Optimizing depression treatment intensity and duration is crucial, given an overburdened mental healthcare system. However, decision-making is challenged by heterogeneous treatment effects. We aimed to investigate these effects, accounting for confounders and population heterogeneity, in a real-world dataset from specialized mental healthcare.

The study included 36,946 participants from mental healthcare providers in the Northern Netherlands. We measured the effects of treatment duration and intensity on time to depression recurrence, using monthly costs as a proxy for treatment intensity. An accelerated failure time model was used, adjusting for confounding via entropy weighting. Non-linear effects were examined using restricted cubic splines to identify turning points, after which linear analyses were stratified. Population heterogeneity was explored through K-means clustering analyses, followed by cluster-specific analyses.

In the high-intensity group (above €360/month), a €1000/month increase in treatment intensity may reduce time to recurrence by 16% (acceleration factor [AF] 0.84, 95% CI 0.77–0.92). Conversely, the same increase in the low-intensity group might prolong recurrence-free time by 9.6-fold (AF 9.6, 95% CI 2.18–42.31). Extending treatment duration by 6 months may reduce time to recurrence by 7% (AF 0.93, 95% CI 0.89–0.97) in the long-duration group, with no significant effect in the short-duration group. Five clusters emerged, three of which comprised only women, with AFs of 0.67, 0.80, and 0.81, respectively, under high treatment intensity.

Increasing treatment intensity appears worthwhile only in the low-intensity group, though residual confounding remains possible.