The 20-item Gambling Harms Scale (GHS-20): Benchmarked to health utility using propensity weighting and control for comorbidities

This is the authors' abstract. We don't add key points for this paper.

Measuring gambling-related harm is crucial for public health initiatives, but existing screens like the Problem Gambling Severity Index (PGSI) have limitations. The 10-item Short Gambling Harms Screen (SGHS/GHS-10) addresses some gaps, but lacks comprehensive domain coverage and robust validation against health utility benchmarks. This study aimed to develop and validate an extended 20-item Gambling Harms Scale (GHS-20). Objectives included achieving better representation across harm domains, selecting items based on unique associations with health utility decrements, and benchmarking the scale's scores against health utility. Data were collected from 2,603 Australian adults who gambled in the past year. Participants completed 31 candidate harm items, the PGSI, the WHOQOL-BREF (measuring quality of life), and the SF-12 (for SF-6D health utility derivation). Item selection utilised latent trait modelling and lasso regression. The final GHS-20 scale was benchmarked against SF-6D health utility using propensity score weighting to balance demographic factors and generalised additive modelling (GAM) to control for comorbidities. Lasso regression identified items providing unique explanatory information, resulting in the 20-item GHS-20, which includes items from all six gambling harm domains. The GHS-20 demonstrated excellent reliability (alpha=.98, omega=.90), correlated positively with the PGSI (r=.78) and negatively with health utility (r=-.38). GAM analysis revealed a significant, near-linear negative relationship between GHS-20 scores and health utility, with statistically significant decrements observed even at low scores. The GHS-20 is psychometrically robust, and can comprehensively assess gambling harm and calculate population harm burden.