![Sample our Health and Social Care journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5938/TOC-Subject-Sample-2021-Health-Social-Care.jpg"})
Abstract
Background: The General Health Questionnaire (GHQ) is a well known screening instrument, but little is known about its psychometric properties.
Aims: The study aimed to evaluate the predictive power of GHQ-28 and compare alternative methods of enhancing its discriminatory power. We examined the detailed sensitivity and specificity of individual items of the GHQ-28 to understand its psychometric properties, and suggested the combination of items that best predict cases with high discriminatory power.
Method: The data obtained in a study of non-psychotic psychiatric disorder in two-phase study of General Practice consulters was analysed. A total of 1670 consecutive patients aged 16 – 65 years were screened using the GHQ-28 and 336 were interviewed using the Structured Clinical Interview for DSM-IV Axis I Disorders (SCID). Cases of non-psychotic psychiatric disorder were defined at interview using DSM-IV and Bedford College criteria. The diagnostic outcomes were dichotomized as cases and non-cases, and were combined with the GHQ data for the logistic regression. The regression was used to select individual GHQ items that exhibit high predictive power. Based on ROC analysis, the performance of the combination of the selected items with high predictive power was compared to the performance of the full 28-item version and the use of stratum-specific likelihood ratios (SSLR).
Results: GHQ items reflecting depression have high specificity while items reflecting somatic symptoms that are non-specific to psychological problems, have high sensitivity. From the GHQ-28 items, eight items were selected using the logistic regression. Five items were selected because of their high sensitivity or specificity values. The remaining three items were selected because they enhance other significant items. Using an area under ROC curve to indicate the level of predictive power, the results showed that by removing the 20 insignificant individual items, the predictive power was increased slightly from 81.6% to 84.4%. When compared to SSLR, both tests seemed to perform better than SSLR.
Conclusion: The study has shown an alternative analysis to evaluate the GHQ's validity. This approach to the GHQ psychometric properties may provide a simple way to enhance and preserve the high predictive power of GHQ. The computational scheme could be easily applied to other screening instruments to assess their validity.