Predictive factors for remission in post-stroke depression: A Markov model cohort study

Post-stroke depression (PSD) follows a dynamic remission-relapse course, yet predictors of state transitions remain unclear. Here, we applied a continuous-time multi-state Markov model to 536 ischemic stroke patients with 1,441 longitudinal assessments. We quantified transition intensities between non-PSD, mild-PSD, and severe-PSD states and identified baseline predictors of deterioration and recovery. The transition intensity from mild to non-PSD was 5.5-fold higher than progression to severe-PSD. Education ≥10 years reduced deterioration risk (HR = 0.63), while functional independence (Barthel index >60) and social support predicted remission. Severe-PSD exhibited an apparent short sojourn time (estimated mean 2.7 months) and frequent observed transitions toward milder states; however, in the absence of mortality data, this estimate may be biased by unobserved competing events and should be interpreted as hypothesis-generating rather than definitive. These findings enable early risk stratification and targeted intervention for PSD management.