Longitudinal Mixed Modelling of Emergency Department Use Among a Sample of Homeless Participants in a Housing First Demonstration Trial

The present Canadian study sought to identify predictors of emergency department (ED) use in a sample of adults who are homeless. Four hundred eighty-three participants were interviewed quarterly for two years, reporting their housing status, income, food security, physical/mental health, and other social determinants of health. Survey data were linked to administrative health and social services data. Predictors of ED visits were analyzed using negative binomial longitudinal mixed modelling. Emergency department visits were positively and reliably predicted by pre-baseline ED use, as well as by concurrent increased social assistance, primary care visits, ratings of physical health, substance use problems, and case management visits. Increased concurrent housing instability and food insecurity, and decreased homelessness history, positively predicted ED visits, but effect reliability was lower. Participant ill health appears to be a primary driver of ED utilization. The findings challenge misconceptions about the appropriateness of ED use among people who are homeless.