Construction and validation of a risk prediction model for hyperuricemia in perimenopausal and postmenopausal women
10.3969/j.issn.1671-8348.2025.08.007
- VernacularTitle:围绝经及绝经期女性高尿酸血症患病风险的模型构建及验证
- Author:
Mei ZHANG
1
;
Yi DIAO
;
Bo WANG
;
Mengqi LI
;
Guitao LI
;
Chuanwanyun DUAN
;
Hui TAO
;
Luming FAN
;
Aifang YE
;
Yong MAO
Author Information
1. 昆明医科大学公共卫生学院,昆明 650500
- Keywords:
perimenopause;
menopause;
hyperuricaemia;
prediction model;
Lasso regression;
random forest;
logistic regression
- From:
Chongqing Medicine
2025;54(8):1804-1810
- CountryChina
- Language:Chinese
-
Abstract:
Objective To develop and compare prediction models for hyperuricemia(HUA)in perim-enopausal and postmenopausal women using Lasso regression,random forest,and multivariate logistic regres-sion.Methods A multi-stage,stratified cluster sampling method was used to select 12 790 subjects from An-ning City,Yunnan Province.Prediction models for HUA were constructed using Lasso regression,random for-est,and multivariate logistic regression.The efficacy of the model was evaluated by accuracy,sensitivity,speci-ficity,F1 score,and area under the curve(AUC).Results LASSO regression analysis screened 19 variables for inclusion in the model,such as age,waist circumference,diastolic blood pressure,BMI,HDL-C,fasting blood glucose(FBG),etc.The accuracy rate was 0.701,the sensitivity was 0.703,the specificity was 0.680,and the F1 score was 0.806.The AUC(95%CI)was 0.770(0.748-0.792).The results of the random forest model show that variables such as creatinine,triglyceride-glucose index(TyG),TG,BMI,TC,Urea nitrogen(Urea),and ALT were relatively important,with an accuracy rate of 0.663,a sensitivity of 0.653,a specificity of 0.738,and an F1 score of 0.774.The AUC(95%CI)was 0.763(0.741-0.785).Multivariate logistic re-gression results showed that 11 variables including creatinine(Cr),TyG,BMI,Urea,and ALT were included in the model,with an accuracy rate of 0.705,a sensitivity of 0.707,a specificity of 0.686,an F1 score of 0.809,and an AUC(95%CI)of 0.771(0.749-0.793).Conclusion The overall performance of LASSO re-gression and multivariate logistic regression models is better.The random forest model has a strong variable screening ability and high specificity,and can be used as a supplement to provide more accurate predictions.
