| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
RESEARCH ARTICLE |
a Stockholm University, Swedish Institute for Social Research
b University of Southern California, Andrus Gerontology Center, Los Angeles
Carin Lennartsson, Swedish Institute for Social Research, Stockholm University, S-10691 Stockholm, Sweden E-mail: carin.lennartsson{at}sofi.su.se.
 |
Abstract |
|---|
 TOP Abstract Theoretical ModelMethodsResultsDiscussionReferences |
|---|
Methods. Baseline data derived from the Swedish Panel Study of Living Conditions of the Oldest Old, a nationally representative sample of persons aged 77 years and older living in Sweden in 1992. The authors used factor analysis to apply a simplifying measurement structure to frequency of participation in 10 leisure activities. They used Cox proportional hazard regression to estimate the relative effects of activity factors and other independent variables on the logged hazard rate of mortality up to 1996.
Results. Analyses revealed 4 domains of activities that lie along 2 basic dimensions: solitary-social and sedentary-active. Among men, only participation in activities that were both solitary and active was significantly associated with reduced mortality risk when health variables were controlled. Among women, none of the activity domains was significant when health variables were controlled. For the entire sample, greater participation in solitary-active activities significantly reduced risk of mortality when all other activity domains and health factors were controlled.
Discussion. Although most of the observed associations between activity involvement and survival are a byproduct of the confound between poor initial health and low activity levels, solitary activities have a positive influence on the survival of very old individuals, especially men, suggesting that nonsocial aspects of activities may promote health and longevity in late old age.
RESEARCH suggests that engagement in meaningful and productive activities, often within the context of friendship, kinship, and organizational participation, is a key component in promoting health and reducing the risk of mortality in later life (see Rowe and Kahn 1998
). For instance, Wolinsky, Stump, and Clark 1995 found that non-kin support (defined as a composite of volunteer work, social contact, religious activities, and social activities) significantly reduced the mortality risk in an elderly sample. Sabin 1993 found in a community-dwelling elderly population that attending religious services, volunteering, and visiting with and talking to friends or neighbors reduced their risk of mortality. Participation in hobbies and recreation, both inside and outside the home, has also been found to delay mortality in older populations (Ljungquist, Berg, and Steen 1996; Welin, Larsson, Svardsudd, Tibblin, and Tibblin 1992). Taken together, these studies suggest an inverse correlation between involvement in activities and risk of mortality among elderly persons, strengthening the veracity of claims that active engagement with life has profound health consequences in later life.
Participation in explicitly physical types of activity has been associated with reduced risk of mortality (Bath and Morgan 1998; Iwarsson, Isacsson, Persson, and Schersten 1998; Kaplan, Seeman, Cohen, Knudsen, and Guralnik 1987; Ljungquist and Sundstrom 1996; Morgan and Clark 1997; Parker, Thorslund, and Nordstrom 1992) as well as maintenance of functional ability among older individuals (Lacroix, Guralnik, Berkman, Wallance, and Satterfield 1993; Mor et al. 1989; Moritz, Kasl, and Berkman 1995; Seeman et al. 1995). Physical activities also minimize several risk factors of mortality, such as cardiovascular incidents (Kaplan, Strawbridge, Cohen, and Hungerford 1996; Rosengren et al. 1990), high blood pressure (Rosengren, Tibblin, and Wilhelmsen 1993), and coronary heart disease (Orth-Gomer, Rosengren, and Wilhelmsen 1993; Simonsick et al. 1993).
Recently, Glass and colleagues showed that social activities (e.g., church attendance, recreation, and group activities), productive activities (e.g., gardening, preparing meals, and shopping), and fitness activities (e.g., sports, walking, and exercise) were independently associated with survival of elderly persons, even after functional disability and history of several diseases were controlled (Glass, Mendes de Leon, Marottoli, and Berkman 1999). The authors speculated that the benefits of activities that "involve little or no enhancement of fitness ... confer survival benefits through psychosocial pathways" (p. 478). If activities benefit health through socioemotional avenues, such as by inculcating an internal locus of control (Menec and Chipperfield 1997), then it becomes important for investigators to differentiate physical from nonphysical benefits of activities. Indeed, research on the health advantages of activities has been criticized for mixing potentially health-enhancing attributes of physical activities with those of more sedentary activities, thus making it difficult to trace the ultimate source of any benefits observed (see Riddoch 2000). Further, when interpreting the association between activities and mortality, one must also account for the possibility that underlying health problems are progenitors of functional inactivity, thereby producing spurious results. Virtually all studies on the subject control for some health factors, although there is little consensus on what aspects of health are most important in this regard.
Leisure and productive activities are often performed within the context of social relationships that may provide health-enhancing benefits. Indeed, studies have revealed significant associations between interpersonal relationships and lower morbidity, improved subjective health, and enhanced survival (Cohen and Syme 1985; Hanson 1988; Orth-Gomer et al. 1993; Ostergren 1991; Seeman, Kaplan, Knudsen, Cohen, and Guralnik 1987; Vogt, Mullooly, Ernst, Pope, and Hollis 1992). Some of the most dramatic findings in this regard relate to the ability of social ties to prolong human life. These findings have been replicated across a variety of population-based samples in the United States (Berkman and Syme 1979; Blazer 1982; House, Robbins, and Metzner 1982) and northern Europe (Kaplan et al. 1987), including Sweden (Hanson, Isacsson, Janzon, and Lindel 1989; Orth-Gomer and Johnson 1987). Even in later life, when the process of selective survival and the force of senescent mortality potentially reduce the health benefits of social relationships (see Markides and Machalek 1984), older individuals who have more social involvement tend to have lower mortality rates than their less involved counterparts (see Seeman et al. 1987, for a review). However, because social connections are often cultivated in the context of leisure activities, it has not always been possible for investigators to distinguish whether benefits derive from the social aspect or from the actual content of the activities.
|
Theoretical Model |
|---|
|
TOPAbstractTheoretical ModelMethodsResultsDiscussionReferences |
|---|
We took as our point of departure Rowe and Kahn 1998 notion that engagement with life is one of the three pillars of successful aging. This construct—itself comprising "maintaining close relationships" and "remaining involved in activities that are meaningful and purposeful" (Rowe and Kahn 1998, p. 46)—is a useful umbrella under which to consider the dimensions that underlie social and leisure involvement of aged persons. In this analysis, we examined activity involvement along these two dimensions to ascertain which attributes of activities are likely to extend the lives of oldest old persons. In addition, we examined these effects first without and then with functional health controlled to determine the degree to which any apparent benefit of activities is the consequence of spurious functional health factors that may predispose persons to being active in the first place.
|
Methods |
|---|
|
TOPAbstractTheoretical ModelMethodsResultsDiscussionReferences |
|---|
, for details). The SLLS dropped participants from subsequent panels when they turned 76 years of age. SWEOLD includes those participants who were interviewed at least once in the SLLS, but who were dropped because of its age ceiling restriction. Thus, SWEOLD includes respondents who were aged 77 years and older in 1992 (Lundberg and Thorslund 1996). Interviews were conducted with 537 such respondents or their proxies (12.6%). The large majority (94.2%) of interviews took place in person, and a small number were conducted over the telephone. The response rate was 95.4% of those deemed eligible for inclusion into SWEOLD. Because the survey designers anticipated that few institutionalized respondents would participate in many of the leisure activities asked about in the survey, only respondents living in the community were asked questions about activities. Thus, after we excluded 69 respondents who were living in institutions and 5 respondents who had missing data on all activity items, the operational sample consisted of 463 people. More than 60% of the sample were women, 27% were aged 85 years or older, and about one quarter (24%) had education beyond grade school. Other demographic characteristics are available from the authors.
Dependent Variable
The risk of mortality between 1992 and 1996 served as the outcome variable in this analysis. Mortality information (date of occurrence) was obtained from the Swedish National Cause of Deaths Registry, which maintains records of death certificates. The event was defined as death occurring between the interview day and the end of 1996. Mortality risk was operationalized as the log of the hazard rate at which a mortality event occurs between time t and t+1, given survival to time t, where intervals of time are defined on a daily basis. We used Cox proportional hazard regression to estimate the relative effects of independent variables on the logged hazard rate of mortality. This method has desirable qualities in that it does not require the specification of a functional form for mortality risk over time and appropriately handles censored observations (Blossfeld, Hamerle, and Mayer 1989; Blossfeld and Rohwer 1995). Exponentiated regression coefficients—the relative hazards—from these models were interpreted as the change in the relative risk of mortality associated with a unit change in the independent variable.
Independent Variables: Social and Leisure Activities
Data regarding the level of engagement in activities were obtained with the following question: "Which of the following activities do you usually do?" A list of 17 activities was then read to each respondent. Response options were coded as follows: 1 = "not at all," 2 = "sometimes," and 3 = "often." To obtain reasonable item-level distributions, we considered only activities that were engaged in by at least 10%, but no more than 90%, of the sample. The distributions of the 10 items that met this criterion are shown in Table 1 . These items have been shown to have test-retest reliabilities (kappa statistic) in the fair to good range, with most over .60 (Bygren 1995). Although it has been acknowledged that some items are lacking in breadth and precision (see Tahlin 1985), they appear to be robust by correlating in predictable ways with other variables, such as occupational status (Karasek 1976).
|
|
We also ascertained the degree of involvement in more formal activities by the reported frequency of participation in the meetings of voluntary organizations, such as those serving retired persons, or those organized by a political party or religious organization. Religious participation was measured as frequency of attendance at religious services. These two indexes ranged from never (0) to at least once per week (5). We computed standardized scores for each variable to ensure comparability with other activity measures. The correlations among the domains of activity involvement are shown in Table 3 . These correlations reveal that activity domains tended to cluster (all were significant), but their magnitudes suggest that the domains were not highly associated (seven correlations were between .2 and .3, seven were between .3 and .4, and one was .45).
|
Control Variables
In our multivariate equations we controlled for initial functional impairment, the presence of heart or circulatory problems, and tobacco use. Other available health and health behavior variables (such as having cancer, body-mass index, and alcohol consumption) were not included in our analyses because they caused no significant elevation in mortality risk with functional impairment controlled.
Functional impairment was operationalized with three subscales—activity limitations, mobility limitations, and motion/strength limitations. Activity limitations reflected the number of the following activities with which respondents reported difficulty: eating, transferring to bed, toileting, dressing, bathing, purchasing food, preparing food, and cleaning the house. Scores for this subscale ranged from 0 to 8. Forty-two percent had at least one functional limitation. The mobility subscale was based on self-reports of difficulty in walking 100 meters, running 100 meters, walking up and down stairs, turning a water faucet on and off, standing up from an armless kitchen chair, and standing without support. Scores for this subscale ranged from 0 to 7. More than three quarters of the sample (78%) had at least some mobility difficulty. The motion/strength subscale was based on the directly observed performance of a series of tasks (Aniansson, Rundgren, and Sperling 1980). Participants were asked to pick up a pen from the floor, touch opposite toes and earlobes (from sitting), lift 1 kg (2.2 lb.), sit on their hands, rotate their wrists, and rise from a chair without using arms. Interviewers recorded if participants performed each test with any difficulty, resulting in a scale that ranged from 0 to 9. Slightly more than half (52%) of the sample had at least some difficulty performing an assigned task. For proxy interviews (n = 34) and telephone interviews (n = 34), we imputed performance scores from activity and mobility scores using a linear estimation technique. Respondents interviewed by proxy or telephone had no higher risk of mortality net of other variables, suggesting that the imputation introduced little bias to the analysis.
We used principal components analysis to derive a composite functional impairment variable from the three subscales. Each subscale loaded strongly (>.6) on a single underlying dimension (80% of common variance explained). Functional disability has often been found to be the most proximate and strongest health factor predicting mortality in older people (Avlund, Damsfaard, and Holstein 1998; Corti, Guralnik, Salive, and Sorkin 1994), because it tends to "accumulate" other symptoms and diseases (Schroll 1997).
A heart or circulatory problem was said to be present if the respondent had experienced a heart attack or had three of the following four conditions or symptoms in the 12 months preceding the interview: chest pain, high blood pressure, heart problem, or dizziness. By this criterion, 22% of the sample experienced a heart or circulatory problem. We constructed a dichotomous variable to represent the presence of this type of condition (0 = "has no heart/circulatory problem," 1 = "has heart/circulatory problem"). Tobacco consumption was operationalized as a dichotomous variable reflecting smoking behavior at the time of the interview (0 = "does not currently smoke cigarettes," 1= "currently smokes cigarettes"). Less than 10% of the sample reported smoking at that time. Additional control variables included gender (0 = "female," 1 = "male"), age in years, and educational attainment (0 = "grade school," 1 = "beyond grade school").
|
Results |
|---|
|
TOPAbstractTheoretical ModelMethodsResultsDiscussionReferences |
|---|
2 (8, n = 463) = 3.61, p = .89, justifying the pooling of men and women into a common model. Table 4 shows the crude relative risks of mortality associated with analytic variables. As would be expected, older, male, and poorly educated respondents had greater mortality risk than their younger, female, and more highly educated counterparts. An increase of 1 standard deviation in the functional disability scale and having a heart-circulatory problem were each associated with a 60% increase in the relative risk of mortality. Smokers had almost twice (90% higher) the risk of mortality of nonsmokers. In terms of involvement in activities, we show for each activity domain the simple risk ratios associated with a 1 standard deviation increase in the level of participation. These ratios indicated that greater involvement in all activity domains was significantly associated with lower relative risk of mortality.
|
|
; Klumb and Baltes 1999), we examined effects of activity involvement by gender. These results (Table 5 ) show that among men only the solitary-active dimension of activities remained a significant predictor when health factors were controlled. Among women, none of the domains of activity was a significant predictor when health was controlled, suggesting that any elevated risk of mortality associated with inactivity appears to be a spurious by-product of the confound between poor health and low activity levels. By contrast, older men who were more active in solitary-active pursuits experienced greater longevity, even after baseline health was taken into account. The most stringent test of our model involved estimating a Cox regression with all activity domains controlled simultaneously. To conserve statistical power given the number of mortality events in each gender group, we estimated these equations for the entire sample only. These multivariate risk ratios are shown in Table 6 . The first column shows the results with gender, age, and education controlled. In this equation, both friendship and solitary-active domains of activity were statistically significant in reducing mortality. When health factors were added as control variables in the second column, the risk ratios revealed that only the solitary-active dimension remained significant, with a 1 standard deviation increase in activity yielding a 20% reduction in the relative risk of mortality. In a test not shown, respondents who were the least active in solitary-active pursuits, defined as being more than 1 standard deviation below the scale mean, had a mortality risk 2.7 times higher than that of the most active respondents, that is, those who were more than 1 standard deviation above the scale mean. This relationship also held after we adjusted for all other activity dimensions, health, and other control variables. Finally, no significant interactions were detected between any of the six activity domains and gender, suggesting that engagement in activities promoted longevity similarly for men and women.
|
|
Discussion |
|---|
|
TOPAbstractTheoretical ModelMethodsResultsDiscussionReferences |
|---|
Older men appear to benefit from participating in solitary but active pursuits, measured with an index that includes hobbies such as carpentry and gardening. This finding deserves a word of caution. Because respondents selected activities on the basis of their capacity to engage in them, we cannot be certain that existing differences in health between active and less active men are responsible for this finding. However, that these results persist when we control for functional ability (that includes a performance-based assessment), mitigates the possibility that the relationship between activities and mortality is spurious. However, even if the observed association were causal in nature, we still would not be able to specify the exact mechanism involved—for instance, whether activities enhance survival by virtue of providing a sense of purpose, increasing physical stamina, or reducing biochemical markers of stress.
The most robust finding in this investigation is the power of the solitary-active domain of activities in the entire sample to retain its predictive strength when all other activity domains are controlled. Activity, even if not explicitly social in nature, has positive effects on the survival of very old individuals. This suggests that psychological dimensions of activities—such as motivation, inner direction, and purposefulness—may be key mediating factors in the promotion of health and longevity in late old age. Our finding leads us to reflect on the subjective basis of "meaningful activity" as applied by Rowe and Kahn 1998 to the promotion of successful aging. It is likely that activities are consequential when they are experienced as significant ways to engage one's free time.
Surprisingly, women who receive visits from friends experience excessive mortality rates. This may be explained by gender norms governing contact with friends during the last stage of life, a period of time when older women may receive greater attention from their friends than do similarly situated older men. Alternatively, women who reduce their contact with the wider world may prolong their lives by "conserving" energy for their most intimate and meaningful relationships (Carstensen 1992).
Formal group involvement—organizational and religious activities—produces no longevity benefits, because their association with mortality is fully explained by the tendency of healthier respondents to be engaged in those activities. This "selection" hypothesis goes against the findings of other studies, particularly those that find religious attendance to be a robust predictor of mortality among elderly persons (Idler and Kasl 1992; Koenig et al. 1999; Oman and Reed 1998). Interestingly, family contact does not predict mortality risk and is consequently omitted from our analysis. This finding is in line with earlier studies of oldest old persons in Sweden showing that family integration tends to have virtually no effect on various kinds of health outcomes of elderly persons. It is also consistent with the current Swedish welfare model that mandates the state, through public services, to assume the main responsibility for providing care needed by elderly persons (Parker 2000; Szebehely 2000).
Several limitations of this analysis warrant discussion. Because the study is observational in design and adjustments for spurious factors are not exhaustive, we must remain cautious in attributing a causal explanation to our findings. The relatively small sample size and number of deaths in this study reduce the statistical power necessary to detect significant associations and preclude the introduction of additional control variables to better account for health selection. Further, the small number of activities representing each of the principal dimensions may not fully capture the variation needed to efficiently predict mortality. Finally, there are no measures of affective ties, emotional support, and social-psychological orientations in the study. This restriction inhibits our ability to identify subjective factors that may serve as intervening mechanisms linking social and leisure activities to mortality outcomes.
In spite of these limitations, there are several strengths to our analysis as well. First, the sample is nationally representative of those aged 77 years and older living in the community in Sweden, thereby permitting statistical inferences to the oldest old population in that nation. Second, response rates are very high for the baseline sample, and follow-up data on mortality are nearly universal and considered highly accurate. Third, the functional health measure includes performance-based indicators that are less likely to be affected by self-reporting bias, thereby reducing the chance that the effect of activity involvement is simply a proxy for physical health (Morgan and Clark 1997). Fourth, the measurement of activities is conceptually broad and empirically detailed enough to capture much of the diversity of social and leisure involvement maintained by elderly persons.
Our approach to studying activities highlights some of the challenges faced by researchers considering the meta-construct of engagement with life and its role in promoting successful aging. Difficulties often arise in operationalizing the complex set of ways that older individuals interact with their social and physical environment to give meaning and purpose to their lives. Indeed, Juster and Stafford (as cited in Klumb and Baltes 1999) asserted that developing a valid and reliable scheme for appraising the salient domains of leisure and productive activities is "one of the most challenging tasks of the time-use researcher". In conclusion, we suggest that developing broader conceptualizations of engagement with life that combine theoretical rigor and desirable measurement properties remains the task of researchers, who face the formidable task of making sense of the large array of possible activities in which older people are engaged. Careful specifications of activity involvement will permit a more informative account of the mechanisms—social, physical, and psychological—that link activities to enhanced survival in later life.
|
Acknowledgments |
|---|
Received for publication May 31, 2000. Accepted for publication April 23, 2001.
|
References |
|---|
|
TOPAbstractTheoretical ModelMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  J. M. Jacobs, R. Hammerman-Rozenberg, A. Cohen, and J. Stessman Reading Daily Predicts Reduced Mortality Among Men From a Cohort of Community-Dwelling 70-Year-Olds J. Gerontol. B. Psychol. Sci. Soc. Sci., March 1, 2008; 63(2): S73 - S80. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. A. Hoppmann, D. Gerstorf, J. Smith, and P. L. Klumb Linking Possible Selves and Behavior: Do Domain-Specific Hopes and Fears Translate Into Daily Activities in Very Old Age? J. Gerontol. B. Psychol. Sci. Soc. Sci., March 1, 2007; 62(2): P104 - P111. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
D. C. Reitzes and E. J. Mutran Self and Health: Factors That Encourage Self-Esteem and Functional Health J. Gerontol. B. Psychol. Sci. Soc. Sci., January 1, 2006; 61(1): S44 - S51. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 Y.-F. Kuo, M. A. Raji, M. K. Peek, and J. S. Goodwin Health-Related Social Disengagement in Elderly Diabetic Patients: Association with subsequent disability and survival Diabetes Care, July 1, 2004; 27(7): 1630 - 1637. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. F. Mendes de Leon, T. A. Glass, and L. F. Berkman Social Engagement and Disability in a Community Population of Older Adults: The New Haven EPESE Am. J. Epidemiol., April 1, 2003; 157(7): 633 - 642. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
V. H. Menec The Relation Between Everyday Activities and Successful Aging: A 6-Year Longitudinal Study J. Gerontol. B. Psychol. Sci. Soc. Sci., March 1, 2003; 58(2): S74 - 82. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
