건강보험 빅데이터를 활용한 생물학적 나이 추정 모형 비교 연구

Abstract

Objectives: The objective of this study is to estimate and compare indicators facilitating objective health assessment by utilizing biological age, a funda- mental component of health metrics, through various estimation methods. Methods: In this study, data from the National Health Insurance Service health examinations were utilized, and various methods for estimating biological age were employed. These methods include multiple linear regression, principal component analysis (PCA), and Klemera-Doubal method (KDM), which are based on statistical approaches, as well as RF and XGB, which are based on machine learning. In this study, ANOVA and regression were performed using the SAS 9.4 program. Results: Among statistical methods, the standard deviation for KDM’s BA-CA is the smallest at 8.6894, while machine learning methods exhibit similar values of approximately 5 for both approaches. Regarding disease diagnosis accuracy, KDM demonstrates the highest accuracy rates in hypertension and dyslipidemia, while PCA excels in diabetes diagnosis. Conclusions: This study can serve as a valuable health indicator, shedding light on the extent of aging within a population.