Predicting Rate of Adverse Effects for Drugs and Identifying Driving Factors

Health Analytics
Written By Alexis Avedisian

Video

Team Information

Team Members

  • Aditya Koduri, MS in Data Science Candidate, Data Science Institute, Columbia University

  • Karunakar Gadireddy, MS in Data Science Candidate, Data Science Institute, Columbia University

  • Yosha Tomar, MS in Data Science Candidate, Data Science Institute, Columbia University

  • Shivani Modi, MS in Data Science Candidate, Data Science Institute, Columbia University

  • Archit Matta, MS in Data Science Candidate, Data Science Institute, Columbia University

  • Faculty Advisor: Adam Kelleher, Adjunct Associate Professor, The Data Science Institute at Columbia University

Abstract

Drugs approved by the FDA before major holidays including Christmas and Thanksgiving have twice the rate of adverse effects. This is due an informal desk clearing nature resulting in a lax review. This behavior is consistent across geographies and is also observed in Europe and China. Time of approval for a drug is the second most important feature driver for adverse effects in the market, surpassed only by clinical rate – the adverse effects seen during clinical trials. Studying the multi-modal text data of FDA clinical trials with NLP is necessary to predict the rate of adverse effects. Other key factors include average payment made by a pharma company to doctors, specific active ingredients used in manufacturing and type of insurance.

Team Lead Contact

Aditya Koduri: ak4592@columbia.edu

Alexis Avedisian
Previous

Do You Trust Me? Development, Implementation and Acceptance of a Machine Learning Model - Opportunities, Challenges and Future Direction
Next

Exploring Environmental Health Risk Factors Associated with Cancer Incidence in New York State