CAP TODAY Pathology/Laboratory Medicine/Laboratory Management
July 2023—Digital pathology, the pathology workforce, and the clinical demand for subspecialty expertise were some of what Compass Group lab leaders took on in their June 6 conversation, with CAP TODAY publisher Bob McGonnagle leading the way. And Stan Schofield, VP and managing principal of the Compass Group, painted a picture of the precarious situation clinical labs are in: “Everyone I talk to says capital equipment is being cut, and staffing costs are increasing dramatically if you want to retain staff. Cost base is going up, reimbursement pressures will continue, and there is no margin left.”
The Compass Group is an organization of not-for-profit IDN system laboratory leaders who collaborate to identify and share best practices and strategies. Here is what they shared last month.
At the Pathology Informatics Summit in May, Alverno CEO Sam Terese gave an excellent presentation on Alverno Laboratories’ decision to go completely digital. Wally Henricks, where is digital pathology today at the Cleveland Clinic?
Walter Henricks, MD, vice chair, Pathology and Laboratory Medicine Institute, and laboratory director, Cleveland Clinic: We use digital pathology for subspecialty case review conferences and consultations. It enables participation from all hospitals in our system. For direct clinical purposes, we do an increasing number of digital consultations, mostly for international and increasingly for domestic. We’ve validated it for frozens at small-volume offsite locations. We haven’t deployed it widely because of limitations in technical staffing at the sites. We’ve validated it for rapid onsite evaluations but haven’t done much because slide interpretations take longer than on a scope, and the communication between cytopathologists and clinicians in a busy environment is not satisfactory. As we are asked to cover rapids at lower-volume sites, we are considering it to support those.
Dr. Henricks
We signed an agreement with a digital pathology and AI company to develop a digital slide library that’s well curated and well annotated. This will be used to support future image-based AI development.
We created a new center in our pathology and laboratory medicine group that’s focusing on AI and machine learning projects.
Can you comment on where the hype and reality are on AI, and what the timeline is on it?
Dr. Henricks (Cleveland Clinic): I tend to be skeptical about the hype. I do think we will see valid applications. I think it will be slower than people expect. I’m oversimplifying here—the way most of these work is you have an idea, build an algorithm, have data on which it’s trained, do your best to train and validate it against the gold standard—and what the gold standard is is an element of that—and implement it. And then many of these will continuously “learn” or at least that’s the idea. But what is acceptable performance? What is it trying to do? So much of that is based on the data being used, not only for the training set but also in the field. How well does it do in the real world compared to the training world?
The regulatory environment is another big piece. In 2019 the FDA published a proposed framework for regulation [of AI/machine-learning–based software as a medical device] with the acknowledgment that some of these algorithms change continuously as they learn [https://bit.ly/04022019-FDA_AI-ML-frmwk]. The FDA’s previous way of handling software as a medical device was to provide guidance about when it’s necessary to do a premarket submission for a change to an existing software product. With AI algorithm changes, the FDA realized that the algorithms under the old version would require premarket submission with every update, which would be very burdensome. Hence the framework. Then the FDA issued an update in 2021 [https://bit.ly/012021-FDA_AI-ML-actplan].
Most significant is the draft guidance issued in April 2023 [https://bit.ly/04032023-FDA_AI-ML-draft] for how machine learning and AI algorithms will be evaluated; it’s called a predetermined change control plan. It’s a fusion of a manufacturer’s general quality systems, which may have been evaluated up front as part of the original premarket clearance or approval for an AI system. The idea, I believe, is to qualify the company and its quality management systems and also qualify the predetermined change plan that the company will be allowed to use with certain controls so as not to have to make a premarket submission every time it changes. That’s going to be the model. The guidance has different sections to look at the details of what might go into the predetermined change control plans, and this idea of a total product life-cycle approach. It is seen as a positive development, I believe, because it’s not realistic to submit a premarket submission every time the algorithms change.
Real-world performance is another point. Laboratories know how to evaluate these things because they are like lab tests, but the rigor in the laboratory world differs from what it is in some areas of the clinical world where these systems are being contemplated.
Johan Otter, where is Scripps at with digital pathology?
Johan Otter, DPT, assistant VP, Scripps Health, San Diego: We got our scanners before we had our histology lab, which is not the way to do it. We’re looking at three scanners that are not up to speed with what we will need for the future, so we have to update our fleet.
Dr. Otter
We are still far away from adopting it for primary diagnosis everywhere. We’re doing a few sign-outs now. We’ve used it for tumor boards across the system.
From a staffing and real-world application perspective, it’s not as simple as it sounds. Our legal department wants us to maintain slides used for primary diagnosis for at least seven years, which requires an enormous amount of storage and is where a lot of the cost is.
Much of the discussion around digital pathology at the Pathology Informatics Summit had to do with the shortage of pathologists. Pete Dysert, what are your thoughts around digital and its relationship to an adequate pathology workforce?
Peter Dysert, MD, chief, Department of Pathology, Baylor Scott & White Health, Dallas: I’d like to think digital pathology will help with staffing, but it remains to be seen. I’ve heard people say recently that there are a thousand open jobs for pathologists around the country, which is probably the most it has been in a while. In our experience, the people who are being trained tend to want a narrower span of responsibility as far as surgical pathology. Most don’t want to pursue the general pathology role.
I seem to be chasing the growing clinical demands for our subspecialty expertise in the form of accreditation and recognition for Baylor’s clinical programs. Through the American College of Surgeons, there are various program plans for what it will consider a recognized center of expertise and its requirements to have a card-carrying, fellowship-trained pathologist in that arena. Almost all of them translate into more conferences for us.
Dr. Dysert
There are similar clinical aspirations in our system to have a recognized center of expertise in areas like bone and soft tissue. The problem is there’s a point you have to cross to be able to afford someone who will do just bone and soft tissue. In the past a lot of us had an interest in that and grew the expertise while we practiced general surgical pathology and in other areas. Nowadays people come into the market and want to land in that spot.
Between the open jobs, the demographics of new pathologists and what their aspirations are, the continued development of clinical centers of excellence that require our services, and the decline in reimbursement, it’s a tough position for me to be in.
CAP TODAY published an article in the May issue on gene panels for oncology, and it’s an interesting discussion about how payment to pathologists, professionally, has been neglected. Patients, drug companies, and people who sell next-generation sequencing instruments and software are benefiting, but pathologists and their laboratories are left holding the bag. Does that resonate with you, Pete?
Dr. Dysert (Baylor Scott & White): Yes, it does. As our clinical colleagues have incorporated what we represent into their outpatient practices—gastrointestinal and genitourinary, for example—the number of those cases being billed goes up tremendously. Payers must look at pathology reimbursement as a zero-sum game, so they turn around and reduce the reimbursement rates for our traditional services.
There hasn’t been an evolution of billing in pathology to keep up with the new roles we’re playing in the clinical care of patients. The disconnect adds to the financial pressure, along with the big, heavily capitalized entities that tend to present what we represent from a contractual perspective as a commodity.