traits (Pedersen & Reynolds, 1998). The correlations between times are not explained by shared environmental effects in any of the age groups. This is consistent with our finding from our mixed effect model where childhood SES, a typical measure of shared environment, does not explain cohort differences in variance or time effects in self-rated health.

Strengths and Limitations
This study was restricted to a somewhat short follow-up time period of 9 years, yet we were able to impose a "synthetic cohort sequential" design such that the width of the age group was the same as the length of the follow-up time. Only pairs where both members responded at any one point in time contributed to information on covariation (i.e., genetic variance); however, it is a strength that we actually can include information on all twins to receive adequate total variance estimates in our Mx models.

The effect of attrition is notable at least for the oldest age group, an expected issue with a longitudinal project. Many in the oldest group participated at fewer than all four occasions. As reported, those who "survived" four occasions of measurement were less variable than those who dropped out. This may very well have to do with terminal decline or selection. Less healthy persons in a population also tend to refuse answering questions both initially and at follow-up occasions (Maddox & Douglass, 1974), another liability in longitudinal studies in general. If anything, our results are attenuated by attrition.

	CONCLUSIONS

TOP Abstract Method Results Discussion Conclusions References

 
Our findings suggest that previous reports of an increase in variance in self-rated health with increased age primarily reflect cohort differences and are not fully replicated longitudinally. Mean levels remain quite stable within age groups over time, and changes with age in level of self-rated health seem to be explained primarily by illness. For variance as well, we found more substantial cohort differences than longitudinal changes, suggesting that the influence of socially mediated and individual-specific environmental effects may be greater than individual differences due to onset of genetically influenced diseases. The phenotypic stability over a 9-year time period is explained almost equally by genetic and environmental factors for all adult ages.

	Acknowledgments

 
SATSA has been supported by the John D. and Catherine T. MacArthur Foundation Research Network on Successful Aging and the National Institute on Aging (Grants AG-04563 and AG-10175). The work on this manuscript was also supported by a scholarship from Erik and Edith Fernström Foundation.

	Footnotes

 
Decison Editor: Charles F. Longino, Jr., PhD

Received for publication January 28, 2004. Accepted for publication January 18, 2005.

	References

TOP Abstract Method Results Discussion Conclusions References

 

• Berg, S. (1996). Aging, behavior, and terminal decline. In J. E. Birren & K. W. Schaie (Eds.), Handbook of the psychology of aging (4th ed., pp. 323–337). San Diego: Academic Press.
• Birren, J. E., Birren, B. A. (1990). The concepts, models, and history of the psychology of aging. In Handbook of the psychology of aging (3rd ed., pp. 3–20). New York: Academic Press.
• Birren, J. E., Schroots, J. J. F. (1996). History, concepts, and theory in the psychology of aging. In J. E. Birren & K. W. Schaie (Eds.), Handbook of the psychology of aging (4th ed., pp. 3–23). San Diego: Academic Press.
• Christensen, K., Holm, N. V., McGue, M., Corder, L., Vaupel, J. W. (1999). A Danish population-based twin study on general health in the elderly. Journal of Aging and Health, 11,49-64.[Abstract/Free Full Text]
• Ferraro, K. J., Kelley-Moore, J. A. (2001). Self-rated health and mortality among black and white adults: Examining the dynamic evaluation thesis. Journal of Gerontology: Social Sciences, 56B,S195-S205.[Abstract/Free Full Text]
• Finkel, D., Pedersen, N. L., Reynolds, C. A., Berg, S., de Faire, U., Svartengren, M. (2003). Genetic and environmental influences on decline in biobehavioral markers of aging. Behavior Genetics, 33,107-123.[Medline]
• Finkel, D., Reynolds, C. A., McArdle, J. J., Gatz, M., Pedersen, N. L. (2003). Latent growth curve analyses of accelerating decline in cognitive abilities in late adulthood. Developmental Psychology, 39,535-550.[Medline]
• Fritzell, J., Lundberg, O. (2000). The Swedish Level of Living Survey: Longitudinal research on life chances over the live course. In C.-G. Janson (Ed.), Seven Swedish longitudinal studies in behavioural sciences (pp. 172–190). Stockholm: Swedish Council for Planning and Coordination of Research (FRN).
• Goldstein, M. S., Siegel, J. M., Boyer, R. (1984). Predicting changes in perceived health status. American Journal of Public Health, 74,611-614.[Abstract/Free Full Text]
• Harris, J. R., Pedersen, N. L., McClearn, G. E., Plomin, R., Nesselroade, J. R. (1992). Age differences in genetic and environmental influences for health from the Swedish Adoption/Twin Study of Aging. Journal of Gerontology: Psychological Sciences, 47B,P213-P220.
• Harris, J. R., Pedersen, N. L., Stacey, C., McClearn, G. E., Nesselroade, J. R. (1992). Age differences in the etiology of the relationship between life satisfaction and self-rated health. Journal of Aging and Health, 4,349-368.[Abstract/Free Full Text]
• House, J. S., Kessler, R. C., Herzog, A. R. (1990). Age, socioeconomic status, and health. Milbank Quarterly, 68,383-411.
• Idler, E. L. (1993). Age differences in self-assessments of health: Age changes, cohort differences, or survivorship? Journal of Gerontology, 48,S289-S300.[Abstract]
• Idler, E. L., Benyamini, Y. (1997). Self-rated health and mortality: A review of twenty-seven community studies. Journal of Health and Social Behavior, 38,21-37.[Medline]
• Leinonen, R., Heikkinen, E., Jylha, M. (1998). Self-rated health and self-assessed change in health in elderly men and women—A five-year longitudinal study. Social Science and Medicine, 46,591-597.
• Liang, J., Bennett, J., Krause, N., Kobayashi, E., Kim, H., Brown, J. W., et al. (2002). Old age mortality in Japan: Does the socioeconomic gradient interact with gender and age? Journal of Gerontolology: Social Sciences, 57B,S294-S307.
• Lichtenstein, P., De Faire, U., Floderus, B., Svartengren, M., Svedberg, P., Pedersen, N. L. (2002). The Swedish Twin Registry: A unique resource for clinical, epidemiological and genetic studies. Journal of Internal Medicine, 252,184-205.[Medline]
• Lichtenstein, P., Harris, J. R., Pedersen, N. L., McClearn, G. E. (1992). Socioeconomic status and physical health, How are they related? An empirical study based on twins reared apart and twins reared together. Social Science and Medicine, 36,441-450.
• Lichtenstein, P., Pedersen, N. L. (1995). Social relationships, stressful life events, and self-reported physical health: Genetic and environmental influences. Psychology and Health, 10,295-319.
• Little, R. J. A., Rubin, D. B. (1987). Statistical analysis with missing data. New York: Wiley.
• Loehlin, J. C. (1996). The Cholesky approach: A cautionary note. Behavior Genetics, 26,65-69.
• Maddox, G. L., Douglass, E. B. (1973). Self-assessment of health: A longitudinal study of elderly subjects. Journal of Health and Social Behavior, 14,87-93.[Medline]
• Maddox, G. L., Douglass, E. B. (1974). Aging and individual differences: A longitudinal analysis of social, psychological, and physiological indicators. Journal of Gerontology, 29,555-563.[Abstract/Free Full Text]
• Markides, K. S., Lee, D. J. (1990). Predictors of well-being and functioning in older Mexican Americans and Anglos: An eight year follow-up. Journal of Gerontology: Social Sciences, 45B,S69-S73.
• Martin, N., Boomsma, D., Machin, G. (1997). A twin-pronged attack on complex traits. Nature Genetics, 17,387-392.[Medline]
• Mossey, J. M., Shapiro, E. (1982). Self-rated health: A predictor of mortality among the elderly. American Journal of Public Health, 72,800-808.[Abstract/Free Full Text]
• Murray, J., Dunn, G., Tarnopolsky, A. (1982). Self-assessment of health: An exploration of the effects of physical and psychological symptoms. Psychological Medicin, 12,371-378.
• Neale, M. C., Boker, S. M., Xie, G., Maes, H. H. (2002). Mx: Statistical modeling (6th ed.). Richmond, VA: Department of Psychiatry.
• Neale, M. C., Maes, H. H. M. (2002). Methodology for genetic studies of twins and families. Dordrecht, the Netherlands: Kluwer Academic Publisher B.V.
• Nelson, E. A., Dannefer, D. (1992). Aged heterogeneity: Fact or fiction? The fate of diversity in gerontological research. The Gerontologist, 32,17-23.[Abstract]
• Pedersen, N. L., McClearn, G. E., Plomin, R., Nesselroade, J. R., Berg, S., de Faire, U. (1991). The Swedish Adoption Twin Study of Aging: An update. Acta Geneticae Medicae et Gemellologiae (Roma), 40,7-20.
• Pedersen, N. L., Reynolds, C. A. (1998). Stability and change in adult personality: Genetic and environmental components. European Journal of Personality, 12,365-386.
• Pedersen, N. L., Steffensson, B., Berg, S., Johansson, B., McClearn, G. E. (1999). The importance of genetic and environmental effects for self-reported health symptoms: A 30 year follow-up considering survival and selection effects. Journal of Aging and Health, 11,475-493.[Abstract/Free Full Text]
• Plomin, R., DeFries, J. C. (1981). Multivariate behavioral genetics and development: Twin studies. In Twin research 3: Intelligence, personality, and development (pp. 25–33). New York: Liss.
• Plomin, R., DeFries, J. C., McClearn, G. E. (1990). Behavioral genetics: A primer (2nd ed.). New York: Freeman.
• Romeis, J. C., Scherrer, J. F., Xian, H., Eisen, S. A., Bucholz, K., Heath, A. C., et al. (2000). Heritability of self-reported health. HSR: Health Services Research, 35,995-1010.
• Rowe, J. W. (1985). Health care of the elderly. New England Journal of Medicine, 312,827-835.[Medline]
• SAS/STAT. (1999-2001). SAS/STAT user's guide (version V8). Cary, NC: SAS Institute.
• Schaie, K. S., Baltes, P. B. (1996). Intellectual development in adulthood: The Seattle Longitudinal Study. New York: Cambridge University Press.
• Seeman, T. E., Crimmins, E. (2001). Social environment effects on health and aging: Integrating epidemiologic and demographic approaches and perspectives. Annals of the New York Academy of Science, 954,88-117.[Medline]
• Svedberg, P., Lichtenstein, P., Pedersen, N. L. (2001). Age and sex differences in genetic and environmental factors for self-rated health: A twin study. Journal of Gerontology: Social Sciences, 56B,S171-S178.[Abstract/Free Full Text]
• Thomas, C., Kelman, H. R., Kennedy, G. J., Ahn, C., Yang, C. (1992). Depressive symptoms and mortality in elderly persons. Journal of Gerontology: Social Sciences, 47B,S80-S87.
• Tibblin, G., Cato, K., Svardsudd, K. (1990). Goteborg Quality of Life Study of men born in 1913 and 1923—Age, sex, job satisfaction and cardiovascular diseases. Scandinavian Journal of Primary Health Care (Suppl), 1,39-45.
• Tibblin, G., Tibblin, B., Peciva, S., Kullman, S., Svardsudd, K. (1990). The Goteborg Quality of Life Instrument—An assessment of well-being and symptoms among men born 1913 and 1923. Methods and validity. Scandinavian Journal of Primary Health Care (Suppl), 1,33-38.
• Undén, A. L., Elofsson, S. (1998). Självupplevd hälsa: Faktorer som påverkar människors egen bedömning [Self-rated health: Factors influencing perceived health status]. Stockholm: Swedish Council for Planning and Coordination of Research (FRN).
• Viken, R. J., Rose, R. J., Kaprio, J., Koskenvuo, M. (1994). A developmental genetic analysis of adult personality: Extraversion and neuroticism from 18 to 59 years of age. Journal of Personality and Social Psychology, 67,722-730.

This article has been cited by other articles:

				S.-T. Cheng, H. Fung, and A. Chan Maintaining Self-Rated Health Through Social Comparison in Old Age J. Gerontol. B. Psychol. Sci. Soc. Sci., September 1, 2007; 62(5): P277 - P285. [Abstract] [Full Text] [PDF]

This Article Abstract Full Text (PDF) Services Download to citation manager Citing Articles Citing Articles via HighWire PubMed PubMed Citation