This study analyzes decisive measures of efficiency of a test, receiver operating characteristic (ROC) analysis and QROC analysis combined with considerations about clinical, health-economic, and ethical aspects when choosing screening instruments.
Analyses of Common Mental Disorders Screening Questionnaire (CMD-SQ) and its subscales SCL-SOM, Whiteley-7, SCL-ANX4, SCL-DEP6, SCL-8, plus combinations, for early detection of psychiatric disorders, are the subject for this analysis. In all, 46.4% of 2,414 new people with continuous sickness absence for more than eight weeks over one year in a well-defined Danish population of 120,000 inhabitants participated in the study. The study was performed as a two phase study. All 1,121 persons in Phase 1 filled in the CMD-SQ. In Phase 2, a random subsample of Phase 1 on 337, the people were further examined by a psychiatrist using SCAN as gold standard. The analyses were performed as weighted analyses on Phase 2.
From 17 analyses it was shown that the efficiency of a test, ROC analyses, and QROC analyses resulted in different optimal scales and cut-points. The random possibility of a positive test or negative test in the population is discussed for efficiency and ROC analyses. QROC analyses correct for this by the relative kappa-values as decisive measures. However, QROC analyses may discard tests of value, all depending on the purpose of the test.
In supplement to test statistics the capacity of services to follow up on screening, ethics, and health economy are issues that should be considered in deciding what rating scale and cut-point should be adopted.