A distributional reinforcement learning model for optimal glucose control after cardiac surgery
This study introduces Glucose Level Understanding and Control Optimized for Safety and Efficacy (GLUCOSE), a distributional offline reinforcement learning algorithm for optimizing insulin dosing after cardiac surgery. Trained on 5228 patients, tested on 920, and externally validated on 649, GLUCOSE achieved a mean estimated reward of 0.0 [-0.07, 0.06] in internal testing and -0.63 [-0.74, -0.52] in external validation, outperforming clinician returns of -1.29 [-1.37, -1.20] and -1.02 [-1.16, -0.89]. In multi-phase human validation, GLUCOSE first showed a significantly lower mean absolute error (MAE) in insulin dosing, with 0.9 units MAE versus clinicians' 1.97 units (p < 0.001) in internal testing and 1.90 versus 2.24 units (p = 0.003) in external validation. The second and third phases found GLUCOSE's performance as comparable to or exceeding that of senior clinicians in MAE, safety, effectiveness, and acceptability. These findings suggest GLUCOSE as a robust tool for improving postoperative glucose management.
Authors
- Zhang-Wei Hong (ORCID: https://orcid.org/0000-0001-8574-6108)
- David J. Reich (ORCID: https://orcid.org/0000-0002-9436-562X)
- Veena Nandwani
- Patricia Kovatch (ORCID: https://orcid.org/0000-0001-8368-1742)
- Ashwin Sawant (ORCID: https://orcid.org/0000-0003-1525-8541)
- Jacob Desman (ORCID: https://orcid.org/0000-0002-5411-6637)
- John A. Kellum (ORCID: https://orcid.org/0000-0003-1995-2653)
- Paul McCarthy (ORCID: https://orcid.org/0000-0002-4896-8374)
- Girish N. Nadkarni (ORCID: https://orcid.org/0000-0001-6319-4314)
- Moein Sabounchi (ORCID: https://orcid.org/0000-0002-6769-8066)
- Lili Chan (ORCID: https://orcid.org/0000-0003-4300-5760)
- Jaskirat Gill
- Monica Kraft (ORCID: https://orcid.org/0009-0002-6965-1090)
- Avniel Shetreat‐Klein (ORCID: https://orcid.org/0000-0002-0309-233X)
- Robert Freeman (ORCID: https://orcid.org/0000-0003-4946-6533)
- Donnie Goodwin
- Ankit Sakhuja (ORCID: https://orcid.org/0000-0002-2045-518X)
- Roopa Kohli‐Seth (ORCID: https://orcid.org/0000-0002-6082-1359)
- Ira S. Hofer (ORCID: https://orcid.org/0000-0001-7456-6697)
- Arpeta Gupta
- Umesh Gidwani
- Matthew A. Levin (ORCID: https://orcid.org/0000-0002-6013-2684)
- Alexander W. Charney (ORCID: https://orcid.org/0000-0001-8135-6858)
- Ana Catarina Costa (ORCID: https://orcid.org/0000-0003-0062-8714)
- Gagan Kumar (ORCID: https://orcid.org/0000-0002-6024-1055)
- Rajeev Sharma (ORCID: https://orcid.org/0000-0003-0733-644X)
- Robin Varghese (ORCID: https://orcid.org/0000-0002-2269-4846)
- Farzan Filsoufi
- D. Powell
- Pulkit Agrawal
- Sanam Ahmed
- Alexandra Carideo
Institutions
- West Virginia University (US)
- Hackensack University Medical Center (US)
- University of Pittsburgh (US)
- Mount Sinai Hospital (US)
- Northeast Georgia Medical Center (US)
- Millennium Engineering and Integration (United States) (US)
- Massachusetts Institute of Technology (US)
- Icahn School of Medicine at Mount Sinai (US)
Publication Details
- Journal
- npj Digital Medicine
- Published
- 2025-05-27
- DOI
- https://doi.org/10.1038/s41746-025-01709-9
- Citations
- 5
- Primary Topic
- Hyperglycemia and glycemic control in critically ill and hospitalized patients
- Type
- article
- Field-Weighted Citation Impact
- 6.47
Funders
- Georgia Clinical and Translational Science Alliance
- National Institutes of Health
- National Institute of Diabetes and Digestive and Kidney Diseases
- National Center for Advancing Translational Sciences