The association between gambling frequency and risk of harm: Analysis using health survey data from England and...

Health economic models can be used to assess the effectiveness and cost-effectiveness of public health policies for gambling. To develop such a model, we must understand how gambling behaviour is associated with risk of experiencing gambling-related harms. This study aimed to: (1) assess the strength of association between gambling frequency and the risk of gambling-related harm and to examine how these associations differ when lottery-only players are excluded; (2) apply the study's findings in a hypothetical policy model aimed at reducing gambling frequency.

Observational study using six waves of cross-sectional data from the Health Survey for England and the Scottish Health Survey.

Survey conducted in England in 2015, 2016 and 2018 and Scotland in 2015, 2016 and 2017.

The sample included 16 648 adults (aged 18 and over) who reported gambling in the past year, generally representative of the populations of England and Scotland.

Gambling frequency was measured using 6 categories which indicated frequency in the past 12 months: (a) 2 or more times a week; (b) once a week; (c) less than once a week, more than once a month; (d) once a month; (e) every 2-3 months; (f) once or twice a year. Risk of gambling-related harm was assessed using Problem Gambling Severity Index (PGSI) score (0-27) and its four categories: no-risk (0), low-risk (1-2), moderate-risk (3-7) and high-risk (≥8). Control variables included age, sex, deprivation, social grade, presence of mental disorder and frequency of drinking alcohol.

Using multinomial logistic regression and zero-inflated negative binomial models we found that gambling at least twice weekly was associated with a statistically significantly higher PGSI score than gambling once or twice a year (incidence rate ratio = 3.528, 95% confidence interval = 2.040-6.103, P value < 0.001). Reducing gambling to guideline levels for people gambling at least twice weekly moved 10% of the sub-sample from higher PGSI categories (low, medium and high risk) to the no-risk category and shifted the distribution of PGSI scores down.

There appears to be a statistically significant association between gambling frequency and risk of gambling-related harm. Data derived from this and similar analyses can be used to model gambling policies which impact gambling frequency.