Estimating Causal Effect of Two-Dose COVID-۱۹ Vaccination on Hospitalization with Machine Learning Techniques: A Propensity Score Matching Approach

In this innovative study, researchers investigated the effectiveness of two-dose COVID-۱۹ vaccination in reducing hospitalization amidst the complex confounding factors present in observational studies. Propensity scores have become increasingly popular for adjusting confounding variables in such studies. While propensity score methods offer theoretical advantages over traditional covariate adjustment methods, their performance in real-world situations remains poorly understood. By employing. Subsequent analysis revealed a significant balance between the vaccinated and unvaccinated groups. The results obtained from both Multiple Logistic Regression and Propensity Score Matching methods indicated that vaccinated individuals were less likely to be hospitalized [adjusted odds ratio (OR), ۹۵% CI using logistic regression: ۰.۲۱ (۰.۱۹, ۰.۳۰), and estimated by propensity score matching using logistic regression and GBM respectively: ۰.۷۲ (۰.۷۰, ۰.۷۴) and ۰.۹۳ (۰.۹۱,۰.۹۵). These findings not only emphasize the effectiveness of vaccination but also underscore the need for a meticulous approach when assessing real-world impacts in complex data environments.

کلیدواژه ها:

propensity score matching ، causal effect ، observational study ، GBM ، logistic regression