Using machine learning to retrospectively predict self-reported gambling problems in Quebec.

BACKGROUND AND AIMS: Participating in online gambling is associated with an increased risk for experiencing gambling-related harms, driving calls for more effective, personalized harm prevention initiatives. Such initiatives depend on the development of models capable of detecting at-risk online gamblers. We aimed to determine whether machine learning algorithms can use site data to detect retrospectively at-risk online gamblers indicated by the Problem Gambling Severity Index (PGSI). DESIGN: Exploratory comparison of six prominent supervised machine learning methods (decision trees, random forests, K-nearest neighbours, logistic regressions, artificial neural networks and support vector machines) to predict problem gambling risk levels reported on the PGSI. SETTING: Lotoquebec.com (formerly espacejeux.com), an online gambling platform operated by Loto-Québec (a provincial Crown Corporation) in Quebec, Canada. PARTICIPANTS: N = 9145 adults (18+) who completed the survey measure and placed at least one bet using real money on the site. MEASUREMENTS: Participants completed the PGSI, a self-report questionnaire with validated cut-offs denoting a moderate-to-high-risk (PGSI 5+) or high-risk (PGSI 8+) for experiencing past-year gambling-related problems. Participants agreed to release additional data about the preceding 12 months from their user accounts. Predictor variables (144) were derived from users' transactions, apparent betting behaviours, listed demographics and use of responsible gambling tools on the platform. FINDINGS: Our best classification models (random forests) for the PGSI 5+ and 8+ outcome variables accounted for 84.33% (95% CI = 82.24-86.41) and 82.52% (95% CI = 79.96-85.08) of the total area under their receiver operating characteristic curves, respectively. The most important factors in these models included the frequency and variability of participants' betting behaviour and repeat engagement on the site. CONCLUSIONS: Machine learning algorithms appear to be able to classify at-risk online gamblers using data generated from their use of online gambling platforms. They may enable personalized harm prevention initiatives, but are constrained by trade-offs between their sensitivity and precision.

Tridimensionality of alcohol use in Canada: Patterns of drinking, contexts and motivations to drink in the definition of Canadian drinking profiles according to gender.

OBJECTIVES: The aim of this paper is to examine whether there is an underlying multidimensional typology of drinking according to gender among a population presenting heterogeneous drinking profiles in Canada. METHODS: Latent class analysis was chosen to analyze the degree of statistical relationship among three indicators of drinking practices: patterns of drinking - i.e., frequency and quantity; contexts; and motivations to drink. Multivariate multilogistic regressions were conducted to explore the composition of each typology by age and education. Participants were selected from the Canadian GENACIS survey (Gender, Alcohol, and Culture: An International Study) and comprised 871 men and 843 women (N = 1,714) aged between 18 and 77 years and being regular alcohol drinkers (consumption at least once a month). Respondents to the GENACIS questionnaire completed questions on use, contexts and reasons to drink as well as socio-economic questions (age and education), adjusted by Canadian province of residence. RESULTS: Six profiles were distinguished among men and five among women. Men and women share four drinking patterns but present distinctive characteristics of drinking. We also observed variability in the relationship according to socio-economic status and gender. CONCLUSION: Our results confirmed the complexity and variability of drinking practices according to gender in Canada and the necessity to focus on gender and social dimensions in order to enhance our understanding of alcohol use. This study also reinforces the idea of adapting promotion strategies and interventions in public health by gender and social status in order to make them more efficient.

Parenthood, alcohol intake, and drinking contexts: occasio furem facit.

OBJECTIVE: The aim of this study was to assess whether the effect of parenthood on alcohol intake varies according to the context in which the drinking act occurs. METHOD: The data were drawn from the Canadian Addiction Survey, a national telephone survey conducted in 2004. The analytical sample included 1,079 drinking occasions nested in 498 female drinkers and 926 drinking occasions nested in 403 male drinkers between 18 and 55 years of age. A multilevel linear statistical model was used to estimate the variance related to the drinking occasion (Level 1) and to the parental role (Level 2). RESULTS: Parenthood was not associated with alcohol intake per occasion. Drinking context variables brought great explanatory power to the study of alcohol intake, but, overall, the effect of parenthood on alcohol intake did not vary according to the context in which drinking occurs. Only one interaction between the parental role and contextual characteristics was found. CONCLUSIONS: Men's and women's alcohol intake within drinking contexts is more likely to be influenced by the immediate context in which drinking occurs than by their parental role. The explanation for alcohol behaviors within the general Canadian population may lie as much in the situation as in the person.

Alcohol consumption: a different kind of Canadian mosaic.

OBJECTIVES: To determine the way Canadians consume alcohol beyond drinking amounts and to verify if the drinking act is similar across provinces given that, in Canada, alcohol policies come under provincial jurisdiction. METHODS: Subjects were 10,466 current drinkers (5,743 women and 4,723 men) aged 18 to 76 years, who participated in the GENACIS Canada study. RESULTS: In Canada, there are three main patterns of consuming alcohol. Maritimers tend to drink more per occasion, report more binge drinking and largely prefer beer. In the Prairies, people tend to drink less, to drink less often during a meal and to favour spirits. Finally, drinkers from Quebec, Ontario and British Columbia drink more often, drink wine more often, drink spirits less often, and drink more often during a meal than drinkers from the other provinces. The same patterns are observed in both sexes, although the differences across provinces are less pronounced among women. CONCLUSION: Knowledge about these three drinking groups should be used to increase the legitimacy and effectiveness of alcohol policies in general.

The importance of drinking frequency in evaluating individuals' drinking patterns: implications for the development of national drinking guidelines.

AIMS: This paper examines the relationship between frequency of drinking, usual daily consumption and frequency of binge drinking, taking into consideration possible age and gender differences. PARTICIPANTS AND DESIGN: Subjects were 10 466 current drinkers (5743 women and 4723 men) aged between 18 and 76 years, who participated in the GENACIS Canada (GENder Alcohol and Culture: an International Study) study. SETTING: Canada. MEASUREMENTS: The independent variable was the annual drinking frequency. The dependent variables were the usual daily quantity consumed, annual, monthly and weekly frequency of binge drinking (five drinks or more on one occasion). FINDINGS: Logistic regressions show (i) that those who drink less than once a week are less likely than weekly drinkers to take more than two drinks when they do drink; (ii) that the usual daily quantity consumed by weekly drinkers is not related to their frequency of drinking; but that (iii) the risk and frequency of binge drinking increase with the frequency of drinking. CONCLUSIONS: Given that risk and frequency of binge drinking among Canadians increases with their frequency of drinking, any public recommendation to drink moderately should be made with great caution.