Psychometric evaluation of a 6-item Chinese version of the Hamilton Depression Rating Scale: Mokken scaling and item analysis

Abstract

Background: It is unclear whether the 6-item Melancholia subscale of the Hamilton Depression Rating Scale (HAM-D₆ ) has better psychometric properties than the parental 17-item version (HAM-D₁₇ ) in the Chinese populations.

Methods: The study was to check the Chinese HAM-D₁₇ and HAM-D₆ for reliability and validity with Mokken scale analysis and item analysis. We reanalyzed a clinical sample of adult psychiatric outpatients (N = 214; age, mean (SD) = 38.3 (10.5) years; 63.1% female) in Taiwan.

Results: Our Mokken scale analysis showed that the HAM-D₆ was a moderate unidimensional scale (H_s = 0.44) while the HAM-D₁₇ was not (H_s = 0.26). The 5 items of the HAM-D₆ had strong invariant item ordering (H^T = 0.58). The HAM-D₁₇ and HAM-D₆ had comparable reliability (α = 0.79) and validity (r = 0.91), and the HAM-D₆ was more homogenous than the HAM-D₁₇ . To predict depression in remission (HAM-D₁₇ ≤ 7), the best cutoff of the HAM-D₆ was 4 (specificity, 87.5%; sensitivity, 100%; positive predictive value, 56.0%; negative predictive value, 100%).

Discussion: The Chinese HAM-D₆ was a reliable, valid, and unidimensional scale of depression.

Keywords: depression; nonparametric item response theory (NIRT); reliability; validity.