Personalized Health Management From The Perspective Of Multi-Source Health Data

The problem of an aging population is a profound,long-term challenge.Given that the disease spectrum is shifting toward chronic disease,the traditional disease-centered treatment mode is evolving to health-oriented and prevention-focused health management.What we now face is the severe contradiction between the people’s ever-growing specific health needs and inadequate health service resources.In response,it is imperative to explore how to respond to personalized health needs and to optimize the allocation of health service resources.In addition,the rising penetration of new-generation information and communication technology to the health care industry,such as big data,the internet of things,artificial intelligence,and blockchain,brings opportunities for innovative practice of smart health,which emphasizes more proactive and personalized health management.Personalized health management is a hot research topic in the health management field,and will be a crucial factor in implementing the Healthy China initiative.This thesis aims to apply systems and integration thinking to examine personalized health management from the perspective of multi-source health data and recommend effective countermeasures and approaches based on new technologies,such as machine learning and blockchain.Our findings are important for providing theoretical and practical guidance for the efficient allocation of health service resources,enriching health management service research,and implementing a healthy aging strategy in China.First,we investigate a health data sharing model based on users’ creditworthiness from the perspective of one type of health data operator,the health data bank.We analyze the basic principles of the health data bank and depict users’ creditworthiness along three dimensions—reliable behavior,honest behavior,and contributing behavior;propose a differentiated service strategy based on creditworthiness;analyze and compare users’ behavioral patterns of sharing personalized health management data under different service strategies through numerical experiments.Our results show that the proposed behavioral incentive strategy based on creditworthiness can effectively increase users’ willingness to share personalized health data,improve the value of data,and realize the sustainability of personalized health management services.Second,to accurately identify the population with similar health needs,a population segmentation model based on unsupervised learning methods is investigated.We construct the health risk indicators from the interactive perspective of a “bio-psycho-social-environmental”model and develop a population segmentation model based on the partition around medoids clustering algorithm(PAM).We use the Gower dissimilarity coefficient to calculate the distance between samples covering continuous and categorical attributes.The algorithm is evaluated in terms of the optimal number of clusters based on average silhouette method and elbow method,clustering stability test according to the hold-out method,visualization of t-SNE dimensionality reduction,and clustering characteristics analysis using ANOVA test and Chi-squared test;the model is applied to stratify the population at risk of heart disease in empirical research.We also offer guidance with respect to health management for high-risk,medium-risk,and low-risk populations,as well as effective allocation of medical resources.Third,we construct a supervised learning-based multi-factor health risk assessment model from the interactive perspective of a “bio-psycho-social-environmental” model to explore the multifaceted risk factors that affect individuals’ health to examine the risk factors for chronic diseases.We analyze the association between these factors from the perspective of the individual’s physical and psychological status,family,society,and environment,etc.;extract risk factor features for chronic disease risk through correlation analysis;incorporate ensemble learning techniques to construct a multi-risk-factor health assessment model based on the XGBoost method.To enhance the interpretability of the machine learning model,we analyze the feature importance rank and use partial dependence plots to visualize the marginal effect of important risk factors based on the structure of the XGBoost algorithm.Finally,we verify the validity of the model and algorithm through empirical analysis.We then study the personalized health management service model,which is based on a digital platform.We analyze blockchain’s technical features and advantages;examine health data storage based on distributed ledger and asymmetric encryption and a consensus mechanism based on the value of health data assets in order to strengthen the security,privacy,and synergy of the digital health platform;and design personalized health management services on the digital platform based on four stages of health data collection: visualization and description,intelligent risk assessment,and personalized health intervention.Two application scenarios are examined: health risk warnings based on smart contracts and dynamic weight management based on personal health data.Finally,we provide strategic recommendations for the effective use of personalized health management data,taking into account the current development status of personalized health management data in China and international experience.The focus is on improving laws and regulations for special health care data;promoting data protection,interoperability,and sharing;developing and promoting unified data standards;innovating and developing data fusion technology;establishing an open sharing platform for health data and the use of health data in a compliant and efficient manner;strengthening international cooperation in the health care field;and fostering digital intelligence ecology for health management.