August 2026—Would artificial intelligence as an aid in ordering laboratory tests and making reports understandable to patients be a worry or a welcome step? In a June 2 online roundtable led by CAP TODAY publisher Bob McGonnagle, a Compass Group member shared what he learned about an Epic pilot, and that launched the following conversation about the clinical laboratory and pathology reports that could be made easier for patients to understand. One participant’s view: “proceed with caution.”
The Compass Group is an organization of not-for-profit IDN system laboratory leaders who collaborate to identify and share best practices and strategies.
Gaurav Sharma, what are you looking for in AI as you look at your system and its needs?
Gaurav Sharma, MD, system vice chair of clinical pathology, division head of regional laboratories, and medical director of the outreach laboratory, Henry Ford Health, Detroit: I serve on our health system’s artificial intelligence committee for the medical group and participate in a regular check-in with hospital physician leaders. At a recent meeting, someone described an Epic pilot that caught my attention. The idea is that when a clinician opens a chart, AI could review the laboratory results, imaging findings, and diagnoses; look for patterns that may be easy to miss; and provide decision support. As the only pathologist on the call, I said, “Tell me more.”
I think this is important because most conversations about laboratory utilization focus on overuse. That matters, but underuse may be an even bigger problem. Order sets do not always capture the full complexity of a patient. Sometimes one or two additional tests can help solve the medical mystery. AI may be able to recognize those situations and guide the clinician.
A lot of AI’s impact on clinical laboratories may therefore happen upstream, when the clinician is deciding what to order. Instead of entering a menu of thousands of tests and trying to choose the right one, clinicians may increasingly have AI helping them navigate the options. I saw a glimpse of that future in this discussion.
The downstream question is different. Patients often receive laboratory results as numbers without much explanation. Could AI provide a simple summary of what the results may mean, rather than just display the values? That is the kind of interpretation a clinical pathologist provides, but AI may make it possible to do this across millions of laboratory results. There are early discussions about using AI to summarize laboratory reports and provide appropriate comments.
Would this be generated by Epic out of an AI-type application in the EHR?
Dr. Sharma (Henry Ford): Yes, as far as I can gather, such a tool will be an Epic-integrated tool. I am interested in whether there will be an automatic ordering of lab tests, which is what we want to avoid.
Diana Kremitske, what is your reaction to the idea that Epic will help patients interpret their laboratory results? It’s a double-edged sword and execution is everything, as was seen with Epic’s sepsis algorithm that proved to be flawed. Many patients want to know what their lab results, and even pathology results, mean for them. How are you coping with that at Geisinger?
Diana Kremitske, MS, MHA, MLS(ASCP), VP laboratory operations, Diagnostic Medicine Institute, Geisinger, Danville, Pa.: My reaction is to proceed with caution. We have to be careful about clinical interpretation by a tool. There could be liability ramifications if it’s clinical interpretation by a tool without a physician overseeing or doing the interpretation. There’s a lawsuit in the state of Pennsylvania involving an AI chatbot tool; the AI tool generated an automated response that appeared to be from a licensed medical professional. Interpreting lab results is to be based on the context of the patient.
There’s room for efficiency, but how we get there will have to be guided by pathology for an understanding of the tools’ pros and cons and where we can use the tools to become more efficient versus using them as a substitute for human performance.
Nkem Okoye, you are in the midst of an Epic install at Northwell. Have you given thought to laboratory mediation on some of the MyChart reporting to patients?
Nkem Okoye, PhD, DABCC, director, clinical chemistry and diagnostic immunology, Northwell Health Laboratories, New York: We’ve experimented with the sepsis detection model from Epic but, as you mentioned, the rate of false alarm is higher than anticipated. Northwell is investing in its in-house models. A couple of investigators at Northwell are developing a clinical deterioration model. It would live in the EHR and look at lab data as well as clinical information and vital signs from the chart to predict patients who would deteriorate clinically. It is not just for sepsis. A couple of Northwell Hospital sites will pilot and refine this in-house model to evaluate its performance relative to the Epic model. The push from Northwell is if we can develop an in-house solution using existing expertise, we should pursue that opportunity.
At the ASCO meeting this year I was struck by how much emphasis companies of all kinds—drug companies, laboratories, and others across the spectrum of diagnostic devices—were putting on patient reporting or understanding. Wally Henricks, does the Cleveland Clinic have an evolving strategy regarding what patient advice is delivered with results? Is commentary offered to help patients understand what they’re looking at?
Walter Henricks, MD, vice chair, Department of Pathology and Laboratory Medicine, and laboratory director, Cleveland Clinic: It is not commentary individualized to a result, and in MyChart we have links to reference material. It is made clear you’re leaving the Cleveland Clinic website to look at this material for an explanation.
Our medical writers have been active in building up reference material on the Clinic’s website, where you can find material on almost any medical condition geared to patient understanding, and there is laboratory content too. You can also search for a number of laboratory tests to find information about a particular test, aimed at patients and the public. We’re emphasizing a patient information portal more than a laboratory result portal. It works well because you get more context that way and there are links to information on conditions patients and others might want to read about.
What do you do for patients on the surgical pathology side, for both secondary consultations in surgical pathology and your primary surgical pathology reports?
Dr. Henricks (Cleveland Clinic): Secondary consultations go back to the pathologist who is requesting the second opinion. Our pathologists issue a personalized letter that includes our interpretation or says if we agree with the provisional interpretation. It gives a description of the workup and any guidance appropriate to that setting or addresses specific comments from the referring pathologist or other physician. That information does not go directly to the patient in those cases.
Our primary surgical pathology reports are available to patients, but they’re written for the ordering physician or provider and we don’t provide an additional non-physician-type explanation. Patients can go to our medical library of online material to look up a diagnosis. We don’t want to get in the way of what their treating physician, or the person who ordered the test or submitted the tissue, is telling them directly. Those relationships are important, especially with regard to interpreting it in the context of what it means to the patient.
Guillermo Martinez-Torres, do you have a comment about what we’re discussing?
Guillermo Martinez-Torres, MD, president and chief physician executive, NorDx, Scarborough, Me.: Dr. Diana Cardona has been tasked by the CAP with the creation of patient-friendly pathology reports. It is something we could implement across all pathology practices and is a tool that would help our patients better understand their individual pathology reports. There’s a lot of nuance within the content of those reports.
Dr. Henricks (Cleveland Clinic): No one would disagree that making the information that is available to patients more understandable or decipherable is a good idea, but every activity comes with effort and a cost. There’s no reimbursement for that, so a pathologist’s time and effort there is uncompensated activity.
Dr. Sharma (Henry Ford): I agree with that and with Dr. Henricks’ earlier point. The less likely model is that pathologists like us will review every case individually and provide patient-level education. That would be difficult to do at scale. We should also prepare for another possibility. AI may first be used in the background, not to interpret the results that were already generated but to identify tests that may not have been ordered. That use case may come first because it addresses a practical problem clinicians already have. Broader interpretation of laboratory and pathology results may follow later.
Dr. Sharma, can you imagine that on a surgical pathology report?
Dr. Sharma (Henry Ford): Yes, I can. For example, a surgical pathology report may contain a diagnosis for which a particular molecular test is considered standard of care. The surgical pathologist may assume the oncologist will order it, while the oncologist may not realize that it still needs to be ordered. AI could run in the background and, perhaps 30 days later, alert both physicians: “This test is recommended for this diagnosis, and it has not yet been ordered.” That could help close an important gap in care.
There’s a shortage of pathologists and many other important people in laboratories, but there’s also a difficult shortage now of oncologists. At the ASCO meeting there must have been 20 or more health systems, all of which had booths in which they were trying to recruit oncologists. And if you talk to oncologists, they’re saying how burned out they are. They’re on call not only for the patient but also the patient’s family, for many hours in the week. It is a big problem in American health care. Jeremy Hart, what do you make of all this as you listen to it?
Jeremy Hart, MD, MBA, assistant vice president, laboratory services, St. Elizabeth Healthcare, Edgewood, Ky.: AI was supposed to bring the human touch back to medicine, and I’m struck by the fact that, even in what we’re talking about here, we’re distancing ourselves further from the patient. That’s a real disappointment for me with all the technology we have. It’s not achieving the human touch.
Also related to AI: In the practice of pathology, AI tends to remove many of the mundane tasks, but as humans, we need mundane tasks. If you remove those, then all you have is complex work, and your brain doesn’t have a chance to recover from that complex work.
Dwayne Breining, what are your views here?
Dwayne Breining, MD, executive director, Northwell Health Laboratories, New York: The trend we’re seeing of being linked more directly to patients is a double-edged sword. You don’t want to get between your clients and their patients, but at the same time the demand is obvious—patients aren’t satisfied with what they’re getting through the usual channels of the massively overworked oncologists, who barely have enough time to speak to the patient. There might be an avenue we can help with here.
At Northwell we’ve been working with radiology, sharing best practices, seeing what they’re doing. They’ve had success working with a company that deciphers the radiology report and puts it in layman’s terms. It’s like an AI chatbot that can simplify it or smarten it up, depending on the audience. We’re talking to that company to see if this could be adapted or created for pathology reports, because all the specialist terminology we use, even though we have gotten pretty good at synoptic reports, is still indecipherable to the lay public. The attraction is it can be a subscription model, it’s not that expensive, and it happens passively in the background. No one has to do extra work. Patients can query and re-query it and go back and forth without burdening our staff. It’s attractive for many reasons if we can get it off the ground.
It strikes me that the close relationship of the patient to the system and patient retention must be a primary goal of anyone in a health care system, and the expert consultation could be invaluable in pursuing that end.
Dr. Breining (Northwell): We’re reevaluating our strategy, which we refresh every couple of years, to figure out the best opportunities, where we want to put our limited resources, but it always comes back to that. One of the targets for success—because we do a lot of business on the open market with the non-Northwell physician population—is when a privateer clinician decides to send lab work to one of the nationals instead of us and the patient is dissatisfied and reprimands them for that decision. There is an opportunity for us to develop that relationship with patients. We have people who reach out to us at the laboratory on a regular basis looking to get an interpretation of lab results. By virtue of being health system and hospital based, we are obligated to be higher touch than those that are working toward business efficiency at the cost of everything else. Maybe we can play to the strengths we have, turn things that have been a liability for us in trying to get the lowest cost per test and contracting and flip that into a positive by building that relationship. But it all comes at a cost. There’s a resourcing equation; you have to do the math and see if it works.
Diana Kremitske, do you have a comment?
Diana Kremitske (Geisinger): It’s an interesting conversation about how you continue to build a relationship with the patient in spite of all this technology. Instead of having a conversation about practice efficiency—many of these tools are put in place for practice efficiency or error reduction—what if we flip the conversation to focus on the patient and how to make the patient’s visit more effective. Dr. Sharma mentioned missing tests. Maybe there is a way to employ AI that helps the patient ask the right questions during the visit, or ask more about their history. How do you empower the patient more with these tools?
Dr. Breining (Northwell): Some peer-reviewed studies were published recently about AI-enabled clinician modules—the model is that a physician assistant or nurse practitioner would use AI for tasks that primary care doctors have traditionally done—and it’s no surprise that AI models outscore the traditional models on outcomes. You have to ask why that is. Physicians, myself included, think we’re smarter than the algorithm sometimes. We may be right four out of five times, but that one time you skip a step in the algorithm, you miss something important and have to go back and redo it.
The AI chatbot-enabled models are outscoring traditional human practice models on things like engagement and empathy. They’re programmed to be engaging and friendly. Maybe we can use this in creating a friendlier patient-to-hospital and patient-to-laboratory interface than we have.
Moira Larsen, what’s your reaction to what you’ve heard?
Moira Larsen, MD, MBA, physician executive director, MedStar Medical Group Pathology, MedStar Health, Columbia, Md.: It is important for patients to have support. With the pathologist shortage and the complexity of the work increasing, I’m not sure I have pathologists who are ready to devote time to building and checking these tools to make sure the AI makes sense for patients. Pathologists want a tool that turns what they say into a synoptic template so they don’t have to go through the clicks, helps ensure they don’t miss key factors in a diagnosis, or does the quantitative analysis of immunohistochemistry for them. I’m trying to identify all the administrative and mundane tasks they do and find other ways to do it and how to support them, because as a whole, pathologists do more of those tasks themselves than any other physician. I need these AI tasks to give my pathologists well-being and help them control their work-life balance and schedules so they’re not working eight to 12 hours a day, six or seven days a week.