Psychometric properties of an automated telephone-based PHQ-9

This study aims to evaluate the psychometric properties of an automated version of the Patient Health Questionnaire-9 (PHQ-9) to further facilitate its use in primary care. We automated the PHQ-9 using a computer telephony modality (interactive voice response system) called telephone-linked communication (TLC).

Eighty subjects were divided into four depression categories: none, mild, moderate, and severe. The automated questionnaire, TLC-PHQ-9, was administered to all subjects five times over a 3-month period, at intervals of 0 (T1), 1, 3, 4, and 4 (T5) weeks, respectively. The Inventory of Depressive Symptomatology (IDS) was administered (paper-and-pencil) at T1 and T5. We examined (1) reliability, (2) validity, (3) sensitivity and specificity, and (4) sensitivity to change.

(1) Test-retest reliability showed substantial agreement between T1 and T2, with a weighted kappa of 0.76 (95% confidence interval [CI] 0.67-0.85). Cronbach's coefficient alpha values ranged from 0.913 to 0.918 for each TLC-PHQ-9 assessment. (2) The weighted kappa of 0.78 (95% CI 0.70-0.87) for T1 and 0.73 (95% CI 0.63-0.83) for T5 showed strong agreement between TLC-PHQ-9 and IDS in all depression categories. (3) TLC-PHQ-9 demonstrated good sensitivity (82.4%) and very good specificity (90.7%) for moderate-plus depression and poorer sensitivity (54.2%) but very good specificity (97.8%) for severe-plus depression. (4) The weighted kappa of 0.53 (95% CI 0.35-0.70) indicated moderate agreement between TLC-PHQ-9 and IDS.

An automated telephony administration of the PHQ-9 appears to be a valid and reliable tool for monitoring depression symptoms and has strong fidelity across patients.