Claritza N. Abreu

AI is disrupting healthcare with unprecedented scope and speed, but our current evaluation, regulatory, and monitoring frameworks are ill-equipped to keep pace to ensure that these tools are safe, effective, and equitable. To address this gap, JAMA hosted a special summit on AI and described key findings in this report. Here are a few of my takeaways: ⚒️ The article starts by outlining key categories where AI tools are being developed and deployed - clinical AI tools (such as supporting diagnosis or treatment); direct to consumer tools (including algorithms built on top of wearables), AI tools to support clinical operations (e.g., revenue cycle management or supply chains), and a category described as “hybrid” which traverses clinical operations and clinical care (e.g.,ambient scribes or AI care navigation). 🦺 But the article, appropriately, spends time diving into some of the unique challenges around evaluating the safety and effectiveness of these AI tools; especially when few of the AI tools listed above fall under the purview of FDA approval. In terms of methods challenges: 🤼‍♂️ Unlike traditional therapeutic interventions, AI tools must consider both the AI software and the broader context and workflow in which it's delivered. An excellent algorithm with a poor user interface or accompanied with ineffective training can diminish the overall effectiveness of an AI solution. Instead of thinking of "AI algorithms" as the unit of analysis, we should be thinking about “AI systems.” Fortunately, studying complex systems and interventions is not new to health care - the field of implementation science and clinical informatics has long been working in this space. Teasing apart the relative impact of the algorithm vs its context and delivery mechanism is already difficult; but it’s even harder when the algorithm is not static but “learning.” ➡️ Inferring causality is also challenge, and the traditional gold-standard of RCTs is often cost prohibitive AND slow. Newer evaluation approaches such as cluster or stepped wedge design may be more applicable. We should also consider other evaluation approaches less used in traditional health care research such as randomized A/B testing. 🕴‍ From a regulatory perspective, the introduction of GenAI creates newer challenges, as these technologies are no longer designed as individual, narrow solutions but are better understood as platforms that can perform a vast and evolving range of tasks. As written about elsewhere, we might consider borrowing from frameworks that license and certify human professionals as we evaluate these platforms. https://lnkd.in/gBUw-XhH And the fact that we had 6 KP attendees (out of 39) participate in this closely curated meeting indicates the level of scientific and pragmatic leadership we have in our organization as it relates to AI! Kristine Lee, MD, David Ouyang, MD, Vincent Liu, Elizabeth A. McGlynn, Andy Bindman, Tracy Lieu

145

9 Comments

Join the TLA Study Group tomorrow for a fireside chat with Jodi Lis. We'll be discussing Learning Engineering in the context of interoperable data standards and will dive into her experiences in global education and digital health. We meet at 3pm EDT and the meeting is free and open to the public. Please comment below if you would like the invite link. About: Jodi Lis specializes in transforming learning through learning engineering, instructional design, learning analytics, and educational technology. She is a learning strategist and educator with over two decades of global experience, having designed and implemented large-scale digital, blended, and hybrid learning interventions across Africa and Asia. She serves as Vice Chair of IEEE ICICLE at the IEEE Learning Technology Standards Committee, is a co-author of the Learning Engineering Toolkit, and has contributed to global competency frameworks in digital health and AI literacy.

11

3 Comments

A great post on how AI should enable staff and teams, not replace. I've begun to think of AI as extending myself and my capabilities, not offloading or replacing my key role in the workflow. The most precious resource we have is time and there's never enough, but I've been able to claw back a little through the use of AI tools and learn a bunch of new skills in the process!

13

1 Comment

This is an excellent #AWS blog from Steven Johnston and James Grant on building Trusted Research Environments (#TREs) in the #cloud! If you're involved in sensitive data research, healthcare analytics, or secure data environments, this is a must-read resource that breaks down: 🔑 Key TRE Components: • Virtual Research Environment • Data Environment • Data Review & Transfer Systems Real-world case studies including UK Biobank, Genomics England and others. Whether you're just starting your #TRE/ #SDE journey or looking to optimise existing environments, the blog provides valuable guidance on balancing security, governance, and research accessibility. #CloudComputing #Research #DataSecurity #Healthcare #AWS #TrustedResearchEnvironments #HealthTech #hdrs

27

The future of IT is unfolding in the cloud, and now’s the time to dive deep. This visual guide, created by Sierra Consulting, highlights the core elements that every professional should master to thrive in the evolving tech landscape. From identifying Job Market Trends and choosing the right cloud platform to leveraging Cost Optimization Strategies, cloud knowledge is shaping how careers take off. As businesses adopt hybrid and multi-cloud architectures, professionals with hands-on experience in security, scalability, and platform flexibility are in high demand. Security in the Cloud is no longer optional; it’s the foundation of trust in today’s digital infrastructure. Whether you’re just starting or pivoting into cloud technology, understanding these pillars empowers you to navigate the revolution with confidence. The cloud isn’t just a place; it’s a career path. Let’s chart it together. #Cloud #ITCareers #TechTrends #HybridCloud #CloudSecurity #JobGrowth #TechSkills #CareerPath #SierraConsulting #CloudFuture #Upskill #DigitalJobs #CloudTech

2

Delighted to share that DexTech Inc. is one of the awardees of this year's Penn AITC opportunity. This $200,000 grant will enable us to develop an initial prototype of our wearable solution that we will test in a future study. Our solution will increase mobility and autonomy, and reduce fall injuries in people with Parkinson's Disease. This consortium has been funded by the National Institute of Aging (NIA). Dextech was one of 10 companies nationwide selected for funding. Our feasibility testing will be conducted with the help of collaborators Brian Wright at DePauw University and Jenae Bennington at The Towne House Retirement Community. We cannot wait to get started.

74

15 Comments

Next week we will dive into how artificial intelligence is streamlining patient interactions. Delighted to share the stage with Kathrina B. And Kevin Olson on real implementations of two highly regarded institutions such as UC San Diego Health and Jupiter Medical Center Perfect timing as we just launched Amazon Connect Health - an agentic AI solution to support patient and clinician experience: https://lnkd.in/d5rukBJc Come join us at Himss Global conference in Vegas next week!

26

1 Comment

Addressing inequities in healthcare outcomes is a key component in improving the wellness of our communities. Boston Medical Center’s Health Equity Accelerator recently launched a framework to help healthcare organizations build strategies to address health disparities – with a promising first implementation addressing several interventional areas – and the topic has been on my mind.   I’m a believer that a more equitable healthcare system starts in the classroom. Elsevier’s Global Content Partners team is privileged to play a role in supporting this by providing inclusive, evidenced-based scientific and healthcare content for students, educators, clinicians, and researchers. This helps ensure our next generations of providers are prepared to champion culturally-competent care in their roles.   Whether through interventions in care delivery, such as the framework introduced by the Health Equity Accelerator, or instilling inclusivity in the practices of our future healthcare professionals, we all have unique roles to play in advancing #HealthEquity. I encourage you to read in detail about the health equity framework and share your thoughts on additional areas where we can step up for the health of our communities.   https://bit.ly/47GoI1R

19

Public health leaders can’t afford to go on autopilot. AI can be a powerful ally, but we need to build systems that challenge us to stay sharp, not systems that think for us. Train for recovery. Promote AI fluency. Design for scrutiny. Here’s why that matters:

4

1 Comment

I often talk about the need to treat #AI as a direct extension of data and not just as another technology. This is especially true for organizations that must rely on internal, non-public, and proprietary data for genAI use cases. Here is the segment I recently recorded for Fed Gov Today with Patrick McGarry and Francis Rose on that point. #BeyondHype #DataGovernance https://lnkd.in/eCJkKDnm

19

AI success is no longer about experimentation. It is about value. CIOs are being asked to lead AI initiatives that create clarity, not complexity. This white paper outlines a practical roadmap for aligning AI to outcomes, scaling responsibly, and measuring value at every stage. If AI is on your agenda for 2026, this is required reading. Download the full white paper and see how value-driven AI leaders are approaching adoption. https://lnkd.in/gZK--_Kd

10