AUTHOR=Song Wenzhu , Gong Hao , Wang Qili , Zhang Lijuan , Qiu Lixia , Hu Xueli , Han Huimin , Li Yaheng , Li Rongshan , Li Yafeng 

TITLE=Using Bayesian networks with Max-Min Hill-Climbing algorithm to detect factors related to multimorbidity

JOURNAL=Frontiers in Cardiovascular Medicine

VOLUME=Volume 9 - 2022

YEAR=2022

URL=https://www.frontiersin.org/journals/cardiovascular-medicine/articles/10.3389/fcvm.2022.984883

DOI=10.3389/fcvm.2022.984883

ISSN=2297-055X

ABSTRACT=Objectives: Multimorbidity (MMD) is a medical condition that is linked with high prevalence and closely related to many adverse health outcomes and expensive medical costs. The present study aimed to employ Bayesian networks (BNs) with Max-Min Hill-Climbing algorithm (MMHC) to detect the direct and indirect risk factors for MMD and to explore their complicated network relationships, thus taking targeted interventions for MMD patients and offering a reference for clinical practice and relevant guideline making.
Methods: The data was downloaded from the Online Open Database of CHARLS 2018, a population-based longitudinal survey. In this study, we included data on demographic background, health status and functioning, and lifestyle. Missing value imputation was first performed using Random Forest. And Elastic Net was employed to select variables highly related to MMD, which were then taken into BNs model construction. The structural learning of BNs was achieved using MMHC algorithm and the parameter learning was conducted using maximum likelihood estimation.
Results: Among 19,752 individuals (9313 men and 10439 women) aged 64.73±10.32 years, there are 9129 ones without MMD (46.2%) and 10623 ones with MMD (53.8%). After variable selection, nine variables were taken into BNs model construction. BNs suggested that age, sleep duration and physical activity constitute direct risk factors for MMD. Sex, education levels, residence, nap, smoking, and alcohol consumption are indirect risk factors for MMD. 
Conclusion: BNs could graphically reveal the complex network relationship between MMD and its related risk factors, and targeted prevention and control could be conducted accordingly, making BNs better application prospects in analyzing risk factors for disease.