Further Exploration of the Psychometric Properties of GamTest: A Rasch Analysis

Abstract

GamTest is a self-rating scale of negative consequences of gambling, included in the popular responsible gambling tool Playscan as part of an overall risk assessment and feedback feature. Two previous psychometric evaluations of this instrument yielded contradictory results: in an online high-gambling population, a five-factor model was supported and the instrument had overall good psychometric properties, but in a low-gambling population, the same factor structure was not supported. Because GamTest is used with both low- and high-gambling populations, more psychometric research is needed to fully understand how the instrument works. The current study examined, for the first time, psychometric performance among a sample of low-gambling respondents using a Rasch analysis. Results indicated that the instrument could be improved by decreasing the scale-steps and removing several problematic items demonstrating misfit. Furthermore, the findings indicated that some items functioned differently depending on gender, and that a shortened, improved nine-item version could not differentiate between different levels of risk. Our findings suggest that the instrument would arguably benefit from being adapted for use in a low-gambling population.

Psychometric properties of the Jonsson-Abbott Scale: Rasch and confirmatory factor analyses

“…A survey (available only in Swedish) was created by the authors (available at https://osf.io/s287j/ ) as part of a larger project to collect data for the psychometric evaluation of gambling-related instruments. The survey included several instruments, of which the following were used in the current study: the GamTest (Jonsson et al, 2017b ; Forsström et al, 2020a , 2021 ) and Problem Gambling Severity Index (PGSI) (Ferris and Wynne, 2001 ) from the Swedish Longitudinal Gambling Study (SWELOGS) (Romild et al, 2014 ). In total, 85 questions were used from SWELOGS, all of which were multiple choices, including the JAS (Jonsson et al, 2017a ).…”

“…Also, JAS had a higher correlation with GamTest, which contained more items and can also be seen to address the risk progression. Furthermore, the overconsumption subscale was highly correlated with both GamTest (Forsström et al, 2020a(Forsström et al, , 2021 and PGSI (Ferris and Wynne, 2001). Further, the other factors can be regarded as markers of the onset of atrisk and problem gambling, as the endorsement of the items occurs in three steps.…”

“…Additionally, the person-item separation map indicates that the items are endorsed on an overall factor-by-factor level (see Figure 1). The JAS had a fairly high correlation with both the PGSI (Ferris and Wynne, 2001) and the GamTest (Forsström et al, 2020a(Forsström et al, , 2021, indicating that it partly addresses a problem gambling construct and can measure the incidence of at-risk and problem gambling. Also, JAS had a higher correlation with GamTest, which contained more items and can also be seen to address the risk progression.…”

“…Different factor structures have been found in low gambling populations (Holtgraves, 2009 ; Forsström et al, 2020a ). Items used in low gambling populations also seem to function in terms of gender, with men and women differing in endorsing items (Forsström et al, 2021 ). This implies that the consequences of gambling and the amount of time and money spent might be less relevant when trying to appraise the risk among individuals with a low level of gambling, which may still be problematic and/or entail a future risk.…”