Proactive Algorithm Monitoring to Ensure Health Equity
JAMA Network Open · 15.12.2023 · 6 (12) · e2345022
DOI: 10.1001/jamanetworkopen.2023.45022 ↗Kurz gefasst
Kommentar in der Zeitschrift JAMA Network Open, online erschienen am 15.12.2023. Verfasst von Mark Sendak und 2 weiteren Autor:innen. Beteiligte Einrichtung: Duke Institute for Health Innovation, Durham, North Carolina. Hauptthemen laut MEDLINE: Health Equity. Die Arbeit wurde bisher 14-mal zitiert (Crossref) und nennt 6 Literaturangaben.
Abstract (Originalsprache)
Health care organizations are grappling with how to discover and mitigate the risks of artificial intelligence (AI) and associated algorithms worsening racial, ethnic, and socioeconomic health disparities.For example, the potential for patient harm has been shown across a variety of conditions, including inequitable access to timely care due to biased pulse oximeters and inequitable access to care management due to label choice bias.1,2 The risk of perpetuating these harms is significant, hence multiple stakeholders aim to hold health care organizations and the health care industry responsible for harm and discrimination, if it occurs.The evidence review by Chin and colleagues 3 developed a set of guiding principles to mitigate and prevent bias in health care algorithms.Their guiding principles add precision and specificity that go beyond broad mandates, such as the White House Blueprint for an AI Bill of Rights and National Institute of Standards and Technology AI Risk Management Framework.However, there may be challenges to integrate these principles with other mandates and operationalizing these guiding principles into practice for health care and technology organizations.To help address these issues, consortia such as Health AI Partnership and Coalition for Health AI aim to curate best practices for health care professionals to use AI safely, effectively, and equitably.4 Translating AI research into policy that then guides best practices clears only 1 hurdle.The other hurdles, which may be even larger, relate to the scarce people, processes, technology infrastructure, and operational supports required for health care organizations to successfully adopt and implement best practices.To do so, we highlight opportunities to close the translation gap from principles surfaced by Chin and colleagues into practice.3 Transparency There is a need to share what is in the so-called black-box of AI algorithms to improve trust and maintain trustworthiness with patients regarding use of data to develop algorithms followed by the use of algorithms in clinical care.Chin et al 3 emphasize the importance of patient and community engagement and go as far as to suggest that "All patients and communities should be informed when an algorithm is used in their care, should be advised about impact of the algorithm on their treatment, and should be provided alternatives if appropriate."Unfortunately, this aspiration is far from reality and likely would be overwhelming.Consider the current situation of AI algorithms being used in everyday life.It happens when someone does an online search, shops online, or engages with a streaming service.Given the impact algorithms may have on health care, there would need to be a stronger regulatory framework, with potential financial implications, to ensure algorithms performed as expected and safely without bias. Life-Cycle ManagementGiven the dynamics of AI and algorithms to continuously evolve, health care organizations must adopt a total product life-cycle approach to mitigate any unintended risks or changes over time as well as cause health disparities.Thankfully, there is consensus among diverse health care settings on the key decision points of algorithm adoption.As described by Chin and colleagues, 3 the process begins with problem selection and formulation and ends with algorithm updates or discontinuation.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Autor:innen und Institutionen
- Duke Institute for Health Innovation, Durham, North Carolina.
- Duke Clinical Research Institute, Duke University School of Medicine, Durham, North Carolina.
Zitieren
APA
Sendak, M., Balu, S., & Hernandez, A. F. (2023). Proactive Algorithm Monitoring to Ensure Health Equity. JAMA Network Open, 6(12), e2345022. https://doi.org/10.1001/jamanetworkopen.2023.45022
Vancouver
Sendak M, Balu S, Hernandez AF. Proactive Algorithm Monitoring to Ensure Health Equity. JAMA Network Open. 2023;6(12):e2345022. doi:10.1001/jamanetworkopen.2023.45022
BibTeX
@article{Sendak2023,
author = {Sendak, Mark and Balu, Suresh and Hernandez, Adrian F},
title = {Proactive Algorithm Monitoring to Ensure Health Equity},
journal = {JAMA Network Open},
year = {2023},
volume = {6},
number = {12},
pages = {e2345022},
doi = {10.1001/jamanetworkopen.2023.45022}
}
EUVIMED