Resumen: Background Data on the course of mental disorders during imprisonment are scarce. Longitudinal studies from high-income Western countries point to improvements of symptoms over time. The aim of the present study was to assess mental disorders and symptoms three years after baseline evaluation at imprisonment and to determine predictors of change in a South American prison context. Methods Consecutively admitted prisoners in Santiago de Chile were assessed at intake and reassessed after three years using the Mini International Neuropsychiatric Interview and the Symptom-Check-List 90 Revised (SCL-90-R). The global severity index (GSI) was calculated with standard deviations (SD) and compared using paired t-tests. The prevalence of mental disorders at baseline and at follow-up were compared using McNemar tests. Analyses of variance were conducted to evaluate whether prespecified socio-demographic variables and disorders at baseline predicted symptom change at follow-up. Results 73 (94%) out of 78 prisoners participated. The prevalence of major mental illnesses was lower at follow-up: 47 (64%) at intake vs. 23 (32%) at follow-up had major depression (p<0.001 22 (30%) at intake vs. 10 (14%) at follow-up had psychosis (p = 0.008). The mean GSI improved from 1.97 (SD 0.65) at intake to 1.16 (SD 0.82) at follow-up (p<0.001). Depression at baseline (F = 9.39; Z2 p = 0.137; β = -0.67; p = 0.003) and working or studying during imprisonment (F = 10.61; Z2 p = 0.152; β = -0.71; p = 0.002) were associated with strong improvement of the GSI at follow-up, whereas psychosis at intake was associated with relatively small symptom improvement (F = 12.11; Z2 p = 0.17; β = 0.81; p = 0.001). Conclusions In a resource poor prison context in South America, mental health symptoms and disorders improve considerably over three years during imprisonment. This applies especially to people with depression at intake. Offers to work or study during imprisonment may improve mental health outcomes.