로지스틱 회귀모형을 사용한 율의 표준화 방법: 국민건강보험공단건강검진코호트 사용

Abstract

Objectives : To illustrate an approach for standardizing rates utilizing logistic regression models that leads to the enhanced reliability of estimation with reduced calculation cost.

Methods : For illustrative purposes, data regarding metabolic syndrome patients in 2013 were extracted from the National Health Insurance Service-National Health Screening Cohort (NHIS-HEALS). The detailed step-by-step calculations of age-sex adjusted prevalence rates of metabolic syndrome were demonstrated by both direct and logistic regression standardization approaches whose results were then compared.

Results : Standardization of rates using logistic regression models facilitated relatively simple calculation that can be easily implemented by using widely employed analytical programs such as R, SPSS, and SAS. Treating age as a continuous variable, the logistic regression approach produced confidence intervals of age-sex adjusted prevalence rates that were much narrower as compared to confidence intervals obtained by the direct standardization.

Conclusions : Standardization of rates utilizing logistic regression models may be a competitive alternative to the direct standardization in terms of computational efficiency and estimation reliability.