In this study we estimate the prevalence of problem gambling using the SOGS-RA for several reasons. First, it allows comparison with several other states including Washington, Minnesota, and Louisiana. Second, it has been found to be a valid and reliable instrument which is based on extensive testing (see Winters et al., 1993a). Finally, the SOGS-RA has been tested using telephone interviews, which is the methodology employed in the current study.
Both the SOGS-RA and the adult version on which it is based, the SOGS (Lesieur and Blume, 1987) were created using the DSM-IIIR classification for pathological gambling (APA, 1987). In order to develop the adolescent version of the SOGS, a research team at the University of Minnesota revised the original SOGS items, with the help of an adolescent focus group, in order to "accommodate adolescent experiences and reading levels" (Winters et al., 1993a, p. 67). A psychometric evaluation of the instrument reported that the SOGS-RA was both a reliable and valid measure of problem gambling for adolescents.
The SOGS-RA consists of a two-part questionnaire which measures a) the frequency and type of gambling activities engaged in by respondents and b) a checklist of 12 signs and symptoms of pathological gambling as described in the DSM-IIIR. In order to estimate the prevalence of pathological gambling, the number of symptoms that a respondent reports are summed to create an overall score which can range from 0 (no symptoms at all) to 12 (respondent experiences all 12 symptoms).
There is not currently a single agreed-upon method for defining level three gambling, no gold standard so to speak. In order to accommodate reasonable variation in definitions of problem gambling and comparisons to other studies, we provide two different estimates of problem gambling. Nonetheless, because the broad method combines frequency of gambling with number of symptoms, we feel it is better than the narrow method for planning preventative and treatment interventions. Both of these classification techniques have been previously used by the developers of the SOGS-RA instrument, and both are reasonably valid and reliable (Winters et al., 1993b; Winters, Stinchfield and Kim, 1995).
The first estimate based on "narrow criteria," uses only the score on the SOGS-RA items to estimate problem gambling. Using this method results in a relatively low estimate primarily because it does not include the frequency of gambling as a criteria. In this method, a SOGS-RA score of four or more identifies an adolescent as a problem gambler. While this ensures a conservative estimate of problem gambling, it is possible that it underreports the number of youth that many would consider problem gamblers. For example, a respondent with a SOGS score of three will not be classified as a problem gambler, even if she gambles every day and reports having trouble in school and with her parents (scored two) as a result of gambling using the narrow criteria.
Estimates reported based on "broad criteria" include measures of gambling frequency in the criteria of problem gambling. Thus, a respondent who gambles every day, and has experienced some problems, is defined as a problem gambler. The broad method is perhaps more instructive in identifying problem gambling because it would identify a heavy gambler who is experiencing some difficulty as a problem gambler, even if the number of symptoms experienced is fewer than four (Winters et al., 1995). This report provides both estimates in order to acknowledge the current variability in defining level three gambling in gambling research. Scoring rules for both narrow and broad criteria are included in Appendix 1.