Level: Beginner
【Presenter】
Kosuke Nakazono, National University Hospital ARO Next Generation Medical Center Technical Staff
Hiroko Watanabe, National University Corporation Kyushu University Hospital ARO Next Generation Medical Center
Masayuki Hirota, National University Corporation Kyushu University Hospital ARO Next Generation Medical Center
Takeshi Toyama, National University Corporation Kyushu University Hospital ARO Next Generation Medical Center
Koshiro Tagawa, National University Corporation Kyushu University Hospital ARO Next Generation Medical Center
Takuya Nagata, National University Corporation Kyushu University Hospital ARO Next Generation Medical Center
Junji Kishimoto, National University Corporation Kyushu University Hospital ARO Next Generation Medical Center
[Announcement summary]
Observational and epidemiological studies that compare groups to estimate therapeutic effects often result in imbalances in the subject's background factors. Careful attention should be paid to the analysis because the imbalance between the groups of confounding factors among the background factors causes the comparability to be impaired. In general, the imbalance of confounding factors can be adjusted to some extent by stratified analysis, but stratified analysis may not be applicable when considering many confounding factors for the sample size.
One of the adjustment methods used in such cases is analysis using a propensity score. Propensity scores are mainly used for stratified analysis, matching, and inverse probability weighting (IPW), and matching can be implemented using JMP add-ins. However, stratified analysis by propensity score and IPW are not implemented in JMP, and some steps need to be taken manually. In the analysis using the propensity score, it is important to automate the procedure because multiple propensity score models are examined.
In this presentation, we will focus on IPW based on propensity score, and introduce the execution method in JMP and the method of automating the procedure with demonstration.
