A comparison of the B-spline group-based trajectory model with the polynomial group-based trajectory model for identifying trajectories of depressive symptoms around old-age retirement.

The life event of retirement may be associated with changes in levels of depressive symptoms. The use of polynomial group-based trajectory modelling allows any changes to vary between different groups in a sample. A new approach, estimating these models using B-splines rather than polynomials, may improve modelling of complex changes in depressive symptoms at retirement. The sample contained 1497 participants from the Swedish Longitudinal Occupational Survey of Health (SLOSH). Polynomial and B-spline approaches to estimating group-based trajectory models were compared. Polynomial group-based trajectory models produced unexpected changes in direction of trajectories unsupported by the data. In contrast, B-splines provided improved insights into trajectory shapes and more homogeneous groups. While retirement was associated with reductions in depressive symptoms in the sample as a whole, the nature of changes at retirement varied between groups. Depressive symptoms trajectories around old age retirement changed in complex ways that were modelled more accurately by the use of B-splines. We recommend estimation of group-based trajectory models with B-splines, particularly where abrupt changes might occur. Improved trajectory modelling may support research into risk factors and consequences of major depressive disorder, ultimately assisting with identification of groups which may benefit from treatment.