Evaluating sociodemographic bias in a deployed machine-learned patient deterioration model
CHARTwatch model performance and impact were generally consistent across measured sociodemographic subgroups. ML-based clinical decision support tools, and associated standardization of care, may reduce existing inequities, as was observed for code status orders among patients experiencing homelessness. This evaluation provides a framework for future bias assessments of deployed ML-CDS tools.
Authors
- Amol A. Verma (ORCID: https://orcid.org/0000-0001-6857-931X)
- Michael Colacci (ORCID: https://orcid.org/0000-0003-4898-6684)
- Arjumand Siddiqi (ORCID: https://orcid.org/0000-0002-2458-9011)
- Muhammad Mamdani (ORCID: https://orcid.org/0000-0001-5199-6344)
- Chloé Pou-Prom (ORCID: https://orcid.org/0000-0002-5719-7786)
Institutions
- St. Michael's Hospital (CA)
- University of Toronto (CA)
- St Michael’s Hospital (IE)
- Public Health Ontario (CA)
Publication Details
- Journal
- JAMIA Open
- Published
- 2025-11-03
- DOI
- https://doi.org/10.1093/jamiaopen/ooaf158
- Primary Topic
- Homelessness and Social Issues
- Type
- article
- Field-Weighted Citation Impact
- 0.00