• AOC Header

    Analytics Oversight Committee (AOC)

The Analytics Oversight Committee (AOC) is the Health Data Oversight program subcommittee that oversees advanced analytic models, including artificial/augmented intelligence (AI) and machine learning (ML), intended for real time or near-real time clinical decision-making.  The AOC evaluates models using the S.A.F.E. criteria: Safety/Risk, Accuracy, Fairness/Bias, and Evidence.  The AOC also oversees the work done by the Health Analytics Core (HAC), which provides data science and biostatistics resources for model development and performance evaluation.

 

The AOC aims to work with requestors to enable innovative, effective, and meaningful solutions to the health system's biggest challenges in support of the clinical strategic plan and organizational vision.

 

Considerations Regarding the AOC Evaluation
An AOC approval does not equate to an overall project/initiative approval. ​Most aspects of an overall project/initiative approval are outside of AOC scope, and can include IT support, project management, regulatory and compliance, data sharing, procurement, clinical and operational implementation, and funding.

Frequently Asked Questions

Models (including devices, algorithms or other technology utilizing AI/ML) that are intended for application in the context of real time or near-real time clinical decision-making and clinical research should be reviewed by the AOC.  If you are unsure, you can complete the AOC Analytics Evaluation/Consult (Staff Only) intake form, and the AOC triage team will make a determination if the model is in or out of scope and advise of next steps.

You can request the AOC to review your model by completing the AOC Analytics Evaluation/Consult (Staff Only) request in ServiceNow.

The AOC currently reviews models based on the S.A.F.E. criteria, as well as the impact to the current standard of care. 

The S.A.F.E. criteria are: 

  • Safety/Risk: Is the model reasonably safe to use on patients? 
  • Accuracy: Is the model reasonable accurate for what it is trying to predict? 
  • Fairness/Bias: Are there any fairness or bias concerns? 
  • Evidence: Is there high-quality evidence that support the use of the model? 

The following are prerequisites for submitting an AOC intake form:

  • Committed Clinical/Business Champion(s) Identified
  • Commitment to Support Request by Executive Sponsor
  • Clinical/Business Workflows Clearly Defined
  • Strategic Alignment
  • If applicable, IRB Approved Protocol Required for Intended Clinical Research Implementation

These prerequisites ensure that the project is ready for AOC consideration. 

After you submit an AOC request, the AOC triage team will review the request and reach out for any additional information.  The review process is outlined here.

AOC process

There is not a standard timeframe for the AOC process from start to finish.  Each model the AOC reviews is unique in its complexity, evaluation stage, and performance.  If requesters have a well thought out implementation plan and the model doesn’t have any S.A.F.E. criteria concerns, the process will be most efficient.  The AOC triage team can provide a status update upon request. 

An approval by the AOC indicates that the model has been found to be safe to implement as presented, including the proposed phase, specific algorithms, and workflows.  Any recommendations outlined in the decision document should be followed.

An AOC approval does not constitute overall project approval, as there may be several steps required for a given initiative or project, including, but not limited to, IT evaluation, budget and resource considerations, strategic alignment and prioritization, compliance and privacy review, and contracting approval.

If the AOC does not approve a request to implement a model, the decision document will outline the area(s) of concern as well as recommendations.  Requestors may be referred to resources to help mitigate the concerns. Requestors can also choose to escalate their request to the Health Data Oversight Committee.  Details will be provided in the decision document.