Latent Variable Decomposition of Transdiagnostic Psychopathology Networks: A Bifactor-S-1 Modeling Framework for Dimensional Symptom Architecture in Treatment-Refractory Clinical Populations

Abstract

Transdiagnostic psychopathology research has increasingly relied on network and factor-analytic paradigms, yet methodological heterogeneity limits cross-study comparability. This study proposes a bifactor-S-1 latent variable decomposition framework integrating Exploratory Structural Equation Modeling (ESEM) with regularized partial correlation networks to resolve the p-factor debate in treatment-refractory adult cohorts (N = 1,847). Utilizing item-level data from the DASS-42, MASQ, and DAST inventories alongside ecological momentary assessment (EMA) time-series, we operationalized symptom-specific residual factors while controlling for general psychopathology variance. Confirmatory network stability analyses via case-dropping bootstrapping demonstrated superior model fit (CFI = 0.963, RMSEA = 0.041) over competing correlated-factors solutions. Results indicate that internalizing and externalizing spectra exhibit hierarchically nested, non-orthogonal residual structures with clinically actionable bifurcation thresholds, offering empirically grounded targets for transdiagnostic cognitive-behavioral intervention protocols.