Personalized Treatment Using Biologics: An Analysis Using Counterfactual Regression Based on Deep Learning
Seongho Eun, Bon San Koo, Ji Seon Oh, Kee-Eung Kim, Byungtae Lee · Journal of the Association for Information Systems · 2021
Observational data, such as electronic medical records in healthcare, have provided new opportunities for clinical discoveries. However, analyses using observational data are challenging due to the absence of a control group and selection bias between treatment groups. We propose a personalized treatment using counterfactual regression based on deep learning, which reduces concerns related to selection bias in a technical way. Using the data on the use of biologics in patients with rheumatoid arthritis, we estimate the individual treatment effect to determine which drugs to prescribe. In addition, we compare the individual treatment effect for each variable to show the relationship between clinical variables and the drug effectiveness. This study contributes to establishing evidence-based guidelines that recommend the optimal drug for each patient, thereby improving treatment responses clinically.