Exploring what we know about retirement: A systematic review of constructs and measures of adjustment and adaptation to retirement and A meta-analysis of the relationship between retirement and depression in later life
Coelho, Claudia Carvalho De Matos Teixeira
Objectives: This thesis includes two distinct pieces of work. A systematic literature review (SLR) aimed to identify, organise and evaluate constructs and measures of adjustment and adaptation to retirement (AAR), as these are currently used in the literature. A meta-analysis (MA) sought to systematically evaluate the evidence of the relationship between retirement and self-reported symptoms of depression in later life. Methods: For the SLR, a comprehensive list of electronic databases, and additional sources, were searched (March-May 2013). The identified constructs of AAR were evaluated in relation to their content, clarity and frequency of use. The identified measures were evaluated in relation to pre-defined psychometric properties and frequency of use. For the MA, an equally comprehensive search strategy was used (December 2012-March 2013). A metaanalysis and systematic assessment of risk of bias were carried out on the studies eventually included. Results: In the SLR, 27 measures of AAR were put forward for evaluation, and 6 distinct clusters of constructs of AAR were identified. AAR is defined inconsistently in this literature. Most measures that are used to assess these constructs are adapted or imported from other contexts (e.g. mental health, well-being, life-satisfaction). The 7 measures that related specifically to AAR (‘retirement satisfaction and role adjustment’) lacked detailed psychometric information. Eight non-randomised studies were included in the MA, 5 cohort studies and 3 cross-sectional studies. Studies were grouped and analysed according to these two design-type subgroups There was evidence of high dispersion of effect sizes, variable risk of bias and methodological and statistical heterogeneity between studies in both sub-groups – cohort (Q=640.728, df =4, p<0.001), cross-sectional (Q=76.611, df=2, p<0.001). Summary effects were therefore not meaningful. Sensitivity and sub-group analyses did not account for high heterogeneity of effect sizes. Conclusions: The SLR concluded that the variability in outcomes of research on AAR found in this literature may be underpinned, in part, by the different constructs and measures that are used. The 27 measures evaluated did not seem, at face-value, to measure the same construct; their psychometric properties also varied. The interpretation of outcomes, and comparisons between studies, in this area is hindered by this inconsistency. The MA concluded that the relationship between retirement and self-reported depressive symptoms seems to be complex and variable. Effect-sizes of individual studies were small, non-significant and highly dispersed, and heterogeneity of true effects was high. These results may be limited by confounding factors in primary studies. This is discussed and contextualised in relation to the use of non-randomised studies in meta-analysis.