Simple rules for detecting depression

• An FFT's ability to detect depressed mood is compared to a unit-weight, a regression, and a naïve maximization model.
• The FFT was competitive with the unit-weight and the regression model, and outperformed naïve maximization.
• The more heavily false alarms were weighted, the better the fast and frugal tree and the unit-weight model performed.
• Simple FFTs and unit weight models offer simple and accurate screening tools for depressed mood.

Depressive disorders are major public health issues worldwide. We tested the capacity of a simple lexicographic and noncompensatory fast and frugal tree (FFT) and a simple compensatory unit-weight model to detect depressed mood relative to a complex compensatory logistic regression and a naïve maximization model. The FFT and the two compensatory models were fitted to the Beck Depression Inventory (BDI) score of a representative sample of 1382 young women and cross validated on the women's BDI score approximately 18 months later. Although the FFT on average inspected only approximately one cue, it outperformed the naïve maximization model and performed comparably to the compensatory models. The heavier false alarms were weighted relative to misses, the better the FFT and the unit-weight model performed. We conclude that simple decision tools—which have received relatively little attention in mental health settings so far—might offer a competitive alternative to complex weighted assessment models in this domain.