A Psychometric Evaluation of the General Health Questionnaire (GHQ-12) in the ELSA Cohort Using Item Response Theory

Background. The 12-item General Health Questionnaire (GHQ-12) is a reliable and valid screening tool for psychological distress and is widely used in epidemiological research. This study aimed to optimise the length of the GHQ-12 without losing information by discarding items that were misfitting the scale following an item-response theory (IRT) model. Methods. The English Longitudinal Study of Aging (ELSA) is a panel study of a representative cohort of participants living in England. As part of the mental health assessment at wave 1 the GHQ-12 was administered. Complete GHQ-12 data were available for 18,552 participants. This sample was aged 18 to 99 years (M = 62.07 years; SD = 11.89) and was 56.55% female. Psychometric properties of the GHQ-12 were investigated using an IRT model with item discrimination and difficult as parameters. Results. In the entire 12-item scale, item discrimination ranged from 1.21 to 3.05, with the item ‘playing a useful part in things’ showing the lowest level of discrimination and the item ‘feeling unhappy or depressed’ showing the highest. The 12-item scale could be reduced to an 8-item scale without loss of test information and without loss of concurrent validity. Items with poor discrimination (“been able to concentrate”, “feeling useful”, “making decisions”, “enjoy day-to-day activities”) were discarded from the scale. All items in the revised scale showed high discrimination and scalability. The concurrent validity of the reduced scale was as good as the longer 12-item scale. Conclusions. The GHQ-12 was found to have good internal consistency, but also unscalable items and items with low discrimination. Further research is needed to validate these findings in other cohorts on the Dementia Platform UK data portal. IRT is a valuable method for scale optimisation, which serves the clinical and epidemiological need of shorter scales with optimised information.