Statistical Methods for Baseline Adjustment and Cohort Analysis in Korean National Health Insurance Claims Data: A Review of PSM, IPTW, and Survival Analysis With Future Directions
10.3346/jkms.2025.40.e110
- Author:
Dong Wook KIM
1
Author Information
1. Department of Information and Statistics, Department of Bio & Medical Big Data, Research Institute of Natural Science, Gyeongsang National University, Jinju, Korea
- Publication Type:Review Article
- From:Journal of Korean Medical Science
2025;40(8):e110-
- CountryRepublic of Korea
- Language:English
-
Abstract:
The utilization of health insurance claims data has expanded significantly, enabling researchers to conduct epidemiological studies on a large scale. This review examines key statistical methods for addressing baseline differences and conducting cohort analyses using Korean National Health Insurance claims data. Propensity score matching and inverse probability of treatment weighting are widely used to mitigate selection bias and enhance causal inference in observational studies. These methods help improve study validity by balancing covariates between treatment and control groups. Additionally, survival analysis techniques, such as the Cox proportional hazards model, are essential for assessing time-toevent outcomes and estimating hazard ratios while accounting for censoring. However, the application of these statistical methods is accompanied by challenges, including unmeasured confounding, instability in weight estimation, and violations of model assumptions.To address these limitations, emerging approaches, such as Doubly robust estimation, machine learning-based causal inference, and the marginal structural model, have gained prominence. These techniques offer greater flexibility and robustness in real-world data analysis. Future research should focus on refining methodologies for integrating highdimensional health datasets and leveraging artificial intelligence to enhance predictive modeling and causal inference. Furthermore, the expansion of international collaborations and the adoption of standardized data models will facilitate large-scale multi-center studies.Ethical considerations, including data privacy and algorithmic transparency, should also be prioritized to ensure responsible data use. Maximizing the utility of health insurance claims data requires interdisciplinary collaboration, methodological advancements, and the implementation of rigorous statistical techniques to support evidence-based healthcare policy and improve public health outcomes.
