Dimensions of the South Oaks Gambling Screen in Finland: A cross‐sectional population study. Issue 3 (1st March 2017)

• Record Type: Journal Article
• Authors: Salonen, Anne H.; Rosenström, Tom; Edgren, Robert; Volberg, Rachel; Alho, Hannu; Castrén, Sari
• Is Part Of: Scandinavian journal of psychology. Volume 58:Issue 3(2017)
• Page Start: 228
• Page End: 237
• Publication Date: 2017-03-01
• Subjects: Diagnostic classification -- gambling instrument -- measurement -- population survey -- problem gambling -- psychometrics; Psychology -- Periodicals; 150
• DOI: 10.1111/sjop.12357
• Languages: English
• ISSNs: 0036-5564

Abstract

Abstract : The underlying structure of problematic gambling behaviors, such as those assessed by the South Oaks Gambling Screen (SOGS), remain unknown: Can problem gambling be assessed unidimensionally or should multiple qualitatively different dimensions be taken into account, and if so, what do these qualitative dimensions indicate? How significant are the deviations from unidimensionality in practice? A cross‐sectional random sample of Finns aged 15–74 (n = 4, 484) was drawn from the Population Information Registry and surveyed in 2011–2012. Analyses were conducted using descriptive statistics, Confirmatory factor analysis (CFA) and multidimensional item response theory (MIRT) models. Altogether, 14.9% of the population endorsed at least one of the 20 SOGS items, but nine items had low endorsement rates (≤ 0.2%). CFA and MIRT techniques suggested that individuals differed from each other in two positively correlated ( r = 0.70) underlying dimensions: "impact on self primarily" and "impact on others also". This two‐factor correlated‐factors model can be reinterpreted as a bifactor model with one general gambling‐problem factor and two specific factors with similar interpretation as in the correlated‐factors model but with non‐overlapping items. The two specific factors may provide clinically useful information without extra costs of assessment.