Developing a brief problem gambling screen using clinically validated samples of at-risk, problem and pathological gamblers


Screening for problem gambling takes place in both clinical settings and in population research. Several short assessment tools for problem gambling have been developed over the past decade for use in these settings. However, the performance of all of these brief screens has been assessed in relation to the longer screens from which they are derived. The purpose of the present study is to identify a subset of items taken from all of the most widely-used problem gambling screens that is effective in capturing the large majority of clinically-assessed at-risk, problem and pathological gamblers. A secondary goal is to examine the performance of existing short screens in correctly identifying clinically assessed individuals. The present study uses data collected in two surveys that included all of the most widely used problem gambling screens and classified respondents based on clinical assessments. The sample includes over 7,000 North American gamblers. The 30 unique problem gambling items were sorted into three dimensions (gambling motivations, behaviors and consequences) and the items most closely correlated with the clinically-assessed {At-Risk}, Problem and Pathological Gamblers in the two surveys and capturing 50% or more of the Pathological Gamblers and 5% or less of the Recreational Gamblers were identified. Once the candidate items in the two surveys were identified, the performance of all two-item, three-item, four-item and five-item combinations was examined to assess capture rates in each sample. All combinations that captured 98% of the Pathological Gamblers, 94% of the Problem Gamblers and 80% of the {At-Risk} Gamblers were considered eligible for further investigation. This consisted of determining the classification accuracy of the most promising combinations of items. Each of the promising combinations was then further examined for its performance across gender, age and ethnicity to assess the level of measurement invariance associated with each combination. Based on performance across both surveys and measurement invariance across major demographic groups, a combination of five items including one motivation item, three behavior items, and one consequences item was identified as the best brief screen for clinically-assessed at-risk, problem and pathological gambling. In contrast, the performance of item combinations that best represented other brief problem gambling screens developed in recent years was unsatisfactory. This underscores the importance of assessing the performance of brief screens in relation to clinical assessments rather than in relation to the longer screens from which they are derived.

Problem with this document? Please report it to us.