Keywords: artificial intelligence, machine learning, healthcare, medicare
Summary:Approximately 20% of Medicare patients discharged from a hospital are readmitted within 30 days, which costs Medicare $15 billion to $18 billion per year. In 2017, 2,573 hospitals were penalized for too many 30-day readmissions. Unplanned admission and readmission performance remain a quality and financial priority for hospitals and healthcare systems. Through the development and implementation of Artificial Intelligence (AI)/Machine Learning (ML)-based models, HealthEC addresses the problems of predicting patients’ unplanned hospital and skilled nursing facility (SNF) readmission and adverse events within 30 days. For each problem, we first develop three baseline machine-learning models: logistic regression, random forest, and support vector machines for trust-building, transparency, and verification. We then develop three different AI-based models: single-layer neural network, multilayer feedforward deep neural network, and recurrent neural network. These models not only identify the patients who are currently in the high-risk band but also those patients who are not currently high risk but are on a trajectory to becoming high risk, where lies the biggest potential for patient impact. Further, the models are developed and validated using a data set of Medicare administrative claims data, including Medicare Part A (hospital care) and Medicare Part B. The proposed work will help transform the complex relationships among patient, disease, and treatment-associated variables into actionable models. Specifically, the AI/ML-based models will enable identification and prediction of the nature and confluence of temporal events, treatment intensities, adverse event, and outcomes that will empower clinicians to improve the timing and quality of patient while lowering costs, particularly for Medicare populations. The HealthEC platform is a testament to the fact that clinicians and medical professionals can work in tandem with ML/A-based models to improve the timing and quality of care, while lowering costs. Currently in the final testing stage, HealthEC will soon be unveiling its solution to these important problems and plans to make these easy-to-use AI-based models available to healthcare stakeholders.