Clinical AI: From Idea to Deployment to Making A Difference
December 17, 2024
Department of Biomedical Informatics

Clinicians and clinical scientists within the Columbia University Vagelos College of Physicians and Surgeons are invited to join an AI Workshop focused on ‘A Toolkit for Implementation’ Dec. 17. This event, hosted by the Department of Biomedical Informatics, will provide a framework to guide researchers through the process for implementation of AI tools in the clinical setting. 

What this workshop is about: In the life cycle of an AI project in healthcare, there are many steps and stakeholders that are involved and not all in a linear fashion. This workshop aims to give an overview of these steps and stakeholders, so that the tool can be successfully deployed and sustained. We will go through a use case of an AI project which has been developed, deployed, evaluated, and transferred across different hospitals in the country. The workshop will also have breakout sessions that focus on different aspects of this cycle, where you can apply these steps to your own project. 

Targeted audience: Clinicians and scientists starting to embark on an AI project with the vision for deploying in the clinical setting. Participants’ AI projects should be well-defined from a computational and modeling standpoint, but deployment plans are not yet well-formed or finalized.

After this workshop, you will have learned:
– when and how to approach executive leadership to support successful deployment 

– what content and communication strategies to consider when creating network of clinician champions and end users
– which stakeholders to involve in your EHR integration strategy
– what governance and legal criteria to consider prior to, during, and post deployment

Registration: When you register (see link below), please submit a 150-word paragraph describing your problem statement, idea, and preliminary data. For instance: “I am building a tool for early detection of endometriosis for patients coming to the Emergency Department with complaints of pelvic pain. If the AI model predicts likely endometriosis, the patient would be referred to specialty gynecology and sent a message to download a self-tracking app. Where I am in the project: I have built the AI model, I am currently looking for the best clinical workflow where to deploy the tool and for whom (nurse, doctor, patient?) and I want to understand and define the barriers to deploying the tool in a busy ED environment.” 

Tentative Agenda

9 -9:30 am – Continental Breakfast 

9:30-10 am – AI Deployment Landscape Overview (Lauren Richter, MD)

10-10:45 am – Use case: Implementing an early warning score in the hospital setting for patient deterioration (Sarah Rossetti, PhD)

10:45-11:15 am – Use Case: Implementing a multi-site tool for precision breast cancer prevention (Rita Kukafka, PhD)

11:15-11:30 am – Break

11:30 – 12:15 pm – Designing Your Toolkit: from idea to deployment to making a difference (Breakout Groups)

Breakout Group A: Executive Leader Stakeholder Engagement
  – Initial and continuous engagement with Executive Leadership
  – Considerations of financial implications and cost/benefit analysis
  – Shared governance considerations

Breakout Group B: Technical Stakeholder Engagement
  – Considerations for computational needs
  – Considerations for data availability needs
  – Bias auditing framework for fairness and model shifts

Breakout Group C: Engagement with Clinical End-Users
  – Strategies for engagement with end users and identifying workflows
  – Training and engaging with end users for the use of AI tool
  – Strategies for sustaining end user engagement with AI tool

12:15-1:00 pm – Toolkit Design Report-Outs and Closing Remarks