Small Area Estimation using Multilevel Regression and Poststratification to Estimate Cannabis Use in the State of Montana

Abstract

Background: Small area substance use prevalence estimates at the county, city, or congressional district level are generally unavailable. In this study, we design a cannabis use survey for the state of Montana and use multilevel regression and poststratification (MRP) to generate county-level population prevalence estimates for past year cannabis use.
Methods: We developed a survey that asks questions about cannabis perceptions and use patterns. We analyzed the survey data specifically for the outcome variable of past year cannabis use using MRP to generate population level prevalence estimates at the county level for the state of Montana.
Results: We received 1,958 responses from our survey. We generated county level estimates by age group for cannabis use over the past year and found that MRP estimates were consistent with prior estimations of cannabis use at the state level and provided the ability to use additional data and validated assumptions to refine and downscale estimations of cannabis use, particularly in counties with low response rates.
Conclusion: Multi-modal survey dissemination was cost effective, but future surveys should intend to recruit a larger and more representative sample to minimize selection bias and improve estimation for demographic sub-groups. Overall, MRP provided a promising methodology for generating small-area cannabis use prevalence estimates, adjusting as much as possible for non-representativeness and non-response.