Psychometric properties and correlates of the Beck Depression Inventory in a community-based and homeless or precariously housed sample

Background: Although homeless or precariously housed populations have a high prevalence of depression, the suitability of existing symptom measurement tools remains unknown. The present study explores the psychometric properties and correlates of the Beck Depression Inventory (BDI) in a community-based sample. Methods: 475 participants recruited from an impoverished neighbourhood in Canada completed the BDI, as well as multiple other clinical assessments. Validity, reliability, and change sensitivity were assessed, and a receiver operating characteristic (ROC) curve was analysed to determine a threshold score for clinical depression. Relationships between BDI scores and psychiatric diagnoses, gender, age, and functional outcomes were studied. Results: A wide range of BDI scores (0–58) was found among participants across psychiatric diagnoses. Convergent validity with psychopathology scores was present (r = 0.659, p < 0.001). Test-retest reliability was excellent (intraclass correlation = 0.915, p < 0.001), as was internal consistency (Cronbach's α = 0.907) and change sensitivity (F = 7.801, p = 0.007). The ROC curve indicated a threshold score = 19 for detecting clinical depression with moderate sensitivity (70%) and specificity (78%). High BDI scores were more strongly associated with poor functioning in older than in younger participants (estimate = -0.003, p = 0.006). Limitations: The generalizability to other vulnerable populations is uncertain. Conclusion: The BDI exhibits sufficient validity and reliability for assessing depressive symptom severity in a community-based, homeless or precariously housed sample, and may aid assessing the severity of clinical depression. BDI scores may also signal functional impairment in older persons in these groups