New Requirements For Algorithms that Involve Patient Care at Duke
All electronic algorithms that may impact patient care at Duke Health are now required to register with the Algorithm-Based Clinical Design Support (ABCDS) Oversight Program.
For information on how to register your model, click here.
The ABCDS Oversight program guides tools through their life cycle providing governance, evaluation, and monitoring. Types of models to be registered include:
- Data Driven Models: a model that builds relationships between input and output data using statistical/machine learning techniques. ML/AI and other statistically-derived models fall under this category.
- Clinical Consensus Models: a formula or set of rules that were derived based on clinical acumen and consensus, the literature, and/or expert recommendations. These algorithms provide the same results on the same inputs.
- Standard of Care / Clinical Guidelines-based Models: Standard of care is typically defined as the level and type of care that a reasonably competent and skilled health care professional, with a similar background and in the same medical community, would have provided under the circumstances. A ‘standard of care’ tool or model would be a tool or model used to guide standard-of-care as defined above and would be supported by evidence in the medical literature, recommended by medical societies, or incorporated into clinical practice guidelines.
Benefits of registering tools include:
- Proactive Alignment with FDA Processes
- Improved Accuracy, Impact and Efficiency
- Quality by Design Meeting the Intent of Use & Institutional Needs
- Quality Control & Assurance
- Model Ownership & Accountability
- Best-Practice Knowledge Sharing