FHC #230: Flashback to healthcare skimping and AI-powered patients

0:11Hello and welcome to Diving Deep, part of the Fixing Healthcare podcast series. I'm one of your hosts, Jeremy Kort, also host of the popular books of medicine podcast and CEO of Executive Podcast Solutions. With me is Dr. Robert Pearl. For 18 years, Robert was the CEO of the Permanente Group, the nation's largest physician group. He is currently a Forbes contributor, a professor at both the Stanford University School of Medicine and Business, an author of the bestselling books, Mistreated, Why We Think We're Getting Good Healthcare and Why We're Usually Wrong, and Uncaring, How the Culture of Medicine Kills Doctors and Patients.

0:39His new book, Chachi PTMD, How AI-Empowered Patients and Doctors Can Help Take Back Control of American Medicine, remains a bestseller. If you want information on a broad range of healthcare topics, you can visit his website, robertperlmd.com. Listeners have been very appreciative of the coverage of generative AI in medicine that we have provided in our Diving Deep episodes. They understand what is possible with today's tools, but they want to know what to expect next year and the one after that. In two weeks, we'll be peering into the future and tackling the complex challenges and opportunities that AI agents could provide doctors and patients today.

1:10To help listeners wrap their heads around the implication of exponential growth, we thought we'd provide a replay of our November 2023 episode about one year after OpenAI released ChaiChiBT-MD and a mere three years ago. In the book, Chai Chi PT MD, How AI-Empowered Patients and Doctors Can Take Back Control of American Medicine, Robbie pointed out that generative AI technology would be doubling in power annually, and that mathematically it means that by five years it would be 32 times more powerful. Comparing the opportunities highlighted in this November 2023 episode to the ones that we will be discussing in a couple of weeks, it's clear that, if anything, the estimates of the past were conservative.

1:45Robbie, two topics listeners continue to ask about are the growing unaffordability of American healthcare and the impact generative AI will have on medical care delivery. Let's start with the first one. What can you tell them? Jeremy, the concern people have about ever higher medical costs, they're well-founded. And the problem is bigger than many individuals realize. Already half of Americans say they can't afford their out-of-pocket costs if they were to become seriously ill, and they'd have to borrow money or sell their house. And it appears that the price for health insurance coverage is likely to rise faster next year than in two decades.

2:31How are insurers and payers responding? To answer that question, Jeremy, let me tell you a story, one from outside the medical world. Not long ago, I opened a new box of cereal. I found a lot fewer flakes than usual. The plastic bag inside, it was barely three quarters full. Now, this wasn't a manufacturing era, Jeremy. It was an example of shrinkflation. Following years of escalating prices, in order to offset higher supply chain and labor costs, packaged goods producers began facing customer resistance. So rather than keep raising prices, what the big brands started to do was to give Americans just fewer ounces, just about everything, cereal, ice cream, flame-grilled hamburgers.

3:19They hoped that no one would notice. But this kind of covert skimping that we've seen in the inflation following COVID-19, that doesn't just happen at the grocery store, the drive-thru lane. This kind of skimping, it's been present in American healthcare for more than a decade. How did we get here? Jeremy, that's a complex question. So let me start at the beginning. Let's go back to middle of the 1960s. Medicare and Medicaid is past. Healthcare costs begin consuming ever higher percentages of the nation's gross domestic product. That's the sum of all the goods and services provided across the nation.

4:04And this is expected. Remember, at the time, there was a huge percentage of Americans who didn't have any medical coverage. In fact, the two populations at greatest risk were the poor and the elderly. The president at the time, Lyndon Johnson. In the United States, it was incredibly rich. We came out of World War II as just about the only large nation in the world unscathed. We manufactured 70 % of the world's steel and 70 % of the nation's autos. In addition, labor unions, they were relatively powerful. And their latest realized that they could get added benefits for their workers if they added health coverage rather than salary.

4:46Health coverage was paid by pre-tax dollars, so they often could get$2 worth of value for every$1 that they received from the employer. Better health seemed like the right place for the government and the businesses of the nation to invest. What happened next? In 1970, medical spending, it took up maybe 6.9 % of the GDP. And that number grew to 8.9 by 1980, 12.1 in 1990, 13.3 in 2000. By 2010, it was up to 17.2%. Now, this trajectory of spending more of your GDP percentage-wise on healthcare, it's normal for industrialized nations. Most countries, they follow a similar pattern. First, productivity rises.

5:38The total value of goods and services increase. Citizens demand better care, newer drugs, better access. And as a consequence, people are willing to pay more and more for health care. And the percent of GDP rises. At least at first, these higher costs equate to better care, longer life expectancy. And that's what happened in the United States from 1970 to 2010. We saw longevity leap nearly a decade as healthcare costs rose as a percentage of GDP. What happened next? Well, beginning in 2010, something very unexpected happened. Both of these upward trend lines, healthcare inflation as a percentage of GDP, and longevity, they unexpectedly flattened.

6:30You fast forward to today, the spending on medical care, it still consumes 17 % of the US GDP. It's the same as in 2010. And life expectancy in the United States in 2020, and I use 2020 because that's pre-pandemic data, it was 77.3 years. That's lower than in 2010 when the number was 78.7. And the most recent data show life expectancy among men post-pandemic, it's now down to 73 years. In essence, half of the decade that increased from 1970 to 2010, it's now disappeared. So far, that all makes sense. What came next? With the passage of the Affordable Care Act in 2010, healthcare policy experts hoped that expansion of health insurance coverage would lead to better outcomes, and that would result in fewer heart attacks, strokes, cancers, and that the assumption was that these fewer life-threatening medical problems, they'd bring down medical costs as they were eliminated.

7:36That's not what happened. Although the rate of healthcare inflation did indeed slow to match the GDP growth, the cost decreases, they weren't from higher quality medical care. They weren't from drug breakthroughs or healthier citizenry. Instead, they were driven by skimping. As a result of this skimping, the United States fell far behind its global peers in measures of life expectancy, maternal mortality, and deaths from avoidable or treatable conditions. Can you give an example to illustrate how skimping reduces medical costs but results in deteriorating public health? Sure, Jeremy. I'm happy to do so.

8:15Let's look at what's called high deductible health insurance. This is a recent introduction. The 20th century traditional health coverage included two out-of-pocket expenses. Patients paid a modest fee up front at the point of care, meaning in the doctor's office or in the hospital, and then they paid a portion of the medical bill, usually totaling a few hundred dollars. Both those numbers began skyrocketing around 2010 when employers adopted high deductible insurance plans to offset the rising cost of insurance premiums. The insurance premium is what the insurance company charges for its coverage.

8:52With this new model of growing percentage of workers, they now pay a sizable sum. And this is coming out of their own pockets, up to$7 ,050 for single coverage and$14 ,100 for families. And that's before any health benefits kick in. Insurers and businesses, they argue that high deductible plans force employees to have more so-called, quotes, skin in the game. That they incentivize people to make wiser healthcare choices. But instead of promoting smarter decisions, these high deductible insurance plans have made care so expensive that many patients avoid getting the medical assistance they need.

9:33Nearly half of Americans have taken on debt due to medical bills. 15 % of people with employer-sponsored health coverage, that's 23 million people. They've seen their health get worse because they've delayed or skipped needed care due to the out-of-pocket costs. Do you think that was the expectation when employers introduced these insurance plans? No, Jeremy. I believe the employers saw this approach as a means of getting employees to fund a piece of the added expense. They didn't see it as something which would harm them in the long run. You know, let's just say an insurer is told by the insurance company that next year's premium will go up 6%, but the company's revenues and profits haven't gone up that fast.

10:18They think they can only afford maybe 3%. A high deductible plan with the out-of-pocket expectations equaling 3 % of the premium would accomplish the goal. It would seem pretty straightforward. But of course, since employees pay far less of the total premium than the employers do, the actual percentage increase in the out-of-pocket expense, rather than being 3 % of the out-of-pocket expense, it's going up now by 10%. And after a few years, that expense now becomes unaffordable. I don't believe that when these plans were put in place, the HR leaders thought they'd have to raise employee costs year over year over year.

10:57But now they find themselves in a bind with no easy way to fund the benefit, but no way to take away from the workers. I'm pretty sure that the employers didn't expect people would delay obtaining necessary care and thereby increasing the likelihood that they would experience ever higher medical costs in the future. This is what the economists call a vicious cycle. What's another example? A second example is Medicaid, the government-run health program for individuals living in poverty. In California, doctors and hospitals are paid dramatically lower rates than with private insurance. It's often 20 % or more below what it costs to actually provide the care.

11:40As a result, the other nation's 90 million Medicaid enrollees have health insurance. They find it difficult to access medical treatment because an increasing number of physicians won't accept them as patients. As such, resolving a medical problem that's relatively straightforward, it can become incredibly difficult to accomplish. And what you see is that people just give up trying. It should come as no surprise that the poorest of Americans die a decade, in some locations, two decades earlier than the affluent ones. This is quite distressing, but let's continue. How else does skimping happen? The next way, Jeremy, it's a little bit complex to understand, but it's equally problematic.

12:24What it involves is a transfer of costs from the government to private payers. To understand how this happens, remember, the U.S. government is the only payer who can legally and unilaterally set prices when paying for health care. Every place else, it has to be negotiated. Now, when Medicare officials decided that they would pay far less than a hospital or doctor would normally bill, the physician and the facility had to begin to make up the difference, charging the private payers even more. The reason is hospitals pay the same amount to the doctors and nurses and for the medications. And it's regardless of how much they're paid by an insurer or by the government to care for a patient.

13:14So let's say that a three-day hospitalization for a particular medical issue costs$10 ,000. Half the people admitted to the hospital have Medicare and half private insurance. If Medicare reimburses only$6 ,000,$4 ,000 less than the cost to provide the care, then the hospital was charged the regular patients$14 ,000 to make up the difference. On average across the United States, private pay patients and the uninsured are charged two to three times more for the same care as government-insured individuals. That's according to the Kaiser Family Foundation report. These higher prices, they generate heftier out-of-pocket expenses for privately insured individuals and massive bills for the uninsured.

13:59They force millions of Americans to forego necessary tests and treatment, or it makes them take on debt that they just can't afford. Robbie, how about one last example? Jeremy, a huge problem is how insurers use the prior authorization process to deny or delay care. Remember that the insurers act as the bridge between those who pay for health care, those of the businesses and the government, and those who provide it, the doctors and the hospitals. To sell coverage, they must design a plan that, on one hand, payers can afford, and on the other hand, that providers of care will accept. When health care costs surge, insurers must either increase premiums proportionally, which payers find unacceptable, or find ways to lower medical costs.

14:49And increasingly, insurers are choosing the latter. Prior authorization is how they skimp in order to reduce costs. Can you tell listeners how this works? Sure. I mean, originally, prior authorization was promoted as a tool to prevent misuse or overuse of medical services and drugs. Prior authorization has become now an obstacle to delivering excellent medical care. What insurers do is require a doctor to call for approval. Any time that individual wants to do an expensive test, a complex procedure, or prescribe a high-priced medication. Insurers know that busy doctors, they'll hesitate to recommend these costly tests and treatments.

15:36They know that these requests will be likely challenged and it will necessitate a huge amount of time for them to respond with letters and more phone calls. By the time they do, patients have grown frustrated and weary of the wait. and often they will abandon the treatment. In fact, it happens nearly one-third of the time. This dynamic creates another vicious cycle. What you see is that costs go down one year because of the skimping, but the medical problems therefore worsen the next year, and they require even more skimping the third year. And we're continuing year over year to regress down this vicious or around this vicious cycle.

16:21Are we at the end of skimping? Unfortunately, no. To the contrary. It's become far worse. What we see is that federal actuaries project that healthcare expenses will rise another $3 trillion per year over the next eight years. They're going to consume nearly 20 % of the U.S. GDP by 2031. And this, it just can't happen. Our country won't be able to afford to spend this much more than today, given the challenges that we're seeing. Ongoing inflation, rapidly rising national debt, Medicare and Medicaid running out of money. As such, it's more plausible that health care's share the GDP, rather than going up to 20 % as projected, will stay around 17%.

17:16Unfortunately, given how resistant to operational improvement medicine has been for the past two decades, this continued leveling out of healthcare costs as a percent of GDP, it won't reflect medical advancements. It won't reflect innovative technologies. But instead, it will result from greater skimping. Robbie, when do you see this beginning? It already has. You know, this year, Medicare decreased payments to doctors by 2%, with another 3.3 % cut proposed for 2024. Since the end of the COVID-19 emergency, funding that ended earlier this year, more than 10 million low-income Americans have been thrown off Medicaid as the states continue to roll back eligibility as a means of lowering health care costs.

18:10And we're seeing insurers using AI now to automate denials for payments, making it harder and harder for doctors to provide the testing, medications and treatments they believe best. What will workers see next year relative to their health care coverage? Of course, the answer to that is that it's going to vary by employer. But at least currently, the competitive job market has business leaders leery of cutting employee health benefits. As such, most people are likely to see only a small increase in their out-of-pocket costs and no major huge shifts. Going forward, however, as the economy shifts and worker availability rises, employees, they should anticipate paying significantly more for their health care.

18:59The truth is, our health care system is grossly ineffective and it's financially unsustainable. until someone or something disrupts that system, replacing it with a more effective alternative, we'll see more and more skimping as our nation struggles to restrain medical costs. And that, Jeremy, will be dangerous for America's health. Let's shift to a more optimistic topic, at least for patients. How do you see generative AI impacting the doctor-patient relationship? Jeremy, I think it will have a big and very positive impact. After decades of doctor knows best, the traditional physician-patient relationship is on the verge of a monumental shift.

19:44Generative AI tools like OpenAI's ChatGPT, Google's Bard, Microsoft's Bing, they're poised to give people significantly more power and control, not just over their personal lives and professional tasks, but over their own medical health as well. Why do you say that? Well, before I explain, Jeremy, let me say that the technology itself, what exists today, it's a toy compared to what will be here in five to 10 years. Remember that generative AI is doubling in power every year. That means it'll be 30 times more capable in five years, a thousand times better in 10 years. Second, keep in mind that tools like the internet, they give people information and knowledge.

20:33but that generative AI provides true expertise. Rather than search engines offering links to articles on a healthcare topic, generative AI will be able to answer very specific questions. Already, it's been able to pass the qualifying exam for doctors. Already, it's been able to handle complex health problems, equal to most physicians. And in the future, these tools will connect with people's electronic healthcare records and home monitors, and they'll be able to update people on their health continuously, day by day. Where are we on the journey? Although we're barely past the starting line, as I just said, Google's MedPOM 2 scored at the expert level 86.5 % on the U.S.

21:24medical license exam. And what we've seen is that just shy of a year, already GPT-4 has replaced the originally chat GPT, and the new tools are being made available to patients, to clinicians, every couple of months. To give you an example of what I see coming, I heard about a Harvard economist, a business school professor, Amitabh Chandra. He was recently admitted to one of Harvard's hospitals with, as he described it, the most intense abdominal pain you could imagine. Several attending physicians told him it was nothing of major significance, most likely something he ate or maybe a gallstone, and they planned to discharge him in the morning.

22:10The patient was certain he had a problem that was more severe than the attending physicians had concluded. And he put his symptoms and all the details of his clinical record into Chachi PT. And according to the AI, Chandra had a 40 % likelihood of a ruptured appendix with just a 2 % chance of a gallstone. Based on this information, he insisted that his doctors order an abdominal CT scan. And the scan confirmed the ruptured appendix. Had he gone home? he might have become septic and died.

22:45That's an amazing story, Robbie. Can you give a glimpse into what's on the horizon? Sure, Jeremy. AI tools will be able to give patients detailed information about their specific medical problems. As we just discussed, they'll be able to integrate health monitors and electronic health records. In fact, already such EHR projects are underway at the Oracle Cerner and Epic Centers. Using this detailed medical information and specific to a single patient, people will be able to reliably and accurately self-diagnose and be able to manage their own diseases. This newfound expertise, it will shake the very foundation of clinical practice.

23:28Already public health experts are touting the concept of clinicians and patients working together through shared decision-making. But even though it's being talked about frequently, it's rarely happening in practice. Gender of AI will alter that reality. What do you see happening to the doctor-patient relationship? The doctor-patient relationship, which, as we both know, dates back more than five millennia, it's always existed on this uneven playing field. Patients relying almost entirely on doctors to help them understand their medical problems and what to do about them. Though patients can, and they do use the internet, to Google their symptoms for a list of possible diagnoses and for potential treatment options, that's not the same as possessing medical expertise.

24:15In fact, sorting through dozens of online services, often with conflicting, inaccurate, and outdated and self-serving information, that proves more confusing to people than clarifying. Nowhere today can web surfers find personalized and credible information based on their age, medical history, genetic makeup, current medications, and laboratory results. But with generative AI, that's all going to change. What do you see in five to 10 years? Well, once generative AI tools become more powerful, they're going to alter this doctor-patient dynamic. They can empower patients and level the playing field.

24:52Already, consumer AI tools could equip users with not just knowledge, but expertise. They allow the average person without any art training to paint pictures. which are hard to discern from a Renaissance master. They can help write songs and record songs that could well have been done by Drake despite never having played an instrument, whoever's programming it or putting the information in place. And it could write computer code without having particularly interest and expertise previously or having attended any IT classes in computer technology. To me, all this is unimaginable. and what future generations will be able to accomplish, it's going to allow people without a background in science or any medical study at all to reliably, rapidly, and safely diagnose their symptoms, manage their chronic diseases, and develop sophisticated treatment plans.

25:52How are these applications capable of doing all this? You know, unlike with search engines that it best provide the links to general articles on a topic. ChatGPT functions more like a digitized second opinion source. It can access millions of medical textbooks, peer-reviewed journals, scientific articles. It can deliver accurate and unbiased medical expertise in layman's language, in response to any question posed, and in fact, in any language that the person wants to speak. The more details as a patient you enter into the application, the better the response. Put in all your medical history and all your medical problems in the chat GPT, you'll get a personalized information response specifically to who you are and how you want to get your care with recommendations that align with the best treatment possible.

26:49And unlike internet sources, the answers provided, they will be chosen based upon built-in financial incentives or advertising models that skew responses similar to what happens with search engines today. You mentioned earlier that generative AI would alter the doctor-patient relationship. White might be an example. With empowered patients knowledgeable about their medical issues, the role of the physician will shift. One way this may happen is by applying a model that teachers have found lead to more educated students. This is the so-called flipped classroom. The flipped classroom can be traced back nearly four decades, but it became popularized in the United States in the early 2000s through the Khan Academy in Northern California.

27:37Students begin the learning process by watching videos and engaging with interactive tools online rather than sitting through traditional lectures. And this is the way that it's flipped. This pre-class preparation, it's often called homework in advance, allows these students to learn at their own pace. It enhances classroom discussions, letting teachers and students dive much deeper into topics than they ever could before this model was put into place. Indeed, students spend most of their time in class applying knowledge and collaborating with each other based upon the knowledge they've already gained to solve problems.

28:17higher level, more challenging problems. That's totally different than just sitting in a classroom, listening and taking notes. And it's been shown to lead to far better outcomes related to education and to thinking. What would that look like in healthcare? When a patient had a new medical problem with some concerning symptoms, rather than trying to figure out what was happening and what it might mean by scheduling an appointment with a physician, the patient instead would begin with degenerative AI tools to understand the symptoms and the likely medical diagnosis. This foundational knowledge would empower patients and accelerate the diagnostic process.

29:04Even before the actual testing, whether that's x-rays or blood work was complete, this pre-consultation phase would allow the patient to comprehend the range of potential diagnoses and the ones which most likely were the cause. And knowing about their symptoms, about the likely diagnosis and best treatments, patients would be better able to understand the questions the physician asked, the steps that the clinician was recommending. And we know today that as much as half of patients leave the doctor's office uncertain about what they just heard or what they should do, this would increase patient expertise and knowledge and more likely lead to better adherence and superior clinical outcomes.

29:59What will happen next in this flipped classroom healthcare model? With the patient's knowledge base already established, the in-office consultation will include not only an in-depth discussion of this new problem, but also the doctor will have the time to provide information on productive health strategies and long-term chronic disease management solutions. Rather than having to start at square one, clinicians will be able to focus on the uniqueness of each individual and personalize the clinical recommendations for that person. This approach will maximize the time that patients and clinicians spend together.

30:39And it will also address the reality, as we just said, that half of the time when patients leave the doctor's office, and I assure what they've been told, they will understand the medical problems at a level that today doesn't exist. What's another way the doctor-patient relationship will be flipped? Let's look at the 60 % of Americans living with chronic disease. Genitive AI combined with wearable monitors will provide real-time feedback to them, and that will help optimize clinical outcomes. Rather than at the start of treatment, driving to the doctor's office, going there on a periodic calendar basis, maybe every three to six months, patients will start by getting daily updates from ChatGPT or some other generative AI technology on their medical progress.

31:30In those cases where generative AI spots a trouble, where the healthcare data about the current situation deviates from the doctor's expectation, the provider will be able to update medications immediately because they can be contacted either by the technology itself or by the patient. In contrast, when patients are doing well, physicians don't have to see the individual for follow-up visits, and that will eliminate wasted time for all. The combination of various changes, what it's going to do is decrease demand on doctors. Better chronic disease management will mean fewer life-threatening medical complications and less need to be seen in the physician's office as frequently.

32:14Providing care to an informed patient will avoid office visits, which are not medically necessary, and it will make the in-person conversations far more valuable. The combination of time-saved, demand-reduced, better in-person interactions, they can increase professional satisfaction and diminish physician burnout. How might this F.L.I.P. model further reduce costs? Jeremy, the idea of starting with the patient rather than the physician opens multiple opportunities. Rather than thinking that the best medical care is in the doctor's office or the hospital, the F.L.I.P. model begins in the home.

32:51An example of what's possible relates to hospital costs and quality improvement. Inpatient care accounts for 30 % of all healthcare costs by continuously monitoring patients who have medical problems like mild pneumonia, controllable bacteria infections, problems that we probably would hospitalize a patient with. Janitor of AI can combine with home monitoring devices, telemedicine access, should there be any concerns, and that would allow individuals to be treated in the comfort of their home safely and more affordably than today. You know, a second example is relative to lifestyle medicine. Janitor of AI already can support preventive health measures and lifestyle changes.

33:39It can reduce the overall demand for in-person clinical care, and it can lower healthcare costs. Studies confirm that focusing on diet, exercise, and recommended screening can reduce the deadliest complications of chronic disease. These are the heart attacks, the strokes, and the cancer by 30 % or more. And decreasing the need for intensive procedures, that's the best way to make American healthcare affordable. That's the best way to improve clinical outcomes and the best way to address what will be a shortage of doctors and nurses that already is getting worse for what would be crisis level in the future.

34:15Can you provide one final example? The U.S. healthcare model often leaves people feeling frustrated and overwhelmed. They feel rushed. On average, clinicians wait only 11 seconds before interrupting patients. Meanwhile, the time constraints, which are placed on doctors from having to see so many patients a day, What they lead to is rushed consultations and misdiagnoses, and these cause an estimated 400 ,000 deaths annually. ChatGPT will help clinicians make the most accurate diagnosis and provide patients with a second opinion. They'll free up the physician to spend more time with patients, further reducing medical error.

34:59And it will allow doctors to go home at night feeling they did the best job possible. and that will decrease burnout and allow them to return to work the next day refreshed. But to reach a destination, manufacturers will need to improve the technology. They need to eliminate the glitches which existed in the application today. And medical culture, it will need to evolve so that clinicians become comfortable collaborating with the empowered patient and being willing to rely on generative AI as a part of the care delivery team. Talk with educators of the Khan Academy. What they'll tell you is that this innovative model results in better educated students.

35:41They'll also tell you how much more satisfied teachers and students are compared to those working in the traditional educational system. And I predict the same will be true for American medicine. We hope you enjoyed this podcast and will tell your friends and colleagues about it. Please follow Fixing Healthcare on Apple Podcasts, Spotify, or your favorite podcast app. If you like the show, please rate it five stars and leave a review. Visit our website at FixingHealthcarePodcast.com. Follow us on LinkedIn, Facebook, and Twitter at FixingHCPodcast. Thank you for listening to Fixing Healthcare Diving Deep with Dr.

36:17Robert Pearl and Jeremy Kaur. Have a great day.