Meeting Highlights: DIA Global Annual Meeting 2023
Needed Collaborations to Illuminate the Future of Generative AI for Patient Benefit
Medicine Development Stakeholders Come Together to Discuss Uncertainties and Opportunities
Lindsay Kehoe
Sridevi Nagarajan
Hoifung Poon
Microsoft Health Futures
Maria Paula Bautista Acelas
“e-Patient Dave” deBronkart
Johan Ordish

enerative artificial intelligence (AI) is heralding a transformative era in healthcare, and ChatGPT and other large language models (LLMs) are prominent among these innovations. The profound impact of these applications spans various domains including improved patient communication, drug discovery, medicine and device safety monitoring, and the integration of diverse data sources. Indeed, the potential of these advancements is already shaping the way we approach our work today while stimulating our imaginations about their possibilities for the future.

Despite this tremendous promise, the adoption of generative AI in healthcare faces multifaceted challenges. Issues of regulation, validation, transparency, trust, ethics, and bias present formidable hurdles. Additionally, finding the right expertise, resources, and overall readiness to embrace these cutting-edge technologies may further complicate the process.

To alleviate these pressing concerns, and to enable addressing a broader range of interests and concerns, experts from around the world—including representatives of health regulatory bodies, industry leaders, statisticians, epidemiologists, patient advocates, academia, technology pioneers, and innovative start ups—participated in a DIA AI Solution Room at the DIA Global Annual Meeting 2023. This discussion built upon the insights shared in the AI DIAmond session Full Exposure: Artificial Intelligence to Advance, Replace, and Add Efficiency for Patient Benefit and served as a platform that fostered a forward-thinking and open dialogue among these experts. During this discussion, participants outlined critical challenges posed by generative AI that demand urgent attention, explored current potential solutions and best practices to overcome the challenges, and envisioned the ideal stakeholder engagement needed to prepare all medicine development stakeholders for the responsible implementation of generative AI for future patient benefit.

The Challenges

This unparalleled moment, characterized by the presence of groundbreaking and potent technologies like generative AI, holds the potential to profoundly influence and expedite innovation for society. However, akin to any influential technology, a host of challenges have emerged, encompassing matters of regulation, validation, transparency, trust, ethics, and bias. Fundamental inquiries abound, such as: How does it work? Is it safe? How do we control the models to ensure responsible application? How can we subject it to rigorous testing? How do we accurately analyze its performance?

smiling DIA crowd posing for a picture
These uncertainties exemplify the need for collaboration, as their mitigation provides a requisite step toward ensuring the responsible integration of this innovation, especially to benefit patients. This list of challenges represents the collective global perspective generated by the international global stakeholders who participated in this Solution Room. Although not exhaustive in its coverage of prevailing challenges, these prioritized challenges and possible solutions (summarized below) serve as an initial foundation for fostering further conversations among key stakeholders.
Clarity and compliance for sustainable AI approaches
Create reliable harmonized regulatory frameworks
Lack of knowledge on effectively integrating AI into existing workflows
Promote collaborative conversations to increase awareness of AI integration
Limitations of training data, assessments of data performance and validation, data restrictions
Define harmonized approaches and industry standards considering regional and local factors
Model generalizability
Increase understanding of data provenance
Improve training and communications across stakeholders
Job security
Establish policies that address concerns
Undue reliance on or avoidance of AI
Build story catalogs on how people are using or have already used large language models
Unclear information on using models
Define data methodology and the accuracy of models
Maximize value to patients
Develop harmonized guidance on how patients can utilize results
Control of data rights, privacy, and communication
Enhance societal awareness and stakeholder collaboration on this subject
Accuracy of data and perception of bias
Enhance “explain-ability” for efficient human verification
Address bias before the widespread implementation of AI
Develop use cases on what has and has not worked

Rising Above the Challenges

Understanding these challenges is paramount to establishing effective avenues for social collaboration to mitigate them. This is the only way we will truly benefit from the accelerated innovation that AI could bring to enhance communications, simplify processes, and make methodologies less expensive and more efficient.

Generative AI is a disruptive technology. It’s of paramount importance to recognize and mitigate potential risks in its applications. However, it’s also important to recognize the risk of not acting fast enough, thereby missing opportunities to produce beneficial innovations for patients more quickly. In the book The AI Revolution in Medicine, Dr. Isaac Kohane commented on the death of a newborn under his watch: “What killed him was then called persistent pulmonary hypertension of the newborn, but I’ve long believed the slow pace of medical research also killed him.”

Engaging multiple stakeholders in these discussions helps ensure that diverse views and opinions are reflected in comprehensive, impactful actions that could help overcome the challenges. During the Solution Room, participants contributed ideas for several key, concrete actions that must be undertaken:

Incentives: Establishing mechanisms that incentivize organizations and industries to facilitate data sharing and to establish reliable data sources will be instrumental in driving progress.

Amplification of Collaborative Initiatives: Fostering a culture of collaboration among stakeholders across domains will facilitate the sharing of expertise, use cases, and best practices, thus enabling collective growth.

Development of Regulatory Guidelines and Frameworks: Clear and adaptable regulations must be established, providing guidance and a foundation for long-term development, to ensure responsible and ethical AI deployment. Additionally, close collaboration and consistency with medical device regulation is key where traditional boundaries of what qualifies as a medical device and a medicine further blur.

Reinforcement of Training and Education: Nurturing a skilled workforce through comprehensive training and education initiatives is imperative to empower individuals to address misinformation and harness the potential of AI effectively.

Ideal Partnerships

As generative AI continues to accelerate in power and complexity, human decision-making and strategic approaches must adapt and evolve to keep pace with its rapid progress and consequent challenges. Crafting meaningful and measurable action plans in response to these challenges requires tailored partnerships that mirror the specific context of the challenge being addressed.

Cultivation of pertinent partnerships and collaborations emerges as a pivotal strategy in this pursuit. They amplify access to valuable resources, heighten awareness, facilitate knowledge dissemination, and furnish other support to propel efforts forward. Beyond these advantages, such collaborations furnish stakeholders with a vital conduit to remain informed of emerging trends and transformative shifts within the generative AI landscape.

In the endeavor to positively steer the integration of generative AI into benefitting patients through transparent decision-making, a broad array of essential collaborations among these varied stakeholders warrants consideration:

  • Patients and caregivers
  • Regulators
  • Digital and Innovation leads
  • Data scientists
  • Information technology sector
  • Pharmaceutical, device, and diagnostic industries
  • Payers
  • Healthcare providers
  • Academia and researchers

True comprehension of the challenges, coupled with meaningful collaboration and well-defined action plans, are pivotal in leveraging the transformative potential of generative AI. By embracing this approach collectively and ethically, we will work toward a future where generative AI-driven enhancements can be transparently and safely integrated into processes and methodologies, yielding greater efficiency and accessibility for every stakeholder.

DIA’s collaborative opportunities will persist as a platform to facilitate transparent dialogues among healthcare stakeholders. This will lead to the cultivation of thought leadership into the realm of adopting innovation and collectively crafting progressive solutions that benefit patients.

Read the Solution Room output.

DIA thanks these attendees for their contributions to this Solution Room:

Jeremy Jokinen (BMS), Adil Alaoui (Georgetown University), Marc Berthiaume (Health Canada), Hoifung Poon (Microsoft), Greg Ball (ASAP Process Consulting), Wout Brusselaers (Deep 6 AI), Alison Cave (MHRA), Nancy Dreyer (Dreyer Strategies LLC), Thomas Pietsch (Parexel), Dave deBronkart (e-Patient Dave LLC), Johan Ordish (Roche), Daisuke Koga (PMDA), Sridevi Nagarajan (AstraZeneca), Cynthia Verst (IQVIA), Munish Mehra (Tigermed US), Michelle Hoiseth (Cytel), Elizabeth Somers (Merck), Rohit Sood (CALYX), Jordan Silva (DocNexus), Michael Meighu (CGI), Ling Su (Lilly Asia Ventures), Lindsay Kehoe (CTTI), Phil Tregunno (MHRA), Sibel Guerler (BMS), and Eli Weinberg (Bain & Company).

Learn More—Get Involved

DIA is seeking contributors to a research project to further understand how AI impacts therapeutic development and demonstrate its potential. To learn more, please contact DIA also invites you to join our new AI in Healthcare Community.