Diagnosing Depression Symptom Profiles: Dichotomous vs. Polytomous Diagnostic Classification Models and Differential Item Functioning Analysis

Objective Polytomous items are often dichotomized for use in dichotomous diagnostic classification models (DCMs), reducing information and distorting symptom estimation. Existing R packages also lack support for differential item functioning (DIF) analyses in polytomous DCMs. This study first examined the impact of dichotomization on depression diagnosis/screening and then investigated gender-related DIF.Method Item responses of 4,772 German participants to the simplified Beck Depression Inventory were analyzed. The generalized deterministic, inputs, noisy “and” gate (G-DINA) and sequential G-DINA (sG-DINA) models were compared regarding model fit, symptom prevalence/profiles, classification accuracy/consistency/agreement, and correlations. The Likelihood Ratio test was used to detect DIF.Results Although G-DINA showed better fit with fewer parameters, sG-DINA more accurately captured depression symptom severity and identified more symptoms per individual. Gender-related DIF was explored in most items across one/more response categories.Conclusion Preserving polytomous response information supports symptom-level profiling for more precise and equitable depression assessment and counseling interventions.