| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
RESEARCH ARTICLE |
1 The Finnish Centre for Interdisciplinary Gerontology, University of Jyväskylä, Finland.
2 Department of Health Sciences, University of Jyväskylä, Finland.
3 Institute of Gerontology, Jönköping University, Sweden.
Address correspondence to Tiina-Mari Lyyra, The Finnish Centre for Interdisciplinary Gerontology, PO Box 35 (Viveca), FIN-40014 University of Jyväskylä, Finland. E-mail: tiina-mari.lyyra{at}sport.jyu.fi
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionReferences |
|---|
Life satisfaction, one of the key aspects of psychological well-being, is a desired subjective feeling closely related to morale and adjustment (Andrews & Withey, 1976
; Diener, Suh, Lucas, & Smith, 1999). There is growing evidence that in addition to physical factors, psychosocial factors are important predictors of mortality, especially in old age. Indicators of subjective well-being, such as positive affect, optimism, and positive self-perceptions of aging, have been found to be associated with survival in older people (Danner, Snowdon, & Friesen, 2001; Giltay, Geleijnse, Zitman, Hoekstra, & Schouten, 2004; Levy, Slade, Kunkel, & Kasl, 2002; Maier & Smith, 1999; Mete, 2004; Ostir, Markides, Black, & Goodwin, 2000; Penninx, Guralnik, & Bandeen-Roche, 2000; Pitkälä, Laakkonen, Strandberg, & Tilvis, 2004), and negative affect, hopelessness, and depressiveness have been found to be associated with mortality risk (Adamson, Price, Breeze, Bulpitt, & Fletcher, 2005; Stern, Dhanda, & Hazuda, 2001; Wilson, Bienias, Mendes de Leon, Evans, & Bennett, 2003). Among samples containing young and middle-aged people, life dissatisfaction has been found to predict job disability, mental disorders, injury mortality, suicides, and total mortality (Koivumaa-Honkanen et al., 2000, 2001, 2004; Koivumaa-Honkanen, Honkanen, Koskenvuo, Viinamäki, & Kaprio, 2002).
In the gerontological literature, life satisfaction has mainly been used as an end point; that is, its determinants and predictors have been studied to the neglect of its possible health effects. Studies of the prognostic impact of life satisfaction in an older population are scarce, although this issue is particularly important in old age when various deficits in health, functional capacity, and social network become more common. In addition, the associations between different dimensions of life satisfaction and mortality remain unclear, and the potential confounders have usually been based on self-reports of various domains of health and functional capacity in studies of subjective well-being and mortality. In the present study, we examined how self-reported life satisfaction and its dimensions predict mortality in octogenarian Swedish twins over a 10-year follow-up. Our underlying hypothesis was that higher perceived life satisfaction would be associated with lower mortality, even after different domains of health and functioning were controlled for.
Poor health, feelings of loneliness, low socioeconomic status, unmarried or widowed status, and lack of social support have been found to be related to life dissatisfaction (Diener, 1984; Hillerås, Jorm, Herlitz, & Winblad, 2001; Mroczek & Spiro, 2005). Personality traits such as extraversion and neuroticism have also been identified as powerful predictors of the level of life satisfaction (Costa & McCrae, 1980) as well as a higher sense of coherence (Sagy, Antonovsky, & Adler, 1990).
Possible pathways have been put forward that link subjective well-being and health. First, it has been suggested that positive emotions promote a more active lifestyle and a motivation toward self-care (Scheier & Carver, 1992). Second, positive affect is also associated with social engagement, which has been found to be related to better survival in older people (e.g., Berkman & Syme, 1979; Lyyra & Heikkinen, 2006). Third, older people with better subjective well-being may have better coping capabilities; optimists usually avoid strategies such as denial and giving up (Segerstrom, Taylor, Kemeny, & Fahey, 1998). Another major research interest concerns different psychophysiological patterns, where the focus is on cardiovascular reactivity or immune response (Friedman, 2000).
The scales most frequently used to measure life satisfaction were originally constructed by Neugarten, Havighurst, and Tobin (1961). The original index, the Life Satisfaction Index A (LSIA), is comprised of 20 items, of which 12 are positive and 8 negative. The Life Satisfaction Index Z (LSIZ) was proposed by Wood, Wylie, and Sheafor (1969) as a refinement of the LSIA, containing 13 of the original 20 items.
The original LSIA is based on a specific conceptual definition of life satisfaction, and several studies have examined its factor structure. Liang (1984), for example, found three first-order factors (Mood Tone, Zest, and Congruence between hopes and reality) and one second-order factor representing general subjective well-being. Shmotkin (1991) found three factors of Zest, Mood Tone, and Congruence. Locating them on a time perspective implies that congruence deals with judgement directed the past; mood tone and zest contain elements of present life and future orientation.
In the present study we sought to build on the current research in two ways. First, we investigated the factorial structure of LSIZ; second, we examined the association between all-cause mortality and LSIZ factors in Swedish octogenarian twins.
|
METHODS |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
), a longitudinal study of same-sex twins, aged 80 and older, living in Sweden. This population-based sample was drawn from the Swedish Twin Registry (Cederlöf & Lorich, 1978). From the eligible 549 twin pairs born in 1913 or earlier, a total of 351 pairs (702 individuals) participated in the baseline measurements conducted during the period from 1991 through 1994, although for 111 pairs one or both members of the pair were unable to complete the tests for reasons of suspected dementia (88 pairs) or major sensory and motor handicaps (23 pairs). The preliminary diagnosis of dementia was based on the performance on a battery of tests for dementia and cognitive impairment, including the Mini-Mental State Examination (McClearn et al. 1997). Mean age at inclusion was 83.7 years. The representativeness of the twin sample in relation to the general population was tested in a separate study by comparing one randomly chosen member of each twin pair to a population-based nontwin sample (Simmons et al., 1997). Furthermore, the prevalence of malignancy in the OCTO-Twin sample corresponds with the figures published by the population-based Swedish Cancer Registry (Nilsson, Johansson, Berg, Karlsson, & McClearn, 2002). This study consists of data on the 320 individuals (235 twin pairs; 126 men and 194 women) who, at the baseline, answered the life satisfaction questions and participated in the clinical tests. Survival data were obtained from the national Swedish population register and were updated until September 4, 2003, to provide a follow-up time of about 10 years.
Measures
The twins were assessed in their place of residence face-to-face by a trained research nurse. Different nurses assessed each twin of a pair to minimize bias. The nurse administered comprehensive batteries of structured assessments to obtain information on the following: sociodemographics, education, health, cognitive and functional capacity, contacts and friends, personality, personal control, and psychological well-being (Haynie, Berg, Johansson, Gatz, & Zarit, 2001). We selected potential confounders from these assessments on the basis of the earlier literature and initial analyses showing them to have an association with mortality in older people or with life satisfaction or both.
Satisfaction with life
We measured satisfaction with life by using the13-item LSIZ (Wood et al., 1969), which, as we stated earlier, is a short form of the LSIA (Neugarten et al., 1961). Instead of the original 2-point agree–disagree score, we used a 5-point scale from 1 = strongly disagree to 5 = strongly agree. A higher score indicated greater satisfaction with life.
Potential Confounders
Number of serious illnesses
We based the number of serious illnesses on medical records obtained from hospitals, outpatient clinics, district physicians, and primary health care centers (Nilsson et al., 2002). A physician did a concurrent review of medical records, including medical history, use of medicine, and self-reported information about diseases. We computed a sum score on the basis of the presence of the following diagnoses: cancer, thyroid disease, diabetes, asthma, heart insufficiency, angina pectoris, coronary infarction, thrombosis in the leg, stroke, brain hemorrhage, multiple sclerosis, and kidney disease requiring dialysis.
Sociodemographic characteristics
These characteristics were represented by the length of formal education in years and whether or not the participant was living alone at the baseline.
Depressive symptoms
We assessed symptoms of depression by using the Center for Epidemiologic Studies–Depression scale (CES-D), which was developed for the purpose of screening for depressive symptomatology in community and epidemiological studies (Radloff, 1977). The questionnaire consists of 20 items, each rated on a 4-point scale indicating the frequency of mood problems during the past week. If one or two items were missing, we imputed them by using the mean of the individual's score. We excluded participants lacking more than two items from the analysis. We calculated the sum score of the items; a higher score indicated more depressive symptoms. In the present study, the internal consistency was good (
= 0.84).
Physical functioning
We measured physical functioning by means of a battery of tests requiring balance, upper body strength, flexibility, and dexterity. We tested balance by normal gait, Romberg's test, and semitandem and tandem stand. We tested the reach and mobility of the upper extremities by having the participant lift a glass and a 1-kg package and perform some basal tasks. We assessed manual ability in various tasks of importance for independent activity of daily living functioning by using a special apparatus, including inserting and turning a key, putting in a plug in a socket, screwing in a light bulb, putting coins in a slot, and dialing a number (Aniansson, Rundgren, & Sperling, 1980; Lundgren-Lindquist & Sperling, 1983). The number of tasks that were completed without difficulty indicated better physical functioning. The sum score showed adequate internal consistency ( = 0.80).
Cognitive functioning
We assessed cognitive functioning by means of three tests based on Thurstone's theory of primary mental abilities, measuring verbal meaning, inductive reasoning, and spatial ability (Thurstone 1938, Dureman & Sälde 1959). We calculated the z score in each cognitive test, and we used the mean of the z scores as an indicator of cognitive functioning. The sum score showed adequate internal consistency ( = 0.80).
Frequency of social contacts
We measured social contact frequency by asking the participants to give a rating on a 7-step rating scale in response to this question: "How often do you see or phone children, siblings, relatives, friends, or acquaintances?" (Response options were 1 = never, 2 = hardly ever, 3 = every year, 4 = every month, 5 = every week, 6 = several times a week, 7 = every day.)
Statistical Methods
Our aim was to study mortality risk at the individual level and, thus, we had to take into account the assumed dependency within the sampled twin pairs caused by influences from genetic background and shared environment. For descriptive purposes, we estimated means and standard errors and we tested the significance of mean differences by using standard methods for complex sampling designs (see, e.g., Levy & Lemeshow, 1999). Similarly, we converted chi-square tests for cross-tables to F tests to correct for the sampling design (Rao & Scott, 1987). We carried out an exploratory factor analysis on the polychoric correlation matrix of the life satisfaction questionnaire items by using PRELIS (Jöreskog & Sörbom, 2001). We used maximum likelihood estimation together with promax rotation to allow for correlation among factors, and we computed factor scores for the rotated loading matrix by using PRELIS and divided them into quartiles with the two middle groups combined for further analysis in the shared frailty model. Factor analysis with item data for men and women separately produced factor scores that correlated highly with corresponding scores from an analysis with pooled data (r =.94,.97, and.88 for the three factors). For this reason, we used the factor scores from the pooled data set in the subsequent analyses to ensure a sufficient sample size for their estimation.
We used the semiparametric Cox regression mixed-effects model (i.e., shared frailty model) to examine mortality risk. Frailty modeling can be used in the analysis of clustered time-to-event data to directly account for a correlation among participants, such as the survival times among twin pairs. We can write the hazard function for fixed effects and a random effect for the twin pairs in the shared frailty model as
|
|
(t) is the baseline hazard function, xij is a p x 1 vector of fixed covariates for individual i in pair j, ß is a p x 1 vector regression coefficient, and Uj is a pair-specific random variable. The latter component in the exponentiated term is an extension of the conventional Cox regression model in that it is modeled differently from the fixed effects.
We can interpret the random variable, U, as excess risk for the mortality for a given twin pair. It is not observable in measurement and its values have to be estimated simultaneously with the coefficients for the fixed effects. In order for the model to be identifiable, the distribution and mean of the random variable are fixed. The logarithm of U cannot take negative values and, thus, the available distributions for the random variable are restricted to distributions with positive values only, such as the log-normal distribution or the more general gamma distribution. The mean is set to zero while estimates of the random variable and their variance, the frailty variance (
2), are parameters estimated in the analysis. A straightforward interpretation can be given to the exponent of the standard error (i.e., square root of variance) of frailty: it indicates to what extent, on average, the excess risk for mortality varies between twin pairs. Whereas the variance of frailty can be used as a summary statistic of excess mortality risk, the estimated random variable coefficients are useful for identifying individual pairs at higher risk relative to the others.
In this study we assumed the logarithm of the random variable to be distributed as N(0, 2), and we used penalized partial likelihood to estimate values of random variables. This method is based on maximizing the partial likelihood function by first treating both fixed and random effects as fixed variables in the usual Cox regression model, and then updating the frailty variance estimate on the basis of the current random-effect estimates. In frailty modeling we adjust the usual proportional hazards partial likelihood with a penalizing term,
|
|
2 is the variance of the frailty variable, and U is the random variable adjusting the model for the within-twin pair correlation. More details of this method with an example can be found in McGilchrist (1991). We constructed shared frailty models with STATA, version 8 (StataCorp., 2003). |
RESULTS |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
|
2 = 102.09, df = 42, p <.001), suggesting that unexplained variance remained. The unrotated factor structure explained about 46% of the variance. Although the interfactor correlations among the three factors of LSIZ were quite high, three-factor solution was supported by statistical tests: Cattell's scree test (Cattell, 1966) indicated that three factors were useful in the case of these data, and when we compared the observed three eigenvalues to random eigenvalues obtained from a parallel analysis (Lautenschlager, 1989), again three factors remained extractable. Finally, in examining the factor configurations, interpretability and results from previous studies guided us to extract three factors.
|
2 = 0.26, p =.045). In this case the exponentiated standard error of frailty was 1.66, which indicates that, in addition to the life satisfaction variables, the average excess risk for mortality varied about 66% above and below the norm between the twin pairs. With the highest quartile as the reference group, the unadjusted HRs for the lowest and two middle quartiles of the Congruence factor for the pooled data were HR = 1.09, CI = 0.78–1.53 and 1.09, CI = 0.74–1.60, respectively. This factor representing satisfaction with past life was not significantly associated with mortality in any of the models (results not shown here).
Multivariate Models for Present Life Satisfaction and Mortality
Table 3 presents the results for zest factor and mortality in octogenarians. Those in the lowest Zest factor quartile had a risk for death that was twofold greater than that of those in the highest quartile when sex, age, number of serious illnesses, and variance of frailty were taken into account (Model 1). We obtained the same results when we controlled for the indicators of sociodemographics (Model 2), depressiveness (Model 3), and functional, social, and cognitive functioning (Model 4). In the full model, adjusted for all the covariates, the risk for mortality for those in the lowest quartile was 1.9 times higher than that for those in the highest quartile. We estimated the variance of frailty in all of the models, but it was not statistically significant.
|
The results indicate that an independent association exists between higher satisfaction with present life in terms of zest and mood and lower mortality in octogenarians.
|
DISCUSSION |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
The basic three-factor structure of the LSIA, which consists of the factors Zest, Mood Tone, and Congruence, has also been confirmed in several earlier studies in samples of older people in the United States and Israel (Shmotkin, 1991 Liang, 1984; Hoyt & Creech 1983; Adams, 1969). Although the LSIZ is a shortened version of the LSIA, the same factors could be identified also here; in this sample of old-old Swedes, the first two factors represented zest and mood, and the third represented congruence. According to Neugarten and colleagues (1961, p. 137) zest refers to taking pleasure "from the round of activities that constitutes [one's] everyday life," mood tone refers to maintaining "happy and optimistic attitudes and mood," and congruence refers to the feeling of "succeeding in achieving [one's] major goals." Liang (p. 614) defines zest as "an optimistic and positive outlook of life at the present and in the future as opposed to the past," mood tone as "happiness, which is a positive affect of cognitive assessment of positive affects," and congruence as the "degree to which one has attained one's desired goals."
Socioemotional selectivity theorists argue that when people perceive their time in life to be limited, futuristic goals become less relevant and present oriented, and emotional goals become more important (Carstensen, Isaacowitz & Charles, 1999; Cheng, 2004). Furthermore, studies in which optimistic attitude and feelings of hope have been found to be related to better survival in old age support our finding. Low satisfaction with life may cause older people to vacillate in their will to live, which may be manifested either indirectly or directly. Psychosocial status may also be an outcome of good health rather than a cause, as the more positive old persons tend to be those who have experienced less illness.
It has been proposed that psychological factors affect mortality directly by influencing immune system functioning (e.g., Segerström et al. 1998) or indirectly by influencing, for example, health-related behaviors. We found that depressiveness was not a mediator between life satisfaction and mortality, even when the items representing mood tone were themselves strongly associated with mortality. If immune system functioning were a mediator, then those people less satisfied in their lives would also have been depressed. In Nordic older samples, especially among women, health behaviors are not highly important mediators, as the number of smokers or drinkers tends to be low.
A growing number of studies have documented the beneficial effect of social support on survival (e.g. Berkman & Syme 1979; Lyyra & Heikkinen 2006), this support is also positively related to life satisfaction (Salokangas, Joukamaa & Mattila 1991). In this study, social functioning measured in number of contacts did not diminish the predictive value of higher present life satisfaction on lower mortality, although the result might have been different if we had used the measures of perceived social support or quality of contacts as covariates. Moreover, the participants in this study had a living twin sister or brother at the baseline.
In addition to social support, sociodemographic variables such as marital and socioeconomic status have been found to be strongly related to life satisfaction (Mroczek & Spiro, 2005) and also to mortality in older people. In this study we used information on whether respondents were living alone or not as a covariate. In older samples, this is more informative, as many of the respondents are likely to be widowed or living in institutions. However, in this sample neither living arrangements nor education, when added into the model, had an effect on the HRs.
Poor cognitive performance and poor physical performance are strong predictors of mortality among older people (e.g., Lyyra, Leskinen, & Heikkinen, 2005). In the present study, satisfaction with present life showed an independent association with mortality, also when adjusted with objectively measured cognitive and physical performance measures. It is also notable that the adjustment for the number of serious diseases did not significantly decrease the predictive value of life satisfaction, despite the fact that the high number of serious diseases itself was an important predictor of mortality. It seems that those older people who reported satisfaction with their present lives lived longer in spite of possible illnesses or poor social, cognitive, and physiological functioning. It is also known that even when they have numerous illnesses, older people often describe their subjective health as good (Heikkinen, Leinonen, & Berg, 1997).
We are aware of some limitations in the present study. First, the study sample was rather small. Studying cause-specific mortality might have been more informative and the results may have been slightly different if we had analyzed disease-related and accidental deaths separately. However, we elected to use all-cause mortality as an end point, partly because it would have been difficult to ascertain the exact causes of death for the older participants.
Despite these potential limitations, the population of this study suited these analyses well, and a sufficient number of deaths occurred during the relatively long follow-up time to enable us to perform the required statistical analyses. In addition, because of the infrastructure of Swedish society, the data on dates of deaths were reliable and conclusive. This study utilized a representative sample of old-old Swedish twins to evaluate the potential impact of life satisfaction on survival. The long follow-up period and the use of comprehensive multidisciplinary measures as potential confounders made these data ideal for the analysis. In order to use twins in these data as individuals, we used a shared frailty model, in which within-pair similarity can be taken into account and the information from both individuals in the twin pair can be used.
The earlier mortality studies with data on twins have commonly used a clustering or randomizing design. This raises the problem that not all the data available can be fully utilized. Applying frailty Cox semiparametric modeling, as in the present study, we found it possible to model the dependence between twins and also to test the significance within the frailty modeling framework. This also allowed us to fully utilize all the available data for the twin pairs. Although generally it is likely that significant dependence will be found between both twins in a pair, we did not find any evidence of this. This may be because we were analyzing all-cause mortality and because of the randomness in the way deaths occur among elderly persons. The latter is a major challenge in the application of any method of survival analysis to elderly persons.
More research remains to be done on the links between life dissatisfaction, psychological processes, and relevant health outcomes. For instance, investigating different coping mechanisms could be a promising way to approach the link between life dissatisfaction and mortality in elderly populations. Future development of time-to-event analyses may open new possibilities to study genetic mechanisms underlying life dissatisfaction and mortality.
To conclude, this study found a strong independent association between higher satisfaction with present life and lower mortality in older people, even when a wide set of demographic, lifestyle, and physical and psychological health and functioning variables were controlled. It would appear that emotional well-being, especially when related to the present, is important for the health of older people.
|
|
Acknowledgments |
|---|
|
Footnotes |
|---|
Received for publication November 23, 2005. Accepted for publication April 3, 2006.
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  A. L. Collins, N. Goldman, and G. Rodriguez Is Positive Well-Being Protective of Mobility Limitations Among Older Adults? J. Gerontol. B. Psychol. Sci. Soc. Sci., November 1, 2008; 63(6): P321 - P327. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
