The SOGS-RA vs. the MAGS-7: prevalence estimates and classification congruence

The purpose of this study was to compare the prevalence rate estimates and congruence in classification status derived from two popular measures of adolescent gambling (SOGS-RA and MAGS-7). Adolescents from three states (Alabama, Mississippi, and Oregon) completed an anonymous questionnaire ( n =1846 high school students total). Results indicate that the prevalence of probable adolescent pathological gambling varied both as a function of instrument and cut-off point utilized for classification (range 1.7%-8.2%). Classification groups (non-problem, at-risk, and problem gamblers) generated by both instruments were found to be associated with reports of gambling frequency, amount of money lost in one gambling occasion, and parental gambling problems. However, concern was raised because the MAGS-7 and the SOGS-RA had little congruence in their three-group classification decisions for specific individuals (e.g., only 20.5% agreement for problem gamblers). To improve clinical utility, an empirical case was made for using the SOGS-RA to generate a fourth group of adolescent gamblers, which we labeled "probable pathological gamblers" (SOGS-RA > or = 6). This group was differentiated from the remaining gambling groups on all the validity indices. The implications and limitations of these findings, as well as future directions, are discussed.