Error Analysis of Machine Learning Predictions on Asthma Hospital Encounters

Abstract

To guide the allocation of finite preventive care resources for asthma management, we recently developed the world’s most accurate machine learning model to predict which asthmatic patients will have poor outcomes of incurring asthma hospital encounters (AHEs) in the following 12 months, where an AHE is an emergency department visit or an inpatient stay for asthma. In this abstract, we present an error analysis of our model’s predictions.