Dr. Robert Califf on the Future of Healthcare

We started out with a study group trying to figure out what to do about heart attack. Because we had access to computing at Duke that other people didn't have in the 1980s, we became a coordinating center for the clinical trials. So, we had the data. That led to a lot of interactions with FDA. I still have a vivid memory of one afternoon, this young FDA person from CBER named Janet Woodcock called me up to yell at me for doing something wrong in a clinical trial. And eventually, I was brought up several times for FDA commissioner in previous administrations. But, things didn't work out, you know, and that was fine. I was having a great career at Duke.

And then, right, in 2016, Peggy Hamburg contacted me. They had a job open for deputy commissioner at the FDA. And she mentioned that she was not going to stay on much longer, that I would be a good candidate for commissioner. But, of course, the president has to ask you. And at that point at Duke, I'd been a vice chancellor, I'd been on the executive committee of the health system, and felt like I'd sort of done what I could do in that setting. So, I literally went up to FDA as a civil servant and spent a year as deputy commissioner for medical products and tobacco. During that time, a lot of people would stop by my office and ask, because Steve Ostroff had become the acting commissioner and they should have a permanent commissioner, so people would stop by and say, 'When are you going to be commissioner?' And of course, it wasn't up to me. And then one day I got a call saying we need to brief the president, that was Obama, on the issue of the error rate of whole genome sequencing.

And my thinking was, 'Well, how many weeks do I have to get ready to brief the president?' They said it's tomorrow at noon.

The next day I was in there and that went well and on the way out he said, 'Why don't you stop by the Oval Office tomorrow?' And so that next day I had a half an hour one-on-one with the president and he asked me to be nominated for the job.

Two things about that. The first is, and it's very important this week, this is a big week for the Research Triangle in basketball, but we literally spent the first half of the half an hour we had bantering about because he hates Duke and he loves UNC. And so we had to get past that in order to get to the nomination. Well, that really the important things. But then on the way out, there was a whole team waiting for me out there who basically their message was, 'It's nice that you've been nominated, but you're not a commissioner unless you're confirmed by the Senate.' So then all the trouble started to go through a Senate confirmation, but that eventually all worked out. And so I was commissioner until the election happened. I think everybody was surprised by the results and a new team was going to come in. So January 20th, the year after the election, I went back to working at Duke, but also half time at Alphabet or Google. When Trump comes into town, I go out of town. Literally January 20th of 2017 and January 20th of 2025, you know, as a political appointee, my time was up and got in my car and drove out of Washington as the Trump folks came in.

As I mentioned, although at heart I'm a clinician, my real, at least I hope, the thing that I bring to the table is a commitment to evidence. And evidence is built on a technology base of being able to deal with data and to analyze it. And so, well before my first FDA stint, a group of us had spent time going out to Google. If you remember back in the 2010s, electronic health records were coming into vogue. Doctors were trying to practice by typing things into the computers and then trying to get their hands on the clinical practice guidelines to get the evidence was a real task. And so, a group of us decided that we needed to find a global company that was well-resourced, that was really good at information, and whose mission might be to make information available. And of course, that led us to Google. And what we had hoped they would do would be to develop essentially just-in-time information for clinicians at the site in the clinic or the hospital. They basically turned us away after a year of our trying to convince them. And then, during my first stint at FDA, they started Alphabet and one of the Alphabet companies was Verily. And they hired Andy Conrad, who was really the leader of our gang who had gone out trying to get them to take this on. And so, when I finished at FDA, called Andy to see what was going on. He said, 'Hey, why don't you come on out here?' They offered me a great job. I just finished being the FDA Commissioner, so I didn't want line responsibility. The title was medical strategy, but it meant basically I was a consultant to pretty much anything that was going on that was of interest.

You know, I would say my parents were very civil-minded. They, you know, my mom was a school teacher, she volunteered a lot. My dad had served in the Battle of the Bulge, actually.

Yep. Coventry had the same parents. It's scary, all right.

So, it was sort of built in and I never quite adjusted to the draft number that I got, which I was very happy to get a high draft number. I wasn't a fan of the Vietnam War, but 1969 they went pretty high.

And so, I'd never done government service. So, when the time came, I felt like I was really ready and I'm glad I lived long enough to do it at a relatively old age. But, what's great about it is when you wake up in the morning and your mission is health of the people of the United States.

And it's clarifying in terms of how you conduct yourself and how you prioritize things. You're not trying to cut corners or do something special to make your little company or whatever do better or your little academic medical center. You're weighing complex issues, trying to do what's best for the country. And I think while nothing's perfect and there's certainly government employees who didn't work hard, but it's hard to fire people in the government or at least it used to be. But, overall, I would put the government workers up, and I worked everywhere. I've done clinical practice, I've been executive in a health system, I've been on the boards of pharma and biologic companies and device companies, and I worked in big tech. I would put the government workers up with anybody in terms of their quality and my dedication to the job. So, it really pissed me off to hear politicians trash government employees. It's just lazy behavior because they're an easy target and they can't fight back.

Hopefully, that'll happen.

I'm an enthusiast, but with reservations. I mean, I think as I said, I feel like I've been lucky in a lot of ways, but I came along at a time when all medical records were handwritten. You know, we were among the first people to enter medical data into computers to make sense of what was going on. And then we went through this phase of, as I said, of electronic health records that were extraordinarily clunky and still are today, although they're getting better and better. And during my time at Google, you know, one of the things that I'm still trying to absorb is you could see all the possibilities. And in fact, what's happening now, we were dreaming about in the late 1970s, early 1980s, but you just couldn't do it. Right. And I think the thing that's very hard for people to fathom, and certainly I'm not a technical expert, this change that's occurred from algorithm-driven procedures in computers to large language models and basic platforms, it's a big change that really enables a lot of things that we had hoped to do that were still clunky in 2020 and 2021. And so many things can be fully automated now. So, now we have the opposite problem. First of all, we don't know exactly where all this is going to go, but if you look at the Anthropic US government controversy right now, there's a similar thing in medicine: when do we need to have the human in the loop? And is the purpose to free up the human condition so that people can interact while the computer is taking care of all the mind-numbing paperwork in the background, or is it to take over what the humans were doing? And as you might guess, I'm a fan of having the computer in the background doing all the stuff that can be automated so that clinicians can actually interact with patients to solve the problems that they have. And it came up at a meeting I'm attending today as part of an effort to use AI to deliver medicines to people in their homes, really as an assist of primary care. I really think we're going to need to free up people because there's just so much unattended misery in the world that we don't even pay attention to now because people are so busy just getting their work done in the health systems. And I think we can make a big difference if we free people up. And when people say, 'No, the computer can do it all,' you know, my answer is watch The Pitt. If you watch The Pitt, it's a very realistic depiction of how irrational and complicated people can be, which to a computer might look like, 'Okay, you have hypertension, take a medicine and your blood pressure comes down.' Well, to a person it may be, 'Why, I got to get to the pharmacy and I've been getting this social media that says these drugs are bad and...'

Well, I was going to say, I think it's very much empowering to some proportion of the population. I don't know what that proportion is, but I've learned that my estimates of the proportion are a lot lower than the real enthusiasts. Right. Let's take the US population. About a third are children. We don't want to give children total autonomy over their health care. About 10% of people have significant depression. There's overlap here, but about an additional 5% will have dementia. It's over 10% of Americans can't read above a second grade level. And even among the people who have good comprehension, a very significant proportion of them, if you give them too many choices, they say, 'Doc, just tell me what to do.' Now, you know, people who run around with people like you and me tend to say the opposite, like, 'Don't tell me what to do. I'll do my own research and I'll ask you questions. Give me good answers.' But that's not the majority of the population. So, we got to be careful about that. And then there's the other factor, which is I think at the core of so many problems that we have in this country today. If you look at who's got expendable income to deal with the technology invasion, half of the expendable income is in 10% of the population. And there's this great quote that came up during the pandemic from Ed Yong, who was I think one of the best writers about the pandemic: 'Technological solutions drip into society's penthouses, diseases seep into society's cracks.' You know, I saw that on Google over and over. It was something that was good. Wealthy, highly educated people just latched right onto it, and the people who probably could have benefited the most had the most difficulty adapting to it and figuring out how to use it. If we're really going to have this accrue to the benefit of the US population, we got to change a lot of things about it.

You know, there's a characteristic of organizations now that I've been in so many that you end up with these sort of, in my view, you end up with these natural cross-cutting tensions. If you work at Alphabet, you've got the Bets and you've got Google. Within Google, you've got YouTube and search and they're like their own things and they're cross-cutting, software implementation is cross-cutting. And so you got a tension there. You could slice the FDA up. I'm not sure it would be any better because the alternative is you have centers that have a lot of autonomy and the laws are somewhat different for drugs, devices, biologics, and food. So, they operate under different rules, but you have a technological base and a fundamental Food, Drug, and Cosmetics Act. I thought you were going to ask is it too much to regulate AI? And I do have a just a quick party line on that.

I really think the way to regulate AI is, first of all, as is already being done, divided into tranches. If it's AI embedded in a traditional device that's a high-risk device like a defibrillator, it should be regulated like traditional devices. If it's AI for decision support, I think the good analogy is food safety. The way food safety works in this country, and it's not perfect, but it's hard to think of a better way.

Well, because it's an easy thing to knock. Okay. You would hope food would be entirely safe, but I don't know about you. I don't, when I go out to eat, I don't really worry that I'm going to get sick. It happens sometimes.

I mean, you got farms, you've got counties, you've got states, you've got the federal government, you've got grocery stores, you've got restaurants, you've got the distribution systems, and they all have to live by something called FSMA, the Food Safety Modernization Act, and there are 10 core rules. So, rules, as you know, are federal statements that have the force of law, but the rules are not highly specific. They're principles that you should live by. And then the work of auditing those rules is split up among multiple entities, but fundamentally it is the responsibility of the farm or the grocery store to make sure it's abiding by those FSMA rules, and it's just checked, like the FDA and the state split up the inspections. It's checked, but not all that often. A farm only gets inspected every three or every five years, and the food's pretty darn safe. Not perfect, but pretty safe. I think because AI is going to be part of everything, if you said we're going to regulate that like traditional devices, you'd need tens of thousands of FDA employees to do that. I think an analogy like, what grocery store or restaurant wants to sell food that makes you sick? Right. We've got health systems which are deploying most of the AI. Since most docs are employed by a health system now, and that's a unit of delivery, and those health systems need to be accountable, but we don't have a consensus on what the fundamental principles should be, and I think that's where we need to focus. And the one that would make the biggest difference in my view is if there was a principle that you got checked on that your AI had to be equitable.

Good point.

You know, I wanted to come back to that, and I work with a food allergy group, and we talk about it a lot.

There we have to divide it, you know, there's fundamental bacteria and viruses in food. That's what we're doing really well with because it's amenable to the systems that we currently have. If you said ultra-processed food is not safe, I'd have to agree with you. We've gone the wrong direction there, but that's a different system. That's a system of putting in laws and regulations and incentives. Remember that we're incentivizing the food industry to the tunes of tens of billions of dollars to grow corn for corn syrup and to feed cows and other animals to produce meat, which is not great for the environment. And then you got food allergies, which I think suffers, and again, this is just my view of it, but I have had a chance to maybe think about it more than most people. I think the issue with food allergies is it falls in between. Because when you talk about like serious food allergies, that's a medical problem. The source of it is the food and we don't have a good way of doing the sort of large, long-term studies that are needed to understand. You know, I think one of the most interesting things that happened is, right, we had all this peanut allergy and everybody thought, well, you should have kids not be exposed to peanuts and it turns out the opposite is true. The best way to prevent peanut allergy is to expose infants to peanuts. That took real quantitative data. So, two things about it. We can make progress now in those big studies because of the same advances in computing. We'll understand what it is about the basics of food and the production of allergies. But the second thing is once you got a food allergy, this is a place where biotech is moving quickly. I don't know what company you're working with, but we're understanding much more about the kallikrein system and other elements of reactivity to food that can be treated beyond just an injection of epinephrine. And some long-term therapies that would be preventative. In between, I think we got this really interesting hot issue now with Maha about other things in the food, environmental exposures, where we have this weird thing going on now with the current administration where they're talking big, but then if you look at what's happening with the EPA, they are unleashing more environmental contamination than we've ever seen before.