Columbia Data Science Society Hackathon

Decoding Devices team members Nikhil Reddy Varimalla and Arusarka Bose (center) with track judges Michael Wong (right) and Dr. Jeannette (left).

The 9th annual Columbia Data Science Society Hackathon was held October 7-8 in New York City on Columbia University’s campus. A $500 prize for the top patient safety idea was sponsored by the Patient Safety Technology Challenge, with funding from the Pittsburgh Regional Health Initiative.

The winning team presented “Decoding Devices: Improving Patient Safety with MAUDE”. MAUDE is the Manufacturer and User Facility Device Experience, a database where the U.S. Food & Drug Administration houses medical device reports.

Decoding team members Nikhil Reddy Varimalla and Arusarka Bose did a deep dive into MAUDE’s data, using it to identify problem areas. Their solution is focused on getting natural language text to collect all the problems faced in a device into one place so action to avoid future harm can be taken.  The solution extracts important text about device malfunctions using LexRank extractive summarization, then summarizing the results using GPT abstractive summarization.

Team member Arusarka Bose, a master’s level computer science student at Columbia shared, “The interest in patient safety stemmed from the tremendous impact that advancements in this field can bring. The prospective of making people's lives better is something that deeply stuck with us.”

As for the future of the project Bose said, “We aim to make it more robust through large-scale deployment so that it can be finally amalgamated into the three agents in the medical realm --- the patient, the hospitals, and the manufacturers.”

Michael Wong, JD, founder and executive director of the Physician-Patient Alliance for Health & Safety, served as the patient safety track judge. Of the experience, he said, “What stuck out to me about the winning team's analysis was that it had real-world applicability. A few years ago, we asked major manufacturers questions that our clinicians wanted answers to about their patient monitoring systems. The winning team’s analysis could help clinicians pick and buy equipment based on the adverse event rates of the manufacturers.”

The Patient Safety Technology Challenge is excited to see the future applications of the team’s solution.