Comparison of risk models for mortality and cardiovascular events between machine learning and conventional logistic regression analysis

Abstract

AimsNon-linear models by machine learning may identify different risk factors with different weighting in comparison to conventional linear models.Methods and resultsThe analyses were performed in 15,933 patients included in the Shinken Database (SD) 2004–2014 (n = 22,022) for whom baseline data of blood sampling and ultrasound cardiogram and follow-up data at 2 years were available. Using non-linear models with machine learning software, 118 risk factors and their weighting of risk for all-cause mortality, heart failure (HF), acute coronary syndrome (ACS), ischemic stroke (IS), and intracranial hemorrhage (ICH) were identified, where the top two risk factors were albumin/hemoglobin, left ventricular ejection fraction/history of HF, history of ACS/anti-platelet use, history of IS/deceleration time, and history of ICH/warfarin use. The areas under the curve of the developed models for each event were 0.900, 0.912, 0.879, 0.758, and 0.753, respectively.ConclusionHere, we described our experience with the development of models for predicting cardiovascular prognosis by machine learning. Machine learning could identify risk predicting models with good predictive capability and good discrimination of the risk impact.

Decreased albumin is associated with elevated N-terminal pro-brain natriuretic peptide and poor long-term prognosis in patients with chronic heart failure

“…[ 31 ] Similarly, another cohort study demonstrated that decreased serum albumin, although in the “normal” range, was still associated with poor long-term outcomes in patients newly visiting the Cardiovascular Institute. [ 32 ] In this study, we also discovered that the decrease of albumin level is related to the severity of heart function and all-cause mortality in CHF patients, which supports the previous conclusions.…”

“…There is clear evidence that the albumin level is negatively correlated with adverse cardiovascular events in patients with cardiovascular and cerebrovascular diseases. [ 29 – 32 ] A study involving 560 patients with ischemic stroke aim to explore the relationship between serum albumin levels and ST and prognosis. The results showed that low albumin level was significantly associated with poor prognosis, and the adjusted odds ratio was 1.972.…”

Comparison of Conventional Statistical Methods with Machine Learning in Medicine: Diagnosis, Drug Development, and Treatment

“…In particular, ML can better predict clinical deterioration in the ward [ 26 ], mortality in acute coronary syndrome [ 27 ], survival in patients with epithelial ovarian cancer [ 28 ], complications of bariatric surgery [ 29 ], and risk of metabolic syndrome [ 30 ]. On the other hand, other studies reported that ML and conventional statistical methods have similar prognostic usefulness in predicting mortality in intensive care units [ 31 ], readmission in patients hospitalized for heart failure [ 32 ], and all-cause mortality and cardiovascular events [ 33 ].…”

Time Series Analysis and Forecasting with Automated Machine Learning on a National ICD-10 Database

“…This study aimed to perform predictive time series analysis using a national ICD-10 dataset of Romania over the period 2008-2018 with AutoTS. Using an AutoTS platform [39,40], we have predicted the incidence of the ten deadliest diseases in Romania, as defined by the WHO (World Health Organization) [41], consisting of ischemic heart diseases, stroke, chronic obstructive pulmonary disease, lower respiratory infections, Alzheimer's disease, lung cancer, diabetes mellitus, road injuries, diarrheal diseases, and tuberculosis. For each affliction, we have selected the most accurate ML model and predicted the monthly counts of new cases for every NUTS 2 region of Romania [42].…”