Predicting cardiometabolic multimorbidity in Chinese older adults via machine learning

Abstract

Aim: Cardiometabolic multimorbidity (CMM) is increasingly prevalent in China’s aging population, posing major public health challenges. Developing machine learning (ML) models for early prediction is essential to inform prevention.

Methods: We used data from 16,970 adults aged ≥ 45 years from the China Health and Retirement Longitudinal Study (CHARLS) across four waves (2011-2018). We used 42 predictors from 2013/2015 (demographics, lifestyle factors, physical measures, and blood biomarkers) to train five ML models: generalized linear model (GLM), gradient boosting machine (GBM), distributed random forest (DRF), deep learning (DL), and a Stacked Ensemble model. The primary outcome was CMM in 2018, defined as self-reported diagnoses of ≥ 2 conditions among hypertension, diabetes, heart disease, and stroke. Models were evaluated using the area under the curve (AUC), Brier score, and calibration curves. Synthetic Minority Over-sampling Technique (SMOTE) and 5-fold cross-validation were used to optimize performance.

Results: The Stacked Ensemble model achieved the best internally validated predictive performance (AUC = 0.755), significantly outperforming GLM and DL (both DeLong’s P = 0.03), with comparable performance to GBM and DRF. Calibration analysis confirmed reliable prediction (Brier score = 0.153). Variable importance analysis based on GBM and DRF identified dyslipidemia, age, triglycerides, high-density lipoprotein cholesterol, waist circumference, self-rated health expectations, pain, cooking fuel type, weight change, and cystatin C as the top-ranked predictors common to both algorithms.

Conclusion: ML algorithms, particularly ensemble models, can effectively predict CMM risk in China’s aging population. The integration of diverse health indicators and self-perceived health measures enhances predictive power.

Keywords

Cardiometabolic multimorbidity / China / aging population / machine learning / prediction

References

[1] Chen Y,Li D.Association of cardiorespiratory fitness with the incidence and progression trajectory of cardiometabolic multimorbidity.Br J Sports Med2025;59:306-15

[2] Fan J,Yu C.China Kadoorie Biobank Collaborative GroupMultimorbidity patterns and association with mortality in 0.5 million Chinese adults.Chin Med J2022;135:648-57

[3] Petrie JR,Touyz RM.Diabetes, hypertension, and cardiovascular disease: clinical insights and vascular mechanisms.Can J Cardiol2018;34:575-84 PMCID:PMC5953551

[4] Beard JR,Cassels AK.The World Report on Ageing and Health.Gerontologist2016;56 Suppl 2:S163-6

[5] Su B,Chen L,Li J.Long-term exposure to PM_2.5 and O₃ with cardiometabolic multimorbidity: evidence among Chinese elderly population from 462 cities.Ecotoxicol Environ Saf2023;255:114790

[6] Ding L,Tan ECK.Smoking, heavy drinking, physical inactivity, and obesity among middle-aged and older adults in China: cross-sectional findings from the baseline survey of CHARLS 2011-2012.BMC Public Health2020;20:1062 PMCID:PMC7336642

[7] Krentz A,Heiss C.International Cardiometabolic Working GroupRising to the challenge of cardio-renal-metabolic disease in the 21st century: translating evidence into best clinical practice to prevent and manage atherosclerosis.Atherosclerosis2024;396:118528

[8] He C,Fu L.Improving cardiovascular risk prediction with machine learning: a focus on perivascular adipose tissue characteristics.Biomed Eng Online2024;23:77 PMCID:PMC11299393

[9] Subramani S,Anand MV.Cardiovascular diseases prediction by machine learning incorporation with deep learning.Front Med2023;10:1150933 PMCID:PMC10150633

[10] Krittanawong C,Bangalore S.Machine learning prediction in cardiovascular diseases: a meta-analysis.Sci Rep2020;10:16057 PMCID:PMC7525515

[11] Mao D,Li Y.Network-based machine learning reveals cardiometabolic multimorbidity patterns and modifiable lifestyle factors: a community-focused analysis of NHANES 2015-2018.BMC Public Health2025;25:2297 PMCID:PMC12224847

[12] Zhu J,Ge Z.Comparative analysis of cardiometabolic multimorbidity predictors in China and the USA: a machine learning approach.Diabetes Res Clin Pract2025;229:112938

[13] Mukherjee M,Thokala HR.Classifying complex multimorbidity using latent class analysis and machine learning to generate insights into clustering of mental and cardiometabolic conditions.PLoS One2025;20:e0335676 PMCID:PMC12622840

[14] Zhao Y,Smith JP,Yang G.Cohort profile: the China Health and Retirement Longitudinal Study (CHARLS).Int J Epidemiol2014;43:61-8 PMCID:PMC3937970

[15] Chen X,Hu PP.Venous blood-based biomarkers in the China Health and Retirement Longitudinal Study: rationale, design, and results from the 2015 wave.Am J Epidemiol2019;188:1871-7 PMCID:PMC6825825

[16] Cui Z,Huang Y.Effects of socioeconomic status and regional inequality on the association between PM_2.5 and its components and cardiometabolic multimorbidity: a multicenter population-based survey in eastern China.Sci Total Environ2024;946:174453

[17] Li D,Yang H.Causal association of modifiable factors with cardiometabolic multimorbidity: an exposome-wide Mendelian randomization investigation.Cardiovasc Diabetol2025;24:241 PMCID:PMC12143095

[18] Yin X,Zhou L,Wang Y.Association between chronic pain and risk of cardiometabolic multimorbidity: a prospective cohort study.Reg Anesth Pain Med2025;50:858-64

[19] Mou P,Guan J,Zhang Z.Extreme temperature events, functional dependency, and cardiometabolic multimorbidity: Insights from a national cohort study in China.Ecotoxicol Environ Saf2024;284:117013

[20] Li J,Wang W.Social risk profile and cardiovascular-kidney-metabolic syndrome in US adults.J Am Heart Assoc2024;13:e034996 PMCID:PMC11963957

[21] Li D,Zhou Z.Role of socioeconomic, lifestyle, and clinical factors in the progression of cardiometabolic multimorbidity in China: a 10-year prospective cohort study.Nutr Metab Cardiovasc Dis2022;32:2383-91

[22] Liu J,Wang L.Association of visceral adiposity index and handgrip strength with cardiometabolic multimorbidity among middle-aged and older adults: findings from Charls 2011-2020.Nutrients2024;16 PMCID:PMC11280108

[23] Sun H,Wang G.Normalized creatinine-to-cystatin C ratio and risk of cardiometabolic multimorbidity in middle-aged and older adults: insights from the China Health and Retirement Longitudinal Study.Diabetes Metab J2025;49:448-61 PMCID:PMC12086583

[24] Lai H,Liao C,He L.Joint assessment of abdominal obesity and non-traditional lipid parameters for primary prevention of cardiometabolic multimorbidity: insights from the China health and retirement longitudinal study 2011-2018.Cardiovasc Diabetol2025;24:109 PMCID:PMC11890515

[25] He L,Tu Y.Correlation of cardiometabolic index and sarcopenia with cardiometabolic multimorbidity in middle-aged and older adult: a prospective study.Front Endocrinol2024;15:1387374 PMCID:PMC11165091

[26] Wang K,Fang W.Comparison of deep learning and traditional machine learning models for predicting mild cognitive impairment using plasma proteomic biomarkers.Int J Mol Sci2025;26 PMCID:PMC11941952

[27] Blough DK,Hornbrook MC.Modeling risk using generalized linear models.J Health Econ1999;18:153-71

[28] Lindsey JK.A review of some extensions to generalized linear models.Stat Med18:2223-36

[29] Rigatti SJ.Random forest.J Insur Med2017;47:31-9

[30] Esteva A,Ramsundar B.A guide to deep learning in healthcare.Nat Med2019;25:24-9

[31] Miao N,Han P.A new ensemble learning method stratified sampling blending optimizes conventional blending and improves prediction performance.Bioinform Adv2025;5:vbaf002 PMCID:PMC11908643

[32] Bunkhumpornpat C,Chouvatut V.FLEX-SMOTE: Synthetic over-sampling technique that flexibly adjusts to different minority class distributions.Patterns2024;5:101073 PMCID:PMC11573909

[33] Dunias ZS,Timmerman D,van Smeden M.A comparison of hyperparameter tuning procedures for clinical prediction models: a simulation study.Stat Med2024;43:1119-34

[34] Mandrekar JN.Receiver operating characteristic curve in diagnostic test assessment.J Thorac Oncol2010;5:1315-6

[35] Alba AC,Walsh M.Discrimination and calibration of clinical prediction models: users’ guides to the medical literature.JAMA2017;318:1377-84

[36] Brier GW.Verification of forecasts expressed in terms of probability.Mon Weather Rev1950;78:1-3

[37] Ruan X,Chen J.Joint effects of triglyceride glucose index and its obesity-related derivatives with estimated glucose disposal rate on cardiometabolic multimorbidity in middle-aged and older Chinese adults: a nationwide cohort study.Cardiovasc Diabetol2025;24:382 PMCID:PMC12492775

[38] Li X,Zhao W,Pan H.The difference between cystatin C- and creatinine-based estimated glomerular filtration rate and all-cause and cardiovascular mortality in populations with cardiovascular-kidney-metabolic syndrome stages 0-3: a prospective cohort study.Am J Nephrol2025;:1-15

[39] Ji RR,Huh Y,Maixner W.Neuroinflammation and central sensitization in chronic and widespread pain.Anesthesiology2018;129:343-66 PMCID:PMC6051899

[40] Xu B,La R.Is systemic inflammation a missing link between cardiometabolic index with mortality?.Cardiovasc Diabetol2024;23:212 PMCID:PMC11191290

[41] Smith L,Soysal P,Koyanagi A.Unclean cooking fuel use and health outcomes in older adults: potential mechanisms, public health implications, and future directions.J Gerontol A Biol Sci Med Sci2023;78:2342-7

[42] Miletic M.Utility-based analysis of statistical approaches and deep learning models for synthetic data generation with focus on correlation structures: algorithm development and validation.JMIR AI2025;4:e65729 PMCID:PMC11969122

[43] Su B,Xie J.Chronic disease in China: geographic and socioeconomic determinants among persons aged 60 and older.J Am Med Dir Assoc2023;24:206-212.e5

[44] Chen J,Zhang Y.Trajectories network analysis of chronic diseases among middle-aged and older adults: evidence from the China Health and Retirement Longitudinal Study (CHARLS).BMC Public Health2024;24:559 PMCID:PMC10882875

[45] Chen C,Yang L.Network analysis of chronic disease among middle-aged and older adults in China: a nationwide survey.Front Public Health2025;13:1551034 PMCID:PMC12016668