As the new ACR CEO, tell us what you see as the greatest opportunities ahead for improving quality and safety in radiology?
In general, the opportunities for radiology are tremendous right now. There is an increasing need for radiology services of all types. People are living to a much more advanced age. We have a higher percentage of patients with health insurance than we've ever had in the history of this country. People are surviving diseases that used to be a death sentence. We also have tremendous opportunities in screening advancements. Years ago, we couldn't detect preclinical disease; now imaging plays a critical role in screening. Along with that comes the opportunity — and obligation — for us to provide safer and higher quality care for our patients.
What are some of the newer opportunities on the horizon related to AI?
We're not fully there yet with AI, but we have the potential for AI to make radiology better, with lower radiation doses, potentially better detection of diseases, improved adherence to guidelines and things like better noise correction in our imaging so that patients don't need repeat imaging or undergo additional imaging. We will also have the ability to mine clinical data to provide a more informed history and, hopefully, narrow the differential diagnosis for imaging findings.
With that said, there are also challenges with all of the above. But I think radiologists are uniquely positioned in using AI to advance our specialty because of our long experience with things like Computer Assisted Detection in mammography. With that technological expertise, we can help all of medicine and patients navigate challenges with AI. We understand the fallibility of AI in ways that others might not yet realize. We already understand the importance of knowing what population an AI algorithm was tested and trained on. We also understand that AI is tested on enriched case sets — meaning it was tested using a dataset that has a higher incidence of disease than you would find in the general population. Of course, that's how AI is currently trained — as are human beings. We use datasets that have more patients with a certain disease, so the algorithm can learn what the findings associated with that disease look like. So, AI algorithms may not perform the same way when they are out in a real practice (where most patients won't have the disease) or if it is used on patients from a different ethnic background than those in the training data, for instance.
Radiologists also understand that there's always (or almost always) a differential diagnosis and the knowledge that AI will perform differently in different populations, for example, in pediatric populations and various ethnic groups. Radiologists already have so much experience with image-based AI that we're able to help everybody else learn how to use these tools, as well as how to be judicious and how to have a little bit of healthy skepticism in their use.
Along with opportunity comes challenge. What are some of the pressing challenges radiology is facing?
It is an interesting and exciting time for our profession, but also a challenging one. We have almost an even distribution of challenges and opportunities, and some of those are actually intertwined (or two sides of the same coin). The ready availability and access to imaging studies is a tremendous benefit for patients, but it also means that there often aren't enough of us in the radiology workforce to take care of all those patients in a timely fashion. Even so, I choose to look at it from a positive lens — that all of these things are happening because people are living longer and living better lives.
In many ways, we could say we are victims of our own success. With more patients surviving serious illnesses like cancer and vascular disease, we have a greater need for follow-up imaging and comparison with previous studies. We also have wonderful new technology that continues to improve. At the same time there is increasing demand for patient imaging, there's too much imaging for the number of radiologists we have. And that mismatch between supply and demand is touching almost everything we do at the ACR — whether it's finding time to volunteer with the College or carving out time for quality and safety initiatives in our own practices.
What is the ACR doing to help radiologists overcome those challenges and continue to improve quality and safety?
A key focus for the ACR is to continue to be a trusted resource for quality and safety. If you have a quality question, or you need to know how a process works or how a change in regulations will affect your practice, we want to be the one-stop shop for answers and easily accessible information. Also, in our accreditation programs and registries, we have to do whatever we can to make it easier, faster and safer — not only for the patients but also for the medical professionals who are trying to submit data and use our systems. We must make sure that our processes are as easy, smooth and fast as we can possibly make them, because healthcare workers are just not going to have the time to deal with any extra burden.
In addition, we have tools like the registries that comprise the
National Radiology Data Registry (NRDR®), which allow practices to benchmark and compare themselves to others and up their game. There are also tools like
RADPEER®, which enables radiologists to monitor their interpretive performance by auditing each other, and it can also facilitate peer learning. We also have the
ACR Learning Network, where practices go through the program as a cohort to share best practices and learn together. Even practices that think they're pretty good, often find out that there are opportunities for improvement. And we have newer programs like
ARCH-AI, which we rolled out in June as the first national recognition program for AI to ensure that AI is being used in a safe fashion in all practices.
If the specialty had one goal to raise its game in terms of quality and safety, what would that overarching focus be?
The goal must be to find a way to balance the challenge of the volume we are facing with our ability to do a high-quality job for our patients. I recently gave a brief talk on this topic at the
Quality and Safety + Informatics Conference (QS+I), and my goal for the session was to offer something actionable as a takeaway. I reviewed the impact of interruptions on both productivity and quality. I took the audience through the literature about the impact interruptions can have on your ability to concentrate and finish your tasks, as well as some strategies to mitigate those issues by simultaneously reducing errors and increasing productivity.
Some of the tips that were proven to be helpful included separating out different radiologist tasks and assigning one “radiologist of the day” who is responsible for answering questions from technologists, referring physicians, non-physician providers and patients. The rest of the group would work in what is called a “sterile cockpit environment,” where they would not be interrupted. There are also other models that use non-radiologists or non-physician providers as the point of contact to decrease interruptions for the radiologists.
What do you think have been the greatest accomplishments to come out of the ACR Commission on Quality and Safety in the last few years?
I think the
ACR Learning Network collaboratives have been a major leap forward in terms of helping practices improve their quality and safety. Participants can be open and transparent and find ways to learn together, which is necessary to improve quality. Again, things are going to change; technology is going to change. So, the quality initiatives, likewise, have to adapt. You need an ongoing commitment to sustained quality improvement as an iterative process.
Tell us something that you're excited about for the future in terms of quality and safety.
I think that a lot of the focus with AI has been on detection, but we have not looked at it closely enough for what it could do for quality and safety. I feel we might see the faster gains there. The ability to mine large datasets — whether it's using language models to analyze reports to create summaries that are more easily understandable for patients or identifying when a scanner is all of a sudden not conforming to a protocol. We might wind up seeing faster gains for AI on the quality and safety side than we do on the interpretive side.
Can you suggest a few steps that radiologists can take to get more active in quality and safety?
Number one, take advantage of the tremendous quality and safety resources and programs that the ACR already has in place, and get involved in your local chapters too. Share best practices with each other in terms of quality and safety. Participate in things like the QS+I Conference and the learning collaboratives. That's a fantastic starting place. Remember you can be part of the change that you want to see in the world. We'd love for people to
get more involved in the ACR and its various committees to help the specialty advance and grow. And remember the ACR is always here to serve as a resource to help our members improve quality and safety in their own practices!
At the RSNA Annual Meeting, ACR launched a landmark artificial intelligence quality registry. Can you tell us a little more about how Assess-AI™ will help ensure the effective performance of AI technology in radiological care?
ACR recently launched
Assess-AI, the world’s first AI quality registry designed to monitor ongoing performance of an array of imaging AI algorithms in real-world clinical settings.
Assess-AI is the newest
NRDR registry, capable of monitoring AI results and collecting an array of contextual information such as patient demographics, modality/exam metadata and output from radiology reports. Its purpose is to provide radiology facilities with analytics on how clinical AI is operating in their own practices over time and compare their results against aggregated national performance benchmarks from other sites using identical or similar products.
Utilizing Assess-AI, clinical sites and AI developers will be able to obtain performance reports of deployed AI derived from local practices. This data will also help developers make improvements to future versions of algorithms. With the rising demand for imaging outpacing the supply of radiologists, AI is seen as an essential tool to help bridge the gap and enable radiologists to maintain high standards of care while meeting increasing demands. ACR, working with leadership at the U.S. Food and Drug Administration and Congress, is positioned to guide the future of AI in radiology.
ACR Chief Executive Officer, Dana H. Smetherman, MD, MPH, MBA, FACR, FSBI, is a diagnostic radiologist specializing in breast imaging who previously served as Chair of the Department of Radiology and Associate Medical Director for the Medical Specialties at Ochsner Medical Center, New Orleans. Previously, Dr. Smetherman served as a member of the ACR Board of Chancellors as well as Chair of the ACR Commission on Breast Imaging, Chair of the ACR Breast Economics Committee and ACR Secretary-Treasurer.