Predicting Acute Onset of Heart Failure Complicating Acute Coronary Syndrome: An Explainable Machine Learning Approach

About one-third of deaths worldwide are due to cardiovascular diseases each year, and the death rate keeps rising.1 Acute coronary syndrome (ACS) as 1 of the most common cardiovascular emergencies is the focus of prevention and treatment of cardiovascular disease.2 ACS includes acute ST-segment elevation myocardial infarction (STEMI), acute non-ST-segment elevation myocardial infarction (NSTEMI), and unstable angina (UA). Patients with ACS are at a very high risk of heart failure (HF). The incidence of HF after discharge for myocardial infarction (MI) is up to 30% at a 1-year follow-up.3 The development of HF after ACS substantially increases the risk of death, regardless of the HF type.4 HF development raises total mortality risk by 3 times and cardiovascular death by 4 times in individuals with a history of myocardial infarction. Therefore, rapid diagnosis and risk assessment of HF among ACS patients are essential to providing timely and cost-effective care.5

To better diagnose and treat ACS, many countries and regions, including China, have established chest pain centers (CPCs) and set up corresponding standardized procedures to enable ACS patients to receive timely and standardized treatment. Although ACS treatment is well established, the treatment effect for ACS patients is hindered by the time cost of hospital admission and prolonged hospitalizations, resulting in increased mortality rates.6 Therefore, accurate prediction of in-hospital HF following admission with ACS possesses great clinical values.

Prior studies have explored the risk factors of HF in patients after discharge for index ACS.7, 8, 9 There are a few studies trying to develop prognostic scores to predict outcomes following ACS in recent decades.10,11 However, these models were limited by their modest predictive power in predicting the incidence and outcome of HF. The complexity of HF pathophysiology and nonlinear relationship with high dimensional interactions among risk factors makes it difficult for clinicians and researchers to deprive a good prediction model using traditional statistical methods. Nonetheless, to the best of our knowledge, no published model has been shown to predict the risk of in-hospital HF among patients admitted with ACS. Indeed, it is challenging to predict patients who are at high risk of developing HF and are most likely to benefit from interventions to prevent HF complications.

The emerging techniques of artificial intelligence (AI) and machine learning (ML) offer a potential solution to analyze the ever-increasing multidimensional data sets and comprehend these intricate relationships to resolve this task.12,13 In this study, we trained and tested a machine learning model to predict the acute onset of HF subsequent to ACS.

留言 (0)

沒有登入
gif