Using Artificial Intelligence to Predict Patient Electronic Health Record Access Points

Electronic Health Records (EHRs) offer benefits to patients and healthcare providers, however, patient records are often split between multiple providers. To understand the distribution of EHRs per patient in multiple counties throughout southeastern North Carolina, data from the Carolina Coastal Health Alliance and the Coastal Connect Health Information Exchange (CCHIE) was provided to researchers. Utilizing Python programming methods, data analysis was performed to find the total unique EHRs available to patients in the study, as well as averages for EHR usage within different age ranges. Machine learning models were developed to predict the multiple aspects of the data set, including the most prevalent EHR provider, total EHRs per patient, and all potential EHRs per patient based on demographic and geographic information. Analysis of results determined that certain models were more successful in the prediction based on statistical measures of accuracy, precision, and recall. Results were cataloged in tables and an interactive mapping tool to highlight discrepancies in the overall data set that may influence results and EHR access for populations in North Carolina.