Psychometric Properties of the Beck Depression Inventory – II Among Adult Psychiatric Inpatients

Purpose: Major depressive disorder (MDD) is a common mental disorder (U.S. lifetime prevalence rate = 16.5%) associated with severe clinical consequences (e.g., functional impairment and suicide). For providers seeking to screen clients for MDD, the well-established Beck Depression Inventory (BDI-I) has been widely utilized due to its demonstrated psychometrics across multiple clinical subpopulations. Yet, the psychometric properties and clinical utility of the revised BDI-II among adult psychiatric inpatients, a group at elevated risk for MDD, is largely unknown; a gap in the literature this paper sought to address.

Methods: Psychiatric inpatient participants (N= 1,904) recruited from a non-profit psychiatric hospital in Texas completed the BDI-II and the Behavior and Symptom Identification Scale (BASIS-24) upon admission. BDI-II reliability was examined via Cronbach’s alpha and construct validity via inter-item correlations and BDI-II total score correlations with BASIS-24 subscales. Factor analyses (exploratory and confirmatory) evaluated BDI-II factor structure. Diagnostic utility was explored using receiver operating characteristic (ROC) analyses of a 575 participant subsample assessed for MDD using the Structured Clinical Interview for DSM Disorders (SCID).

Results:  The BDI-II sample mean was 24.93 (SD=13.20), indicating that on average, participants experienced moderate depression at admission. Analyses of psychometric properties yielded Cronbach’s alpha of .93, and significant (p<0.01) BDI-II correlations with the BASIS-24 Depression (r=.79) and Overall subscales (r=.81). BDI-II factorial structure was explored by splitting the sample into two random halves. Maximum likelihood estimation extraction with direct oblimin rotation conducted on one random half decomposed the items into three factors with eigenvalues greater than one (8.57, 1.51, 1.20) although scree plot and principal components parallel analyses supported a two factor model (i.e., Cognitive and Affective/Anhedonia) that accounted for 48.03% of variance. The pattern and structure matrices indicated all items loaded onto one of the two factors (>.35); inter-factor correlation was -.68. Generalized least squares with bootstrapping confirmatory factor analysis (CFA) was conducted on the second random half due to non-normality of data; factors were correlated, correlated errors were not permitted. Chi-square tests were significant (p<0.001) due to the large sample size (Hoelter’s critical N=285). The RMSEA=.06, SRMR=.06, GFI=.93, and Adjusted GFI=.91 indicated good model fit. As the RMSEA of the null model was lower than .158 (RMSEA=.07), incremental indices of fit such as TLI and CFI were not valid. ROC analyses on the SCID subsample (72% prevalence rate of major depression) revealed low AUC (95% CI) = .65 (.61 - .71) for the BDI-II and produced cutoff scores with poor balance between sensitivity/specificity.

Implications: This large-scale psychometric investigation of the BDI-II with psychiatric inpatients revealed good internal consistency and convergent validity but suboptimal MDD diagnostic performance. Factor analyses decomposed the BDI-II into two factors (Cognitive and Affective/Anhedonia), mirroring the Cognitive and Somatic/Affective factors previously reported for psychiatric outpatients by the instrument authors. Study findings support BDI-II utility as an inpatient screening measure of depressive symptomology but call into question its diagnostic capabilities. Social workers within inpatient settings are recommended to use the BDI-II to assess depression severity but to consider alternative screening methods for MDD.