Construction of Logistic regression and decision tree prediction model for the risk of mild cognitive impairment in type 2diabetes patients
10.3760/cma.j.cn211501-20230511-01140
- VernacularTitle:2型糖尿病患者轻度认知功能障碍发生风险的Logistic回归和决策树预测模型的构建
- Author:
Ning ZHANG
1
;
Yan CHEN
;
Hongmei ZHANG
;
Yujiao SUN
;
Lili SHEN
;
Miao DING
;
Zhou ZHANG
Author Information
1. 南京大学医学院附属鼓楼医院内分泌科，南京 210008
- Keywords:
Diabetes mellitus, type 2;
Mild cognitive impairment;
Influencing factors;
Prediction model
- From:
Chinese Journal of Practical Nursing
2023;39(32):2514-2520
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To analyze the influencing factors of type 2 diabetes patients with mild cognitive impairment by Logistic regression and decision tree analysis, so as to provide reference for the prevention of these patients.Methods:The cross sectional investigation and convenience sampling method was used for the observational study, patients with type 2 diabetes hospitalized in the Endocrinology Department of the Drum Tower Hospital Affiliated of Nanjing University Medical School from April 2019 to August 2022 were selected as the objects, they were divided into the cognitive normal group and the mild cognitive impairment group, single factor analysis and Lasso analysis were used to screen variables. Logistic regression and decision tree analysis of diabetes with mild cognitive impairment were established and evaluated respectively.Results:Logistic regression analysis and decision tree analysis both showed that age, years of education, and insulin sensitivity index were effective early warning indicators of mild cognitive impairment with type 2 diabetes ( Z = - 9.39, 12.21, - 4.62, all P<0.05), and the decision tree model analysis showed that the number of years of education had the highest correlation with mild cognitive impairment. The different influencing factors of the two models were peripheral neuropathy, abnormal bone metabolism, and lower limb macroangiopathy. The specificity (62.7%) of the Logistic regression model was lower than that of the decision tree model (81.6%), and the sensitivity (77.3%) was higher than that of the decision tree model (54.9%). The AUC of the logistic regression model was 0.763 (95% CI 0.737-0.790), and the AUC of the decision tree model was 0.743 (95% CI 0.715-0.771). There was no difference in the predictive performance of the two models ( Z = 1.05, P = 0.295). Conclusions:The prediction ability of Logistic regression analysis model is similar to that of decision tree model. The Logistic regression analysis model can be used to screen out meaningful main effect early warning indicators, and further analysis of the correlation between indicators and research outcomes, as well as the interaction between various research factors, using a decision tree model, providing a reference for the prevention and control of diabetes patients with mild cognitive impairment.
