Based on the vulnerability–stress model, we aimed to (1) determine new onset of depression in individuals who had not shown evidence of depression at baseline (5 years earlier) and (2) identify social, psychological, behavioral, and somatic predictors.

Longitudinal data of N = 10 036 participants (40–79 years) were evaluated who had no evidence of depression at baseline based on Patient Health Questionnaire (PHQ-9), no history of depression, or intake of antidepressants. Multivariate logistic regression models were used to predict the onset of depression.

Prevalence of new cases of depression was 4.4%. Higher rates of women (5.1%) than men (3.8%) were due to their excess incidence <60 years of age. Regression analyses revealed significant social, psychological, behavioral, and somatic predictors: loneliness [odds ratio (OR) 2.01; 95% confidence interval (CI) 1.48–2.71], generalized anxiety (OR 2.65; 1.79–3.85), social phobia (OR 1.87; 1.34–2.57), panic (OR 1.67; 1.01–2.64), type D personality (OR 1.85; 1.47–2.32), smoking (OR 1.35; 1.05–1.71), and comorbid cancer (OR 1.58; 1.09–2.24). Protective factors were age (OR 0.88; 0.83–0.93) and social support (OR 0.93; 0.90–0.95). Stratified by sex, cancer was predictive for women; for men smoking and life events. Entered additionally, the PHQ-9 baseline score was strongly predictive (OR 1.40; 1.34–1.47), generalized anxiety became only marginally, and panic was no longer predictive. Other predictors remained significant, albeit weaker.

Psychobiological vulnerability, stress, and illness-related factors were predictive of new onset of depression, whereas social support was protective. Baseline subclinical depression was an additional risk weakening the relationship between anxiety and depression by taking their overlap into account. Vulnerability factors differed between men and women.