Mercy’s Experience with NLP of EHR Data for Real World Insights - Kerry Bommarito

Kerry Bommarito and Nick White from the Data Science team at Mercy described how they use NLP of EHR data to gain real world insights. Mercy has 40 hospital in the Midwest US with 900 physician practices and 45,000 coworkers serving millions of patients annually. The Data Science team specializes in predictive analytics, machine learning, social determinants of health, geo-mapping and NLP: and recently in facilitating Mercy's COVID-19 analytics and operational response.

Mercy was an early adopter of EHRs and added medical device scanning and longitudinal device surveillance data for analytics, and then embraced NLP. They designed their Epic EHR to be unified, with the same configuration across hospitals, clinics, providers, and patients, to provide a single patient record across the care continuum. Other challenges beyond unification were accessing device-specific data (where experience with UDIs helped), and making use of the unstructured data buried in clinical and patient notes. 

Kerry commented that NLP is helping Mercy enhance the completeness and accuracy of its medical record data, and reducing provider burden and documentation time. NLP lets Mercy get to pertinent clinical, social behavioral data that was previously inaccessible. Other applications of the NLP-extracted data include developing predictive risk models, and helping with regulatory submissions.