It is important for doctors and patients to know what factors help recovery from depression. Our objectives were to predict the probability of sustained recovery for patients presenting with mild to moderate depression in primary care and to devise a means of estimating this probability on an individual basis. Participants in a randomized controlled trial were identified through general practitioners (GPs) around three academic centres in England. Participants were aged >18 years, with Hamilton Depression Rating Scale (HAMD) scores 12-19 inclusive, and at least one physical symptom on the Bradford Somatic Inventory (BSI). Baseline assessments included demographics, treatment preference, life events and difficulties and health and social care use. The outcome was sustained recovery, defined as HAMD score <8 at both 12 and 26 week follow-up. We produced a predictive model of outcome using logistic regression clustered by GP and created a probability tree to demonstrate estimated probability of recovery at the individual level. Of 220 participants, 74% provided HAMD scores at 12 and 26 weeks. A total of 39 (24%) achieved sustained recovery, associated with being female, married/cohabiting, having a low BSI score and receiving preferred treatment. A linear predictor gives individual probabilities for sustained recovery given specific characteristics and probability trees illustrate the range of probabilities and their uncertainties for some important combinations of factors. Sustained recovery from mild to moderate depression in primary care appears more likely for women, people who are married or cohabiting, have few somatic symptoms and receive their preferred treatment.