Statistical Analyses

We conducted all analyses with Stata14. First, we compare mean levels using one-way analysis of variance to assess gender differences in education and secondary healthcare use at retirement (T0). We then apply Generalised Estimating Equations models (Liang and Zeger, 1986) with repeated measurements and autoregressive correlation structure. This type of multilevel model separates the variances within the individuals from the variance between individuals and therewith acknowledges that observations within an individual are more similar than those between individuals. As the outcome variables are binary, we apply a logistic distribution function to analyse secondary healthcare use during the retirement transition. The results are displayed as odds ratios (OR) with 95% confidence intervals (CI).