Background There is a lack of studies that model patient-specific recovery patterns in depressive symptoms after stroke. Patient- specific recovery patterns provide insight into how patients progress overtime, critical time periods and time varying factors that influence these recovery patterns. This study used a multilevel modelling approach to indentify trajectories of depressive symptoms after stroke and investigated the effect of socio-demographic factors and baseline stroke severity on recovery patterns after stroke.

Data Secondary data analysis of 592 stroke patients who participated in a stroke outcomes study (SOS2) was conducted. The General health questionnaire (GHQ) 28 depressive and anxiety subscales were assessed at multiple time points over a year (3, 6, 13, 26 and 52 weeks). Average recovery trajectories for anxiety and depression subscales were characterised using multilevel modelling approach. Random intercept and random slopes models were preferred based on change in deviance. The predictors were age, sex, marital status, education, ethnicity and presence of hemianopia, occupation, house ownership, smoking, baseline stroke severity measured by Barthel index and River mead mobility index.

Results The mean age of the sample was 70.34 (SD 11.89) and 253(43%) were female. The average anxiety and depressive trajectories showed that most change occurred in the initial-13 week period. Female gender (B = 1.029; SE = 0.288), younger age (B = -0.035; SE = 0.015), presence of hemianopia (B = 1.251: SE = 0.497), increased baseline stroke severity (B = -0.073; SE = 0.029), living in a council/tenant house (B = 1.026; SE = 0.315) were associated with increased anxiety symptoms.

Conclusion Multilevel modelling has indicated that most change in depressive symptoms after stroke occurs in the initial 13 week period. Gender, place of residence, age, presence of hemianopia and baseline stroke severity are significant predictors of depressive symptoms after stroke