Gambling and homelessness: A systematic review and meta-analysis of prevalence.

INTRODUCTION

There is growing concern internationally about co-occurring gambling and homelessness. We systematically review prevalence estimates in help-seeking and community samples.

METHODS

Adopting PRISMA guidelines, we searched CINAHL Plus, Cochrane Library, Embase, Ovid MEDLINE, PsychINFO, Proquest Central, PubMed, Scopus, Web of Science, and Google Scholar for relevant peer-reviewed articles in English. Primary outcomes examined in narrative and quantitative syntheses included prevalence of: (i) gambling in persons experiencing homelessness; (ii) harmful gambling in persons experiencing homelessness; and, (iii) homelessness in persons experiencing harmful gambling.

RESULTS

Searches identified 917 records after removing duplicates. After screening, 45 articles providing 54 prevalence estimates across 12 countries were included, with help-seeking (k = 37) and community based sample (k = 8) estimates pooled separately. Gambling prevalence (all timeframes) in help-seeking samples of persons experiencing homelessness is low (28.7%, 95% CI: 17.3-41.7, k = 14) compared to the general population (approximately 60-80%). However, harmful gambling prevalence (including problem, pathological, and disordered gambling) in help-seeking samples of persons experiencing homelessness is high (16.5%, 95% CI: 10.2-24.2, k = 20) compared to the general population (approximately 1-7%). Additionally, homelessness prevalence is high in help-seeking samples of persons experiencing harmful gambling (23.6%, 95% CI: 18.4-29.2, k = 4) compared to the general population (<1%). Meta-analysis found high between-study heterogeneity and risk of bias from small samples sizes.

CONCLUSIONS

There are high rates of harmful gambling in persons experiencing homelessness and, concurrently, high rates of homelessness in persons experiencing harmful gambling. Improvements in sampling and measurement are needed to strengthen robustness and generalizability of prevalence estimates, which can potentially inform the scale and targeting of clinical interventions, support services, and policy responses.