Subscribe to The Podcast by KevinMD. Watch on YouTube. Catch up on old episodes!

Internal medicine physician Shiv K. Goel discusses his article “Claude for Healthcare vs. administrative burden: a physician’s review.” Shiv contrasts the two hours he spent fighting a prior authorization with the promise of Anthropic’s new “Claude for Healthcare,” an AI system designed to handle claims and verify coverage in minutes. The conversation explores the race between Anthropic and OpenAI to dominate medical AI and the potential for these tools to liberate physicians from paperwork. Shiv warns, however, that without physician input, these efficiencies could simply be used to increase patient quotas rather than improve care. Discover whether the AI administrator is the solution to burnout or a new threat to the profession.

Partner with me on the KevinMD platform. With over three million monthly readers and half a million social media followers, I give you direct access to the doctors and patients who matter most. Whether you need a sponsored article, email campaign, video interview, or a spot right here on the podcast, I offer the trusted space your brand deserves to be heard. Let’s work together to tell your story.

PARTNER WITH KEVINMD → https://kevinmd.com/influencer

SUBSCRIBE TO THE PODCAST → https://www.kevinmd.com/podcast

RECOMMENDED BY KEVINMD → https://www.kevinmd.com/recommended

Transcript

Kevin Pho: Hi, and welcome to the show. Subscribe at KevinMD.com/podcast. Today we welcome back Shiv K. Goel, internal medicine physician. Today’s KevinMD article is “Claude for health care versus administrative burden: a physician’s review.” Shiv, welcome back to the show.

Shiv K. Goel: Thank you. Thank you for having me, Kevin.

Kevin Pho: So today we are going to talk about Claude for health care. It is the new product from Anthropic, which is one of the large language models. I know ChatGPT also has ChatGPT Health. So they are going into the health care space. Tell us for some context, for those who aren’t familiar, what Claude for health care is and why you decided to write this review on KevinMD.

Shiv K. Goel: Claude for health care is a large language model by Anthropic that has been adapted for health care tasks like summarizing the chart, helping draft documentation, and supporting insurance and revenue cycle workflows such as prioritization and appeals.

In my article “Claude for health care versus administrative burden: a physician’s review,” I walked through how this tool performs when pointed directly at the real bottlenecks physicians face, like eligibility checks, prior authorization narratives, appeal letters, and clerical chart work, rather than abstract AI demos. For example, I take common high-friction admin tasks from our internal medicine practice and evaluate how Claude generally saves time and cognitive load and where it risks adding steps or errors.

I present a scenario in which Claude drafts strong, payer-friendly justifications in seconds. But I also warn that if health systems simply treat this as free labor, they may use it to stack more visits or tasks onto already overextended clinicians and overburdened doctors.

Kevin Pho: So how does it work? For those who are familiar or didn’t watch the demos, you mentioned that example with prior authorization. Tell us about the step-by-step approach about how it worked.

Shiv K. Goel: They haven’t launched the model they have developed yet, but they have announced what we need to do. Before, it used to take hours and hours on the phone with the people at the insurance company to get a prior authorization, whether it was for a procedure or for medications. You kept faxing them details.

Now with Claude, you have all the information on the cloud. You submit it to the system and they are the ones who take care of it in the background. If they need anything, they automatically create everything in it and send it to the insurance companies. Because the large language model has been trained on exactly how each insurance company wants it, and every insurance company has its own criteria that keep changing, they have every possible outcome in mind. So instead of us taking so long and being on the phone, the AI does it in the background. They have a collaboration with the average insurance company, and now a lot of them are using AI assistants. So as long as it meets their criteria and the clinicians are submitting the documents, it creates the authorization in seconds.

Kevin Pho: So Claude would have access to the patient’s chart, or you would have to upload the patient’s chart to Claude. In the background, it will take whatever is relevant from that patient’s chart to process the pre-authorization. It could be for an imaging study or a drug, and automatically send it in the background to the insurance company. Is that having that right?

Shiv K. Goel: Yes. Yes, exactly.

Kevin Pho: Do they have any success rates in the demo or the pilot projects that show how much more successful an AI prior authorization tool would make a clinician?

Shiv K. Goel: Right now, I haven’t personally used it because they just launched it, but it looks very promising. If it can be done, they have to navigate HIPAA laws and regulations, but it seems like an easy process to get it done as long as we are mitigating it through the right tools.

My reason for writing the article is not about whether AI is making our life easy, because it definitely is, or if it is going to replace physicians. I don’t think it can replace physicians. It is about what happens after we implement all those things and our administrative burden is much less so that doctors can work as doctors. A physician’s mind is more valuable than just working on admin stuff. I wonder if that leads to more workload on physicians. Would the hospital systems or the big bureaus want you to see more and more patients because they think now you have more time? Or would they have you do other work? That is something I wanted to prioritize. Are we really thinking about whether it is making our life easy or is it making our life more tough?

Kevin Pho: I think that this topic has come up in the past, especially with those ambient AI scribes. They have been shown to decrease the cognitive burden, and physicians are already so burnt out. But exactly as you said, when it comes to these administrators, they only see the numbers. They see physicians spending less time in a chart and maybe having less cognitive burden, and they are going to think: “Why don’t we just load up on more patients?” It is important and I would say imperative for health care administrators not just to look at the numbers but to realize that their physician workforce is already burnt out. The answer for this increased productivity gain that AI gives them isn’t simply to load them with more patients.

Shiv K. Goel: Yes, exactly. I think this impact will also depend on the governance and the workflow design. It depends on whether physicians really have a voice in implementation. It feels like every tech company, whether it is Claude or ChatGPT, is creating tools for health care and for physicians without them having a say in the system.

The main reason why physicians were burning out is because of so much documentation and admin work while seeing 20 to 30 patients on top of that in order to make a living. Now if AI is there to take away that administrative burden, they should have more time to talk to the patients. This will really help in health care because a lot of times we can hardly spend five minutes with the patients. In the background before and after, we are spending almost 30 to 40 minutes just on the administrative tasks for that patient.

So instead of increasing the load, which doesn’t lead anywhere because patient outcomes suffer. I have worked in the inpatient setting for a long time, around 15 plus years, and have worked in outpatient settings for almost six years now. I have worked in a university academic setting as well, and I have seen it everywhere. It is the same thing. We are doing just a band-aid approach to the health care system. It is most likely a triage system. A patient comes with symptoms like uncontrolled diabetes, and we give them extra insulin or increase the dose. But we never have the time to really address the issue of why their diabetes is uncontrolled or why they have diabetes to begin with. We ask if there is something we can do and then we refer to the primary care physician. We think the primary care physicians are going to take care of the preventative disease, but they are also overburdened. They have hired two, three, or four nurse practitioners who take care of those patients and who basically just keep refilling the same script again and again.

So the main question is how we can navigate this. Why don’t physicians have a voice in creating the systems that are for them? Instead of creating something that is just imposed on us, we should be voicing our opinions in creating it. The second issue is the health care system itself. We must focus on the root cause issues of the problem so patients can really get benefits, which can result in a decreased hospital admission rate, lower ER rates, and decreased morbidity and mortality. If the AI can save us 40 minutes per patient, even if we spend 20 minutes with the patient finding out what is causing the issue, educating the patients, and giving them handouts, it helps. AI can help us generate educational handouts we can give to them. We can not just save 20 minutes for ourselves, but we will also improve the health care system in general. People will benefit and we will not feel overburdened and burned out.

Kevin Pho: We have to be careful that we don’t let what happened to us with electronic medical records happen again. Physicians didn’t have a say, and you had all these systems just imposed on us that didn’t really take into account clinical workflow. I agree with you exactly that physicians need to be involved, especially with this AI boom. So how can we physicians do that? How can physicians get involved with all these new AI tools?

Shiv K. Goel: I think this is where we all have to unite and voice our concerns. I have seen that there are so many companies that are so-called consulting physicians for this. It is not about consulting a physician who is just in a studio or advertising their own supplement company. They need clinicians who are working day and night and who understand the problems that are happening. Development should go side by side. Those tech companies are not going to come to us. They are not coming to us asking what we think and what we want to create. We have to voice it. Until we all stand up and voice it, it is not going to change. Otherwise, we can just sit and have something come to us and direct us what to do.

Kevin Pho: Is there anything that worries you about this increasing reliance on large language models like Claude for health care, for instance, regarding hallucinations or misinformation? Even these insurance companies are using their own AI tools as well, and you are just going to have a fight between AI models in the background.

Shiv K. Goel: Yes, actually. AI should only be used as a tool to provide possible outcomes, not for making the diagnosis. It should not be saying that this is what it is. I recently saw the Trump administration launch AI agents. According to them, in six months they are going to get FDA approval and in three years they would be for cardiovascular diseases. They are not consulting; they are telling us what to do. That is where a real problem lies. We have seen recently so many cases where a teenager killed himself because he asked the AI how to kill himself. He fell so much in love with that AI that he felt heartbroken. There are so many other cases that have been happening.

AI is not a human. AI is basically collecting all that data and providing it to you. It is a good tool that can make your administrative work or reading work or creating scripts for many things easy for you. We need to use it as a helper. AI cannot replace physicians. People have a fear inside them. The more fear they have, definitely it is going to replace the ones who have the fear. It cannot, and we shouldn’t even let this happen. AI is not a human. It can’t read the emotional healing part of the patient. A lot of times it happens when a patient walks into the room, or I walk into a patient’s room, it is not just what the patient is saying, it is what the whole picture is telling me. The space in between the silence and the patient’s emotional resonance give me more ideas about what might be the real problem. If your patient says: “Yes, I am OK,” but the whole response sounds like “I am not really OK, but I am saying I am OK,” that is telling.

There are a lot of things that we have to consider. Plus, it is not always about symptoms. It is more about how we understand the patient. A physician who has spent 20 plus years seeing thousands and thousands of patients from different cultural backgrounds, different ethnicities, and different takes on life understands their belief systems. We understand how and why they do things, why things affect them, and why things don’t affect them. Ten different people with the same symptoms will have ten different diagnoses. That is where clinician judgment is needed to see what exactly I should be doing and if I should be prescribing this patient.

Those large language models shouldn’t be touched by technology companies all the time because they could change the whole system. That is the point. If we create AI as agents to implement things, that means we are telling the doctors what to do. If we don’t voice our concerns, then those AI agents will be telling us what to do. If the doctors feel that we are too busy with our work to make a living, then we will be working under an AI. At the end, who will be suffering? The patients. People are the ones who are going to suffer because it is a revolution that is happening. The race is not between two AIs. The race is between who will be the next AI God and who will control health care. Because health care is everything. Health is related to your mind, body, and even spiritual health, your social health, and your emotional health. If we take up mental health, psychiatry is now going into AI as well. So while at the same time people feel excited, like I felt so excited that day when I heard a tool could get prior authorizations done in seconds, it also made me feel that we need to wait a second and ask what is going to happen then.

Kevin Pho: We are talking to Shiv K. Goel, internal medicine physician. Today’s KevinMD article is “Claude for health care versus administrative burden: a physician’s review.” Shiv, let’s end with some take-home messages that you want to leave with the KevinMD audience.

Shiv K. Goel: Yes. AI like Claude for health care will not fix medicine by itself, but it can be a powerful tool if physicians stay at the center of how it is designed and deployed. The goal isn’t to make us faster typists. It is to remove the administrative noise that keeps us from thinking, listening, and connecting with patients. If we pair these tools with thoughtful leadership and strong boundaries, we can claim time, reduce burnout, and build a care system that serves both patients and clinicians far better than what we are living with now. But we have to stand up and voice our opinions. Otherwise, the same system that is designed for us will be designing us.

Kevin Pho: Shiv, as always, thank you so much for sharing your perspective and insight. Thanks again for coming back on the show.

Shiv K. Goel: Thank you for having me, Kevin.