The goals of this project are to: 1 ) Calculate the prevalence rates for overall and PCS ED use in a managed-care population in central Massachusetts and identify associations between outcomes and patient and practice-level factors;2) Create predictive models using administrative claims data and calculate ED risk scores and evaluate the performance of models predicting overall and PCS ED use;and 3) Expand these models by adding patient characteristics from EMRs, neighborhood characteristics, and practice characteristics. This project will break new ground in predictive modeling of healthcare utilization. No prior studies have published methods for predicting ED use or creating an ED risk score. Our method will include the critical components of risk adjustment necessary for making predictions about future health services use. We also propose the innovation of combining data from linked claims and EMRs with practice and neighborhood characteristics. Our study will determine the predictors of different measures of ED use and create a predictive ED risk score that could be used to identify patients at risk of future ED use. Developing an accurate predictive model/risk score for ED use will enable higher-risk patients (and/or their providers) to be targeted for educational and care management interventions. Additionally, ED prediction models could help in creating performance measures to enable rewarding providers for providing better care and access to care for their patients.
There are no published ED risk score models, which we will create in this study. Creating ED risk score models is important because if providers and payers could accurately evaluate the risk of ED use in a population, they could target high-risk patients with educational and care management programs and try to prevent future ED visits. An ED risk score model could also be used to set credible and fair risk-adjusted targets for expected use for panels of patients against which actual use can be judged.
| Lines, Lisa M; Rosen, Allison B; Ash, Arlene S (2017) Enhancing Administrative Data to Predict Emergency Department Utilization: The Role of Neighborhood Sociodemographics. J Health Care Poor Underserved 28:1487-1508 |
