The classification accuracy of four problem gambling assessment instruments in population research

Abstract

Improved methodology was used to re-examine the weak correspondence between problem and pathological gamblers identified in population surveys and subsequent classification of these individuals in clinical interviews. The SOGS-R, the CPGI, the NODS and the Problem and Pathological Gambling Measure (PPGM), as well as questions about gambling participation and expenditures, were administered to a total of 7272 adults. Two clinicians then assessed each person's status, based on comprehensive written profiles derived from these questionnaire responses. Instrument classification was then compared to clinical classification. All four instruments correctly classified most non-problem gamblers (i.e. had good to excellent sensitivity, specificity and negative predictive power). However, the PPGM was the only instrument with good classification of problem gamblers (i.e. excellent sensitivity and positive predictive power). The CPGI and SOGS-R had weak positive predictive power and the NODS had only adequate sensitivity and positive predictive power. Improvement in the classification accuracy of the CPGI occurred when a 5+ cut-off was used and when a 4+ cut-off was used with the SOGS. In general, the classification accuracy of the NODS, SOGS and CPGI is better than prior research suggested but overall accuracy is still modest. With adjusted cut-offs, all three instruments are reasonably congruent with clinical ratings.

Keywords::

• problem gambling
• measurement
• psychometrics
• classification
• screening instruments

Funding

This project was funded by the Ontario Problem Gambling Research Centre. The authors have no affiliations with the gambling industry and the vast majority of their research funding has come from research organizations or government agencies. However, they have conducted several small research projects for some entities in the gambling industry.

Notes

^1. Throughout this document the term ‘problem gambling’ is inclusive of the term ‘pathological gambling’. Problem gambling is the preferred term because it has more neutral etiological connotations and because it is inclusive of less severe forms of the disorder that are nonetheless associated with significant negative consequences.

^2. Although comparison with clinical assessment is typically used to judge an instrument's accuracy, it must be recognized that clinical assessment can also be flawed and there is no unambiguously true ‘gold standard’.

^3. False positives are not considered problematic when these instruments are used for screening purposes in clinical settings.

^4. The lower cut-off for the SOGS is based on the approach adopted by the developers of the SOGS-R (Abbott & Volberg, 1996) and the cut-off for the NODS is based on the original report of the instrument's development (Gerstein et al., 1999). While a score of 3 to 7 has conventionally been used to designate ‘moderate risk gambling’ on the CPGI, a score of 3 or higher has been used by some authors to denote ‘moderate problem gambling’ in recognition that the 8 or higher cut-off may be too stringent (something found in the present study).

^5. The lower cut-off for the SOGS is based on the approach adopted by the developers of the SOGS-R (Abbott & Volberg, 1996) and the cut-off for the NODS is based on the original report of the instrument's development (Gerstein et al., 1999).

^6. Research by the present authors has found that requiring a minimal amount of gambling involvement (e.g. $49 or more in a typical month) significantly reduces the frequency of false positives (i.e. people who receive an instrument designation of problem gambling in the absence of any reported gambling involvement).

^7. Many researchers adopted a cut-off of 3 or more on the CPGI in preference to the cut-off of 8 or more recommended by the instrument's developers because the higher cut-off was too stringent and yielded too few problem gamblers for analysis.

Additional information

Notes on contributors

Robert J. Williams

Robert J. Williams is a professor in the Faculty of Health Sciences at the University of Lethbridge and a Research Coordinator for the Alberta Gambling Research Institute. For the past 10 years, his work has focused on gambling, an area in which he is an internationally recognized expert. Dr Williams is a leading authority in the areas of Internet gambling; best practices in the population assessment of problem gambling; prevention of problem gambling; the socio-economic impacts of gambling, the proportion of gambling revenue deriving from problem gamblers; the prevalence and nature of gambling in Aboriginal communities; and the etiology of problem gambling.

Rachel A. Volberg

Rachel A. Volberg has been involved in research on gambling and problem gambling since 1985. She has directed or consulted on numerous gambling studies around the world, including national prevalence surveys in the United States, Australia, New Zealand, Great Britain, Norway and Sweden. Dr Volberg is an authority in the areas of best practices in the population assessment of problem gambling; the socio-economic impacts of gambling; the proportion of gambling revenue derived from problem gamblers; and the epidemiology and etiology of gambling and problem gambling.