My Recent Podcast

My most recent podcast guest was David Sontag, Professor of Computer Science at MIT. David has been working at the intersection of AI and Healthcare for many years. He has great insights into the current pain points of the healthcare system, both for providers and consumers, and how AI will address them.

So, check out my conversation with David.

Desperately Seeking Holism in Healthcare

In reflecting on the conversation with David, one of my key takeaways is that at the moment, machines and doctors aren't working well together. In fact, we are getting the worst of both worlds. We're not getting the human touch, nor are we getting good, informed decision making from humans, who have become increasingly specialized and hurried.

It’s clear that AI has the promise to make healthcare more humane and better if it is designed and used correctly. I find it ironic that the machine will rescue us from what has become a factory approach to healthcare that we have been veering towards for the last decade.

Here’s an example of one of my recent interactions with the American system, but I’m sure you have your own versions of similar encounters. A couple of months ago, my neighbor’s two German Shepherds charged down my driveway, chasing my Labrador Retriever around my house. I ran after them, tripped on a hard pavement, and got a nasty gash above my left knee. I was driven to the local Emergency Room, where I was told that “the provider” would see me shortly. What a strange term, I thought. Whatever happened to doctors?

Sure enough, four providers processed me, one after another, each asking many of the same questions, before handing me to the next. One took my blood pressure and vitals. Another asked me about allergies. Another asked me whether I wanted an anesthetic before the procedure. Yet another asked whether I preferred sutures or staples, and recommended staples since the wound was in a high movement area.  The last provider punched staples into my knee, and told me to come back in roughly ten to fourteen days to get the staples removed.

I felt like I had gone through an assembly line, not unlike my car in a repair facility.  But this didn’t seem particularly worrisome at the time since it was just a gash, and not a life threatening situation.

When I returned to have the staples removed, it hurt like crazy and bled. Unfortunately, the gash opened up the next day, so I went back to ER since they are not permitted to offer advice remotely. This time around a new provider told me not to worry since the wound wasn’t infected, and scar tissue would eventually fill up the gash.

Two months later, I can walk normally, but I still can’t flex my knee completely without pain. It is impossible to tell whether I just got unlucky, or whether I accepted the recommendations too readily. For example, I can’t but help ask myself whether I should have opted for self-dissolving sutures, which have worked well for me in the past. Or whether the staples were removed too early. Or whether the staple remover botched the job. But the more important point is that no one is keeping score of interventions and outcomes. None of the providers learned anything from the data about my bad outcome. The system will do the same thing again and again, just like a factory.

And mine was a simple case. The stakes are a lot higher in life and death situations, where it is important to keep track of things like toxicity of treatments and other kinds of outcomes associated with interventions. But even in my non-life-threatening case, AI would help both the patient and provider do better. Imagine the AI telling me something like “based on the data about the laceration, I wouldn’t  go for staples since it is sufficiently far away from the kneecap where the amount of movement favors staples. And it could hurt like hell when they remove them.” Or, when I went to get the staples removed, imagine the AI looking at my knee and saying “it’s too early to remove the staples, come back in four days.”

And on the provider side, imagine them using AI tools to explain the situation, choices, and reports to the patient,  who feels confused and unable to interpret all the markers at the moment. Providers can also use tools like GPT to connect the markers and patient history into a coherent whole for the patient.

Talking about markers, doctors understand health using markers they can interpret, which are things like cholesterol, PSA levels, A1C levels, glucose, etc., which can be used to prescribe simple therapies. But a machine isn’t restricted to such markers, and can think differently.  For example, in my conversation with cardiologist Eric Topol, the author of Deep Medicine, we talked about the fact that AI can tell a person’s sex based on a retinal scan, something that humans are unable to do. But the eyes contain a lot more information, such as about cardiovascular health, and can predict heart problems. There’s a lot in images that humans miss. Similarly, there may be hundreds of molecules other than things like PSA levels that are associated with the diagnosis and prevention of prostate cancer. Unless we’re looking for them, we will never discover how they relate to things that we care about, like cancer, diabetes, or strokes.  As Lord Kelvin famously said, you can’t understand something if you can’t measure it. AI can measure and analyze a lot more than humans. It’s driving a lot of AI research in healthcare.

The Brave New World of Healthcare

It is ironic that AI will bring holistic individualized healthcare to everyone. I say ironic because we don’t associate holistic thinking with machines, but with humans. And yet, in healthcare, humans need AI to see the bigger picture and connect the dots for us in order to provide informed and individualized care. At the moment, they’ve become cogs in a business machine.

Such a capability will use much denser and wider of data about us. Denser data will come from sensors such as wearable devices. Such data will be increasingly augmented with data about our genomics and biology, which includes molecules, images and other data modalities such as smell.

The results of interventions will also be more objectively measurable and quantifiable. Imagine an AI that says something like “based on a million other people with identical cholesterol levels, genomics and prior history, if you continue to eat burgers and fries, you’ll live for another seven years plus or minus two, but if you switch to a plant-based diet, it will be fourteen years plus or minus one.”

It’s a whole brave new world of healthcare emerging, powered by AI working together with doctors.

Home, Home Again

I was in Kerala a few days ago for a professional event. Since I was in this part of the world, I decided to touch base with my roots and spend a few days in the valley of Kashmir before heading back to New York. Being here always gets that timeless Pink Floyd song called “Time” going in my head:

Home, home again

I like to be here when I can…

It’s saffron season here. Just look at the flowers in the picture. Each purple flower yields three tiny red strands of paradise, a phrase that will resonate with John Mayer’s “Waitin on the Day” fans. The soil and climate produce what is arguably the best saffron in the world. It’s growing in my back yard here. I add it to the local tea, called Kahva, along with pounded almonds and whole cinnamon. If you’ve never had Kahva, you’re missing a little sliver of paradise.

Unfortunately, Indian cities and towns are the polar opposite of paradise. They have turned into cesspools of congestion, pollution, and trash, and Kashmir is no exception. Wherever my fellow Indians go, trash is not far behind. My brother and I made a quick trip up to the pristine mountain range surrounding the valley. The views were breathtaking, as you can see in the pictures. But even there, close to the ridge line, there’s no getting away from discarded bottles of soda, packaging, and cigarette butts. I’ve included a picture of a bottle of Sprite near the peaks of the Pir Panjal mountain range. What a shame, and disrespect for nature.

It will take a lot to change India’s norms, but it should start before the problem gets even worse. Apparently, Americans would routinely trash public spaces until a few generations ago, until the norms changed. So, it’s possible in India, and it’s a good time to start.