Can an AI health record help you get more out of a doctor visit?

Dr. Steven Charlap built an AI health record that interviews you, reads your old records and gives you the questions your doctor may not have time to ask.

It can, if you treat it as preparation and keep your doctor in charge. Dr. Steven Charlap's Medome gathers your history, medications, wearable data and past records, then gives you a list of possible explanations and the questions to bring to your appointment. He calls it a second layer on top of your doctor, and he's clear that the doctor still makes the call.

Steven is a surgeon and the co-founder of HealthDrive. He started building this after losing his brother, a cardiologist, to a misdiagnosis. When he came on the show, I wanted to know how a patient actually uses something like this, and what doctors think when that patient walks in prepared.

Why don't doctors want more information about their patients?

Steven's team pitched doctors first, and the answer shocked him. The doctors said they didn't want more information. They didn't have time for the information they already had. Steven reads that as a proxy for a payment problem: nobody pays them to process more. Then came the part he called chilling.

That additional information might include something important. And we might miss it and be held liable, so we prefer not to know.

Dr. Steven Charlap, recounting what doctors told him, 07:12

So he turned the product around and put it in the patient's hands. The patient is the one person with a reason to read all of it.

What goes into an AI health record?

As much of your life as you're willing to share. Steven listed previous medical records, medications and supplements, your socioeconomic situation, a three-generation family history, the nutrition in what you eat, recordings of your doctor appointments, environmental factors, symptoms and a journal. Wearables feed in too. I told him about my Oura ring and my scale, and he said all of it gets pulled in. You can photograph a medication bottle and it goes straight into your profile.

Steven calls the result a precision patient profile. His argument is that the more a model knows about you, the more dots it can connect that a busy clinician can't.

Can AI notice a health problem before symptoms show up?

That's the claim Steven is building toward, and he described it as picking up signals before symptoms. His example was small on purpose. Your daily activity on a wearable drifts down a little, or you lose three pounds. Three pounds isn't an emergency. In the context of everything else the system knows about you, it might be worth a question.

According to Steven, his team has built over sixty proprietary databases with 150 million training data points. One of them, he says, has two million rows of symptoms that appear anywhere from three months to ten years before a diagnosis. The system is meant to flag only what matters in your context, because a tool that flags everything is useless. These are his figures about his own product, and he told us he wants more validation before he'd ask anyone to trust their life to it.

What does the AI hand back before an appointment?

Steven walked us through the flow. The app runs a full medical interview. Then it thinks for about five minutes and asks a second round of branching questions based on everything you've shared. After those, it spends seven to thirty minutes on a full analysis. He says it looks for:

• Discrepancies in your record, like a condition on file that you never mentioned.
• Abnormal test results that nobody followed up on.
• Medication and supplement combinations that may interfere with each other.
• Possible diagnoses, with the workup and treatment for each.
• A set of questions to ask your doctor when a visit is clearly needed.

The abnormal results point came with a story. The lawyer who handled the closing on Steven's house had a mole removed. A year later he read his own record and saw it had come back as melanoma. Nobody had told him.

How do doctors react when a patient comes in prepared?

It depends on the doctor. Steven thinks good doctors should welcome it, though a long note is still one more thing to read. He pointed to two studies, one in the US and one in the UK, which he says found that doctors who see a list of possible diagnoses before taking a patient's history are eight to twelve percent more likely to get it right. His team is working with the UK group on an NHS-funded early cancer detection study.

The app also lets you record the visit and gives you a transcript, since Steven says people forget 60 to 80% of what was said in an appointment. Then it offers a second opinion on whether it thinks the doctor got it right. That's the trust but verify idea he keeps coming back to.

What could you possibly miss based on what you're hearing? What else should you ask to make sure that you don't miss something?

Dr. Steven Charlap, 16:44

Who is this for, and what does it cost?

Steven's answer to who needs it was blunt. He says about a hundred million Americans don't have regular access to primary care, the number of primary care doctors is shrinking, and many of the ones left are moving to concierge practices that charge $2,500 to $15,000 a year. His long-term picture is AI-first primary care, with a person checking the AI and a doctor brought in when it's truly needed.

For now, he said, you can start with a free chat that tells you whether what's bothering you is simple or not. If it isn't, it asks you to register for a full intake, with seven days free and no credit card. After that the site lists $14.50 a month, or $120 a year, which works out to $10 a month. He also built 81 AI coaches. The therapy coach is trained in motivational interviewing and cognitive behavioral therapy, and it's capped at 40 minutes a day so people don't get too attached. When I asked about weight loss, he said the coach won't hand you a plan right away. It first works out what motivates you and how you handle plans, so the plan is one you'll follow.

He was also candid about limits. He mentioned a recent study he'd read finding that food photo apps miss about 30% of the fat and 20 to 30% of the calories in prepared foods, and said his team is still working on that feature. I trust a founder more when he tells me where his product falls short.

Steven's advice to men was to stop ignoring their health, and to women, not to let the men in their lives ignore theirs. His reasoning: you insure your car and your house and you get the air conditioner checked before August. Your body deserves the same attention before something breaks. This post reflects Steven's description of his own product and isn't medical advice. Bring what you learn to a qualified clinician.

Cut from His Brother Was a Cardiologist. He Still Died From a Misdiagnosis.