Historically, the informatics community has led innovation in health IT. Early EHR systems were created by clinicians and informaticians to test how technology could support health and healthcare. These grassroots systems enabled rapid experimentation and shared learning.

Today, innovation in healthcare is more constrained. While consumer technology has enabled anyone with an idea to create and distribute apps, commercial EHR systems have limited similar bottom-up innovation in healthcare. To move forward, we need systems that are:

• HIPAA and ONC compliant
• Integrated into existing clinical workflows
• Powered by both evidence-based and cutting-edge AI technology

To meet this need, we are building Health Universe — a flexible platform designed to support innovation in oncology informatics. Health Universe enables developers and clinicians to create, deploy, and manage apps within healthcare settings. It supports ONC regulatory requirements for decision support and integrates with existing workflows using established standards.

What Health Universe Offers:

• AI-driven workflow integration that anticipates clinical needs based on context
• Predictive and evidence-based apps using state-of-the-art tools
• An innovation sandbox for governance, piloting, and organizational app development
• A collaborative community and app marketplace for knowledge exchange and reuse

___________________________________________________________________________________________

New Focus: Building a Cancer-Specific AI Evaluation Framework

But how do we know if the technology is performing as intended? How do we monitor the performance over time, or advance new metrics that focus not just on information, but action and clinical outcomes?

At the recent CI4CC AI Collaborative Steering Committee kickoff, we refined the initiative’s direction to focus on advancing the evaluation of AI in oncology. Rather than solely building tools, the group reached consensus on contributing to the scientific literature on AI evaluation, specifically within cancer care.

This effort focuses on task-specific evaluations, similar to the MedHelm initiative at Stanford, and adapts them to the unique needs of oncology, including:

• Complex imaging and disease tracking
• Genetic and molecular data integration
• Variability in clinical context across institutions

Evaluation Framework Highlights:

We hope to move beyond static metrics, to ones that are specific to oncology tasks, focus on action-based evaluations (did the clinician change their plan, based on the AI agent?) and consider how to build agents to help track outcome-oriented evaluations (time to treatment, survival, prognostic accuracy, etc). 

Targeted Use Cases for Evaluation and App Development:

To keep the scope manageable, the group will focus on 4 key use cases: 

1. Cancer Registries Support
   Leverage AI to automate abstraction and coding. Collaborate with SEER and CDC, which are both undergoing major programmatic updates.
2. Clinical Guidelines and Decision Support
   Address gaps left by industry limitations. Explore guideline digitization via collaborations with NCCN member institutions.
3. Cancer Screening Optimization
   Improve risk stratification and compliance in primary care through AI-enabled tools.
4. Administrative and Documentation Tools
   Automate cancer case summarization and reduce documentation burden for oncologists.

As we develop an evaluation framework, we can test it on these use cases first, and expand the evaluation framework as we learn more. 

___________________________________________________________________________________________

Getting involved

If you are interested in learning more, or participating in the use cases (or evaluation), let us know at 📧 Doug.Fridsma@healthuniverse.com 

Together, we can redefine how AI improves cancer care—through collaborative innovation, rigorous evaluation, and meaningful clinical impact.