A new pattern for cardiology visits

A 27-year-old patient booked a visit with me last week via our recently launched self-booking site (I’m proud to have built it). His blood pressure had been in the 140s since he could remember, and he had a family history of early heart attacks. Concerned by both of these facts, he had decided to get, on his own initiative, a direct-to-consumer lab panel and found that his Lp(a) was quite high. No referral, no PCP intermediary, just a calendar slot, a Stripe payment, and a structured email sent soon after booking, listing his concerns. He had already read extensively about Lp(a) and wanted to discuss what to do given his age, family history, and values. His questions were better organized than most of the consult requests I used to receive from internists.

A few days later, a second patient, a 40-some-year-old man, booked a visit and sent an elaborately constructed document before our video visit. His CAC score was 170, localized to the LAD. To his frustration, his LDL-cholesterol had plateaued at 123 mg/dL for twenty months despite four years of disciplined lifestyle and introduction of a dizzying list of nutritional supplements. His Lp(a) was 52 nmol/L. He was curious about non-statin options primarily because of a family history of statin intolerance as well as the muddied picture on the internet. The document itself was a Claude-summarized PDF, which included color graphics depicting his health story compressed into two pages, peppered with current guideline citations, and capped with a structured list of questions he wanted us to tackle. He had not been referred to me by anyone. He had asked Claude which cardiologists were doing prevention work in his region and Claude recommended us! (I incidentally joked that given how much we’re spending at our startup on the Claude Max 20x subscription, a little reciprocation was welcome.)

I keep seeing this. The pattern is consistent enough that I want to write about it before I fully understand it. It’s not how I was taught cardiologists are to play a role in care, and certainly at odds with how value-based care plans would like our involvement.

My read on what's happening

Most of these patients don’t have a PCP. Some have aged out of the relationships they had in their twenties, or moved to a new city and faced an 8-month waitlist, lost continuity due to insurance changes, and just never bothered to re-establish, given there were no pressing concerns. Some have actively given up on the model after years of fifteen-minute visits that produced no useful action on their cardiovascular risk. Some never had one to begin with; the younger end of this cohort grew up in a world where the model of “your PCP coordinates your care” was already broken, and they’ve simply built a different model, shared by their friends and coworkers. The framing is not that they’re bypassing primary care. It’s that primary care, as a continuous relationship, is already absent for many of them.

I have some thoughts on what might be driving this.

Information democratization via LLMs. Patients can now compress what used to be six hours of poking around PubMed and lay-summary sites into a twenty-minute back-and-forth that returns a synthesized, mostly accurate, mostly current picture of what’s known about their situation. The output is not always right. In this particular PDF, Claude was a bit too deferential to the patient’s wishes, interpreting them as rigid dogma rather than a starting point for understanding. Some of the medication recommendations seemed implausible. A little too much “Book smarts”. But it is good enough, particularly for organizing thoughts, so that a patient who arrives after spending that twenty minutes has a usefully different starting point than one who hasn’t.

Primary-care training and bandwidth mismatch. Most primary care physicians are not trained in advanced lipid management and guidelines and new medications are a moving target. They do not order Lp(a). They do not think in ApoB terms. They are not, structurally, going to spend thirty minutes thinking about a four-year lipid trajectory and what it means that the curve has flattened despite the patient’s adherence. Lipid has been a hot topic again for Pharma – and the number of new drugs that have both been released in the last 10 years and are coming in the next 5 puts to shame the prior 40 years. None of this is a failure on the part of individual PCPs, who are managing impossible panels under time pressure. It is a structural mismatch between what prevention-focused patients want and what the primary-care visit can deliver. Patients who have done their homework can feel that mismatch within five minutes, and increasingly they are routing around it.

Direct-to-consumer lab availability. A patient can now order their own Lp(a), ApoB, NMR lipoprotein particle counts, fasting insulin, hs-CRP, and a dozen other advanced biomarkers without a PCP order. Function Health, Marek Health, InsideTracker, Quest’s consumer-facing arm, LabCorp OnDemand, BloodOne — the list grows quarterly. The implications are larger than most physicians have processed. The historical gatekeeping function of “your doctor decides what to test for” is dissolving for any patient willing to pay a few hundred dollars out of pocket. By the time these patients arrive for my visit, they often have a richer biomarker profile than their PCP would have obtained over five years of annual physicals. The lab work is the entry into the system and the cardiologist visit is where they want it interpreted.

The prevention-curious cohort is bigger than medicine has acknowledged. Whether driven by Peter Attia, Eric Topol, Mark Hyman, longevity culture, the post-pandemic interest in metabolic health, the rise of Facebook discussion groups around specific biomarkers, or some combination of all of the above. There is now a substantial population of educated adults whose mental model of cardiovascular risk is more quantitative and more biomarker-driven than the model the average PCP carries. They aren’t necessarily right about everything. They are right about enough things, though, that the conversation deserves a different format than a fifteen-minute visit can hold.

A US phenomenon, or a leading indicator

A few things make me think it might be US-specific. We allow direct specialty booking without PCP referral on most commercial insurance plans. We have a long-running cultural acceptance of self-directed health research, going back to the early days of WebMD. Our healthcare system is fragmented enough that being your own care navigator is a rational response to a system that will not navigate for you. Patients have learned to do this because they have had to.

A few things make me think it might generalize, but likely in different ways. The information-asymmetry collapse that LLMs are producing is not US-specific. The supply gap between specialists and the population that wants their attention is universal in the developed world. Cardiology workforce shortages are not just an American story: the UK, Canada, Germany, and Australia all report similar dynamics. What’s different is the system architecture: whether patients can act on their information by booking specialists directly, or whether NHS-style gatekeeping prevents this. So the question may be less about whether the motivation is American and more about whether the expression of that motivation through direct booking is.

I’d be curious to hear from clinicians outside the US who are seeing similar patient self-direction even within gatekept systems, or who are not.

The likely trajectory

I think it grows. I’m not certain, but the trajectory seems clear.

LLMs become more capable and more accurate, but far from infallible. The younger cohorts who treat asking an AI for medical context as obvious will age into the years where cardiovascular risk matters. Direct-to-consumer labs will continue to expand the menu of self-orderable biomarkers. Specialty supply is not going to catch up with demand on any reasonable time horizon. Patients will continue to route around bottlenecks. The Reddit and Facebook groups and podcasts that have built the prevention-curious cohort will continue to produce more of them.

The interesting question is what cardiology does with this. The existing infrastructure of preventive cardiology is organized around assumptions that are slow to change: referral patterns, payer incentives, slot scarcity, in-person defaults, that were designed for a different patient flow than the one I am now seeing. There are no lucrative procedures to be done for these patients and the imaging that might be useful is unlikely to be owned by individual practices. So perhaps most cardiologists don’t want to see these patients.

We are building our startup, Atman Health, around the assumption that this trend continues and broadens, that autonomous and semi-autonomous AI agents will increasingly mediate the front of preventive care, and that the cardiologists who matter in this world are the ones who engage with the resulting patient flow rather than dismissing it.

For now, I just want to note that something is happening in this segment of my practice that is genuinely different from what I saw two years ago. Patients are arriving with their lab values, family histories, LLM-mediated initial understanding, and willingness to skip the referral chain entirely. They are arriving without us having done much to invite them, which suggests the supply of such patients is not the limiting factor. What’s limiting is the supply of clinicians willing to engage with how these patients arrive.