Development of a prediction model for the incidence of type 2 diabetic kidney disease and its application based on a regional health data platform
10.3760/cma.j.cn112338-20240117-00024
- VernacularTitle:基于区域健康数据平台开发2型糖尿病肾病发病风险预测模型及其应用
- Author:
Lijia LIU
1
;
Xiaowei CHEN
;
Yexian YU
;
Meng ZHANG
;
Pei LI
;
Houyu ZHAO
;
Yexiang SUN
;
Hongyu SUN
;
Yumei SUN
;
Xueyang LIU
;
Hongbo LIN
;
Peng SHEN
;
Siyan ZHAN
;
Feng SUN
Author Information
1. 北京大学公共卫生学院流行病与卫生统计学系，北京 100191
- Keywords:
Diabetes mellitus, type 2;
Diabetes kidney disease;
Prediction model;
Cohort study
- From:
Chinese Journal of Epidemiology
2024;45(10):1426-1432
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To construct a risk prediction model for diabetes kidney disease (DKD).Methods:Patients newly diagnosed with type 2 diabetes mellitus (T2DM) between January 1, 2015, and December 31, 2022, were selected as study subjects from the Yinzhou Regional Health Information Platform in Ningbo City. The Lasso method was used to screen the risk factors, and the DKD risk prediction model was established using Cox proportional hazard regression models. Bootstrap 500 resampling was applied for internal validation.Results:The study included 49 706 subjects, with an median ( Q1, Q3) age of 60.00 (50.00, 68.00) years old, and 55% were male. A total of 4 405 subjects eventually developed DKD. Age at first diagnosis of T2DM, BMI, education level, fasting plasma glucose, glycated hemoglobin A1c, urinary albumin, past medical history (hyperuricemia, rheumatic diseases), triglycerides, and estimated glomerular filtration rate were included in the final model. The final model's C-index was 0.653, with an average of 0.654 after Bootstrap correction. The final model's area under the receiver operating characteristic curve for predicting 4-year, 5-year, and 6-year was 0.657, 0.659, and 0.664, respectively. The calibration curve was closely aligned with the ideal curve. Conclusions:This study constructed a DKD risk prediction model for newly diagnosed T2DM patients based on real-world data that is simple, easy to use, and highly practical. It provides a reliable basis for screening high-risk groups for DKD.
