Min Chen, Florida International University – Algorithms Detecting Strokes

On Florida International University Week:  Can an algorithm save your life?

Min Chen, The Endowed SunTrust Bank Professorship Holder & Associate Professor, Department of Information Systems and Business Analytics, explores how this might be the case.

Dr. Min Chen’s research examines information technology innovations, healthcare analytics, and issues relevant to the economics, organization, and regulation of the U.S. health care system. Her research has been published widely in high impact peer-reviewed journals across disciplines, recognized with competitive awards, and featured in media outlets. Prior to joining FIU, Dr. Chen worked as an economic consultant at Charles River Associates and advised on antitrust issues in a range of industries. Dr. Chen holds a B.A. in Economics from the Renmin University of China, as well as a Master’s degree in Public Policy from the University of Chicago and a Ph.D. in Managerial Economics & Strategy from the Kellogg School of Management at Northwestern University.

Algorithms Detecting Strokes

Stroke is among the most common, dangerous – and misdiagnosed – medical conditions.

Timely detection is key. Patients treated within the first hour of symptom onset have a much higher survival rate and lower risk of long-term brain damage. This is known as the “golden hour.” 

However, patients often miss this crucial early diagnosis window, and this is particularly true for certain groups such as Blacks, Hispanics, women and older adults on Medicare,.

Our research demonstrated that a machine learning algorithm, employing data from hospital records and social determinants of health, can help diagnose a stroke quickly, with 83 percent accuracy.

Notably, this can happen even before lab tests or diagnostic images are available. The strength of this algorithm also lies in its ability to continually improve its performance as it analyzes more data. 

We used data from suspected stroke patients to train the model, like their age, race, and how many chronic conditions they might have. We got our hands on this data from a whole bunch of hospitals in Florida—over 140,000 unique patient visits, in fact.

But we didn’t stop there. We combined this with social determinants of health data from the American Community Survey. Now, if you’re wondering what that means, these are non-medical factors that play a huge role in health outcomes. Think along the lines of income and whether a person has stable housing.

Read More:
[Journal of Medical Internet Research] – A Machine Learning Approach to Support Urgent Stroke Triage Using Administrative Data and Social Determinants of Health at Hospital Presentation: Retrospective Study