Affiliation:
1. Gies College of Business, University of Illinois at Urbana-Champaign, USA
2. National Chengchi University, Taiwan
3. University of Hawaiʻi at Mānoa, USA
Abstract
In this article, the authors provide an overview of the potential and challenges of machine learning for healthcare decision support. They first discuss the healthcare decision support ecosystems, including (1) beneficiaries, (2) health data, and (3) models. They then introduce the three main challenges of the healthcare decision support systems: data complexity, decision criticality, and model explainability. From there, they use unplanned intensive care unit readmission predictions in tackling the three main challenges of machine learning-based healthcare decision support systems. They investigate the data complexity issue by adopting dimension reduction techniques on patients' medical records to integrate patients' chart events, demographics, and the ICD-9 code. To address the decision criticality issue, they perform an in-depth deep learning performance analysis, and they analyze each feature's contribution to the predictive model. To unpack the model explainability issue, they illustrate the importance of each input feature and its combinations in the predictive model.
Reference45 articles.
1. Explainable Artificial Intelligence (XAI): Concepts, taxonomies, opportunities and challenges toward responsible AI
2. An Empirical Derivation of the Optimal Time Interval for Defining ICU Readmissions
3. Patients readmitted to the intensive care unit during the same hospitalization
4. Doctor AI: Predicting clinical events via recurrent neural networks.;E.Choi;Proceedings of Machine Learning Research,2016
5. Learning low-dimensional representations of medical concepts.;Y.Choi;AMIA Joint Summits on Translational Science Proceedings AMIA Summit on Translational Science,2016