Simulation and Modeling for Improving Access to Care for Underserved Populations

Abstract

This research, through partnership with seven Community Health Centers (CHCs) in Indiana, constructed effective outpatient appointment scheduling systems by determining care needs of CHC patients, designing an infrastructure for meaningful use of patient health records and clinic operational data, and developing prediction and simulation models for improving access to care for underserved populations. The aims of this study are 1) redesigning appointment scheduling templates based on patient characteristics, diagnoses, and clinic capacities in underserved populations; 2) utilizing predictive modeling to improve understanding the complexity of appointment adherence in underserved populations; and 3) developing simulation models with complex data to guide operational decision-making in community health centers. This research addresses its aims by applying a multi-method approach from different disciplines, such as statistics, industrial engineering, computer science, health informatics, and social sciences. First, a novel method was developed to use Electronic Health Record (EHR) data for better understanding appointment needs of the target populations based on their characteristics and reasons for seeking health, which helped simplify, improve, and redesign current appointment type and duration models. Second, comprehensive and informative predictive models were developed to better understand appointment non-adherence in community health centers. Logistic Regression, Naïve Bayes Classifier, and Artificial Neural Network found factors contributing to patient no-show. Predictors of appointment non-adherence might be used by outpatient clinics to design interventions reducing overall clinic no-show rates. Third, a simulation model was developed to assess and simulate scheduling systems in CHCs, and necessary steps to extract information for simulation modeling of scheduling systems in CHCs are described. Agent-Based Models were built in AnyLogic to test different scenarios of scheduling methods, and to identify how these scenarios could impact clinic access performance. This research potentially improves well-being of and care quality and timeliness for uninsured, underinsured, and underserved patients, and it helps clinics predict appointment no-shows and ensures scheduling systems are capable of properly meeting the populations’ care needs.