search for




 

Marginalized two-part model for censored medical cost
Journal of the Korean Data & Information Science Society 2022;33:1085-94
Published online November 30, 2022;  https://doi.org/10.7465/jkdi.2022.33.6.1085
© 2022 Korean Data and Information Science Society.

Ae Jeong Jo1 · Eun Jin Jang2

1Department of Medicine, Hanyang University
2Department of Information Statistics, Andong National University
Correspondence to: This work was supported by a Research Grant of Andong National University.
1 Research assistant professor, Department of Medicine, Hanyang University, Seoul 04763, Korea.
2 Associate professor, Department of Information Statistics, Andong National University, Andong 36729, Korea. E-mail: ejjang@anu.ac.kr
Received October 12, 2022; Revised October 27, 2022; Accepted October 27, 2022.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/3.0) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.
Abstract
A regression model can be applied to identify risk factors affecting the medical cost or to compare the medical costs of various treatments. A large proportion of patients have zero costs due to no treatment and nonzero medical costs are highly skewed to the right. In addition, the medical costs of some patients are not fully observed because they are not followed until the endpoint of interest, thus the medical costs are right censored. In this study, we proposed a marginalized two-part model using inverse probability weights for the medical cost considering skewed medical costs with inflated zero cost and censoring. As a result, simulations and real data analysis show that the performance of the model with inverse probability of weighting have better performance even if the skewness and the proportion of zero costs and censored costs increases.
Keywords : Censored medical cost, inverse probability of weighting, marginalized model, two-part model.