/
PHAISE Task Force Meeting 14
Edit

PHAISE Task Force Meeting 14

Tuesday, April 21, 2026

Recording

Quick Recap

The meeting focused on updates from the Responsible AI initiative in the Department of Medicine and discussions about AI in health research. Paul and Chirag presented on their program's mission to integrate data science and precision medicine into clinical education and research at Johns Hopkins, highlighting their new REACH dataset containing de-identified EHR data from 8 million patients. The group discussed potential collaborations across different schools, including opportunities in global health and digital health systems. Jinsung, with PHAIS, then presented plans for an upcoming AI and Health conference featuring keynote speakers and panel discussions on AI implementation in clinical settings. The conversation ended with a discussion about the previous AI Research Day event and potential future formats, including suggestions for more targeted collaborative events focused on specific data sets like REACH.

Next steps

PHAIS

  • Share conference website and details with the group, and coordinate with PHAISE/PHASE leadership about participation in the upcoming AI and Health Conference, including identifying a representative for the Friday afternoon working group if Greg is unavailable.

Collaboration

  • Paul/RAISE team: List all courses that leverage the REACH dataset on the REACH website, including recommendations for relevant courses (e.g., Jesse's course in SPH).

  • Paul/RAISE team: Announce when the REACH dataset will be available to the School of Public Health and School of Engineering, expected to be a matter of weeks after the upcoming Thursday symposium.

  • Ahmed/Paul/RAISE team: Continue work on the Discovery Award analyzing all clinical notes across the REACH dataset, including development of a playbook for efficient LLM analysis of large-scale clinical notes.

  • Ahmed/RAISE team: Consider adapting the AHA AI course modules for hospital leadership for deployment at Johns Hopkins, including scenario-based learning.

  • PHAISE leadership: Consider organizing more targeted collaborative events or hackathons focused on the REACH dataset, combining research and education, as suggested by Brian and others.

  • PHAISE leadership: Consider providing non-monetary incentives (e.g., proposal consultations, priority support) for new collaborative projects emerging from research day/poster sessions, as suggested by Kadija.

Summary

Responsible AI Launch at Hopkins

Paul and Chirag discussed the launch of Responsible AI in information sciences, education, and medicine at Johns Hopkins, focusing on integrating data science and precision medicine within the health system. They explained their mission to develop physician leaders in data science, offering educational programs and support across different levels of clinical practice, from medical students to faculty. The initiative aims to transform Johns Hopkins into a learning health system by leveraging tools like the REACH dataset and analytics platforms to accelerate clinical research and improve patient outcomes.

REACH Dataset Access and Collaboration

Paul described the REACH dataset, a de-identified collection of clinical data from 8 million patients over 10 years, currently accessible to the School of Medicine with plans to expand access to the School of Public Health and School of Engineering within weeks. He announced that on Thursday, there would be an update about when these other schools can access the dataset without needing a corresponding clinical PI. Gregory highlighted the importance of comparing clinical data with public health cohort data to address selectivity and bias, suggesting opportunities for triangulating different data sources. Joseph expressed interest in potential collaborations with colleagues in the Department of International Health, particularly regarding digital health systems and global health projects. Chirag mentioned existing global health applications of the REACH dataset, including prediction models for tuberculosis and diabetic retinopathy detection tools validated in both the US and Bangladesh.

EHR Systems Implementation Discussion

The group discussed electronic health record (EHR) systems and their implementation globally, with Paul highlighting collaborations with the Center for Population Health Informatics and the Odyssey Evidence Network. Gregory shared insights from his experience in Uganda, noting the challenges of electronic systems disconnecting clinicians from patients. Nilanjan suggested cross-listing courses between the School of Medicine and School of Public Health to avoid duplication and promote relevant training opportunities. Paul agreed to list all courses leveraging REACH data on the website and proposed creating practicum courses that bring together students from different disciplines to work on clinical problems using the REACH dataset.

Clinical Data Implementation Strategies

Paul explained Raise's core strength lies in clinically acquired data, including medical records, images, and EHR data, with capabilities to link patient identifiers through various platforms. Chirag highlighted the process of implementing models into EHR systems, emphasizing the steps of creation, validation, and scaling, which can take significant time due to IT processes and validation requirements. Ahmed discussed a discovery award proposal focused on analyzing clinical notes across data resources to create a data model for efficient AI analysis, while Paul noted the need for better research labels within the data for more precise cohort discovery. Ahmed also mentioned a multi-module AI course being developed for hospital leadership through the American Hospital Association, with plans to potentially deploy it at Johns Hopkins.

AI Health Conference Planning Meeting

The meeting focused on two main topics: the upcoming AI and Health Conference organized by Jinsung, and reflections on the previous AI Research Day event. Jinsung presented details of the conference, which will feature keynote speakers, panel discussions, and working groups on AI implementation in healthcare, with particular emphasis on public health applications. The group discussed the success of the AI Research Day poster session format and considered ways to enhance future events, including potential targeted collaborations with other departments and the possibility of organizing themed hackathons focused on specific data sets.