Development and validation of an international risk prediction algorithm for episodes of major depression in general practice attendees. The PredictD Study

Background: There are no valid risk assessment tools for the prediction of future episodes of major depression. We aimed to develop a risk prediction algorithm for European general practice attendees and test it in a non-European setting. Methods: General practice attendees were interviewed at baseline, 6 and 12 months in six European and one Latin American country. We measured 39 known risk factors for DSMIV major depression made according to the Composite International Diagnostic Interview. Findings: We recruited 10,048 people who were 66% of all attendees approached. Subsequent response rates were 89.5% at six and 85.9% at 12 months. 6190 European attendees were not depressed at recruitment and form the population in which we developed an 11 factor risk prediction algorithm. Five factors were immutable past events or personal characteristics: age, sex, educational level, lifetime screen for depression and family history of psychological difficulties. Six were existing mutable factors: physical health and mental health scores on the Short Form 12, DSM IV anxiety and panic disorders on the Patient Health Questionnaire, reported difficulties in paid or unpaid work and experiences of discrimination. The prediction algorithm performed with an Area under the Relative Operating Characteristics Curve (AUROC) of 0.798 (95% CI 0.776, 0.820) The standardised effect size for difference in predicted probability of depression between those who became depressed and those that did not was 1.23 (95%CI 1.12, 1.33). Application of the prediction algorithm in Chilean attendees resulted in an AUROC of 0. 724 (95% CI 0.685, 0.763). Interpretation: The PREDICT risk algorithm compares favourably with similar risk assessment tools for cardiovascular events and has potential in prevention of major depression.