PGSI and DSM-IV in the 2007 British Gambling Prevalence Survey: reliability, item response, factor structure and inter-scale agreement

Abstract

Data from the 2007 British Gambling Prevalence Survey were used to examine the psychometric properties of the two problem gambling scales used in the survey: the Canadian Problem Gambling Severity Index (PGSI) and a DSM-IV-based scale. Analysis was based on those who reported any gambling in the past 12 months (between 5483 and 5528 participants for most analyses). The PGSI gave evidence of high internal reliability, uni-dimensionality, and good item-response characteristics. Several PGSI items showed extreme male to female endorsement ratios and a possible conclusion is that the PGSI is under-estimating the prevalence of problem gambling among women. The DSM-IV-based scale showed only satisfactory internal reliability, evidence suggesting bi-dimensionality, and poor performance of at least two items: those relating to gambling-related crime and ‘chasing losses’. The results also suggest that, for population survey purposes, problem gambling/non-problem gambling might best be viewed as lying on a continuum.

Keywords:

• screening instruments
• psychometrics
• prevalence
• problem gambling
• survey

Acknowledgements

We wish to acknowledge the support of the Gambling Commission who commissioned the British Gambling Prevalence Survey of which the work reported in this paper was part. Our thanks are also due to Pat Evans of the Alcohol, Drugs, Gambling and Addiction Research Group at the University of Birmingham who prepared draft and final versions of this paper.

Notes

1. Based on deprivation in the seven domains of income, employment, health deprivation and disability, education, skills and training, barriers to housing and services, crime, and living environment, each respondent in England was assigned to one of five approximately equal sized categories according to postcode area. In the analysis reported here the highest two quintiles (most deprived) were combined and compared with the lowest two categories combined (low deprivation).