Integrating Machine Learning in Clinical Decision Support for Heart Failure Diagnosis: Case Study

Heart failure is the leading cause of hospitalization in people older than 65. Accurate referrals can reduce the devastating impact of heart failure. Timely diagnosis of heart failure from other cardiovascular conditions based only on symptoms is a major challenge. Machine learning has demonstrated potential for overcoming the diagnostic challenges of cardiovascular diseases. Many research papers are now focusing on application of artificial intelligence methods applied to diagnosis of heart failure, where databases continue to be a limitation. The current study used a dataset of 368 patients (297 patients with diagnosed heart failure, 71 control subjects) from an upper middle-income country, containing information on subject population characteristics, symptoms and laboratory test results. Manual feature selection was performed, focusing on clinical symptoms that are easily measurable. Four common machine learning methods were tested and compared: Decision Tree (DT) algorithm, Random Forest (RF) algorithm, Support Vector Machine (SVM) and Naïve Bayes (NB) algorithm. Models were developed through a holdout process of training-validation and testing. Our final model was a Decision Tree, achieving an AUC of 94.3%, with the advantage of being fully intelligible and easily interpreted. The performance achieved suggested that intelligible machine learning models can enhance symptom-based referral of heart failure.