Predicting survival in heart failure: a risk score based on machine-learning and change point algorithm

Outcomes and predictors of one-year mortality in patients hospitalized with Acute Heart Failure

“…Thus, despite several medical therapies have been proven to improve survival in patients with HF over the past decades, the mortality after AHF hospitalization is still very high. A large number of literatures have identified post-discharge mortality predictors and proposed several predictive models for patients with AHF based on the past registries [11] , [12] , [13] . Due to the significant advancement of HF management, a contemporary registry is needed to provide evidence on clinical outcomes and prognostic predictors in patients hospitalized for AHF, especially in Thailand where the evidence relies on a registry conducted over a decade ago.…”

Mortality Prediction in Patients With or Without Heart Failure Using a Machine Learning Model

“…In addition to MARKER-HF, other machine learning-based algorithms for predicting the prognosis of patients with HF have been developed. 1 , 14 , 15 Jing et al used a machine learning-based algorithm to generate a mortality risk prediction model based on data from the electronic health record of 26,971 patients with HF. This model which used 26 variables performed slightly better (AUC = 0.77) than a model using linear logistic regression (AUC = 0.74).…”

“…This model which used 26 variables performed slightly better (AUC = 0.77) than a model using linear logistic regression (AUC = 0.74). 14 Kim et al 1 developed a machine learning model incorporating 27 continuous and 44 categorical variables from patients included in the Korean Acute Heart Failure registry. The AUC for predicting 1-year mortality was 0.71, which was comparable to that of the MAGGIC-HF score (AUC = 0.711).…”

“…In particular, individuals at high risk can be more intensively monitored, referred for advanced therapies, or counseled regarding end-of-life issues. 1 Identifying patients at high risk of mortality may also help improve the efficiency of clinical trials that include death as a primary or secondary end point. 2 In many settings, however, assessment of a patient’s risk of dying over a finite period of time is imprecise, particularly when it is based on clinical judgment or a single variable.…”

Predicting mortality in patients with heart failure based on machine learning approaches

“…This suggests that it is feasible to use ML models to build clinical decision aids related to HF to diagnose disease subtypes and monitor their progression in the future. Although ML shows promise in diagnosing and screening for HF, our knowledge of how ML might be used in other settings, such as patient risk assessment, is still in its infancy [8][9][10]. Many of the death risk prediction models for HF patients currently in use were created using conventional statistical techniques, like regression modeling [11][12][13].…”