Columbia Data Science Society’s Data Science Hackathon

Boosting Patient Safety was the winning team taking home a $500 prize for their innovative approach to analyzing a data set on adverse drug events.

Accessing and analyzing data is a critical component in understanding the scope of the patient safety crisis. Acknowledging the complexity of the issue is a prerequisite to innovation. The Patient Safety Technology Challenge, funded by the Pittsburgh Regional Health Initiative, was thrilled to sponsor the BIG IDEA: Patient Safety Technology Award at Columbia Data Science Society’s 8th Annual Data Science Hackathon.

The Data Science Hackathon was an all-day event held on campus at Columbia University on Saturday, February 11th, 2023. The ten teams competing for the Patient Safety Technology Award were provided same data set on adverse drug events and patient safety information to investigate. The winning team, Boosting Patient Safety, used data visualizations of the top 10 medicinal products in adverse conditions and the frequencies of patient reactions to analyze adverse drug reactions to predict the severity of adverse reactions based on drug attributes. Team members, Sai Rithvik Kanakamedala, Saili Myana, Shreya Oak, Vedangi Wagh achieved a validation accuracy of 79.9% with the help of XGBoost Classifier.

The Patient Safety Technology Award was judged by Columbia faculty in the computer science and industrial engineering and operations research departments. According to judge Yongwhan Lim, Boosting Patient Safety Team’s methods were “more rigorous compared to other projects and it may lead to further discovers that might be worth looking into.” JudgeYi Zhang added that the team “paid close attention to whether the team performed the tasks from a user perspective.” This is an important aspect to consider because for data to be applied it has be understood and useable first.  

The Boosting Patient Safety Team “hopes that one day, the adverse reaction of a particular drug can be successfully predicted by inputting patient demographics into a user-friendly application.” It is exciting to know that their project is a first step towards this goal.