The prospective findings of this study found that more frequent contacts with friends and relatives significantly predicted higher QoL in older adults from the Czech Republic, Poland and Russia. Moreover, it appeared that depressive symptoms play a relevant role in this relationship.
Although the nature of the hypothesised relationship between SN and QoL has not been investigated in population-based samples of older adults from CEE, findings were consistent with UK-based evidence. The magnitude of the effect of contact frequency with friends and relatives on QoL in our sample was remarkably similar to those observed for older English adults [12, 14].
Despite the fact that depressive symptoms have not been specifically examined as a mechanism for the relationship between SN and QoL, our findings are congruent with research in the area. There is evidence of depressive symptoms mediating the relationship between social support and health-related QoL [28, 29]. Similar methods were used to assess the mediating role of depressive symptoms in these studies, even when exposure and outcomes were different from this study. Longitudinal studies focusing on eudemonic wellbeing have shown an association with depression, where greater symptoms were associated with diminished QoL . As depressive symptoms were assessed as a covariate and not a mediator in the mentioned study, the comparability of the results is limited. Nevertheless, the evidence is consistent regarding the negative association between depressive symptoms and QoL and is compatible with the hypothesised pathway effects in this study.
Apart from depressive symptoms, the available literature suggests other important mechanisms as to why older adults with more frequent interactions with friends and relatives tend to have improved QoL. Frequency of interactions, a structural aspect of SN, have been associated with Cohen’s and Wills’ main hypothesis . This hypothesis posits that being embedded in SN provides a positive effect on the individual. It has been suggested that these structural aspects can enhance QoL . The independent association of frequency of contact with QoL of this study, seen in both the regression and path analyses, fits with the mentioned hypothesis. However, it may not be the only mechanism that could explain this relationship.
Another plausible explanation is that people who have more frequent contacts can rely on higher levels of social support from their network. Face-to-face social contact is significantly related to adequate levels of perceived instrumental and emotional social support . In turn, higher perceived social support is related to higher levels of QoL . Functional aspects of SN such as social support, have been associated with Cohen’s and Wills’ buffer hypothesis . Structural aspects of SN like frequency of contact can influence functional aspects of them, related to social support, which can ultimately have an independent effect on QoL. Extensive evidence has suggested a positive and perhaps bidirectional association between social support and depressive symptomology . In this study, as the direct effect from SN and QoL remained strong after accounting for depressive symptoms, social support may also be an important explanatory mechanism for the strong associations observed in this population. A social support pathway was not incorporated into the path model because variables regarding social support were not available at the time.
This study encompassed some limitations. A potential source of information bias may arise from the measurement of variables. Although there is evidence of CASP-12v.3 being a better fit for this population , it could not be used due to lack of the variables for that version at baseline and follow-up in the HAPIEE study. Moreover, while contact frequency with friends and relatives has been operationalised as a marker of one’s SN , this measurement did not originate from a validated instrument. Another limitation of this study is the similarity between some items in the CASP-12 and CES-D scales, which could become a potential source of over-adjustment. For example, responses to the CASP-12-item scale, ‘I feel full of energy these days’ were very likely to coincide with those given to the CES-D 20 and 10 items, ‘I feel that everything that I did was an effort’. While the items were designed to capture self-realisation and somatic symptoms in CASP and the CES-D scales, in turn, there appeared to be an overlap between these two items in our data. However, the sensitivity analysis of the main path model also found that the share of the total effect due to depressive symptoms at baseline was considerable at 44%.
A recurring limitation of longitudinal studies is attrition . As the analyses required data from participants who took part in both waves, this made selection bias more likely in our sample. From the total of individuals with data on QoL at baseline, about half of them had at least one missing variable. Although the analytical sample was sufficiently large, there was some evidence of less favourable characteristics among the excluded individuals compared to the analytical sample. These differences, although statistically significant, were not large in magnitude. A potential imputation of missing data was considered, however, it was decided not to perform it as both longitudinal and cross-sectional results have high levels of agreement, as well as the sensitivity analyses of excluded individuals (Supplementary Tables 2, 3 and 4). Due to these reasons, imputation would have not changed our results. Attrition also could have played a role in the increase of CASP-12 score at wave 2 compared to the baseline. Although the analysis was made on complete cases, if attrition was more likely to occur on less healthy individuals, the healthier individuals left could have increased their QoL at follow-up, given that there are theories that support an increase of QoL over time at old age. The socioemotional selectivity theory  posits that people draw more positive elements from their relationships as they get older, which could also explain higher levels of QoL at follow-up compared to baseline levels. Non-random attrition could undermine the generalisability of the results. The survival effect also needs due consideration. People with poor health and QoL are more likely to die [3, 37], leaving healthier people who might be more prone to increase in their QoL at follow-up, as explained by the socioemotional selectivity theory.
There are aspects of the path analysis that also need to be addressed. Due to the temporality of the measurements and the number of points in time where data was collected, it was only possible to construct a half-longitudinal design . This type of path analysis is limited by the concurrent measurement of both exposure and mediator, which may lead to bias estimates. To reduce the potential bias of concurrent measurements, the sensitivity analysis (Supplementary Fig. 1) using only depressive symptoms at wave 1, to avoid measurement concurrency of the mediator and the outcome, was carried out. In this latter analysis, the indirect effect of depressive symptoms on QoL, although smaller, remained strongly significant. Still, the indirect effect may not be correct due to a model´s misspecification. Depressed individuals tend to engage less frequently with their SN as a result of their condition . Given that one diagnostic criterion for depression is impairment of social functioning , a path model that hypothesises a relationship between depressive symptoms and QoL that is mediated by contact frequency is also conceivable. This possibility was deemed as unlikely due to the substantial evidence discussed earlier reporting associations and models in a similar fashion as our path model [19, 20, 28, 29]. The assumption of stationarity, a stable degree of change over time, cannot be tested with only two points of measurements. Although the half-longitudinal designs have limitations, adjustments for baseline elements in both the mediator and outcome were considered to reduce these limitations . As the relationships between SN, depressive symptoms and QoL are complex, future research on older CEE adults should be replicated using data with a longer follow-up time where these temporal sequences can be evaluated.
Finally, there are other aspects of SN that are related to QoL that were not measured. Structural elements such as size of the SN and diversity need to be considered for a comprehensive assessment of the role of aspects of SN on QoL. Since there is evidence that these other aspects are especially important at older ages [8,9,10], it is feasible that our findings may not have fully captured the relationship between SN and QoL for older adults in our sample. Similarly, non-structural elements, such as quality of SN, have been posited to have a large influence on QoL , although for older population there is evidence of a weaker association compared to younger population . However, they represent a key feature of SN and should be considered in further research. Although our exposure was a combination of two elements, this decision was made based on the literature and psychometric properties of the combined exposure. Moreover, a separate analysis of the exposures was made (Supplementary Table 5) and it showed significance for both exposures. Also, the possibility that residual confounding influenced our results cannot be discarded.
There are some strengths of the study that should be considered. The longitudinal design enabled us to provide the first report on the nature of this prospective relationship for older adults in CEE. The large sample size provided enough power to detect small effects in the examined relationships. Furthermore, QoL was measured using the CASP-12v.1, which is an instrument validated to measure eudemonic wellbeing in older populations. In terms of methods, the path analysis is an appropriate method to approach our mediational hypothesis.
Findings suggest that there is an opportunity to increase QoL with strategies that promote social engagement and participation in ageing populations from CEE. This study, therefore, draws attention to particular strategies and programmes to promote mental health and wellbeing in later life. Although individual-level interventions remain important, our study suggests that interventions should have a comprehensive approach that is embedded within the interpersonal and community spheres of older adults’ lives .