RESUMEN
INTRODUCTION: We sought to identify latent profiles of polysubstance use patterns among people who inject drugs in three distinct North American settings, and then determine whether profile membership was associated with providing injection initiation assistance to injection-naïve persons. METHODS: Cross-sectional data from three linked cohorts in Vancouver, Canada; Tijuana, Mexico; and San Diego, USA were used to conduct separate latent profile analyses based on recent (i.e., past 6 months) injection and non-injection drug use frequency. We then assessed the association between polysubstance use patterns and recent injection initiation assistance provision using logistic regression analyses. RESULTS: A 6-class model for Vancouver participants, a 4-class model for Tijuana participants and a 4-class model for San Diego participants were selected based on statistical indices of fit and interpretability. In all settings, at least one profile included high-frequency polysubstance use of crystal methamphetamine and heroin. In Vancouver, several profiles were associated with a greater likelihood of providing recent injection initiation assistance compared to the referent profile (low-frequency use of all drugs) in unadjusted and adjusted analyses, however, the inclusion of latent profile membership in the multivariable model did not significantly improve model fit. DISCUSSION AND CONCLUSIONS: We identified commonalities and differences in polysubstance use patterns among people who inject drugs in three settings disproportionately impacted by injection drug use. Our results also suggest that other factors may be of greater priority when tailoring interventions to reduce the incidence of injection initiation. These findings can aid in efforts to identify and support specific higher-risk subpopulations of people who inject drugs.