1 Introduction

Questions about whether and to what extent individuals in their advanced old age remain satisfied with their life are of twofold interest, namely for understanding the particular dynamics of very old age, as well as for our general knowledge on the adaptivity of subjective well-being (SWB) under adversities which profoundly challenge the human adaptational system in advanced old age, such as the onset and accumulation of chronic health conditions (e.g., Kolberg 1999; Ferraro 2006; Freedman and Martin 2000; Fried 2000). In this article, we focus on adaptation of life satisfaction (LS) to chronic physical multi-morbidity in a sample of individuals aged 80 years and above.

Health constraints in advanced old age do not only differ from those typically experienced in younger ages by increased frequency of occurrence, but also in coming as irreversible losses due to chronic conditions, and by increasing co-occurrence of such conditions (e.g., van den Akker et al. 1998). Thus, among the very old chronic health aggravations may have increased impact on subjective evaluations of one’s life in general, that is, on LS, which is commonly considered a cognitive component of SWB (e.g., Diener et al. 1999). However, the older people get, the more physical degradation is expected and “on time” (Neugarten 1970), hence, very old people may perceive health problems as a “natural” part of the aging process, which could mitigate the effects of worse health on LS judgments. Furthermore, major life-span developmental theories of resilience suggested psychological mechanisms to cope with and maintain positive SWB under age-related health loss (e.g., Brandtstädter 1999; Baltes and Baltes 1990; Heckhausen and Schulz 1995). Without going into detail here, such reasoning boils down to consider health aggravations as highly relevant, but potentially complex aspect of LS in advanced old age.

Therefore, our first and basic study objective is to examine the impact of chronic physical multi-morbidity on intra-individual change in LS in advanced old age, to add to the limited body of empirical studies concerning LS in this age-group. Doing so, we limit our analysis to the impact of physical health loss, leaving out mental conditions. Considering impacts on evaluative judgments of LS, mental conditions, meaning some kind of interference of psychological functioning (including SWB), should be distinguished from physical conditions. In particular dementia, the most prevalent chronic mental condition in very old age (e.g., Fried 2000; Kolberg 1999), degrading the individual’s cognitive abilities, may not only lead to less positive evaluations, but rather disrupt the ability to evaluate as such.

Empirical studies of the impact of physical multi-morbidity on LS in very old age are still rare. Numerous studies addressed the relationships between physical health and LS, many of which involved “young–old” (up to about the mid-seventies), but few only “old–old” ages. Overall, indicators of objective physical health conditions showed only minor-to-moderate positive correlations with LS judgments (Diener et al. 1999; Smith et al. 2001; Veenhoven 2011). Also, the effects of physical multi-morbidity on LS have been studied rarely (Schüz et al. 2009).

1.1 Adaptation to Increasing Chronic Morbidity: Reduced Responsiveness of Life Satisfaction?

Our second study objective is to check for adaptation effects in the linkage of chronic multi-morbidity and intra-individual change in LS in very old age. As has been argued elsewhere (Schilling and Wahl 2006), concepts of psychological adaptation to chronic and progressive health adversities refer to processes wherein the physical impairment may have a reduced impact on psychological outcomes over time. Going back to “classic” definitions, such as outlined by Bevan (1965), adaptation originally implied the notion of a process that refers to changes in the individual’s psychological response to novel conditions, which may result in reduction or redemption of this response. With respect to adaptation to increasing chronic morbidity, this means that LS may be most severely corrupted across an “early” phase of accumulating chronic conditions, but “on the long run” with persisting and accumulating health constraints, LS worsening may attenuate or even turn into “recovery” to the level of LS held before the onset of the chronic phase. Considering psychological theory, such adaptation dynamics of LS under worsening health conditions could result from active psychological coping (e.g., Folkman and Moskowitz 2004). Also, adaptation may be expected from theories of happiness built on the idea that subjective evaluations imply some kind of a “rollback” towards pre-adaptational levels due to dynamical adjustments of subjective target states to the actual status of living conditions (e.g., Fujita and Diener 2005; Michalos 1985).

Due to this view of adaptation, accumulating exposure to chronic conditions may go in hand with less worsening and even some degree of restorative improvement of LS. As indicator reflecting such accumulated exposure, we consider the status of chronic multi-morbidity “reached” by the onset of the measurement. That is, we follow the simple reasoning that accumulation of also implies accumulated exposure to chronic conditions, meaning that the more chronic conditions developed within an individual, the more of such exposure is effective in promoting psychological adaptation. Thus, chronic multi-morbidity, in terms of the number of chronic conditions coexisting within an individual, may be taken as a kind of proxy measure of the quantity of health loss experiences, which “habituate” the individual to the consequences of chronic health degradation.

Empirical findings that address this particular view of adaptation of LS under chronic health aggravations are surprisingly rare. Reviewing the literature with regard to such adaptation effects, Easterlin (2003) concluded that health constraints tend to exert outlasting negative effects on LS, which may be restored by adaptation only partially. However, with this conclusion largely drawn from cross-sectional relations of health with LS measures, the evidence still appears rather inconclusive with regard to the intra-individual effects of adaptation. Thus, more longitudinal analyses of “change in (intra-individual) change” are needed to reveal whether LS changes follow the adaptive pattern of attenuation and restoration across the course of accumulating chronic health conditions.

1.2 Loss of Functional Competence as Mediator of Chronic Health Effects on Life Satisfaction?

Our third study objective was to examine the mediating effect of functional competence in the linkage of chronic multi-morbidity and intra-individual change in LS in very old age. With respect to very old age, functional competence, in terms of ability to conduct activities of daily living (ADL), could be considered as a fundamental aspect mediating the consequences of constrained health on subjective evaluations of one’s current live. Research on the oldest-old evidenced the importance of ADL measures as predictors of various aspects of SWB and depression (e.g., Freund and Smith, 1999; Schieman and Plickert 2007; Smith et al. 2002). In advanced old age, when the accumulation and progression of chronic conditions disable physical functionality, declines of functional competence tend to come frequently and as irreversible losses, impeding individual life conduct and preventing attainment of goals which may be a prerequisite of positive SWB. Moreover, these losses are frequently accompanied by losses in assisting social networks, such that among the very old, interference of ADL-independence may push hard on well-being and may promote negative evaluations of life. For the majority of those at younger ages, in contrast, basic abilities to conduct ADL are not substantially constricted and taken for granted by the most. Thus, irreversible loss of functional competence could be regarded as a rather unique experience of old age, meaning deprivation of a basic need which may degrade LS judgments.

However, considering functional competence as mediator of chronic health impact on LS raises the question as to whether this “functional pathway” also mediates LS adaptation, in that adaptation of LS basically follows from the adaption of functional competences to chronic health aggravations. That is, very old individuals exposed to an accumulation of chronic conditions may be increasingly forced to learn to conduct ADL under poor physical health, for example, by use of assistive devices (Wahl et al. 1999). With LS judgments corrupted by loss of functional competences, such stabilization and restoration of ADL performance may then promote also some stabilization or restoration of LS. Thus, the question if the impacts of chronic health burden on the LS of the very old are “cushioned” by adaptation dynamics could be further differentiated by asking whether such LS adaptation is simply a consequence of “primary” adaptation of functional competences to increasing health constraints, or whether it rests upon more specific adaptation dynamics in subjective evaluations of one’s current conditions (such as considered, e.g., in the set point model of happiness; Fujita and Diener 2005).

1.3 Research Aims and Hypotheses

Summing up the above considerations, this study addressed three research objectives.

  1. 1.

    Impact of chronic multi-morbidity on intra-individual change in LS: we hypothesized that, in general, the amount of chronic morbidity given at the baseline of our observational period predicts subsequent changes in LS.

  2. 2.

    Adaptation of LS to chronic multi-morbidity: we hypothesized that changes in LS are related with chronic multi-morbidity at baseline according to a curvilinear, U-shaped pattern, indicating largest LS aggravations among those with relatively medium accumulation of chronic conditions, but no pronounced tendency of LS decline towards both ends of the observed chronic health continuum.

  3. 3.

    Functional competence mediating the linkage of chronic multi-morbidity and LS: we expected that change in LS is driven by aggravations of functional competence, which may be triggered by chronic health conditions. We also checked if this mediation pathway explains LS adaptation, in that changes in functional competences are related to the baseline multi-morbidity according to the U-shaped pattern, showing largest ADL loss under medium accumulation of chronic conditions, which is”transferred” on the change in LS as well.

2 Methodology

2.1 Data

Data from the German sub-sample of the multi-national European research project ENABLE-AGE (Iwarsson et al. 2007) was used. The sample consisted of community dwelling very old people (born 1912–1922), living in single households in the urban area of Heidelberg-Mannheim. These eligibility criteria were chosen due to the ENABLE-AGE study focus on the relationships between the home environment and healthy aging indicators, which targeted the very old living alone, described as group particularly sensitive to environmental press (Iwarsson et al. 2007). Participants were drawn randomly from official registers, stratified by age (80–84, 85–89) and gender to guarantee sufficient sampling of 85+ and male (about 25%) persons. At baseline measurement 2002/3 (T1) home interviews were conducted with N = 451 (354 female; 78.5%) subjects, and 12-month follow-up home interviews were conducted 2003/4 (T2) with N = 323 (251; 77.7%). Drop-out analyses revealed small to medium effects of sample selectivity due to age, education and income comparing those who did not participate at T2 (128; 28.4%) with all T1 participants. For more details on the sample and the ENABLE-AGE study see Iwarsson et al. (2007) and Oswald et al. (2007). In 2005 (T3), another 16-month follow-up was conducted by means of phone interviews, with N = 300 (232; 77.3%) surveyed successfully. After having established good personal relationships during home visits at T1 and T2, telephone surveys at T3 were perceived as helpful and less exhausting for the participants, resulting in an extremely low drop-out rate (7.1%).

2.2 Measures

2.2.1 Chronic Multi-morbidity

We computed a sum-score of the number of self-reported chronic conditions (NCC) due to a disease list used in ENABLE-AGE. The list included 44 diseases selected from the ICD-10 manual (World Health Organization 1992) by trained physicians participating in the ENABLE-AGE study consortium. For the compilation of the list, geriatric expertise was obtained to ensure inclusion of diseases which could be expected to occur most prevalent within the age-group under study. The subjects were asked to indicate whether or not they suffered from the specified disease within the previous year. For the present study, we excluded seven diseases marking acute conditions which are curable even in very old patients (such as fractures) to compute the NCC for use as indicator of chronic multi-morbidity at the T1 baseline measurement. Table 1 shows the 37 conditions included in the NCC score, as well as the self-reported prevalence of the conditions in the sample.

Table 1 Self-reported chronic conditions: list of diseases and frequencies of occurrence
Full size table

2.2.2 Functional Competence

ADL-independence was assessed at all three measurement occasions by means of the ADL Staircase (Sonn and Hulter-Åsberg 1991). This instrument is an extension of Katz’s ADL Index (Katz et al. 1963), comprising five personal ADL items (feeding, transfer, going to the toilet, dressing, bathing) and four instrumental ADL items (cooking, shopping, cleaning, transportation). The overall score of ADL is computed as number of independently conducted activities (theoretically ranging between 0 and 9). Several studies on community-living older people demonstrated validity and reliability of the instrument (Iwarsson 2005; Iwarsson and Isacsson 1997; Sonn and Hulter-Åsberg 1991). In our sample, Cronbach’s α amounted to .71.

2.2.3 Life Satisfaction

In ENABLE-AGE, LS was assessed by means of a single-item self-evaluation measure asking for present satisfaction with life to be rated on a 0–10 numerical scale (0 = “absolutely dissatisfied”, 10 = “absolutely satisfied”). This easily applicable and economic single-item measure is used in many population-based social science surveys (e.g., Inglehart et al. 2008). Considerable methodological literature concerning its reliability and validity has been published, resulting in some consensus concluding that single-item measures provide sufficiently valid and reliable LS measurement, and do not perform markedly worse in measuring LS as multi-item questionnaires (Frey and Stutzer 2002; Veenhoven 1996).

2.3 Statistical Modeling

Apart from basic statistics, we analyzed the data by means of latent growth curve models (LGCM, e.g., Stoel 2003). With regard to the three research objectives listed above, two models were analyzed as depicted in Fig. 1. In Model 1 (M1), we modeled intra-individual trajectories in LS across the three measurement occasions and included NCC and NCC-squared as predictors of the latent slope component, hence regressing the individual rates of LS change on the chronic multi-morbidity measure due to a curvilinear, quadratic function. We also modeled the latent level-slope covariance by means of a directed path to be consistent with the model’s general rationale to predict change from baseline status. In Model 2 (M2), we extended M1 by inclusion of latent level and slopes of the ADL measures, regressing also the ADL slope component on NCC due to a quadratic function and adding the latent ADL slope component as predictor for the LS slopes. Thus, the overall effect of chronic multi-morbidity on change in LS is split-up in M2 into an indirect path, mediated via ADL slope, and a direct path.

Fig. 1
figure 1

Latent growth curve models M1 and M2

Full size image

All LGCM analyses were computed with Mplus software (Muthén and Muthén 2006) and by use of bootstrapping for confidence interval estimation, on which we relied in testing for the significance of path coefficient estimates. Bootstrap confidence interval estimation does not assume symmetry of parameter sampling distributions, which may be violated in case of skewed distributions of the variables involved and particularly in testing indirect paths involving products of coefficients, hence it is particularly recommendable for tests of mediator effects (MacKinnon et al. 2004; Shrout and Bolger 2002), which we conducted in M2 analysis. Thus, bootstrap testing seems more reliable here than application of symmetry based “conventional” tests, and we report significances indicating that a zero value of the coefficient lies outside the 95 or 99% bootstrap confidence interval. Model fit was assessed by use of the root-mean-square error of approximation (RMSEA) and the standardized root mean squared residual (SRMR), following recommendations to assess RMSEA ≤ .05 and SRMR ≤ .05 as good, RMSEA ≤ .08 and SRMR ≤ .10 as acceptable fit (see e.g., Schermelleh-Engel et al. 2003, for a comprehensive overview on definition of and guidelines to interpret model fit indexes). To deal with the missing values, which resulted in our database particularly from longitudinal dropout, FIML estimation was applied, which is at present the highly recommended method of missing data treatment (Schafer and Graham 2002).

3 Results

3.1 Descriptive Findings

Means, standard deviations, and bivariate correlations for the study variables are shown in Table 2. Regarding the mean level changes across the three measurement occasions, a general decline in LS appears statistically significant with p < .001 (repeated-measures ANOVA, T1-T2- and T2-T3-constrasts also significant with p < .001), whereas no such pronounced tendency appears for mean level change in ADL (repeated-measures ANOVA: overall change and T1-T2-contrast significant with p < .05, T2-T3-contrast not significant). Notably, NCC was only marginally skewed (skewness = .42), with only 9.6% reporting none of the diseases, but 10% reporting 8 or more (maximum = 12). Thus, the respondents in this sample were not at the “ceiling” of chronic multi-morbidity when entering the survey. Also, all the ADL and LS means appear rather far away from the theoretical minimum of the scales, with less than 5% of the subjects scoring below a scale value of 5 for both, ADL and LS, and at all measurement occasions. This finding is notable with regard to the potential adaptation effects on changes in ADL or LS, because it precludes a “bottom-effect” which may produce such a pattern, in that the individuals with high accumulation of chronic conditions at T1 had already reached the lowest levels of ADL-independence or LS and could not worsen.

Table 2 Means, standard deviations, and correlations of study variables
Full size table

With regard to the correlations listed in Table 2, baseline NCC appears correlated with all ADL and LS measures at only low (.10–.30) effect size levels. Correlations between ADL and LS measures vary from insignificant minor to low sized values, with all ADL measures correlating highest with the T3 measure of LS. The autocorrelations of the repeated-measures indicate rather low rank-order consistencies of ADL and LS, which can be taken as evidence for considerable intra-individual changeability in both variables. This arises also from the intra-class correlations, indicating the share of the inter-individual variance in the total (inter- plus intra-individual) variance of the measures, which were .60 and .37 for ADL-independence and LS, respectively.

3.2 Latent Growth Curve Modeling Results

3.2.1 Initial Analyses

Prior to M1 and M2 computation, we run simple LGCM for ADL and for LS to check for the suitability of the linear trajectories of the intra-individual changes.Footnote 1 Regarding the intra-individual dynamics estimated in these LGCM, the latent slope factor scores for LS ranged between −1.476 and −.058, the mean slope score was −.450. Thus, for all respondents a tendency of decline in LS across the study period was estimated. The ADL slope factor scores ranged between −1.432 and .339, the mean score was −.167. Negative ADL slope values were estimated for 76.8%, hence the vast majority of respondents declined in functional competence.

3.2.2 Model 1

We computed M1 to address our first and second research objective. M1 fitted the data well (RMSEA = .043, SRMR = .038, with χ2 = 5.45, df = 3). The path coefficient estimates, as well as the latent LS slope R 2 are shown in Table 3. Consistent with our hypothetical expectations, the NCC appears linked with the latent LS slope with statistically significant path coefficients. LS slope appears only marginally linked with LS level and the M1 effects together account for about 29% of LS slope variance. Thus, NCC at baseline “explains” a rather substantial share of inter-individual variance in subsequent intra-individual change in LS. Also, the effect of NCC-squared is significant and reveals a curvilinear relation as expected. Notably, the NCC scores have been mean-centered prior to the computation, such that the negative value of the linear NCC coefficient indicates a tendency of LS decline at the average level of NCC. Thus, the positive sign of the quadratic NCC coefficient indicates a slow-down of this LS declines with less or more chronic conditions reported at T1.

Table 3 Model M1 and M2 results: path coefficients, intercepts, covariances, and R 2
Full size table

3.2.3 Model 2

We computed M2 to address our third research objective, regarding the effect of intra-individual change in functional competence on LS changes. M2 fitted the data well (RMSEA = .042, SRMR = .038, with χ2 = 23.09, df = 13) and again, the path coefficient estimates and the R 2 for the latent slope components of ADL and LS, as well as the resultant estimates of the indirect mediation paths from NNC on slope LS via slope ADL are shown in Table 3. Due to our hypothetical expectation the path coefficient leading from ADL to LS slope is significant with positive sign, indicating that intra-individual development in LS was closely related with intra-individual changes in functional competence. Notably, the R 2 for the LS slope factor spread to .706 in M2, showing a rather large increase of “explained” LS slope variance due to the addition of the ADL slope component.

However, the effects of baseline chronic multi-morbidity on the ADL slope component are rather weak, with the NCC-squared effect significant, but only about 6% of ADL slope variation accounted for by M2. Notably, the negative sign of this quadratic effect implies a kind of inverse U-shaped relationship: given the negative ADL slope intercept (i.e., the change in ADL predicted for those with mean NCC), M2 predicts lowest loss of functional competence for those with medium NCC and larger losses towards both ends, low and high, of the NCC range (the overall mean of ADL level was −.178, thus, average loss of functional competence indicates that the most declined in ADL). This might indicate some kind of adaptation pattern, in that developing some chronic conditions people learn to dampen functional declines, such that those with medium NCC may show less functional loss compared to those with few NCC, but higher accumulations of chronic conditions overburden the functional competencies, leading to more ADL loss again.

With this only weak NCC effect on the ADL slope, the indirect mediation pathway from chronic multi-morbidity on intra-individual LS changes was insignificant and not confirmed, whereas the direct effects of NCC on LS slope still appear substantial. Thus, the effects of baseline level of chronic multi-morbidity on subsequent LS support the notion of other causal pathways, apart of the changes in individual functional competences.

To show that the curvilinear effect of baseline NCC on the LS slope component largely confirms the expected pattern of attenuation of LS decline across accumulation of chronic conditions and to explore this effect more in-depth, we visualized the relationship implied by the coefficient estimates in Fig. 2. The figure shows the latent LS slope scores resulting from M2 computation plotted against the values of NCC at baseline, and the quadratic regression curve due to M1 (thin curve) and M2 (bold curve) coefficient estimates (computed by inclusion of all estimated effects of the LS slope regression equation conditional on the mean of the LS level component). The main message which could be taken from the figure appears twofold. On the one hand, the curves predicting the LS slope from the NCC effects due to M1 or M2 coefficient estimates indicate an adaptation dynamic as expected hypothetically, in that, starting on the left with no chronic condition, increasing LS declines are expected, the more chronic conditions have been reported at baseline, but by reaching a level of about 5–6 conditions, this tendency attenuates, such that with approaching the very high levels of chronic morbidity at the right, less LS decline is predicted. This tendency appears particularly pronounced for the M2 curve, which shows the effects of NCC controlled for the simultaneous impact of ADL changes. On the other hand, however, apart of some LS slope values plotted on the left hand side, representing the slopes of the “healthy” with few chronic conditions at T1, most individual LS slope values are located below the zero line, meaning decline of LS. Moreover, the upswing of the curves in the right hand “high morbidity region” represents only few cases. Noticing that only 19.3% of the respondents reported seven or more chronic conditions, it seems that the “adaptation tendency” does not apply to the most, but only to those with very high accumulation of chronic conditions. Thus, the curvilinear effect of NCC may be taken as finding suggesting adaptation effects, such as hypothesized, or as artefact of fitting the quadratic curve to the few cases with extreme values of NCC. Our data, containing only few with very high NCC, cannot unambiguously rule out one of these possibilities.

Fig. 2
figure 2

Latent slope scores of life satisfaction by number of chronic conditions

Full size image

4 Discussion and Conclusions

In general, this study confirmed that for individuals in advanced old age, chronic multi-morbidity is a detrimental factor of happiness, denoting a condition which worsens individual judgments of LS. Taking into account that chronic conditions occur frequently within that age group, such that it could be experienced as expectable concomitants of reaching advanced old age, such negative impact on individual evaluations of one’s life appears not as self-evident, as it may at the first sight. Moreover, the findings also revealed some support of our hypothetical assumption of an adaptation effect concerning the impact of accumulation of chronic conditions on the LS of individuals at advanced old age. That is, increasing chronic multi-morbidity may not strictly and unavoidably be accompanied by continuous loss of LS. Rather, aging individuals may be able to stabilize at least partially their well-being under worse chronic health conditions, as suggested by the curvilinear effect of NCC on the intra-individual development of LS, indicating an attenuation of the LS worsening with greater accumulation of chronic conditions. This adaptation effect may be explained as consequence of psychological mechanisms of coping, self-regulation, or set point adjustments (Folkman and Moskowitz 2004; Brandtstädter 1999; Baltes and Baltes 1990; Fujita and Diener 2005), though it would exceed the scope and aims of this study to evidence the operation of these mechanisms as such.

However, these conclusions on adaptive potentials of very old people should be drawn from our results with caution only, noticing that the quadratic curve representing this NCC effect in Fig. 2 indicates such adaptive attenuation under a rather high level of chronic multi-morbidity, which occurred for only few cases in our database. Thus, with respect to our hypothetical expectation of such an adaptation effect, the findings leave some ambiguity and might be taken as tentative confirmation, to be interpreted in regard of the study limitations as discussed below. Adaptation of LS appeared as stabilization under high accumulation of health constraints, but no evidence for restorative increases was found and LS worsened among the most of our sample across the study period.

Overall, these findings may be considered relevant with respect to previous research on adaptation to adverse health in general and in particular in old age. A multitude of studies have addressed the effects of physical health aggravations and functional loss on SWB, and from reviews of these, it could be concluded that SWB is more closely linked with self-reported health ratings as compared with objective measures of physical health (e.g., Diener et al. 1999; Easterlin 2003; Michalos et al. 2000). Low effects of objective health indicators on measures of SWB have been interpreted in terms of adaptation, meaning that people “recover” from losses of well-being resulting from initial reactivity to physical health problems, whereas subjective health ratings may to some extent reflect the individuals’ psychological adaptation to health conditions. However, even with respect to objective health indicators it has been concluded that adverse health changes have negative effects on SWB, and that “there is less than complete adaptation to deteriorating health” (Easterlin 2003, 11178). Our study adds to this general conclusion, in that our findings suggest that adaptation to accumulating chronic morbidity provides some stabilization of LS, rather than restoration of the LS level held before the chronic phase. Thus, people cannot fully maintain their LS when exposed to chronic conditions, but at least they can prevent excessive dissatisfaction when more and health problems occur.

With respect to research on LS in old age, the notion of a “stability despite loss paradox” has become widespread across past decades (Kunzmann 2000), meaning absent age-related decline of LS across old age, which was taken as evidence of resilience and successful adaptation to the psycho-social losses and health aggravations frequent in the old age period (Staudinger 2000; Baltes and Carstensen 1996; Brandtstädter and Greve 1994). However, such absent age-related decline was largely found in cross-sectional studies, which did not (or with low sample sizes only) include very old age (Schilling 2006). In contrast, longitudinal findings including very old study subjects evidenced substantial losses in LS noted increasingly with the “old–old” age period (Mroczek and Spiro 2005; Schilling 2005) or with close distance to death (Berg et al. 2011; Gerstorf et al. 2008; 2010; Palgi et al. 2010), which may indicate the impact of increasing adversities, particularly health loss in the end-of-life period. Our findings of “incomplete” adaptation of LS to chronic multi-morbidity add to this research. On the one hand, the emergence of chronic morbidity in late life narrows people’s LS, contradictory to the notion of a paradox. On the other hand, there may occur at least some “stability despite repeated loss”, preventing continuous dissatisfaction under accumulation of chronic conditions.

Apart from the effects of chronic multi-morbidity, our findings demonstrate that in our sample of advanced old individuals, loss of functional competences to conduct their everyday life independently was crucial in corrupting LS. This adds to previous findings of ADL measures as crucial predictors of SWB in very old age (e.g., Freund and Smith 1999; Schieman and Plickert 2007; Smith et al. 2002). Notably however, our hypothetical expectation that loss in functional competence mediates the effects of chronic multi-morbidity on LS change was not confirmed. That is, we found only weak relationship between baseline multi-morbidity and the ADL slope component, such that the respective indirect “mediation pathway” appeared irrelevant as well. The linkage between intra-individual changes in the ADL measure and in LS appeared as rather strong linear effect, working independently of the amount of chronic multi-morbidity reached by the onset of the study period.Footnote 2

Such independence may appear counterintuitive, as it could have been expected that the more chronic conditions co-occur within an individual, the more loss of functional abilities these may cause. In contrast, our results underscore that loss of everyday functional competence is weakly linked with chronic health problems in a curvilinear way, which suggests that people with a medium NCC experience less loss of functional competence, as compared with people with low or high NCC. Accounting for a rather small proportion of change variation in ADL, this effect should not be overstated.

The overall linkage between NCC with ADL slopes may be weakened because various chronic conditions differ in terms of functional declines they provide, and interact in providing with personal and environmental aspects (e.g., Stuck 1999; Wahl et al. 1999). Theoretical frameworks such as the Disablement Process Model by Verbrugge and Jette (1994) considered the “translation” of chronic disease into loss of everyday functional competence as a complex process, involving constraints of physical function which are caused by the disease and could hamper the person’s competence to conduct particular everyday activities. It depends on personal conditions and environmental circumstances if a given physical constraint is crucial for conducting a particular task in everyday life. Thus, chronic health problems could generally be considered an origin of processes leading to functional decline in old age, such that increasing chronic burden comes with worse functional competence, producing a negative “cross-sectional” association between levels of chronic multi-morbidity and functional competences at a fixed point of observation time in very old samples. This was revealed in our analysis in terms of the significant negative correlation between NCC and the latent ADL level component. However, this process employs multiple confounding variables, such that the NCC reached by the onset of an observation period does not simply add up to the amount of intra-individual decline in functional competence across this period.

Overall, our findings underscore the substantial role of loss of everyday functional competences in corrupting individual evaluations of one’s life. Notably, in our sample loss of functional competence hit the vast majority of participants regardless of the chronic health level “reached” by the onset of our observational period. Thus, it provides a crucial corruptor of LS at the end of the human lifespan, even for those not affected by much chronic conditions in a medical sense. Loss of functional competence may mark a limit to psychological adaptation to worsening conditions of living in advanced old age. Concluding from this result, under conditions implying loss of functional competence, such as advanced old age, it may be considered as crucial source of inter- or intra-individual variation in LS, as well as in other aspects of SWB. Apart of the aging perspective, research on SWB should consider functional competences as basic need, able to affect happiness crucially, if it is not met.

Several limitations of this study deserve consideration. First, our use of NCC as indicator of accumulation of chronic health burden may be questioned. This sum-score summarizes conditions which may differ largely in terms of severity of health aggravation and medical treatment options, and it is based only on self-reports of the presence of these conditions, without any objective and detailed information on the time of diagnosis, medical treatment received, and severity and “chronicity” of individual disease progression. Thus, the NCC measure does not cover all in-depth information on various facets of chronic burden present at the time of measurement and may be also affected by subjective knowledge biases regarding one’s own medical conditions. However, our trust in NCC as reliable indicator rests on the ENABLE-AGE project’s data collection procedure, which involved intense interviewer training to ensure the respondents understanding of the questionnaire items and to pay attention for and protocol answers which appeared unreliable. As the interviewers reported that the respondents’ knowledge on the conditions asked for in the disease list seemed very good, the NCC measure may at least not be crucially biased by false responses. Thus, we would argue that the NCC can be taken as coarse indicator of “more or less” chronic multi-morbidity, such as considered theoretically in our study hypotheses, without claiming that it also gives an in-depth and exhaustive picture of the study subjects’ chronic health burden.

Second, our theoretical considerations of adaptation in terms of a process attenuating the relationship between the NCC “reached” at baseline measurement and the subsequent intra-individual development in LS may be questioned. This approach bears on taking the number of chronic conditions as indicator of the quantity of chronic health loss experiences relevant for individual habituation to chronic health degradation. However, also aspects such as speed and duration of chronic health aggravations deserve consideration as potential aspects triggering the individual adaptation process (Schilling and Wahl 2006). For example, in our previous work on adaptation to chronic vision loss, we focused on aspects such as time of living under the chronic condition and the degree of progression of the chronic impairment (Schilling and Wahl 2006; Schilling et al. 2011). These aspects could be seen as quantities relevant for adaptation, however, with regard to chronic morbidity in general, rather than a particular chronic disease, multiple information covering the respondents medical history previous to the study period would be needed to be merged into some indicators of duration and progression of chronic morbidity across multiple chronic conditions, depending also on the validity of the operational rules of merging these information. Thus, the strategy chosen in this study provides could be taken as pragmatic approach to model adaptation to chronic morbidity in general, when other indicators of exposure to chronic health loss are costly and hard to operationalize.

Third, we could not model in our analyses the effects of simultaneous intra-individual change in NCC across the study period on the LS changes, because the disease list was applied only at the baseline measurement of the ENABLE-AGE study. It would be desirable to extend the longitudinal scope of this analysis, in order to analyze LS adaptation with respect to duration, speed, and severity of chronic conditions. Such in-depth analysis needs an enlarged longitudinal database to monitor development of chronic health aggravations across longer observation periods.

Fourth, as multi-item questionnaires commonly are assumed to perform better in terms of reliability and validity, it may be seen as a limitation to our findings that no multi-item LS measures were available in our longitudinal database. Thus, we use a single-item measure of LS. However, as was already noted above, the single-item 11-point LS scale is widely accepted in research (e.g., Easterlin 2003; Inglehart et al. 2008) and psychometric findings suggest that it does perform well enough to be used (Veenhoven 1996).

Summarizing the findings of this study, chronic multi-morbidity in terms of NCC and loss of everyday functional competences are crucial sources of worsening happiness in advanced old age, which together accounted for a large portion of the LS latent slope factor variance. However, both variables should not be considered as close concepts, meaning strongly linked indicators of the very olds’ physical health, but rather as distinct aspects of increasing health risks in advanced old age. With respect to chronic multi-morbidity, the accumulation of chronic conditions “reached” at a given point in time exerts pressure on individual LS and promotes more negative evaluations of one’s current life, but this evaluation may also be driven by psychological adaptation mechanisms, able to absorb this pressure in order to prevent boundless dissatisfaction under extreme chronic burden. However, LS worsened in our sample of very old, which contradicts the widespread notion of a “stability despite loss paradox” of SWB in old age and adds to findings that evidence general decline in LS across the old–old age period (Mroczek and Spiro 2005; Schilling 2005). Thus, our findings suggest limits of psychological adaptation to worsening conditions of living in advanced old age.

Finally, in terms of an applied perspective, our data support a dual intervention strategy as particularly useful in very old age. First, the crucial impact of losses in functional competence on intra-individual development of LS, as evidenced in this study, underscores the relevance of training and rehabilitation of everyday functional competence (Stuck 1999). Second, in terms of very old patients’ multi-morbidity, psychological consultation and education in psychological management strategies could be successful as well, in enhancing psychological adaptation potentials to maintain a positive view of life (Steverink et al. 2005).