An Examination Of The Psychometric Properties Of The CPGI In Applied Research (OPGRC# 2328) Final Report 2007 Total funds awarded: $34,980.00 Dates of period of support: November 2005 to August 2006 Sponsoring Organization: Centre for Addiction and Mental Health Principal Investigator: Emily Cripps, Ph.D., C.Psych. Psychologist, Anger and Addiction Clinic & Law and Mental Health Program Centre for Addiction and Mental Health Phone: 416-535-8501 x2970 Email: emily_cripps@camh.net Co-investigators: Lorne Korman, Ph.D., C.Psych. Clinic Head, Anger and Addiction Clinic Centre for Addiction and Mental Health Phone: 416-535-8501 x6568 Email: lorne_korman@camh.net Jane Collins, B.Sc. Clinical Project Leader, Concurrent Disorders Program, Centre for Addiction and Mental Health Phone: 416-535-8501 x4451 Email: jane_collins@camh.net Contract Period: November 2005 to August 2006 Page 1 of 10
Table of Contents Abstract 3 Introduction 4 Method Participants 5 Measures 6 Procedures 6 Data Analysis 6 Results Comparing Study Samples 6 Psychometric Properties of the CPGI 6 Confirmatory Factor Analysis of the CPGI 7 Study Limitations 8 Discussion 8 References 9 List of Tables Table 1. Means and standard deviations of CPGI items and total score 7 List of Figures Figure 1. Confirmatory Factor Analysis of the CPGI 7 Page 2 of 10
Abstract The Canadian Problem Gambling Index (CPGI; Ferris & Wynne, 2001) was developed in part to address concerns about the South Oaks Gambling Screen SOGS). Although the SOGS has been the most commonly used measure of problem gambling in research studies, several problems with its use have been identified. Most notably, evidence suggests that the SOGS over-estimates the rate of problem gambling in general population surveys (Stinchfield, 2002). Although several studies have since employed the CPGI in research, none thus far have attempted to replicate Ferris and Wynne s findings regarding the psychometric properties of this new measure. Further, the psychometric properties of the CPGI have not yet been investigated in a clinical sample. Accordingly, the goal of the current project was to evaluate the psychometric properties of the CPGI in a clinical sample of individuals with clinically-significant problems with anger and/or gambling problems. Overall, the CPGI demonstrated strong psychometric properties within this applied clinical sample. These results replicate and extend the work of Ferris and Wayne, and indicate that the CPGI is appropriate for use in both clinical and non-clinical populations. Keywords: Pathological gambling, gambling behaviours, assessment, psychometrics. Page 3 of 10
Introduction The South Oaks Gambling Screen (SOGS) (Lesieur & Blume, 1997) is currently the most commonly used measure of problem gambling. The SOGS is a 20-item questionnaire that was designed to identify gambling problems among hospitalized patients. Although designed as a measure for clinical settings, the SOGS is now used frequently in research and is arguably the most popular assessment instrument in problem gambling research. Despite the popularity of the SOGS, many problems with its use have been identified. Stinchfield (2002) identified some theoretical concerns regarding the SOGS. He noted that almost half of the items relate to borrowing money to finance gambling and argued that this places too much weight on a single aspect of problem gambling. Stinchfield also stated that the lifetime focus of the SOGS items is problematic because it can result in individuals who have histories of problem gambling but no current difficulties being identified as problem gamblers. He noted that some have attempted to resolve this issue by modifying the instructions to reflect a different time period but expresses caution regarding this solution because the effects of such modifications on the psychometric properties of the SOGS have not yet been determined. Ladouceur at al. (2000) obtained evidence suggesting that another possible problem with the SOGS is that respondents tended to misunderstand the meaning of several SOGS items. Stinchfield (2002) has also conducted research regarding problems with the use of the SOGS. He examined the reliability, validity, and classification accuracy of the SOGS in the general population among problem gamblers. He found acceptable reliability estimates in the gambling sample (Chronbach s alpha =.86) but observed lower reliability in his general population sample (Chronbach s alpha =.69). Stinchfield found evidence of acceptable validity in the form of higher scores in the clinical sample and high correlations between SOGS scores and DSM-IV diagnostic criteria. However, while the SOGS demonstrated good to excellent classification accuracy in the clinical sample, the measure yielded a 50% false positive rate in the general population sample. This suggests that the SOGS cannot discriminate effectively between problem and nonproblem gamblers when used with general population samples. This shortcoming may be related to the fact that the SOGS was developed for use in clinical settings. Several other authors have addressed the issue of over-classification and Shaffer, Hall and Vander Bilt examined these five studies in a 1997 meta-analysis. Shaffer et al. concluded these studies demonstrate a clear over-classification bias, with the number of pathological gamblers identified by the SOGS is about twice as high as is obtained using the DSM-IV diagnostic criteria. Addressing some of the shortcomings of existing measures of problem gambling, Ferris and Wayne, working with the Canadian Centre on Substance Use, developed and validated the CPGI. This measure was designed specifically in response to dissatisfaction with the existing measures of problem gambling for use with general Page 4 of 10
populations. The CPGI is a 31-item measure which, using nine of the items, produces scores in several categories: non-gambling, non-problem gambling, low risk gambling, moderate risk gambling, and problem gambling. The remaining CPGI items provide information about components and correlates of problem gambling, such as type of gambling, loss of control, problem recognition, and adverse consequences. The authors designed the CPGI to be consistent with the DSM-IV criteria for Pathological Gambling and have demonstrated acceptable reliability (i.e., internal consistency and test-retest reliability estimates greater than.78), validity (i.e., high correlation between CPGI classification and DSM-IV diagnosis), and classification accuracy (i.e., correctly identifying individuals as having a problem with gambling). At the present time, the CPGI has been used in prevalence studies in several Canadian provinces (Smith & Wynne, 2002; Wiebe, Single & Falkowski-Ham, 2001; Wynne, 2002) and is currently being used in other prevalence studies (Korman et al a and Korman et al b). These studies have focused on the results of the CPGI rather than examining the use of the measure itself, and as such did not examine or report on the psychometric properties or experience of using the CPGI. While the CPGI s psychometric properties were examined by Ferris and Wynne (2001), these findings have not yet been replicated. Accordingly, one goal of the current project is to evaluate the psychometric properties of the CPGI in another sample of individuals. The proposed research will also extend the examination of the CPGI s properties from the general population in Ferris and Wynne s research to a clinical sample of problem gamblers. Further, given that the CPGI was developed for use in the general population, it is important to describe the experience of using it with problem gambling clients in a clinical setting, as well as its application in applied clinical research. The goal of the current project is to evaluate the psychometric properties of the CPGI in a clinical sample in order to extend the existing findings from a general population sample. It is hypothesized that the psychometric properties of the CPGI in an applied clinical setting will be comparable to those observed by Ferris and Wayne (2001). Methods Participants: The participants of the current investigation were from one of two samples. The participants of a randomized clinical trial were recruited from the Assessment Service of the Addiction Program at CAMH, the Problem Gambling Service at CAMH, the Metro Area Addiction Referral Service, community gambling treatment agencies, community social service agencies, and from ads placed in local newspapers. 84 individuals completed the assessment portion of the study, which included completing the CPGI. The participants of an investigation of the prevalence of domestic violence in problem gamblers were recruited from the Problem Gambling Service of CAMH, other community Page 5 of 10
treatment agencies, and newspaper advertisements. 258 participants completed the CPGI during this study. Measures: The Canadian Problem Gambling Index (CPGI; Ferris & Wynne, 2001) was administered to all participants. It is a valid, reliable and widely used measure of gambling activity in many countries (Ferris & Wynne, 2001). Procedure: Participants were recruited via newspaper advertisements and study flyers posted in community gambling treatment agencies. Study advertisements were placed in the Metro free daily paper. The Metro is the highest circulation free daily paper in the GTA and is distributed in the TTC and corner news boxes. Treatment-seeking problem gamblers were recruited from problem gambling treatment services in the GTA. The CPGI, along with other clinical measures, was self-administered in the randomized clinical trial and the prevalence study. Data Analysis: We computed the means and standard deviations for the CPGI items and total. We further computer the average inter-item correlation (AIC) and Cronbach s f o r t h e CPGI. We also conducted a confirmatory factor analysis (CFA) of the CPGI with AMOS 5.0 (Arbuckle, 2005), using maximum likelihood method of estimation. Goodness of fit was assessed using the Confirmatory Fit Index (CFI), with values >.90 indicating acceptable fit; and Root Mean Square Error of Approximation (RMSEA), with values >.1 indicative of poor fit, <.08 acceptable fit, and <.05 close fit (Bentler, in press; Hu & Bentler, 1999; Ullman, 1996). Results Comparing the Study Samples In order to ensure the legitimacy of combining the two study samples, we compared the means of the CPGI items and total scale, and the covariance matrices of the CPGI items. There were no differences across the two study samples upon the CPGI items (all ts < 1.9, all ps >.06), or the CPGI total score (t =.43, p >.05). In a multigroup CFA, there was no significant decrease in fit when the covariances among the CPGI items were restricted to be equivalent across the two study samples ( 2 = 45.78, df = 36, p >.05). Thus, results indicated that item means and covariances were not significantly different across samples. All subsequent analyses were based upon the full sample. The Psychometric Properties of the CPGI The means and standard deviations for the CPGI items and total score are displayed in Table 1. Page 6 of 10
Table 1. Means and standard deviations of CPGI items and total score. Mean Standard Deviation CPGI Item 5 2.03.90 CPGI Item 6 1.48 1.02 CPGI Item 7 2.08.89 CPGI Item 8 1.31.89 CPGI Item 9 2.05.88 CPGI Item 10 1.38.86 CPGI Item 11 1.47 1.02 CPGI Item 12 1.85.92 CPGI Item 13 1.73.96 CPGI Total 15.37 5.04 The average inter-item correlation (AIC) was.29 and Cronbach s Alpha was.78. The sample yielded 16 moderate risk for problem gambling (CPGI scores in the range of 3 through 7) and 315 problem gambling. Confirmatory Factor Analysis of the CPGI The one-factor model of the CGPI provided a good fit to the data ( 2 = 55.43, df = 27, p <.01; CFI =.95; RMSEA =.06 (90% CI.04 -.08). All model parameters were significant (see Figure 1). Figure 1. Confirmatory Factor Analysis of the CPGI CPGI.65.42.54.63.56.41.45.67.48 CPGI_5 CPGI_6 CPGI_7 CPGI_8 CPGI_9 CPGI_10 CPGI_11 CPGI_12 CPGI_13 e1 e2 e3 e4 e5 e6 e7 e8 e9 Page 7 of 10
Study Limitations The CPGI data used in this project was collected in an applied clinical sample. As a result, the psychometric analyses of the CPGI will be specific to this population. Discussion In order to address some of the shortcomings of existing measures of problem gambling, Ferris and Wayne (2001) developed and validated the Canadian Problem Gambling Index (CPGI). The CPGI is a 31-item measure which, using nine of the items, produces scores in several categories: non-gambling, non-problem gambling, low risk gambling, moderate risk gambling, and problem gambling. The remaining CPGI items provide information about components and correlates of problem gambling, such as type of gambling, loss of control, problem recognition, and adverse consequences. The authors designed the CPGI to be consistent with the DSM-IV criteria for Pathological Gambling and have demonstrated acceptable reliability (i.e., internal consistency and test-retest reliability estimates greater than.78), validity (i.e., high correlation between CPGI classification and DSM-IV diagnosis), and classification accuracy (i.e., correctly identifying individuals as having a problem with gambling). At the present time, the CPGI has been used in several prevalence studies; however, these investigations have focused on the results of the CPGI rather than on examining the psychometric properties of the measure. While the CPGI s psychometric properties were examined by Ferris and Wynne (2001), these findings have not yet been replicated. Given that the original validation study of the CPGI used a non-clinical population, a goal of this project was to evaluate the psychometric properties of the CPGI in a clinical sample. The participants in the current research had clinicallysignificant problems with gambling and/or anger. The results indicated acceptable reliability, which is consistent with Ferris and Wayne s data. Further, consistent with the exploratory factor analysis of Ferris and Wayne, the current results indicated that a onefactor solution was a good fit to the data. Overall, the CPGI demonstrated strong psychometric properties within an applied clinical sample. These results replicate and extend the work of Ferris and Wayne, and indicate that the CPGI is appropriate for use in both clinical and non-clinical populations. Page 8 of 10
References Ferris, J. & Wynne, H. (2001). The Canadian Problem Gambling Index: Final Report. Submitted to the Canadian Centre on Substance Abuse. Ottawa, Ontario: CCSA. Ladouceur, R., Bouchard, C., Rheaume, N, Jacques, C., Ferland, F, Leblond, J & Walker, M. Is the SOGS an accurate measure of pathological gambling among children, adolescents and adults? Journal of Gambling Studies, 16 (1), 1 24. Lesieur, H. & Blume, S., 1987; The South Oaks Gambling Screen (SOGS): a new instrument for the identification of pathological gamblers. American Journal of Psychiatry, 144, 1184 1188. Shaffer, H.J., Hall, M.N. & Vander Bilt, J. (1997). Estimating the prevalence of disordered gambling behaviour in the United States and Canada: A metaanalysis. Boston: Harvard Medical School. Smith, G.J. & Wynne, H.J. (2002). Measuring gambling and problem gambling in Alberta: Using the Canadian Problem Gambling Index (CPGI): Final Report. Prepared for the Alberta Gaming Research Institute. Stinchfield, R. (2002). Reliability, validity, and classification accuracy of the South Oaks Gambling Screen (SOGS). Addictive Behaviors, 27(1), 1 19. Toneatto, T. & Millar, G. (2004). Assessing and Treating Problem Gambling: Empirical Status and Promising Trends. Canadian Journal of Psychiatry, 49(8), 517 525. Wiebe, J., Single, E. & Falkowski-Ham, A. (2001). Measuring gambling and problem gambling in Ontario. Canadian Centre on Substance Abuse, Responsible Gambling Council. Page 9 of 10
Wynne, H.J. (2002). Gambling and problem gambling in Saskatchewan: Final Report. Prepared for the Canadian Centre on Substance Abuse. Page 10 of 10