Officials at Chapel Hill-based University of North Carolina Health Care decided to embark on a project to see if big data could help refine the health system's risk scoring for 30-day hospital readmissions. Their research paid off.
System Vice President and Chief Analytics Officer at UNC Health Care's enterprise analytics and data sciences division Jason Burke explained to Information Management how UNC Health Care utilized predictive modeling, machine learning and clinical research to reduce readmissions.
The health system adopted Epic to standardize its EHR platform and created a separate enterprise data warehouse environment, Carolinas Data Warehouse, to combine EHR data with laboratory, pharmacy and registration data.
The health system then utilized predictive analytics to see if it could identify patients that were most at risk to return to the hospital within 30 days of discharge. Using EHR information, patients' zip codes and socioeconomic data, researchers developed risk scores to assign to patients.
Here are two key findings from the initiative, according to Mr. Burke.
1. Researchers found advanced analytic methods could improve scoring. "The answer was a resounding yes. We found that those analytic methods provided a dramatic improvement in our ability to predict risk," said Mr. Burke. The models proved to be 30 percent more accurate at identifying patients who were later readmitted to the hospital, according to the article.
2. However, researchers found that utilizing patients' zip codes and socioeconomic data may not always improve the accuracy of the predictive models. "We didn't find that sociodemographic data improved the models, but it did reinforce our understanding of the relationship between the longitudinal clinical data and sociodemographic information," said Mr. Burke.
More articles on quality:
Allegheny Health Network opens one-stop-shop for addicted mothers
31 cases of whooping cough detected in Wisconsin county
Antibiotic resistant bacteria detected in polluted air