Registration & Risk Assessment
ABCDS Oversight manages a registry of all algorithmic solutions under consideration for use in patient care at Duke Health. To be included in the registry and start the governance process, complete a registration form and e-mail it to us at abcds@duke.edu.
Not sure if your algorithmic solution falls within our scope? Send us an email or book our office hours!
Our Implementation Subcommittee will review the information in the form, complete a risk assessment (more details below), and communicate the corresponding evaluation pathway:
- Full review (high risk)
- Fast-track review (moderate risk)
- Registration only (low risk)
For low-risk tools, the Clinical and Model Owners are responsible for performing independent validation and testing.
Risk-Based Evaluation
For each registered algorithmic solution, an independent Review Committee is assigned to perform evaluations at checkpoints throughout the lifecycle — from procurement or development to general deployment, monitoring, and maintenance.
During registration, each algorithmic solution is assigned either a Full Review or a Fast Track Review evaluation pathway.
At each evaluation checkpoint, project teams present results from the most recently completed and upcoming phases.
The outcome of each review results in one of the following decisions:
- Approval
- Approval with contingencies
- Re-Review
- Decline
If the algorithmic solution is not approved to move into the next stage of the process, details of and a timeline to resolve gaps will be provided.
The Review Committee will evaluate each algorithmic solution for ethical and quality principles at every evaluation checkpoint.
The Review Committee will evaluate each algorithmic solution for ethical and quality principles at every evaluation checkpoint.
Guidance for Reducing Disparities
The Bias Management Framework is guidance that we provide to our community in order to understand how biases may be introduced into algorithmic solutions and/or associated workflows.
Regulatory Resources
To understand how current regulatory policy may apply to your AI solution, please step through the questions in the FDA Digital Health Policy Navigator.
For detailed information Software as a Medical Device (SaMD), please refer to the FDA’s 2022 Clinical Decision Support Software Guidance.
The Office of Regulatory Affairs and Quality and the ABCDS Oversight Committee can help answer any questions project teams may have about the regulatory context.
Publications
- A framework for the oversight and local deployment of safe and high-quality prediction models. Bedoya et al., 2022
- Translating ethical and quality principles for the effective, safe and fair development, deployment, and use of artificial intelligence technologies in healthcare. Economou et al., 2023