Exploring EMR data reveals effective treatment patterns
Exploring electronic medical records (EMRs) can help healthcare providers learn more about typical treatment patterns for specific situations, according to new findings published in Artificial Intelligence in Medicine.
“Recently, to reduce hospitalization costs and improve care services, studies on clinical data mining using EMRs with advanced analytics have developed rapidly; it is believed that they hold the key to achieving these goals,” wrote first author Jingfeng Chen, Dalian University of Technology in China, and colleagues.
The authors aimed to extract treatment patterns from a real-world EMR dataset using a multi-view similarity network fusion (SNF) method. The dataset included data from more than 8,000 cerebral infarction patients who received care at one of three traditional Chinese medicine (TCM) hospitals in China. The SNF technique combined three similarity measures: a content-view similarity, a sequence-view similarity and a duration-view similarity.
The researchers then pre-processed all applicable EMR data, tagging medical terminology and manually proofreading for any possible mistakes. The EMR data was found to primarily contain demographic information, laboratory indicators, diagnostic details, doctor’s orders or the treatment outcome.
Overall, the multi-view SNF method was more effective than any single-view similarity measures. The team was able to compare different treatment patterns with each other as well as a variety of official guidelines, determining which actions resulted in the most desirable results.
“The extracted typical treatment patterns by combining doctor order content, sequence, and duration views can provide data-driven guidelines for AI in medicine and help clinicians make better decisions in clinical practice,” the authors wrote.
Chen and colleagues also looked ahead, detailing what the next step may look like in terms of academic research.
“In future work, we will study similarity measure methods based on multi-source heterogeneous data (patient demographic information, diagnostic information of Chinese and Western medicine, laboratory indicators and wearable device data) in the era of big data to promote precision treatment recommendations,” they concluded. “Moreover, considering medical practice, designing a clinical decision support system based on the proposed method is also of crucial importance.”
Michael has more than 18 years of experience as a professional writer and editor. He has written at length about cardiology, radiology, artificial intelligence and other key healthcare topics.