Since Mercy’s 2008 implementation of the electronic health record, the organization has been collecting vital clinical data. Data that, if used effectively, could improve a patient’s quality of life. However, 80% of all healthcare data sits “trapped” in an unstructured state, such as provider notes. Mercy needed to develop a way to access this data for greater healthcare insight. That’s where the use of an artificial intelligence called natural language processing (NLP) became an important tool. Mercy extracted data from both form fields and provider notes to help measure medical device safety and positively impact patient outcomes.
Mercy Technology Services (MTS), the IT arm of Mercy, worked with Linguamatics, an NLP architecture, to bridge the gap between extracting cardiology data from physician and other clinical notes. Using more than seven years of deidentified EHR data for 100,000+ chronic heart failure patients with implanted cardiac resynchronization therapy (CRT) devices, data was extracted, transformed and then loaded into Linguamatics. Using this data, a team of MTS’ data scientists developed sound analytical processes evaluated on statistical rigor of precision and recall. The result: an analytical process that contains real-world clinical data for each heart failure patient.
Using “real-world evidence,” Mercy developed a reliable and accurate method (with an F-measure of 0.94) to extract relevant data from clinical notes. From 35.6 million data entries, this process identified 3.4 million symptoms with high accuracy affecting how heart failure patients receive care. Heart failure is a chronic and progressive syndrome. Yet, by using advanced data analytics such as NLP to unlock vast amounts of healthcare data, Mercy and MTS are helping its providers make better data-driven decisions that could have big impact for heart patients, while also helping manufacturers gain insight that could improve device safety and effectiveness.