Dr. Roy Ziegelstein on Trust, AI, and the Future of Evidence-Based Medicine

1. What safeguards are in place to ensure that AI-generated insights maintain clinical accuracy, trust, and transparency for healthcare providers?

I don’t think the term “AI-generated insights’ is quite right, since the term “insight” implies a clear and deep understanding and perception that relies on human experience and discrimination. I would say that at DynaMed and Dyna AI, insight is provided by oversight… human oversight of content to ensure accuracy, readability, and engagement. Human oversight to ensure that AI-generated information is clinically relevant. Human oversight to make certain that it is the kind of information that clinicians want when caring for other human beings.

At Dyna AI, a clinical assessment team provides ongoing evaluation to assess the accuracy and quality of responses. Dyna AI responses additionally provide the opportunity for users to provide their feedback, much as Waze collects information from users to modify their content. Indeed, responses to more than 3,000 customer queries by our Dyna AI team to date have exceeded a 95% benchmark for quality.  

If queries are not adequately covered by our evidence-based content corpus or have not undergone a full clinical quality evaluation, Dyna AI will not answer the query or will provide information directing users to an alternate resource. References used to generate Dyna AI responses are always fully provided and available with one-click access directly to the source material, mitigating the risk of lack of contextual understanding.

2. How do you plan to balance speed and rigor in content development, especially in high-stakes fields like infectious disease, oncology, or cardiology? 

Speed and rigor in content development are both critical, and never more so than today. Information is of limited use unless it incorporates the most currently-available evidence, however at the same time trust is something that is non-negotiable given what’s at stake. At DynaMed and Dyna AI, speed is ensured by our proprietary systematic literature surveillance system that selects the best and most recent evidence for all medical specialties, including infectious diseases, oncology, and cardiology. Once selected, the evidence is critically and meticulously appraised and summarized by a team of clinically-active, academically-renowned expert healthcare professionals, and then entered into our content which is updated on a daily basis, ensuring both speed and trustworthiness. With editors who are intimately involved in national and international professional societies, who create and revise core clinical guidelines, and who lead clinical and bench research in their healthcare fields, we are able to ensure that clinicians are provided with trustworthy, accurate information on advances in diagnostic techniques, new drug and device approvals, clinical trial results, and evolving treatment guidelines.

I think it’s also important to note that for evidence-based information to be useful at the point of care, it must not only be current and accurate, but also concise, easily and quickly digestible, and free from conflicts of interest. DynaMed and Dyna AI ensure that the only vested interest writers and editors have is their interest in providing current, accurate, trustworthy information to clinicians at the point of care that is free of industry bias, even when unintentionally introduced. That is essential for trustworthiness.

3. You serve a diverse global audience. What steps are being taken to ensure cultural and regional relevance in clinical recommendations?     

Car manufacturers must ensure global usability by designing and adapting cars for specific regions. For example, they may need to switch the steering wheel if a car is built for a country that drives on the left side of the road. However, it is the driver who must ensure that the car is driven in such a way that it is relevant to existing laws in that country, province, state or region.

Much is the same with DynaMed and Dyna AI. We ensure global usability by incorporating recommendations from national and international clinical practice guidelines and evidence from clinical research studies performed throughout the world, making certain that content is inclusive of all conditions whether seen in the U.S. or in other countries. However, Dyna AI clearly notes that responses “should not be solely relied upon for medical practice. Interpretation and application are subject to the judgment of a healthcare professional.” To put it another way, it is the healthcare professional who – like the driver – must ensure that the information is applied in such a way to the care of the patient that it is relevant to existing cultural and regional issues. Indeed, I would go even further and note that it is critical for healthcare professionals to apply the information we provide to the care of the unique needs of the individual patient before them regardless of country or setting.  

4. How do you collaborate with medical societies, hospital systems, or academic institutions to co-develop trusted content?

DynaMed and Dyna AI work with a variety of partners to co-develop trusted content. One example of this is our relationship with the American College of Physicians (ACP). The ACP – which has 161,000 members (including me) – identifies experts from their membership who review our content and make suggestions for improvement. This ensures that our topics are accurate and clinically useful at the point of care. We also work with the ACP to have access to their guidelines as they are being readied for release so that we can have them incorporated into DynaMed as quickly as possible. We also have a relationship with the American Academy of Pediatrics (AAP). In order to provide the highest quality content for our users, we license content from the AAP to be included in DynaMed topics. Similarly, we work with the National Comprehensive Cancer Network (NCCN) to provide access to their chemotherapy regimens. We have excellent working relationships with academic institutions and hospital systems that allow us to partner with them on innovative work. An example of this is the collaboration we have done in having institutions be beta testers for Dyna AI. As Editor-in-Chief, I look forward to adding to and enhancing these relationships for the benefit of DynaMed and Dyna AI users and the patients they care for.

5. How will Roy Ziegelstein background in medical education at Johns Hopkins influence your approach to clinician engagement and continuous learning?   

I have dedicated my professional life to educating learners at the bedside, in the classroom and lecture hall, in journal articles, and in medical textbooks. I appreciate the sacred obligation to provide accurate healthcare information, however I also understand the importance of making the process of obtaining that information engaging and interesting. If information is rigorous, accurate, and free from conflicts of interest, but does not engage the intended audience, it won’t be used. One thing I intend to focus on at DynaMed and Dyna AI is ensuring that our content not only accurate, current, and trustworthy, but also engaging and even enjoyable to use.

More than 50 years ago now, a group of investigators conducted a fascinating study published in the Journal of Medical Education in which an actor calling himself Dr. Myron L. Fox delivered an animated and entertaining lecture on a topic he knew nothing about to a sophisticated audience of health care professionals, educators, and graduate students… who gave the talk overwhelmingly positive evaluations. I know that sounds unbelievable so you should look for the YouTube video of the Dr. Fox lecture online. You’ll be amazed.

The point here is not that educators should teach on subjects they know nothing about. Indeed, a subsequent study found that while the so-called “Dr. Fox effect” engages an audience, it does not translate into actual learning. But to me, it’s important to also note the opposite: accurate, evidence-based, high-quality information provided in a manner that is not engaging will similarly not translate into actual learning… whether at the bedside, in the classroom or lecture hall, or when delivering evidence-based clinical decision support online at the point of care.

6. What metrics or outcomes will define success for editorial and technological leadership? 

It is important to us that independent healthcare research agencies such as KLAS that evaluates the performance of healthcare technologies honored DynaMed as “Best in KLAS for Clinical Decision Support” in 2025 and in two of the three previous years. Of course, the continuous feedback from our customers is also critical to ensuring the success of DynaMed and Dyna AI and to making any needed editorial and technological changes to make our products the best in class for those who are on the front line providing care for patients.

But the bottom line is that I want DynaMed and Dyna AI to be #1 in the industry. That’s my goal, and if it weren’t my goal… and if I didn’t think it possible… and if I didn’t think it deserved to be… I wouldn’t have taken on this role.

Get in touch with our MarTech Experts.