Machine Learning and the Pursuit of High-Value Health Care
Ishani Gangulia b, William J. Gordonb c d, Claire Lupoe, Megan Sands-Lincolnf, Judy Georgeg, Gretchen Purcell Jacksonh i, Kyu Bak Louis Rheej, David Westfall Batesk l
- aAssistant Professor of Medicine, Harvard Medical School, Boston, Massachusetts, USA
- bDivision of General Internal Medicine and Primary Care, Brigham and Women’s Hospital, Boston, Massachusetts, USA
- cInstructor in Medicine, Harvard Medical School, Boston, Massachusetts, USA
- dMedical Director, Health Innovation Platform, Mass General Brigham, Boston, Massachusetts, USA
- eResearch Assistant, Division of General Internal Medicine and Primary Care, Brigham and Women’s Hospital, Boston, Massachusetts, USA
- fCenter for AI Research and Evaluation, IBM Watson Health, Cambridge, Massachusetts, USA
- gResearch Health Data Scientist, IBM Watson Health, Cambridge, Massachusetts, USA
- hAssociate Professor of Surgery, Pediatrics, and Biomedical Informatics, Vanderbilt University Medical Center, Nashville, Tennessee, USA
- iVice President and Chief Science Officer, IBM Watson Health, Cambridge, Massachusetts, USA
- j
Papers open inside SciRah from the original publisher or repository; SciRah never hosts them.
Abstract
The United States faces unsustainable growth in health care costs despite limited return on this investment in the form of health outcomes, prompting efforts to improve the value of health care. Artificial intelligence has the potential to enable high-value decision-making from the perspectives of the patient, clinician, and health system, but it also could worsen value. This article assesses these opportunities and challenges, separates reality from hype, and proposes policy approaches for contemporary artificial intelligence — specifically, machine learning — to contribute to rather than detract from high-value care. The conclusion is that it is critical to consider value — in particular, the cost component — in all machine-learning work and to let humans and algorithms each do what they do best.
TL;DRWith thoughtful deployment, machine-learning tools have the potential to help patients, clinicians, and health systems make health care decisions that promote better outcomes at lower cost. (Semantic Scholar)