Evaluating the Beck Depression Inventory-II Fast Screen in middle-aged and older adults using random forests and factor analytic methods.

= 6.36) enrolled in an observational research study. Two random forests (RF)-derived scales were compared to the existing BDI-FS in predicting current major depressive episode. Scales were compared using confirmatory factor analysis, measurement invariance, and receiver operating characteristic analysis. Overall, results were comparable across scales. Models based on RF-selected items did not demonstrate better model fit than those based on the BDI-FS; however, scores derived from the RF-selected items exhibited higher reliability. In addition, models based on the RF-derived scales and the BDI-FS showed invariance across demographic and clinical variables. Scores derived from all scales, including RF-derived scales and BDI-FS, showed excellent classification performance in receiver operating characteristic analysis. Future research should further investigate these RF-derived scales, with the BDI-FS, to evaluate their reliability and validity across other diverse populations. (PsycInfo Database Record (c) 2026 APA, all rights reserved).