1 Increases in life expectancy and ageing populations have given rise to a greater interest in old age and ageing. In Europe, the proportion of the population aged 65+ is projected to reach 30% by 2060, and the number of adults over 80 years of age is projected to triple (Eurostat, 2008). In light of these demographic changes research on successful ageing has become of greater importance.
2 Thoughts on old age as a time of positive growth are by no means new. As early as 44 B.C., the Roman philosopher Cicero argued that old age is a time of pleasure and self-fulfilment (Cicero, 1923). These thoughts did not reach the scientific arena until much later, and gerontological research was for a long time focused on the negative consequences of ageing, e.g. the risk of chronic illness, loneliness and decline. During the last few decades, new perspectives have emerged (Gergen/Gergen, 2001), focusing more on positive outcomes and their determinants in old age. This “new gerontology” (Sheidt/Humpherys/Yorgason, 1999) has been given a number of different labels, e.g. “positive ageing”, “productive ageing”, “ageing well” and “optimal ageing” to name a few. “Successful ageing” is, however, the term most frequently used in research (Hung/Kempen/de Vries, 2010; Peel, 2004), and it is also the one considered to have had the greatest impact on academic gerontology as well as on political arenas (Dillaway/Byrnes, 2009).
3 The successful ageing terminology has not been without critics, though. One major concern is the lack of agreement in defining what successful ageing really is. Even though the concept has a history of almost half a century (e.g. Havighurst, 1963) there is still a lack of consensus among researchers regarding its constituents (Bowling, 2007; Depp/Jeste, 2006). Inspired by the influential view held by Rowe and Kahn (1987) – that the ageing process can be characterised as either “successful” or “usual” depending on health, functional and cognitive capacity and productive engagement – numerous studies have investigated successful ageing as a state of high functional and social capacity. Definitions within this biomedical perspective have included good health, no chronic illness, independence in ADL, high cognitive functioning, good physical strength and doing voluntary work (e.g. Andrews et al., 2002; Hank, 2011; Chavez et al., 2009; Ford et al., 2000; Lamb/Myers, 1999; McLaughlin et al., 2010; Roos/Havens, 1991; Strawbridge et al., 1996). Others have adopted more psychosocial perspectives, placing focus on adaptation to the ageing process and associated psychological resources. This line of research has often been influenced by the lifespan developmental approach (Baltes, 1987; Baltes/Baltes, 1990), which argues that development is a lifelong process of balance between gains and losses. Definitions within this perspective have been broader and more subjective, including ageing satisfaction, positive emotions, lack of loneliness, good mental health and social support (e.g. Freund/Baltes, 1998; Jopp/Smith, 2006; Vaillant/Mukamal, 2001). Bowling (2007) in her comprehensive review on the topic argues that this lack of unity in definition is mainly a reflection of researchers choosing a definition according to their research discipline rather than on a theoretical basis, and calls for more interdisciplinary and multidimensional approaches.
4 The need for multidimensionality is supported by research on definitions of successful ageing by the elderly themselves, showing that the views of older adults are far more multidimensional than the approaches taken by academics (Hung et al., 2010). When asked, older adults rate health and physical functioning, social relations, cognitive functioning, activities, independence, adjustment and coping, well-being and life satisfaction, personal growth and spirituality, financial well-being and positive attitude as important dimensions for successful ageing (Bowling, 2006; Duay/Bryan, 2006; Fisher/Specht, 1999; Guse/Masesar, 1999; Hilton et al., 2009; Knight/Ricciardelli, 2003; Phelan et al., 2004; Tate et al., 2003; Torres/Hammarström, 2009; Von Faber et al., 2001). This clearly supports the notion that according to older adults « successful ageing includes a broad set of circumstances that include, but transcend, health » (Bowling, 2007:272).
5 Another frequently discussed matter within research on successful ageing is who should be characterised as ageing successfully. According to researchers’ criteria – regardless of definition – only a minority of older adults tend to be successful (e.g. Bowling/Iliffe, 2006; Roos/Havens, 1991; Strawbridge et al., 1996), whereas a majority of the elderly rate themselves as ageing successfully (Bowling, 2006; Tate et al., 2003; von Faber et al., 2001). Strawbridge et al. (2002) addressed this question directly by comparing the percentage of successful agers according to the criteria put forth by Rowe and Kahn (1987) – i.e. absence of disease and disability, high physical and mental functioning and active engagement in life – and the percentage rating themselves as ageing successfully. What the authors found was that about half of the sample strongly agreed that they were successful agers; less than one-fifth, however, was objectively judged to be so. Another interesting finding was that many of those living with chronic conditions regarded themselves as successfully ageing, a finding replicated by Bowling (2006), whereas one-third of those who were objectively functioning optimally did not consider themselves to be ageing successfully.
6 What is the reason for this discrepancy between subjective evaluations and objective measures? The general approach among researchers in defining successful ageing has been to define successful ageing as obtaining optimum scores on all selected criteria. This dichotomisation into successful vs. unsuccessful on the aggregate level can be considered problematic for several reasons. Firstly, it is obvious that few older adults achieve unmitigated success in all areas of life, and this approach will inevitably continue the focus on a small elite. Secondly, it is a poor representation of the views of the group under consideration. It appears that even though many older adults face some decline in health and functioning, the subjective experience of ageing well is rather stable. Recent work by Pruchno et al. (2010) provided support for a model of successful ageing that incorporated objective as well as subjective components, and they also found that age was associated with objective but not subjective success. Thirdly, it may give an unreasonably negative image of ageing by classifying as “unsuccessful” older adults who have experienced a minor decline in functioning or have some manageable disease, but are otherwise in good health. Recent studies on the prevalence of successful ageing as defined by Rowe and Kahn (1987, 1997) in a national sample of older adults in the US (McLaughlin et al., 2010) and 14 European countries (Hank, 2011) illustrate this quite clearly: the percentages of successful agers were only 11.9% and 8.5% respectively. In the study by Hank (2011), however, 42.6% of respondents were free from major disease, 83.7% had no disability, 48.5% had high cognitive functioning, 57.3% had high physical functioning and 27.1% were actively engaged, yet only 8.5% of respondents fulfilled all conditions. There have been calls for less rigid definitions (Bowling/Iliffe, 2006; McLaughlin et al., 2010), but the nature of these relaxed classifications remains unclear. Lastly, several cross-sectional studies (Hank, 2010; Pruchno, 2010; Weir et al., 2010) indicate that different components of successful ageing are weakly correlated and also evince different age-related patterns, but the interrelationships among these over time are not well understood.
7 A less addressed, but no less important issue, is the lack of focus on changes in research on successful ageing. Although scholars within the gerontological field (Hofer/Piccinin, 2010; Hofer/Sliwinksi, 2001; Spiro, 2007) have highlighted the crucial role of direct assessment of intra-individual changes and inter-individual differences in within-person change for understanding development and ageing, this has not been well reflected in research on successful ageing. Considering that the intention of the division into successful vs. usual ageing was to draw attention to the degree of variability in the ageing process (Rowe/Kahn, 1987), it is surprising that few attempts to detect different patterns of changes have been made. The absolute majority of studies, including those with a longitudinal design, have treated successful ageing as a state of being at a given point in time, rather than a process that may take different forms between individuals. As such, one can question the current approach as a tool to describe heterogeneity, since the information it has yielded is basically that there is a small group of people who appear not to have changed, whereas most people have faced some changes – which in many cases does not seem to bother them as much as it does researchers. Besides, measuring ageing as a static state has added little to our understanding of how and why individuals change over time, which in turn has been labelled a major research priority within research ageing (Sliwinski/Piccinin, 2010). In addition, the limitations associated with cross-sectional comparisons limit their utility to answer questions regarding changes (Hofer/Sliwinski, 2001). An exception is a study by Liang et al. (2003), who identified different types of trajectories of functional change among older adults in Japan, and interpreted these in terms of successful, usual and pathological ageing, and also investigated how psychological factors and demographic variables predicted trajectories. As the authors note, however, functional status is but one of many aspects of successful ageing, and conclude that there is a need to analyse multiple trajectories of successful ageing, and their interconnections.
8 Therefore, this study seeks to shed further light on successful ageing as a diverse process. More precisely, we will study changes in domains that are defined by older adults as important aspects of successful ageing, examining how changes in these domains can be explained by differences in factors known to be predictive of successful ageing, and how changes in the domains are related to changes in general well-being. This leads to the following four research questions: (1) Do sample-level trajectories of the dimensions rated as important for successful ageing by older adults change over time? (2) Do these trajectories vary between individuals? (3) Do demographic variables account for some of the variability in the ageing trajectories? (4) Are changes in the different domains related to changes in general well-being? Answers to these four questions will help further our understanding of how different aspects of (successful) ageing are related to general well-being, and also lay the ground for defining successful ageing as a diverse process rather than a static condition that a majority of older adults fail to achieve.
9 Data were derived from the Longitudinal Aging Study Amsterdam (LASA), which is an on-going longitudinal, multidisciplinary cohort study on a wide range of factors related to physical, cognitive, psychological and social functioning in older adults in the Netherlands (Huisman et al., 2011). The LASA cohort is based on a nationally representative sample of older adults aged 55-85 years (years of birth 1908-37), based in three geographical regions in the Netherlands. The LASA sample was originally recruited for the study Living Arrangements and Social Networks of Older Adults (LSN) in 1992 (Knipscheer et al., 1995). Data were used in this study from follow-ups that were carried out in 1992-1993 (N= 3107), 1995-1996 (N= 2545), 1998-1999 (N= 2076), 2001-2002 (N= 1691), and 2005-2006 (N= 1257). For each follow-up, on average 81 of the respondents were re-interviewed, 12% had died, and 8% could not be interviewed for some other reason (refusal, illness, could not be contacted). For the present analysis only participants who were recruited in 1991 with at least responses in the last wave (2005-2006) were included. The sample size used in this study was therefore 1257. Survival over the 14 years between the first and last wave was related to almost all study variables, leaving a study sample that was on average nine years younger, with fewer chronic diseases (1.1 versus 1.6), better cognitive functioning (MMSE scores 27.7 versus 25.9) and emotional functioning (CES-D scores 6.8 versus 8.5) and larger social networks (14.8 versus 11.1 network members) than the complete sample. However, the remaining study sample was sufficiently heterogeneous in the relevant measures to answer our research questions (see also Table 1). Data were collected by means of face-to-face and computer-assisted interviews, medical interviews and written questionnaires. For each observation the interviewers were trained and supervised. The interviews were tape-recorded for quality checks, and took between one and a half and two hours.
10 The main domains of interest are domains of successful ageing according to older adults themselves and a general measure of well-being. The following domains have been found to be important aspects of successful ageing according to older adults: health and physical functioning, social relations, cognitive functioning, activities, independence, adjustment and coping, well-being and life satisfaction, personal growth and spirituality, financial well-being and positive attitude (Bowling, 2006; Duay/Bryan, 2006; Fisher/Specht, 1999; Guse/Masesar, 1999; Hilton et al., 2009; Knight/Ricciardelli, 2003; Phelan et al., 2004; Tate et al., 2003; Torres/Hammarström, 2009; Von Faber et al., 2001). LASA provides valid information on the following domains which were included in the analytical models: health, physical functioning, cognitive functioning, social relations, activity participation, coping, financial health and different domains of well-being.
11 Physical health was conceptualised as consisting of two dimensions: number of chronic conditions and self-rated health (SRH), with the former thought to reflect an objective evaluation and the latter a subjective. Number of chronic conditions was measured through self-reported presence of the following seven chronic diseases or disease events: chronic nonspecific lung disease, cardiac disease, peripheral arterial disease, diabetes mellitus, stroke, arthritis and cancer. The selection of these was based on their prevalence (>5%) in the 55+ age group in the Netherlands (Kriegsman et al., 1996). The self-reports of chronic conditions have been shown to be fairly accurate as compared with general practitioners’ reports (Kriegsman et al., 1996). SRH was assessed by the question “How is your health in general?” with response categories ranging from (1) “Excellent” to (5) “Poor”.
12 Physical functioning was measured in terms of functional limitations by asking participants to rate their ability to perform the following activities: walk up and down a staircase of 15 steps without resting, use own or public transport and cut own toenails (van Sonsbeek, 1998). Response categories were (0) “no difficulties” to (3) “difficulties in all three activities”.
13 Cognitive functioning was measured with the Mini Mental State Examination1 (Folstein et al., 1975). The MMSE is a 20-item scale widely used to assess general cognitive functioning. Scores ranged from 0-30, with higher scores indicating better cognitive functioning.
14 Social relations were measured with participants’ frequency of contact with the nine most important network members with whom they had regular contact other than their partner, with the question “How often are you in touch with…?”. Answers were coded as total number of daily contacts.
15 Activity participation consisted of two different domains of activities: social and leisure activities and physical activities. Participation in social and leisure activities was assessed by asking participants for how many hours during the last year they had engaged in visiting activities or meetings of different types of organisations (e.g. senior associations, organisation for leisure or hobby, church), and how often they participated in seven types of leisure activities (e.g. going to the cinema, doing sports, shopping for pleasure). Physical activity was measured by asking participants for how many minutes during the last two weeks they had been walking outside, bicycling, doing work in the garden, doing light household work, doing heavy household work and engaging in sports activities, with a maximum two sports.
16 Coping was assessed in terms of participants’ sense of mastery, i.e. the extent of perceived control over their own lives and ability to manage on-going events and situations. Mastery was measured with a five-item abbreviated version of the Pearlin Mastery Scale (Pearlin/Schooler, 1978). Response categories ranged from (1) “Strongly agree” to (5) “Strongly disagree”. The total score ranged from 5 to 25, with a higher score indicating a stronger sense of mastery.
17 Financial health, or having sufficient income, was measured by asking respondents to evaluate their income with the questions “Are you satisfied with your income level?” and “Are you satisfied with the living standard you enjoy on your income?”. Response categories ranged from (1) “Dissatisfied” to (5) “Satisfied”, with a sum score between 2 and 10.
18 General well-being was assessed as a general factor with six indicators from three different domains: satisfaction with life, positive emotions, and absence of loneliness. These domains are viewed as reflecting outcomes of successful development (Freund/Baltes, 1998). Satisfaction with life was measured with the question: “Taking everything into account, have you been satisfied with your life lately?” Response categories ranged from (1) “Dissatisfied” to (5) “Satisfied”. Positive emotions were assessed with four items from the validated Dutch version of the Center for Epidemiologic Studies Depression Scale (Beekman et al., 1997; CES-D; Radloff, 1977). The four items were “I felt I was as good as other people”, “I felt hopeful about the future”, “I was happy” and “I enjoyed life”, with response categories ranging from (0) “Rarely or never” to (3) “Mostly or always”. Loneliness was measured with the De Jong Gierveld Scale (De Jong Gierveld/Kamphuis, 1985), a widely used 11-item scale to assess level of loneliness. Total scale scores ranged from (0) “Never lonely” to (11) “Very lonely”.
19 Age, gender, level of education and marital status were included as demographic variables since these have been found to be important determinants of successful ageing (Hank, 2011; McLaughlin et al., 2010; Roos/Havens, 1991). Level of education was measured in numbers of years of completed education, and varied from 5 (less than primary school) to 18 (college or university). Marital status was included in the analytical models as a time-varying variable, and response categories were recoded as (1) “No partner” (widowed, never married, divorced) and (2) “Partner”.
20 The analyses comprised several steps. First, to answer the research questions whether domains of successful ageing change over time and whether these trajectories vary between individuals, a latent growth model was specified for each domain of successful ageing and general well-being. Latent growth models (LGM) are a specific type of structural equation models that enable researchers to study correlates and predictors of individual change over time (Willet/Sayer, 1994). With LGM, observations for one or more observed variables that are measured on two or more occasions over time are represented as so-called growth parameters, reflecting the starting level (or intercept) and amount of linear (slope) or non-linear (e.g. quadratic slope) change. The variance of each parameter is also estimated, which provides information about the significance of the differences between individuals (inter-individual differences). A significant intercept indicates that the starting level is different from zero; significant variation of the intercept indicates that people have different starting levels. Accordingly, a significant slope indicates that people do change over time (intra-individual change), and a significant variation of the slope (or quadratic slope if there is non-linear change) indicates that people differ in rate of change over time (inter-individual differences in intra-individual change). For each domain, one or two (depending on the number of dimensions per domain) LGMs were estimated to determine whether there was significant change over time on average, and whether there was significant variance in level and slope. The analytic strategy was to first fit a linear model, and if this gave a poor fit add a quadratic parameter to account for non-linear change. Since general well-being was constructed using six indicators (i.e., four items for positive affect, one for loneliness and one for satisfaction with life), an exploratory factor analysis (EFA) was first performed at baseline to test if a one-factor solution gave an adequate fit. Next, we checked whether the assumption of factorial invariance, i.e. invariance of measurement of the concept of general well-being, was satisfied by constraining the factor loadings on the subsequent waves to be equal to the factor loadings of general well-being on the first wave.
21 The next step was to include in each LGM a series of explanatory variables, i.e. age, gender and level of education, and the time-varying variable marital status, in an attempt to answer question three, i.e. whether demographic variables account for some of the variability in the successful ageing trajectories. The estimated growth parameters’ intercept and slope(s) were regressed on these demographic variables to assess the effect on level and change in these domains. Since marital status could have varied on each occasion, the effect of marital status was assessed by estimating the direct effect of marital status on the observed measures (see Figure 1 for a diagrammatic representation of a Latent Growth Model with predictor variables).
Figure 1. Latent Growth Model for health, controlled for age, gender and education, and time varying marital status
Unobserved variables are enclosed by ovals, observed variables by boxes.
22 The final step was to see to what extent changes in the domains of successful ageing were related to changes in general well-being. This was done by means of Parallel Growth Models (PGM). In a PGM, two LGMs are combined and intercepts and slopes of both models (i.e. general well-being and one domain of successful ageing) are permitted to co-vary. A significant covariation between intercepts and slopes of two parallel processes indicate that starting level and change in one of the domains of successful ageing were correlated to starting level and change in general well-being. All models were estimated with Mplus version 5.12. Incompleteness was treated using full information likelihood (FIML) statistical methods available in the Mplus program, which means that missing values were substituted by estimates of the missing value based on all the data available.
23 To provide more insight into the characteristics of the study sample, descriptive statistics are provided for the baseline (1992) observation (Table 1). The sample (N= 1257) consisted of more women than men. Ages ranged from 55 to 85 years and the mean age was 65. More than two-thirds of the respondents (71.5%) were married or cohabiting at baseline, and about 50% were still living with a partner 14 years later. Means and variances of the observed indicators for the domains of functioning are presented in Table 1.
Table 1. Descriptive statistics (observed variables) of the study sample (N=1257) at baseline (1992)
24 The EFA of the general well-being factor at T1 gave a good model fit, χ2 (8)= 26.11, p < .001, CFI= .99, TLI= .99, RMSEA= .04, leading to the conclusion that the indicators used reflected a general factor of well-being in the sample. In order to ensure factorial invariance of the factor general well-being, CFAs were performed on all subsequent waves (T2-T5), with factor loadings constrained to be equal to loadings at T1. CFAs in all waves showed adequate model fit (with all GFIs and TLIs higher than .97 and all RMSEAs lower than .08), thus meeting criteria for strong factorial invariance.
25 Maximum likelihood estimates of the mean and variance of the levels and slopes of each of the domains of successful ageing and general well-being are shown in Table 2. All LGMs showed an excellent fit, indicating that the models gave an adequate representation of the data. Table 2 can be interpreted as follows: the estimated means reflect the average baseline level; the variance indicates variations in the mean across individuals. A slope reflects annual rate of linear change; a quadratic slope reflects annual rate of non-linear change. The total amount of change in a non-linear model at a certain time point (Ti) can be calculated as follows: total change since baseline at Ti= Ti * point estimate of the slope + Ti2 * point estimate of the slope 2. For example, total change at T5 (= 14 years after baseline) for number of chronic diseases is= 14*.112 + 142 * -.002= 1.176, indicating that the mean number of chronic diseases in the study group increased from 1.140 to 2.316 in 14 years.
Table 2. Estimates for Latent Change Models of successful ageing domains (Slope reflects amount of change per year)
26 As can be seen in Table 2, all domains of successful ageing changed over time. On average, there was an increase of 1.18 in the number of chronic diseases over the total study period of 14 years. In that same period, SRH, which was coded as 1 (“excellent”) to 5 (“low”), became 0.27 points higher, indicating a lower level of subjective health. The number of functional limitations increased by 1.09 and cognitive capacities fell by 1.64 scale points of the MMSE. The number of social contacts per day decreased by 0.60 contacts per day from 3.99 contacts per day to 3.39 14 years later. Also, a significant reduction in activity participation was found for the two types of activities. Reductions in coping style indicated that people felt that their control over their own life and ability to manage on-going events and situations decreased over time. In contrast, satisfaction with income increased slightly but significantly from 8.74 to 8.81, indicating that people became slightly more satisfied with their income at higher ages. On average, general well-being did not change over the 14 years of study.
27 Conclusions on inter-individual differences in level and change in domains of successful ageing can be based on the significance of the variances of the means and slopes. As can be seen in Table 2, there was significant variation in all means and slopes of all domains of successful ageing, indicating that there were differences in starting levels and in amount of change over time between individuals. With respect to general well-being we conclude that on average there was no change. However, since slopes vary significantly between individuals, declines in well-being in some individuals were equalled out by growth in well-being in other individuals.
28 Estimates of the impact of age, gender, level of education and marital status on level and change in successful ageing and general well-being can be found in Table 3. Since marital status was a time-varying variable, the estimated effects on level are reported at each wave. Age was not included in LGMs with non-linear change because of poor model fit.
29 In LGMs with only linear change, we could estimate the effect of age on baseline performance and rate of change. From the first column it follows that older people had fewer contacts per day, were less physically active, and had lower levels of mastery but equal levels of general well-being, social activities and SRH. Older people showed more marked declines in activity participation (both social and physical) and mastery. Also, for general well-being a significant effect on change was shown, indicating that general well-being changed slightly faster at higher ages.
30 Women had lower levels of physical functioning (more chronic conditions, lower SRH and more functional limitations), fewer contacts per day, a lower level of mastery, and lower levels of physical activity than men, but higher levels of cognitive functioning. Also, general well-being was lower in women than men. No significant effects of gender on slopes in domains of successful ageing were found, indicating that changes in these domains were not related to gender. General well-being, however, changed more strongly among women than men.
Table 3. Latent Growth Models with demographic variables age, gender, level of education and marital status
31 More years of education was related to better SRH, fewer functional limitations, higher levels of cognitive functioning, fewer social contacts per day, more time spent on social and leisure activities, and higher satisfaction with income. Level of general well-being was unrelated to more years of education. Level of education was not related to the rate of change in any domain, except number of social contacts per day. There was less decline in number of social contacts for people who were more highly educated.
32 Having a partner was related to fewer chronic conditions, higher levels of cognitive functioning, higher number of contacts per day, better satisfaction with income and higher levels of general well-being on all or most of the observations.
33 In order to answer the last question, whether changes in domains of successful ageing were related to changes in general well-being, nine PGMs were estimated (one PGM for each combination of a successful ageing domain and general well-being), while intercepts and slopes of successful ageing were permitted to co-vary with intercept and slope of general well-being. A significant correlation between intercepts indicates that a high level in one domain of successful ageing is related to a high level in general well-being. A significant correlation between slopes indicates that change in one domain of successful ageing is related to change in general well-being.
34 There was a significant correlation between the intercepts of all domains of successful ageing and general well-being. Strong correlations existed between mastery (r= .66) and SRH (r= -.56), indicating that higher levels of mastery and better SRH were related to higher levels of general well-being. Moderate correlations with general well-being were observed among number of chronic conditions (r= -0.28), number of functional limitations (r= -0.28) and financial health (-0.28). Time spent on social and leisure activities (r= 0.16) and cognitive functioning (r= 0.09) were only weakly correlated with general well-being.
Table 4. Correlations between intercepts and slopes of general well-being and domains of successful ageing
35 There was a very strong correlation between change in mastery and change in general well-being (r= .97), and between changes in SRH and change in general well-being (r= .76). Strong correlations between change in general well-being and changes in physical activities (r= .59), changes in social and leisure activities, and changes in the number of contacts per day (r= .43) were also observed. Little or no correlation with general well-being was observed for satisfaction with income, number of functionnal limitations (r= .13), number of chronic diseases and cognitive functioning (r= -.02).
36 The purpose of this article was to study the debated concept of successful ageing from a developmental perspective, and from the perspective defined by older adults themselves. Additionally, we studied successful ageing as a dynamic concept, rather than a static outcome. A longitudinal analysis of the domains rated as important for successful ageing may be a helpful tool for those interested in refining the concept to better fit the views held by the group under consideration, and a necessary one for those interested in what factors are associated with late life changes as opposed to static positions. Investigating the associations between changes in successful ageing domains and changes in well-being enables a life-span perspective to be adopted on the relations between subjective and objective aspects of ageing well, and also underscores the importance of differrent aspects of successful ageing for quality of life in old age.
37 Applying LGMs to longitudinal data enabled modelling of population and individual change in multiple domains of successful ageing over 14 years, helped identify predictors of individual differences in change, and showed how changes in the different domains were related to changes in well-being. On average there was change in all domains over time, except for the domain of general well-being, which was stable. However, these population-level trajectories gave a more nuanced picture of late life changes than cross-sectional age group comparisons.
38 The general pattern of change showed that ageing is associated with some decline in multiple functions, even in a survivor sample like the one used in this study, a fact that supports the need for relaxed definitions of successful ageing, unless the concept is to be used to describe only a small minority of older adults. At study entrance mean age was 65 years; on average there were already some chronic conditions and functional limitations present, and these increased over time. On the other hand, the direct assessment of changes over time showed that even though there was significant decline in most domains, this decline was small, and of an accelerated nature only in three domains (chronic conditions, functional limitations and cognitive functioning). For example, the number of contacts per day went from 3.99 to 3.48 over 14 years, and SRH deteriorated by .23 points on a scale from 1 to 5. In part these results may reflect the biased sample of survivors only, but it also indicates that among those who do reach old age the average loss of functions is not tremendous.
39 There were individual differences in level as well as in rate of change in all domains of successful ageing, confirming the great heterogeneity that characterises the ageing process and the importance of using methods that allow consideration of individual differences as well as average trends of change (Spiro, 2007). Regarding predictors of individual differences, the results in this study highlight the importance of relating between-person differences to intra-individual change in research on ageing in general and successful ageing in particular. An interesting finding was that background variables accounted for individual differences in level to a much greater extent than in change. Age predicted lower levels and faster decline in some domains; this supports the notion that advanced age is accompanied by increased limitations to the individual’s capacity, but also raises the question if the meaning of successful ageing varies depending on age, given that many people of advanced age and with several limitations consider themselves successful (von Faber et al., 2001). Regarding gender, education and marital status the associations across domains were mixed. In many but not all cases men, the married, and those with more education held a favourable position regarding levels, but neither gender nor education predicted faster decline other than in a few domains. These results indicate that rate of change in later life is less affected by gender and SES than many cross-sectional studies have found, and they also highlight the role of cumulative inequalities in terms of resources for ageing successfully. It appears that women – although they do not show a faster decline – are in many domains in a more vulnerable position to start with, and this is also reflected in general well-being. A contextually sensitive perspective on successful ageing needs to acknowledge that people enter old age in different positions and with different resources. Only a developmental perspective is able to detect to what extent rate of change is affected by background or other variables. What is decline for one person may be stability or even growth for another. If we aim at promoting successful ageing in older population groups, it is important to detect the difference between what predicts genuine decline and what predicts entry into old age in a more or less favourable position.
40 The PGMs on the associations between changes in the different successful ageing domains and changes in well-being showed that losses in physical, social and mental functioning were associated with lower levels of well-being. This confirms that changes in domains defined by older adults as important aspects of successful ageing are in fact reflected in level of general well-being. The strongest correlations were found between changes in SRH and well-being (r= -.76) and mastery and well-being (r= .98), whereas the weakest correlations with change in well-being were found for functional limitations, cognitive capacity and number of chronic diseases. This further supports the distinction between objective and subjective dimensions of successful ageing (Pruchno et al., 2010) and suggests that well-being in old age depends more strongly on one’s subjective evaluations of health and the possibilities for controlling one’s own life than on objective measures of functioning. It appears that subjective satisfaction in advanced age is possible despite decline in objective aspects of health and functioning. This finding, however, is not new and has previously been labelled the “disability paradox” (Albrecht/Devlieger, 1999), referring to the fact that many people with serious disabilities still experience good quality of life. It suggests that people have the capacity to use different mechanisms of adaptation to changing circumstances, and that this is a more crucial factor in successful ageing than absence of disease and disability (Inui, 2003).
41 Despite its various strengths, such as the large population-based sample, the wide variety of objective and subjective measures of successful ageing and the long follow-up period, this study also has limitations. Firstly, the average change trends reported here might reflect more advantageous levels and more positive change models due to survivor bias. Some of the measures used were also perhaps not optimal in terms of capturing broad constructs (e.g. SRH, social relations, SES). Studies using more sensitive measures might be needed to validate the relations reported here. Caution is also required regarding distinctions between subjective and objective aspects of successful ageing, since results reported here rely on self-reports only. Some domains were thought to reflect more objective dimensions, but they were not objectively assessed measures. Also, self-rated successful ageing was not directly measured in this study. A general well-being domain was instead created as a criterion to evaluate other domains against. This choice was made because some lay reports indicate well-being to be an “overarching” dimension of successful ageing, because the dimension reflected a distinct positive outcome rather than just the absence of decline, and because it was thought to be least sensitive to age (i.e. not by definition favouring younger subjects). It should be noted, though, that other researchers may find other criteria more important, and also that direct conclusions on self-rated successful ageing cannot be drawn. Finally, the selection of background variables was in essence data-driven, as we selected predictors that appeared to have a significant effect on successful ageing, rather than theoretically relevant predictive variables. Future studies should further explore the different domains of successful ageing by alternative longitudinal approaches, e.g. latent class growth analyses.
42 Despite the limitations of the study and although the concept of successful ageing is surrounded by disagreement, the results presented here show that it has potential to serve as a useful basis for multidimensional and holistic studies on the process of ageing. The findings highlight the importance of studying true change rather than drawing conclusions based on cross-sectional data, and of distinguishing not only between subjective and objective aspects of ageing well but also between the context of changing and the context of being. The fact that decline in domains generally considered as crucial for successful ageing – absence of disease and high functional and cognitive capacity – showed the weakest correlations to decline in well-being also gives empirical support to much of the critique of the successful ageing paradigm. If we are open to defining successful ageing as something more than achieving a (socially and politically) desirable image of ageing (Dillaway/Byrnes, 2009; Rozanova, 2010), we need to take into account that doing well is not the only path to feeling good.