Predicting the outcome of heart failure against chronic-ischemic heart disease in elderly population – Machine learning approach based on logistic regression, case to Villa Scassi hospital Genoa, Italy

Abstract

Totally 167 patients were admitted at cardiology ward in Villa Scassi hospital, Genoa, Italy. We worked with two control groups: heart failure 59 patients (mean age: 71.37 ± 13.27 years) and chronic-ischemic heart disease 108 patients (mean age: 68.85 ± 11.3 years). Nine parameters: Hb, Serum Creatinine, LDL, HDL, Triglycerides, ALT, AST, hs-cTnI, CRP were evaluated onset to hospitalization. We aimed to identify significant independent predictors relative to the outcome of heart failure versus chronic-ischemic heart disease and select combination of biochemical parameters in logistic regression-based model that would provide on average excellent discrimination to the outcome of heart failure versus chronic-ischemic heart disease in elderly population. Applying 20-fold repeated stratified cross-validation, 4:1 train/test ratio split, we have found that model: pHF=eα+β1Hb+β2SerumCreatinine+β3AST+β4hs-cTnI+β5CRP1+eα+β1Hb+β2SerumCreatinine+β3AST+β4hs-cTnI+β5CRP, pHF: probability of heart failure, provides best discrimination of the outcome of heart failure against chronic-ischemic heart disease, having learned coefficients: α,β1,β2,β3,β4,β5 upon training set.

Keywords

• Heart failure
• Chronic-ischemic heart disease
• Machine learning
• Logistic regression
• Diagnostic
• Prediction