Gambling screens and problem gambling estimates: a parallel psychometric assessment of the South Oaks Gambling Screen and the Canadian Problem Gambling Index

In 2005 the Northern Territory of Australia conducted its first population-based gambling and problem-gambling prevalence survey, administering both the South Oaks Gambling Screen (SOGS) and the Canadian Problem Gambling Index (CPGI) to the same sample of respondents. Using a sub-sample of regular gamblers (n=361), the respective problem gambling screens were subject to psychometric testing that included dimensionality, internal consistency, external validity, classification validity and screen order effects. Analyses were conducted for all regular gamblers stratified by gender. The CPGI produced a significantly lower prevalence estimate than the SOGS as well as lower rates of false-positives as measured against external criteria. Consistent with other studies, dimensionality analysis revealed a multi-dimensional factor structure for the SOGS and a single dimension for the CPGI. The CPGI displayed stronger correlations with external criteria and stronger internal consistency than the SOGS. A gender effect was observed, with both screens performing better for females. In addition, screen order significantly affected problem gambling prevalence estimates, although only for males and all persons. As a group, the psychometric analyses revealed that the results produced by the respective gambling screens are heavily context dependent, both in terms of methods of application and the characteristics of target populations. The key message of the paper is that post-hoc psychometric testing of gambling screens is essential in understanding the limitations of problem gambling prevalence estimates and to qualify and guide their interpretation when applied in general population surveys