Modeling methodology for early warning of chronic heart failure based on real medical big data. (1st August 2020)

• Record Type: Journal Article
• Authors: Zhou, Chunjie; Li, Ali; Hou, Aihua; Zhang, Zhiwang; Zhang, Zhenxing; Dai, Pengfei; Wang, Fusheng
• Is Part Of: Expert systems with applications. Volume 151(2020)
• Publication Date: 2020-08-01
• Subjects: Heart failure -- Early warning -- Social network -- Risk factors -- Medical big data; Expert systems (Computer science) -- Periodicals; Systèmes experts (Informatique) -- Périodiques; Electronic journals; 006.33
• DOI: 10.1016/j.eswa.2020.113361
• Languages: English
• ISSNs: 0957-4174

Abstract

Highlights: Construct a medical social network to describe the similarity among risk factors. The personal health can be denoted as a probabilistic combination of risk factors. Propose a division algorithm to divide the network into a high and low risk group. Propose a heart failure risk prediction method, which accuracy can reach almost 90. Set four tests to measure the effectiveness of our method. Abstract: Heart failure (HF) is among the most costly diseases to our society, and the prevalence keeps on increasing these days. Early detection of HF plays a vital role in saving lives through adjusting lifestyles and drug interventions that can slow down disease progression or prevent HF. There are many cardiovascular risk factors associated with HF, and they often coexist. In this paper, we assess the predictive value of pathological factors for early HF detection through a social network based approach. We use electronic health records (collected from the project HeartCarer) and compute the similarity of risk factors. The similarity values are used to construct an unweighted and a weighted medical social network. The constructed medical social network is further divided into a HF high-risk group and HF low-risk group using a group division algorithm. Patients in the high-risk group will be suggested for early screening. To evaluate the prediction value of our method, we perform four experiments based on real world data. The results demonstrate the high effectiveness of our method on heart failure risk assessment, with the best accuracy close to 90%.