Exploring Racial and Gender Differences in Opioid Use Disorder and Designing NLP-based Diagnosis Tools (DSI-SRP)
This DSI-SRP fellowship funded Chen Jin to work in the laboratory of Dr. Alvin D. Jeffery, Ph.D. in the School of Nursing during the summer of 2023. Chen is a rising senior with majors in Economics and Medicine Health and Society (MHS), and a minor in Data Science.
Opioid Use Disorder (OUD) has become an epidemic in the United States, significantly damaging the quality of life. Early diagnosis and intervention of persistent opioid use are essential to addressing this pervasive issue. Their research aims to develop a diagnostic decision support tool that helps clinicians identify potential OUD cases.
Using VUMC’s clinical notes on patients with chronic pain — those more susceptible to OUD — Chen and their mentor have obtained each patient’s “key terms”(lemma) and “standardized concepts”(CUIs) through different text-based data extraction methods based on natural language processing (NLP). Because some evidence suggests clinicians use different words and phrases for different races and genders of patients, our research delves into these nuances by comparing the top-ranked “key terms” and “standardized concepts” between all races and genders, and our initial findings have highlighted some significant disparities.
In collaboration with experts in the field, they constructed scoring systems based on these “key terms” and “standardized concepts.” Their goal was to determine the most effective algorithm for OUD detection. Moreover, they aimed to compare the efficiency of models that incorporate race and gender data with those that don’t. As they moved forward, they intended to experiment with and fine-tune relevant machine learning models, pitting each patient against their manually crafted scoring systems to discern the most suitable models for diagnostic purposes. This research will ultimately pave the way for a more tailored and efficient approach to OUD detection through NLP, making healthcare more personalized and effective for all.
In addition to receiving support through a DSI-SRP fellowship, this project was supported and facilitated by the DSI Data Science Team through their regular summer workshops and demo sessions.