Abstract:
OBJECTIVE: To determine the criterion validity of the Sinhala version of the General Health Questionnaire-12 (GHQ-12) and GHQ-30 employing different sampling designs and scoring methods. METHODS: This was a descriptive cross-sectional study including 374 patients. The GHQ-30 was completed by the participants using likert scale and then converted into standard score. The ‘Clinical Examination’ was done blindly to the GHQ score as the reference standard. Total study sample was considered as a representative sample taken consecutively. Case-reference design included 126 cases and 126 randomly selected controls based on reference standard. Test result-based designs included two groups of positive and negative GHQ, based on the optimal cut-off level. Cut-off levels were determined by using three criteria. In addition stratum specific likelihood (SSLR) ratio also considered. RESULTS: Applying consecutive sampling design, for the GHQ-12, the optimal cut-off levels were 9/10 using likert score and 2/3 using standard score and for the GHQ-30, 25/26 using likert score and 6/7 using standard score. The optimal cut-off level depends on the different sampling designs employed in addition to criteria for determining cut-off levels. The SSLR of>1 was useful for determining optimal cut-off level. Irrespective of the scoring methods, application of case-reference design tends to be overestimation of the specificity with high threshold values and test result-based design tends to be overestimation of the sensitivity, compared to consecutive sampling. Using likert scoring method, the sensitivities were higher than standard scoring method. CONCLUSIONS: The optimal cut-off levels depend on the sampling design and the scoring method employed and criteria to determine cut-off levels.
