Within-individual variation of depression symptom measurement instruments in primary care: a retrospective cohort study

Depression symptom measurement instruments such as PHQ-9, HADS and BDI are used to aid in the diagnosis and management of depression. Variation in within-individual depression symptom measurements may affect their clinical utility but estimates of this variation are currently almost non-existent. Data on PHQ-9, HADS and BDI results and sociodemographic, lifestyle and comorbidity covariates were extracted from the CPRD database. A minimum of four measurements in the same individual was the only inclusion criterion. Within-individual measured variation was calculated as a coefficient of variation (CV) using a linear regression random effects model. RESULTS: 195,171 participants had at least four measures of PHQ-9 after impossible values were excluded. For HADS, this figure was 8231 participants and for BDI 215, making this by far the largest study of variation of depression symptom measurement instruments to date. The overall total coefficient of variation (CV_T) (95% confidence interval) was 0.412 (0.411 to 0.413) for PHQ-9, 0.370 (0.365 to 0.374) for HADS and 0.763 (0.701 to 0.826) for BDI. In contrast to studies of variation of other tests, variation decreased with increased mean patient score (i.e. variation decreased with worsening disease severity). Estimated within-individual variation in this analysis of real-world data is very high and has not previously been reported. Variation decreases with patient mean score. The high within-individual variation of depression measurement instruments and the higher estimated variation in less severely affected patients have important implications for the diagnosis, monitoring and clinical decision-making for depression.