Predicting Low Resource Intensity Emergency Department Visits in Children

Abstract Objectives Interventions to reduce frequent emergency department (ED) use in children are often limited by the inability to predict future risk. We sought to develop a population-based model for predicting Medicaid-insured children at risk for high frequency (HF) of low-resource-intensity (LRI) ED visits. Methods We conducted a retrospective cohort analysis of Medicaid-insured children (aged 1–18 years) included in the MarketScan Medicaid database with ≥1 ED visit in 2013. LRI visits were defined as ED encounters with no laboratory testing, imaging, procedures, or hospitalization; and HF as ≥3 LRI ED visits within 365 days of the initial encounter. A generalized linear regression model was derived and validated using a split-sample approach. Validity testing was conducted examining model performance using 3 alternative definitions of LRI. Results Among 743,016 children with ≥1 ED visit in 2013, 5% experienced high-frequency LRI ED use, accounting for 21% of all LRI visits. Prior LRI ED use (2 visits: adjusted odds ratio = 3.5; 95% confidence interval, 3.3, 3.7; and ≥3 visits: adjusted odds ratio = 7.7; 95% confidence interval, 7.3, 8.1) and presence of ≥3 chronic conditions (adjusted odds ratio = 1.7; 95% confidence interval, 1.6, 1.8) were strongly associated with future HF-LRI ED use. A model incorporating patient characteristics and prior ED use predicted future HF-LRI ED utilization with an area under the curve of 0.74. Conclusions Demographic characteristics and patterns of prior ED use can predict future risk of HF-LRI ED use in the following year. Interventions for reducing low-value ED use in these high-risk children should be considered.