EUVIMEDEUVIMED entdecken
Crossref

Predicting cardiovascular-kidney-metabolic multimorbidity in Chinese adults with overweight and obesity using machine learning: an internal evaluation

Xin Li, Zhitong Li, Yexin Ji, Yuanyuan Yan, Xiaoqin Chen, Linlin Gao, Baoling Wei, Chenxia Zhang, Ruixue Duan, Shiwei Liu

Metabolism and Target Organ Damage · 24.07.2026 · 6 (3)

DOI: 10.20517/mtod.2026.60 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Metabolism and Target Organ Damage, online erschienen am 24.07.2026. Verfasst von Xin Li und 9 weiteren Autor:innen. Die Arbeit nennt 46 Literaturangaben.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Aim

To develop and internally evaluate machine learning (ML) models for predicting incident cardiovascular-kidney-metabolic (CKM) multimorbidity in Chinese adults with overweight or obesity, and to identify key predictors.

Methods

We included 4,244 participants from the China Health and Retirement Longitudinal Study (CHARLS) with overweight/obesity [body mass index (BMI) ≥ 24 kg/m2] and with zero or one CKM disease group at baseline (2015). CKM multimorbidity at follow-up (2018) was defined as coexistence of ≥ 2 disease groups (cardiovascular, kidney, metabolic). A two-stage feature selection [least absolute shrinkage and selection operator (LASSO) with bootstrap stability analysis] identified predictors from 56 sociodemographic, lifestyle, psychological, clinical, and environmental variables. Seven ML algorithms were compared; performance was assessed by area under the curve (AUC), calibration, Brier score, decision curve analysis, and SHapley Additive exPlanations (SHAP) interpretation for internal model evaluation.

Results

During 3-year follow-up, 648 (15.3%) participants developed incident CKM multimorbidity. Nine predictors were selected: age, depression, pain, health expectation, weight change, dyslipidemia, hypertension, heart disease, and kidney disease. Artificial neural network (ANN) and logistic regression showed the best discrimination (AUC: 0.758 and 0.757) and acceptable calibration in internal testing (Brier score: 0.113 and 0.114). SHAP analysis identified hypertension, dyslipidemia, and depression as top contributors. A nomogram was developed for preliminary risk stratification, but external validation is required before any clinical application.

Conclusion

In this internal evaluation, ANN and logistic regression showed moderate discrimination and acceptable calibration for predicting CKM multimorbidity in Chinese adults with overweight/obesity. Logistic regression performed comparably to complex algorithms yet is simpler; however, external validation is still required before use.

Zitieren

APA

Li, X., Li, Z., Ji, Y., Yan, Y., Chen, X., Gao, L., Wei, B., Zhang, C., Duan, R., & Liu, S. (2026). Predicting cardiovascular-kidney-metabolic multimorbidity in Chinese adults with overweight and obesity using machine learning: an internal evaluation. Metabolism and Target Organ Damage, 6(3). https://doi.org/10.20517/mtod.2026.60

Vancouver

Li X, Li Z, Ji Y, Yan Y, Chen X, Gao L, et al. Predicting cardiovascular-kidney-metabolic multimorbidity in Chinese adults with overweight and obesity using machine learning: an internal evaluation. Metabolism and Target Organ Damage. 2026;6(3). doi:10.20517/mtod.2026.60

BibTeX
@article{Li2026,
  author = {Li, Xin and Li, Zhitong and Ji, Yexin and Yan, Yuanyuan and Chen, Xiaoqin and Gao, Linlin and Wei, Baoling and Zhang, Chenxia and Duan, Ruixue and Liu, Shiwei},
  title = {Predicting cardiovascular-kidney-metabolic multimorbidity in Chinese adults with overweight and obesity using machine learning: an internal evaluation},
  journal = {Metabolism and Target Organ Damage},
  year = {2026},
  volume = {6},
  number = {3},
  doi = {10.20517/mtod.2026.60}
}
Kostenlos starten →