Algorithm-Based Clinical Decision Support (ABCDS) Oversight
What Is ABCDS Oversight?
The Algorithm-Based Clinical Decision Support (ABCDS) Oversight initiative at Duke Health establishes governance and evaluation of algorithms, including artificial intelligence (AI), intended to be used for patient care. The initiative promotes innovative, safe, fair, and high-quality patient care by instituting pre- and post-deployment checkpoints designed to maintain transparency, quality, and accountability.
Get Started
Have you deployed, or are you working on, an algorithm to be deployed at Duke Health?
Access resources on the ABCDS Oversight process, including registration guidelines, review requirements, and key documentation.
Our Mission
Out of our primary focus on patient safety and high-quality care, our mission is to guide algorithm-based clinical decision support (ABCDS) tools through their lifecycle by providing governance, evaluation, and monitoring.
Our Scope
The scope of ABCDS Oversight includes all algorithmic solutions intended to be used for patient care, including:
- Diagnosis, cure, mitigation, treatment, or prevention of a disease or condition
- Communication, documentation, or outreach pertaining to individual patients
- Clinical operations or planning purposes pertaining to the care of patients
Our governance applies equally to internally developed solutions and those procured from external sources. Unsure whether your tool falls in scope for ABCDS governance? Please send us a message at abcds@duke.edu.
Our Guiding Principles
Algorithmic solutions must enhance clinical care by improving patient outcomes, ensuring patient safety, and increasing patient satisfaction. They should also contribute positively to the clinician experience by reducing burnout and improving overall workflow efficiency. Additionally, these solutions should optimize healthcare operations to better serve both patients and providers.
Algorithmic solutions should be designed and implemented to ensure fairness and reduce disparities in patient care. This means the solutions must be unbiased and trustworthy, aligning with the commitment to delivering consistent and just care for all Duke patients, regardless of their background or circumstances.
The deployment of algorithmic solutions should be accompanied by sufficient human and technical resources to support their effective use. This includes ensuring that the solutions are user-friendly, reliable, and sustainable over time. Adequate training, support, and resources must be provided to facilitate smooth adoption and long-term satisfaction among end users.
All algorithmic solutions must comply with applicable regulatory standards and best practices throughout their entire lifecycle. This includes adherence to legal, ethical, and safety guidelines to ensure the solutions meet all required regulatory and quality standards.
Algorithmic solutions must have clearly defined clinical and technical ownership. The development process should be transparent, with comprehensive documentation of the data, methodologies, and decision-making processes involved. This ensures accountability and allows for clear understanding and evaluation of how the solutions are developed and implemented.
Executive Leadership
Michael Pencina, PhD
Chair, ABCDS Oversight
Chief Data Scientist, Duke Health
Eric Poon, MD, MPH
Chair, ABCDS Oversight
Chief Health Information Officer, Duke Medicine
Nicoleta Economou-Zavlanos, PhD
Founding Director, ABCDS Oversight
Director, Duke Health AI Evaluation & Governance
Contact Us
Collaboration opportunities
To learn more about the Duke Algorithm-Based Clinical Decision Support (ABCDS) Oversight initiative or for general queries, send us an email or book our office hours.
Media and communications
For media and communications queries, contact Rabail Baig, Director of Communications
Email: health.ai.eval.gov@duke.edu