The Informed Patient in the Age of AI

Participants:

• Dhruti Ramchandani, City of Hope
• Laura Beilke, Triage Cancer
• Lindsey Whyte, Leukemia Research Foundation

 

Welcome and thank you for joining us for the second webinar in our Leukemia Awareness Month series. Today's topic is Informed Patient in the Age of AI. My name is Lindsey Whyte and I'm the Director of Programs and Partnerships at the Leukemia Research Foundation. We're glad you're here.

 

In a moment, I'll introduce our speakers, Dhruti Ramchandani and Laura Beilke. Before we do that, we'd like to acknowledge the supporters of our program and the Leukemia Awareness Month series. They are our 2026 advocate partner, Autolus, our series supporters, Cycle Pharma and Johnson & Johnson, and our sustaining supporters, AbbVie, Astellas, Merck, Novartis, and Syndax. Thank you all for your generous support.

 

The Leukemia Research Foundation's mission is to cure leukemia through innovative research funding and to support patients and families. The Foundation was established in 1946 and has raised over $120 million in support of our mission since that time. We have funded research grants to over 800 investigators worldwide. Our support programs for leukemia patients and their loved ones include information and resources, education programs, financial assistance, and a directory of other helpful organizations and links on our website.

 

For today's program, all participants will be muted, but we will welcome your questions in the Q&A box at the bottom of your Zoom screen. Please note that you should check the box for anonymous if you prefer your question to be anonymous. Otherwise, your name may show. If you already submitted a question at registration, please know that we have your questions and we will do our best to cover as many as we can.

 

After today's program, you will be sent a brief evaluation through email. Please take a moment to complete the evaluation so we can improve future programs. Also, this webinar will be recorded and a link to it will be sent to all registrants for anyone who needs to leave early or would like to watch the replay with captioning and a transcript on their own time. Please note that AI scribes and automated note-taking tools, which we'll be talking about later, will be removed from this session.

 

On that note, I'm going to share a few of the learning objectives for today's webinar. If you were hoping to learn about treatments, specific treatments for your form of leukemia, I highly recommend that you register for our next series, which is Leukemia Q&A, which takes place in October. And I'll be sending a link to register for those programs after today's program. But in the meantime, today's program, as I mentioned, focuses on AI in healthcare, and the objectives are for participants to learn where AI may be used in leukemia care, understand their privacy rights, and how to protect your personal information, how to understand whether online health information is trustworthy, to learn some tips about how to ask your care team important questions, how to use AI and digital tools more safely, and to be a better advocate for yourself or your loved one.

 

So we are grateful today to have our speakers, Laura and Dhruti. Laura Beilke is a staff attorney at Triage Cancer. In this role, Laura ensures that people facing cancer have the information and resources they need to make informed decisions about their care, their work, and financial wellbeing. Specifically, Laura contributes to the legal and financial navigation program, helping individuals diagnosed with serious or chronic medical conditions and caregivers navigate complex systems related to health insurance, employment, and disability benefits. She also develops and updates educational materials on key topics impacting the cancer community. As a frequent speaker at national and regional events, Laura delivers trainings to both the cancer community and healthcare professionals, translating legal concepts into practical tools for everyday use. Laura earned her Bachelor of Arts in neurobiology and psychology from the University of Wisconsin Madison and her Juris Doctor from Brooklyn Law School and is a member of the State Bar of Wisconsin.

 

-Dhruti Ramchandani is the program manager in the Department of Supportive Care Medicine at City of Hope in Duarte, California, where she leads cross-functional initiatives to strengthen patient and caregiver education, enhance health literacy practices, and improve supportive services across a comprehensive cancer center. She brings nearly two decades of experience designing, implementing, and evaluating health education, wellness, and disease prevention programs across diverse clinical settings. Dhruti's academic training includes a Master of Aging Services Management from the University of Southern California's Leonard Davis School of Gerontology and a Master of Public Health in Sociomedical Sciences from Columbia University. She also holds a Bachelor of Science in Business Administration from the University of California Berkeley and a Master Certified Health Education Specialist. In addition, she is a certified meditation and mindfulness instructor. In her role at City of Hope, Dhruti partners with interdisciplinary clinical teams to integrate evidence-based educational practices, develop population-specific interventions, and continually further the impact of patient and caregiver education on patient care. Her work spans across a range of disciplines, including geriatric oncology, mental health, caregiving, advanced care planning, pain management, and digital education. She routinely provides professional training. Her contributions have been presented nationally and internationally, and she has chaired and served on committees focused on health literacy, inclusion, and patient education excellence.

 

Ladies, thank you so much for joining us today. Please, let's get things started now talking about managing through this age of AI.

 

- Thank you, Lindsey, for that wonderful introduction and for having me here today. Just making sure my screen is up and clear.

 

- Yes.

 

- Okay. Thank you. So at this point, with so much going on in AI, I feel we're really learning together, but I'm glad to be here to share some of what I'm seeing in the field, seeing in healthcare, and hope you'll feel a little more confident in navigating through it.

 

So let's start with a couple of basics and one question, which is what do we actually mean when we say AI? At its simplest, artificial intelligence is a technology that looks at a lot of data, it spots patterns, and it uses those patterns to help with the task or to produce something. So that task might be predicting something, it could be sorting information, it could be recommending something, summarizing large data sets, or even creating something brand new. So for example, the first draft of this presentation was generated using AI, Copilot specifically. And on the slide, you can see four everyday examples in terms of finding patterns, think of software that scans thousands of images or lab results, and then it'll flag the ones that look unusual, that stand out in some way so that an expert can take a closer look at those.

 

In terms of predicting, an AI tool might estimate someone's risk of something, kind of like weather forecasting, but for health. And with summarizing, an AI scribe, which we'll talk a little bit about later, might listen to a visit and draft a note which the clinician can then review and update as they need to. Your provider may have already asked you in a clinic visit if you're okay with them recording your visit, and this is probably why.

 

And in terms of creating, that's the chat tool that writes a paragraph or an image when you type in that request, "Can you please draft this email?" "Can you please put together a flyer?" So you probably already use AI without thinking about it. When your phone suggests the next word, when you're drafting a text message or when your streaming service like Netflix or Hulu recommends a show, that is all AI noticing patterns of what you like to watch. So here's the most important part though, and it's really, it's at the bottom of this slide and it's the key message to remember. It's that AI is a tool. It can be fast, it can be genuinely helpful, but it doesn't know your full medical story. It doesn't know your goals or what matters to you. So you always want to remember that your care team is learning this and you know this. You're the expert in your, in your care in what you want. So as we go through today, just think of AI as a helper, as your personal assistant, but not as a decision maker for you.

 

And if you haven't already, you're likely to hear a handful of AI terms in the news and maybe in your doctor's office. So let's just quickly build a small vocabulary together and don't worry about remembering it, it's just really so we can get a little more comfortable with the terminology around it. So machine learning is AI that learns from examples. Imagine showing a program thousands of photos that are labeled cat or labeled dog until it can start to tell them apart on its own. In healthcare, these examples, these photos may be images or they may be components of lab data. Generative AI is AI that creates something new. So we're talking about new emails, drafting those emails, drafting this PowerPoint presentation, images, even audio or summarizing. When someone says they asked a chatbot to write an email, that is exactly what this is, generative AI.

 

A large language model is the technology behind many of those chat tools. It works by predicting the most likely next words in a response, and that's why it can sound actually pretty natural sometimes. An AI assistant is any tool that helps with tasks, questions, or finding your way around. That could also be like a chatbot on a hospital website that helps you schedule an appointment or an AI scribe that's embedded into the electronic health record that your provider has up. It's listening to the conversation with your provider and it's drafting the visit note which the clinician can then review. And the one I really want you to remember in terms of safety around AI is the term hallucination. You've likely heard it before. That is when AI gives an answer that sounds completely believable but is wrong or even made up. For example, you might ask a chatbot to show you studies on a treatment and get a nice list of articles, and some of them might not even exist. And the tool isn't lying on purpose. It's not a problem. It's not doing something wrong. It just isn't built out yet that way. So it's predicting things for you, things that sound right, things that line up with what you're looking for. So keeping in mind with, keeping this issue in mind with AI and remembering sounding confident is not the same as being correct. You may know people like this. AI is just like this. It sounds very confident, but we need to double check, fact check, especially anything that is health related.

 

And you might hear people describe an AI tool as open source or closed source, so we'll quickly unpack what that means because the labels can be confusing. So open source means some parts of the AI model are shared publicly, so other people can examine it, look at it, or run it on their own system. So think of it like a patient education booklet on leukemia. It's freely available to everyone. Different hospitals can use it, update it, or adapt it for their patients. The great part is that many people can learn from it and make it better. The downside is that the quality and accuracy might vary depending on who makes changes and how carefully those updates are reviewed.

 

On the other hand, closed source means a company or organization keeps the details of that AI model private and manages it themselves internally. So that would look like educational materials that are created, maintained, and updated by a single cancer center. The organization is responsible for keeping information current, accurate, and safe to use. The benefit here is you're going to find greater consistency and there's accountability for that information. And the trade-off is that everyone can't fully see how the backend system was developed or updated.

 

And the bottom line for this is that neither one is automatically safer or better. The label alone doesn't tell you whether a tool is accurate, whether it keeps your information private, or whether it's right for your situation. So as a patient, it's always important to ask a couple of questions thinking about is the information accurate? Is my information protected? On the side, there's a note that data submitted goes into their system. So thinking about how is that going to be used, who is it going to, and is it, is it appropriate for what you're using it for? Those questions will matter far more than the label of open source or closed source.

 

And now let's talk about where you might actually run into AI in healthcare. The short answer is maybe before, during, and after a visit, and sometimes you might not even see it. So going through the six areas on the slide, patient access, this is sort of the front door. Think about appointment reminders, a phone system that routes your call, or a portal message that gets sorted to the right team. And then during visits, a clinician might use an AI scribe that gets, that gets all of the information and drafts the visit note so that the clinician, your provider, can better engage with you instead of looking at a keyboard. I've personally noticed that transition in my healthcare. And it could also be a tool that helps with translation or captioning, and we've used that on patient education tools.

 

In regards to clinical information, there may be software that can help spot patterns and lab results, pathology slides or imaging. It could flag something for a specialist to review. And keep in mind that AI is not making the decisions for your healthcare. It's supporting how your clinician can review the information that's in front of them. In research, AI can sift through huge data sets and help match people with possible clinical trials, operations. A lot of AI works quietly behind the scenes on things like prior authorizations, insurance claims. And in patient education, there are tools being developed that can explain medical terminology. You can leverage chatbots to help you prepare questions or organize information. And one important caveat, and it's on the slide, is that not every organization uses every type of AI. So your hospital might use some of these, all of them, or maybe even none of them. So if you're curious, it's completely fair to ask your care team what they're using.

 

So let's zoom in on one place where a lot of healthcare AI is showing up in the electronic health record. If you've ever logged into your patient portal to see test results or send a message, you've seen a small window into the EHR. This is the digital record that your healthcare uses to document or manage your health information. So we're talking about visit notes, lab results, medications, images and messages. More and more, AI is being built right into these systems, which is why some people are calling this a front door again.

 

And here's how it tends to work following the three steps that are on this slide. So step one, information is coming in. That could be a portal message that you send, a note from your last visit, new lab results, or some type of image. Step two, AI may jump in and assist. For example, it might sort incoming messages so that urgent ones get seen faster. It might summarize a long chart documentation before an appointment to support the provider in being prepared for your visit, or it could draft a reply to a portal message. And then step three, and again, a key part here is that a person is going to review it. Healthcare professionals stay responsible for decisions and information about your care. So if you ever get a portal reply and wonder whether a computer AI tool wrote it, you're allowed to ask that. A simple question like, do you use any AI tools or who looks over what AI generates? This is perfectly reasonable. And that's what this final statement on the slide is. Ask what the tool does and who reviews the result.

 

And here we have five questions you can ask at any time that you feel AI might be playing a role in your care. You don't need any technical background to ask them. The responses should be clear. The first one is, what is the AI tool being used to do? Is it drafting a note, reading an image, sorting messages? Knowing its purpose can help a lot. The second one, what information does it use? Is it looking at your lab results? Is it looking at visit notes? Is it following your conversation with your doctor? Third is how was it tested? Did it include people like me? And this one is important because a tool that worked well in one group of patients might not work as well in others. So it's fair to ask whether it was tested with people of your age, background, or diagnosis. The fourth question is who reviews the result before it affects my care? You want to hear that a qualified person, an expert in that area is reviewing it. And number five, what happens if the result is incomplete or wrong? Every tool can make mistakes. So a good answer here will tell you that there's a backup plan, there's a safety net in place, and most likely that's going to be the expert that's reviewing the information. And if you feel that these questions are not framed in a way you'd feel comfortable sharing them, you can always do some variation of them. It could be something like, "I noticed my visit summary was written with the help from an AI tool. Can you tell me what it does and who's checking it?" It's short and friendly, perfectly reasonable again, and these questions help you build that awareness and keep up accountability around AI use. These questions will also cover the purpose, the data, the evidence, that human review component, and the safety net. And just a side note is that these are not official questions from any organization. They're a practical patient advocacy framework, and they're based on common recommendations from healthcare AI experts, patient trust research, and responsible AI guidance. So these are just put together as a framework to support your learning and your advocacy as you are encountering AI in your healthcare.

 

So that leads to the question, should you use AI tools yourself? AI can be genuinely helpful for getting ready and understanding things, but it definitely should not be making any medical decisions for you. So looking at the left side of the slide first, ways AI may help you. It can explain a term in plain language. Say you see the word neutropenia for the first time on your lab report. You could ask a chatbot what that means in everyday words. It'll summarize a handout for you. If you get something from your provider, if you get something in your portal, you can cut and paste that in and ask it to create a summary. It can also organize your own notes. So if you are taking notes, whether they are written or typed up, you can plug them into this AI chatbot and ask it to organize them and prioritize based on importance and see what it comes back to you with. AI can also help you draft questions for your next visit. Again, if you have concerns and symptoms, if you're plugging some of that into a list, it can help you prioritize and it can help you compare information from trusted sources. So we'll talk a little about trusted sources in a minute. Then moving over to the right side, what AI should not do. It shouldn't diagnose you on its own. It shouldn't replace your leukemia care team or override your treatment plan in any way, and it should never be the reason that you delay care. If you have a fever or new symptoms and a chatbot tells you this is probably nothing, please call your care team anyway. Following the guidance that you receive from your care team is always the right thing to do. Also, be careful about sharing identifying health details. Unless a tool has been approved for that kind of use, perhaps one that your hospital provides, don't type in your name, your birthdate, or medical record number. A great habit on the bottom of the slide is to turn AI answers into questions for your healthcare team. So for example, instead of the chatbot said I should switch medications, you can ask your care team. You can say, "Well, I read that some people switch medications to support X, Y, and Z symptom issues that I'm having. Is this something that you might consider for me?" So it can inform you enough to pose the question, but not overriding the expertise of your healthcare professional.

 

And whether information comes from a website, from social media, from any app or an AI chatbot, here are five checks to run before you rely on it. So the first check is the source. So here we're really talking about trustworthy information and verifying that you're not getting hallucinations or misinformation from uncredentialed sources. So looking at the source who created or reviewed this content, a page from the National Cancer Institute or a major cancer center is different, very different from an anonymous post in a comment thread. And open source AI is going to pull from all of these, and that's where the hallucinations can happen. Second is the evidence. Does it point to research, clinical guidelines, or qualified experts? If a post promises a miracle cure but can't point to any studies from an academic organization or research journal, then that's a big red flag. The third is the date. Is it current enough for your question? Leukemia treatment has changed in recent years and an article from 10 years ago may be out of date. So this is important even within a trustworthy organization. Some nonprofits just don't have the staffing to keep online information current. The fourth one is privacy. What information is the app or site collecting and who is it shared with? It's worth a quick look at the privacy settings before you enter anything personal if you want to or feel that you need to at all. Rule of thumb is do not put any private information into an AI chatbot. The fifth is fit. Does it actually apply to your diagnosis and your situation? Information can be accurate and trustworthy and still not be right for you. For example, you may get advice on one type of leukemia which may not apply to another. So generally, trustworthy sources are going to include government health agencies, hospitals, universities, professional medical societies, and peer reviewed journals, the National Cancer Institute and Medline Plus both have helpful guides on this. When in doubt, bring it to your care team. They would much rather talk through it with you than have you worry alone or have you sitting with the wrong information. The bulk of my professional work is in health literacy, which is being able to find, understand, and use health information to make decisions that are right for you. And this is one area where AI can be a very helpful sidekick.

 

So as long as you stay in the driver's seat, let's look at four ways that it can help in terms of health literacy. And here's some, there's some sample prompts that you can use as well. So to help understand something, you could plug in explain X term in plain language. Using the word plain language is going to activate the tool to put medical terminology into easy to understand language. If you don't understand it, you can write back to it again and say, "I still don't understand. Please break it down further. Give me an example of this," and it will help you to engage with whatever information you're getting from your clinical care team. You might paste a word in from a pathology report, ask what it means and, or ask it to explain it as if you were telling a friend about your specific issue. It can help you prepare, so help me write questions for my next visit. You could say, "I'm starting a new treatment next week. What questions should I ask my doctor about side effects?" It can help you organize. "Put all my notes into a short checklist." So after a busy appointment, you could type in your scribbled notes and you could ask for a simple to-do list, and then be sure to check. So what sources support this answer? This one is, again, very important. If a tool can't point to trustworthy sources, treat the answer with caution and remember that some tools may even make up the sources, so just quick verification. One big safety reminder at the bottom. Again, I've said it a couple of times already, but I think it can't be said enough. Don't include names, birth dates, MRNs (medical record number), or other identifying details into a public AI tool. You can still get very helpful answers by keeping it general, like I'm a 60-year-old with this diagnosis and I'm looking for X, Y, or Z information.

 

And then I'll just touch on advocacy for a minute here. And I'll also just be very clear that advocating for yourself doesn't mean that you're challenging or pushing back on your doctor's expertise. It really means that you're being an active partner in your care. You're asking questions, you're sharing what matters to you. You're making sure that whatever plan is set forth makes sense for you and makes sense to you. So here's a practical before, during, and after approach. Before the visit, take a look at your portal or any handouts that you've received. Write down your top three questions because in the moment they're so easy to forget. This is a great place to use AI to help you organize your thoughts. Appointments can go quickly and staying focused can be key. And if it helps, you can bring a support person. If you have someone that can join you, a family member or a friend who can take notes or be a second set of ears, that can be incredibly helpful. During the visit, it's okay to say, "Can you explain that in plain language?" Ask what happens next, such as when will you get results back or when is your next appointment going to be? And you might not have heard of Teach-Back. It's something we do a lot on the staff side in terms of helping make sure that the care team is explaining things clearly and that the patient and caregiver are understanding. But in the reverse role, you can very easily just ask your provider, "Can I just share that back to you in my own words to make sure I understood?" So for example, letting them know they've explained a medication regimen to you. "I'll take this medication twice a day with food and call you if I get a fever over 100.4 degrees. Did I get that right?" And just doing a crosscheck of information.

 

And after the visit, reviewing the visit summary in your portal or the printout, and if an instruction is unclear, just sending a message, ask your care team. Keep track of your symptoms and know exactly who to contact, including after hours. And the closing message on this slide is the one I hope you'll take home, is that you don't need to be an AI expert to be a strong advocate for yourself and for your care.

 

Lastly, I will go over shared decision making. This is when you and your care team really collaborate and make decisions together, bringing their medical expertise and intertwine them fully with your values and priorities. So AI can help you prepare for that conversation. And of course it can't have it for you, but here are four areas to think about and you can use them as a simple framework. So thinking about choices, what options are available? Sometimes there's more than one path, including watching and waiting. There's a standard treatment or a clinical trial. Consider benefits. How might each option help? For example, one option might aim for a longer remission while another might be easier to manage day-to-day. Consider the risks, what side effects or downsides should I know about? And that could be anything from fatigue to more hospital visits or needing to take time away from work. And remember the fit. How does each option fit my goals and daily life? Maybe you care a lot about being able to keep working, traveling to see family, grandkids maybe, or limiting trips to the clinic. Each of those things matter and your care team needs to hear them.

 

So here's a practical example. You could ask an AI tool, help me list questions about the benefits and risks of treatment options for my diagnosis, and then you could use that tool, you could use that list and bring it to your appointment. You could even plug in things, all the things that are important to you and say, compare this with what will happen in my day-to-day life if I choose this specific treatment and see how that balances out. It can help set you up to have this conversation. It'll help you get organized, but all decisions will happen with between you and your care team. So as the slide says, your values and your care team's clinical judgment belong at the center. Now I will hand it off to Laura to share the legal and financial perspective and take all of this to the next level.

 

- Thank you.

 

- And thank you. While Laura brings up her slides, I just want to remind participants that if you have questions, please enter them in the chat. Or sorry, not the chat, the Q&A box at the bottom. I see we have a couple of questions that have come in, but hopefully we'll have a few minutes at the end of this to go through the questions. Thank you.

 

- Thanks everyone. And again, I am Laura. I'm a staff attorney at Triage Cancer. And just to remind you all, we provide free education on legal and practical issues that may impact individuals diagnosed with cancer and their caregivers. And we have a whole host of resources on these topics, whether written materials, educational events. You can also speak with our staff attorneys one-on-one, and all of that information can be found on our website at triagecancer.org.

 

You may or may not have realized, but AI already touches your care before you even see your doctor. We know that providers and hospital systems are using AI to answer the phone, helping with booking appointments and routing you to the right person. When it comes to medication refills, AI is also being used. Refill requests make up a significant number of calls into a lot of primary care offices, and AI is now handling much of that, collecting your medication and pharmacy info, then sending it to a nurse or doctor to approve. One state's even gone farther, Utah just became the first to let AI approve certain routine refills on its own, only for stable long-term medications like blood pressure pills with a doctor able to override it. It is early and narrow, but it's a sign of where things may be headed.

 

You may have also noticed your doctor typing less during visits. That's often AI listening in the room and drafting the note afterwards. But in all of this, AI handling speed and paperwork, they're not handling the decisions. A person is still deciding your care. I also think it's worth asking your doctor if you can use an AI note taker during visits for your own records that can help you stay more present during your visit and not worried about trying to take all the notes and let you have a clear record of what was talked about to review in the future or even possibly share with loved ones. But you should always ask for consent before using that AI tool because there are state laws about recording visits without the consent.

 

When we think about using AI in your actual medical care, radiology and pathology are probably where it's furthest along. There are places where they're using AI to help review scans or slides. It's always working alongside the radiologist or pathologists, not instead of them. In one of the bigger studies on this, AI-assisted mammogram screenings actually caught more cancers than the usual approach of two radiologists reviewing independently. There's also work happening around catching problems early. Blood cancer patients are especially vulnerable to infection, and some hospitals now use AI to watch vital signs and lab results continuously looking for those early signs of sepsis hours before a care team might otherwise notice. For example, at John Hopkins, that kind of system was tied to a real drop in deaths when the alerts got acted on fast.

 

And then there's dosing. This is newer and only happening in some very small pilot studies, but instead of giving everyone the standard dose of a drug, AI helps adjust the dose based on how a specific patient is actually responding. Same idea with treatment selection, using a tumor's genetic profile to help figure out which therapy is most likely to work for that particular cancer. Here's the thing though that I want you to walk away with. And none of this means better just because it's newer or it's been tried somewhere else. One of the most widely used AI tools for predicting sepsis looked great in its original hospital and then performed a lot worse once it was tested somewhere else. So that's not a reason to dismiss any of this. It just means that this is still an evolving tool within the healthcare space.

 

AI is also being used in research. It's helping match patients to clinical trials, and that used to be done by hand. And honestly, a lot of eligible patients got missed simply because no one had time to check. So once a trial is actually running, AI is also being used to look at the results, potentially catching patterns or safety signals that might take a research team months to find on their own. AI is helping to analyze patient data in aggregate, tracking outcomes, measuring quality, spotting trends across thousands of records at once.

 

Now let's talk about the usage of AI that probably touches your life the most, insurance. AI has primarily worked its way into three parts of our health insurance process. Before you get a treatment or a test approved, there's a decent chance AI is reviewing that prior authorization request. Once care happens and a claim gets submitted, AI is often part of processing that too. And if something gets denied and you appeal, AI is increasingly involved in that review as well. This isn't a small trend either. A survey covering insurers across 16 states found that 84% of them are already using AI somewhere in the prior authorization process, which tells you this is close to the norm at this point, not some experimental edge case. Many people are concerned about how this use of AI is impacting clinical care. Six in 10 physicians surveyed said they're worried AI is behind a rise in denials they're seeing for their own patients. There's also a lawsuit working through federal court right now that alleges that an insurer used an AI model to override doctors' own treatment decisions, and that model has a 90% error rate. But I do want to be careful with that number since that is an allegation made in an active lawsuit, so not something a court has confirmed, but it really highlights a challenge patients are facing as a result of AI.

 

Several states, including New York and California, have tried to be more transparent and add guardrails into this situation, and they now require insurers to disclose when AI was involved in a denial decision, and some require an independent human specialist to review an AI influence medical necessity denial before it's finalized.

 

Now, the good news is that even if a patient is denied a prior authorization or a claim, they have the right to appeal that denial. The details do differ based on the type of insurance, but at a minimum, in almost all types of insurance that there are, someone would have two chances to appeal. The first is an internal appeal where they go back to the insurance company and ask them to reconsider. Some people are successful at this stage. However, there is also a high chance that AI is being used here as well. If someone is denied at the internal stage, then they have the ability to file what is called an external appeal. And this is where you are going to an independent entity that is completely separate from your insurance company and they look at the information and decide if the care is medically necessary and if the insurance company should cover it. And what is so great about the external appeal is that it is binding on the insurance company. So if that external review that is completely independent from your health insurance company, if they decide that your care should have been covered by your health insurance, your health insurance now has to cover it.

 

Now, this does put a lot of things on people's plates, especially at a time when they may not be feeling the best, but we really think it's worth it to go through the appeals process. And the reason I say that is that when someone files the external appeal, between 40 to 60% of the time, they ultimately end up winning. The challenge is that people don't know about this process. Of the 48 million claims that were denied in the marketplace in a single year, 99.9% of those denials were never appealed, even to the first stage. But if you remember, I just told you that when people do file the external appeal, they win roughly half the time. So knowing the appeals process and educating people around it is really a key priority for us, but I also don't want to minimize that there are a number of hurdles and some of those hurdles can actually be cleared with the help of AI.

 

So with the appeals process, especially when we're talking about a prior authorization denial, there are deadlines that come with it, but there are ways to expedite actually hearing a decision from your insurance company, and that is by filing an expedited or urgent appeal. This is the fastest way to challenge a denial when waiting for the usual process could harm the patient's health. And getting information from the provider and potentially the help of AI to show the need for the urgent appeal could end up getting someone access to care that they need and not having to delay that care. The key to overturning these denials is evidence, clear, strong proof that the care is essential for the patient's health. So in other words, that the care is medically necessary. And most of the time, patients aren't going to be able to prove that without the help of their healthcare team. So it is very helpful when the healthcare teams and the patients work together to gather this evidence. And the evidence, as you can see, can be pretty wide ranging.

 

So maybe it's a letter from the healthcare provider explaining why the prescribed drug, or it could be a medical records or test results. If insurance has been covering the service previously, that documentation could also be helpful. Maybe it's a second opinion from an expert. It could include a personal narrative from the patient themselves, or maybe it's literature and study results. So anything that can support why a particular service is medically necessary needs to be gathered, and AI might be very useful in gathering some of this information. Some people may not have the support of their healthcare teams and may need to do their appeal on their own and find themselves in the position of having to find this evidence of medical necessity on their own. One of the challenges can be that some journals or other sources require a subscription or a payment to access that information, but there are credible resources that can help support a medical necessity appeal that are free to access. For example, PubMed is a free online database from the US National Library of Medicine that lets you search millions of citations and abstracts from biomedical and life science journals. It's a trusted source for finding peer-reviewed research, clinical studies, and reviews on health medicine and related fields with links to full texts and articles when available. And I've included some other options or ways that someone could find the information needed to show why something is medically necessary or maybe not considered experimental or investigational.

 

And if using AI to help gather this evidence, you can be clear about pulling information only from the legitimate sources like these and providing citations to verify. Another hurdle in this process is something called a peer review. And peer review is a process during the internal review stage where an insurance company will have a healthcare provider talk with the prescribing provider. So the health insurance company doctor is going to talk to my oncologist. They chat for a little bit, and there's really a couple of challenges here. The first is that many times the things that are being appealed are highly specialized medicine, and oftentimes the doctor at the insurance company is being put up against these very highly specialized providers are not the same specialty. And so it can be really helpful for your provider to ask to speak to another oncologist, for example.

 

There is also a challenge in that people are confused and believe that the peer review is the same thing as an external review, and it is absolutely not the same. So even if you've done a peer review and an internal appeal, you still have the right to that external appeal. And where AI might be helpful at this stage is we often hear that appeals is a time commitment from providers as well as the patients. So potentially AI could help the doctor prepare for the call. It can pull together the specific clinical criteria that the insurer applied, match your chart against those criteria, and serve as the medical guidelines that back up your treatment. It's, again, never substituting the clinical expertise of the provider, but it's just helping pull the necessary evidence together faster.

 

At Triage Cancer, we also have an AI tool available called the Triage Health Appeals Navigator, which is an AI tool that will write your appeal letter for you. This is a tool that can be used by the patients or anyone who is helping someone with their appeal. Healthcare professionals can also use this tool to help draft appeals letters on behalf of the patients. Soon that's also going to be coming with the Triage Health Appeals Navigator is it's going to provide reminders of when these deadlines in the appeals process are coming up, because if someone misses out on one of those deadlines, they may not have any more opportunities to appeal that decision. So staying on top of deadlines is extremely important, and soon this tool will be able to be a reminder for those deadlines. We also have coming soon a state-specific appeals toolkit, which will explain the steps to appeal depending on the type of plan that you have and the state that you live in.

 

But it's not just in appeals that we at Triage Cancer have embraced AI as a way to get people to the information and answers that they need faster. Next month, we will be officially launching our latest free resource, which is Lane. Lane is an AI-powered chat bot that is trained on our Triage Cancer material and resources that we have created. So there are guardrails around what information is providing, what is provided to you. So if you ask it a question, it's only going to answer with information that we have already created and reviewed. If you asked a question beyond our scope of expertise, Lane would actually provide that to you and tell you that we don't have or we might not be the best resource to get you this information. Where if you went maybe on, let's say, ChatGPT where it's pulling information from all across the world, you don't know where that information is coming from, and it may do one of those hallucinations that was talked about earlier where it just creates an answer thinking that is the best answer for you. So that is where it is important to know the distinction of what tool you are using and where that information is being gathered from. Again, though, Lane does not replace us or a legal and financial navigation program, but it just compliments it. So if you wanted to check out Lane, it is already live on our website if you want to have a chat.

 

Everything we've talked about today does come down to one idea, trust but verify. AI is already part of your care in scheduling, in documentation, in research, and especially in your insurance. Some of it is genuinely helpful and some of it may be not. I think the point isn't to be suspicious of every tool your care team or your insurer uses, it's to know it's there and to ask about it when it matters. That means asking your care team directly whether AI played a role in your visit notes, your treatment recommendations, or a coverage decision. It means knowing that if you get a denial, you have the right to ask whether AI was involved, and in a growing number of states, your insurer actually has to tell you. And it means remembering that most people who get denied never appeal, even though appeals win far more often than people expect. It also means being thoughtful about the tools you choose to use yourself.

 

If you take a lab report or a doctor's note and you put it into something like ChatGPT to help you understand, it's worth knowing that the tool isn't bound by the same privacy protections your doctor or your insurer is. Read what you're agreeing to, ask your team if they offer something built for this instead. You don't have to become an expert in how AI works to protect yourself. You just have to know it's part of the process now. Ask a few direct questions and use the resources built to help you navigate it, whether it's your care team, a patient navigator, or even a tool like Lane.

 

You've probably heard the word HIPAA before, but let's actually define it. HIPAA stands for the Health Insurance Portability and Accountability Act, and it's the federal law that protects the privacy and security of your health information. But HIPAA doesn't protect health information everywhere it goes. It only applies to what's called a covered entity and their business associates. So covered entities are your doctor, your hospital, your health insurer, and similar organizations directly involved in your care or coverage. A business associate is any vendor those covered entities work with to handle your information on their behalf, like a billing company or a scribe tool your doctor's office contracts with directly.

 

Here's the part that catches people off guard. HIPAA does not extend to a tool you choose to use yourself. So if you take a lab report and you paste it into ChatGPT or a similar consumer AI tool to help you try and understand it, you're not a patient interacting with a covered entity anymore. You're a customer of a tech company, and that company can generally use, store, or share what you typed under its own privacy policy, not HIPAA. You also have a federal right under HIPAA to access and get a copy of your own medical records generally within 30 days of asking, and in most cases, for no more than a reasonable cost-based fee. That's true whether the records exist on paper or inside an AI-assisted system.

 

We are also seeing a state law develop in this area, as we talked about some states, including New York and California, now require insurers to disclose when AI was involved in that coverage denial. Keep in mind that if a health app or an AI tool is free, ask yourself how it's making money. Often the answer is your data, so read what you're agreeing to before you upload something personal, and when in doubt, ask your provider whether they offer a tool built specifically to be HIPAA compliant instead. So given everything we've covered, here are some questions worth asking. With your care team, you could ask whether AI is being used to document or summarize your visits and ask whether it plays any role in your treatment recommendations. With your insurer, you can ask directly whether AI was involved in an authorization, a claim, or a denial decision. You also have the right to ask who the licensed reviewer was who actually signed off on that decision and what specific clinical criteria they applied. The last question matters more than people realize because a lot of algorithmic denials come back with vague generic language, and so asking for the specific criteria forces a real answer.

 

And then anytime you're offered an AI tool directly, whether that's from your provider, your insurer, or something you find yourself, ask where your data goes, who else can see it, and whether it's actually HIPAA covered or just a consumer product operating under its own privacy policy. And as you know, we've been talking about some states that are trying to make regulatory efforts. If you would like to stay up to date with the laws as they change and to hear about the updates like with our state specific appeals toolkit, you can do so by either signing up for our blog, our newsletter, or even tuning into our podcast. And I will turn it back to Lindsey.

 

- Thank you so much. And I apologize to everyone. We've run over a little bit on time, so much great information today. We did respond, I think, to most questions through the Q&A. Unfortunately, if everyone else was participating and listening and probably furiously taking notes like I was, but if you missed some of the responses to the Q&A, you can always reach out to me. I'm happy to help with just kind of helping out, but you also will be able to watch this video when we share it again. I'm just going to wrap up real quickly because we're out of time. Thank you for the great questions. I think this was a very informative webinar. I hope you all agree. Please don't forget that you will be receiving an invitation to provide feedback in your email later today, so I welcome any feedback you have and suggestions for future programs. I'd like to thank our speakers for all their work in preparing for today's program. And I'd also like to thank, once again, the partners and sponsors, Autolus, Cycle Pharma, Johnson & Johnson, AbbVie, Astellas, Merck, Novartis, and Syndax. Thank you all for helping to make these webinars possible. And thank you for your participation. I hope it was as informative for you all as it was for me, and we hope to see you at a future program. Don't forget, we have our Leukemia Q&A series coming in October, which you will be hearing about through your email. Okay? Thanks everyone. Thank you, Dhruti and Laura so much for all your work in putting this together. We appreciate you.