The Biggest Public Health Trends of 2026: Part 2

Once again, happy new year! I think that in 2026, we’ll see an intensification of four existing public health trends.

4. Increasing use of GLP-1 drugs.

1 in 8 American adults use a GLP-1 drug for weight loss or conditions like type 2 diabetes. We can expect that number to increase in 2026, now that Wegovy is available in tablet form, for only $149 per month, and the cost of injectables is dropping. Last week I discussed some of the pros and cons of these drugs; others are summarized nicely here.

3. Growing resistance to MAHA.

In 2026 we can expect greater resistance to MAHA policies and practices, owing to increasingly coordinated objections from scientific and medical authorities, coupled with growing internal schisms. On Monday, for instance, Senate health committee chair Bill Cassidy (R-LA.) criticized the CDC’s revised vaccine schedule, predicting that it will “make America sicker”.

(The discussion of MAHA that I posted last week inadvertently contained some draft material. My apologies. I’ve posted a cleaner version here.)

2. Further integration of AI into health care.

AI is transforming biomedical science, clinical practice, and health care administration. Whether you find this encouraging or terrifying – or, like me, a little of both – it’s hard to keep up with the many changes.

AI is unprecedentedly good at gleaning patterns from large datasets. Train AI on millions of mammograms and it can outperform radiologists at detecting breast cancer. Given millions of chemical fragments, AI can design medicines that treat drug-resistant strains of gonorrheal and staph infections. Broadly, experts predict rapid improvements in AI-supported imaging and new drug design during 2026.

Some of these developments seem like unqualifiedly good news. A December study showed that AI outperforms traditional methods of determining when a stroke has occurred and whether it’s reversible. This is a boon for emergency care, and it’s hard to see a downside.

In most cases, the benefits of AI introduce risks, and we need to be thoughtful about how the new technologies are integrated into health care.

Take breast cancer screening. AI is increasingly used in radiology as a “second pair of eyes”, but studies show that like other skilled professionals, radiologists are susceptible to automation bias – i.e., the tendency to favor suggestions from AI-driven systems.

What happens when AI disagrees with human experts? In a recent study, radiologists at all levels of experience showed some degree of automation bias, preferring AI’s recommendations over their own, even when AI was wrong.

Sadly, automation bias can be dangerous even when AI gets it right. Consider the overdiagnosis problem. Not all abnormal clusters of cells would be labeled “cancer” by a human expert, and some grow so slowly that biopsies and treatment would create needless anxiety, expense, and risk. Human judgment is still needed to determine when minor abnormalities and low-grade cancers call for intervention.

Meanwhile, the AI market forges ahead. There’s already a waitlist for Clairity’s AI-driven mammography screener, approved by the FDA last May. Technologies like this, and many others, are being purchased by increasing numbers of hospitals and clinics. More than a dozen universities have responded by offering online certificate programs for health care professionals.

(Out of curiosity, I clicked Johns Hopkins’ “Download Brochure” button. This led me to a request form that asked for my name, phone number, and email address. Within 15 minutes, someone called me trying to sell the program. Thanks, I said, I just want to look at the brochure. As of this morning I’ve received two emails, two texts, and three additional phone calls. I feel bad for the program administrators; I didn’t mean to lead them on. I just wanted to see the damn brochure.)

Last year, an AMA survey found that 66% of physicians were using AI, nearly twice as many as in the previous year. I think this is an encouraging trend, to the extent that AI is used responsibly. By all accounts, “responsible use” means something like “monitored by qualified humans”. I find it concerning that roughly 1 in 6 adults regularly use AI chatbots for medical advice, and that even more rely on bots for mental health support

Finally, experts predict rapid growth this year in the use of agentic AI, which reduces administrative burdens by planning and carrying out activities more or less independently. This too seems like a clear win if humans monitor each stage of the process (e.g., scheduling tests, interpreting results, creating reports, making recommendations to patients, etc.)

What could go wrong here? To take just one example, a study published last April showed that when hypothetical patients with medical concerns were labeled as Black, homeless, or LGBTQIA+, nine large language models more frequently advised urgent care, invasive interventions, and/or mental health evaluations. The LLMs were inherently biased. We would not want them playing an agentic role in an institutional setting.

In sum, AI will be play an increasingly powerful and varied role in public health this year. Some of the technology in the works, like the AI stethoscope (basically, a small ECG device that feeds data to an AI-driven cloud app) is exciting because it promises better tools. Most of the new technologies need to be integrated cautiously, with human oversight.

1. Expansion of personalized medicine.

Why is personalized medicine so popular and continuing to grow? Here’s a little anecdote that answers the question:

Last week I fell and hurt my shin on a hardwood floor. I sat there for a couple of minutes, regretting my clumsiness and weighing options.

What I wanted most was immediate relief. Not for the pain, which was bearable, but so that I could continue training for spring marathon season without interruption. (Hoping to run on an injured leg is nuts, I know, but what would you expect from someone who thinks running marathons is fun?)

I didn’t call my primary care physician, because I doubted I could get an appointment anytime soon, and my interactions with her tend to feel rushed and impersonal anyway. But I didn’t want to pay for an ER visit or a specialist. So, I consulted with Dr. AI.

Together, the good doctor and I hypothesized that I’d sustained a periosteal contusion of the lateral tibial tuberosity – in other words, a bruise to the membrane covering my upper shin bone, but off to the side, in a place that’s not load-bearing.

This was confirmed the next day, when I ran 13 miles without pain, though the injury remained highly sensitive to touch.

Why 13 miles? I’m experimenting with a scientifically-based training plan developed by Running World. The plan called for 12 miles that day, a distance I had tweaked on the basis of a conversation I had once with a running coach. Although I didn’t do this, I could’ve also consulted my COROS PACE 3 sports watch for information about whether the injury affected my cadence, stride length, left/right balance, et cetera.

Details aside, this anecdote reflects most of the raisons d’être for personalized medicine:

• long wait times for medical appointments

• the impression that doctors don’t pay enough attention to individual needs

• AI’s capacity to support self-diagnosis

• the availability of data-driven, personalized plans for optimizing health and physical performance

• the convenience of wearable technologies that provide real-time physiological data

The trends I’ll discuss below pertain to concierge medicine, health and longevity coaching, personalized biotracking, and gene editing – all expected to grow rapidly in 2026.

Concierge medicine

All medicine is meant to be “personalized”, or tailored to the patient’s individual needs. What many of us experience instead are long delays before appointments can be scheduled, long wait times in doctor’s offices, and then what feels like rushed, impersonal care.

Concierge medicine addresses some of these frustrations. For an out-of-pocket fee, typically $2000 to $5000 per year, you get more personalized, immediate service from a primary care provider.

If I had such a service, I could’ve made a same-day or next-day appointment to have my shin examined, and, if needed, get an X-ray that would be immediately reviewed.

The market for concierge medicine is expected to grow 10% per year over the next five years, owing to a shortage of primary care physicians and the lure of more personalized attention.

Still, it’s not for everyone. I haven’t signed up yet. Concierge primary care is expensive and, for the moment, most of the health problems I experience are likely to require specialists (e.g., for running injuries) or over-the-counter remedies (e.g., for traveler’s stomach).

Health and wellness coaching

Not everyone I have in mind here would call themselves “coaches”. I just need a term to describe the rapidly expanding cadre of experts (or “experts”) who sell individually-tailored recommendations on nutrition, sleep, longevity, gut health, workouts, and so forth. (Not all of their recommendations constitute “medicine” – primary prevention is often emphasized - but here again, to keep things simple, I just want to use one term.)

What most coaches have in common is close attention to data. Numerical data in particular. The results of lab tests, self-report assessments of diet, sleep, and other lifestyle variables.

Health and wellness coaching is a multibillion dollar industry and rapidly expanding. For instance, I learned this week that one can become certified as a longevity coach. (I was curious about these programs, but I didn’t want a repeat of the Johns Hopkins experience, so I poked around and found a nice overview from one of the providers here.)

One criticism of health and wellness coaching is that it’s excessively commercialized and, like concierge medicine, contributes to disparities in health care that favor the affluent. I think this criticism is unfair. Experts should be paid for their time. Moreover, coaches varying in popularity from Peter Attia on down offer a ton of free content. I particularly like Kat Fu’s Longevity Vault newsletter for its detailed discussions of the biological mechanisms underlying sleep, healthy living, etc.

As with concierge medicine, the decision about whether you need the individualized support of a health or wellness coach should be, well, individualized. I don’t partake (apart from the free content), because I consider myself healthy, and I think I know where my lifestyle needs improvement. I should eat less sugar, for instance, and avoid falling on hard surfaces. But sure, there’s always room for improvement...

Personalized biotracking

I’m discussing each trend in this newsletter separately in spite of substantial overlap. Your health coach, for instance, is likely to use AI to help analyze biomarker input.

The lure of personal health data seems to be irresistible. Pretty much any substance that your body produces (cells, hair, blood, spit, pee, etc.) can now be sent to a lab for testing.

In fact, if you wish, you can just gather data on everything. For instance, Function (co-founded by Mark Hyman) offers analyses of 160+ lab tests per year, in an attempt to detect every possible health problem as early as possible. Included are widely respected tests as well as ones of dubious value such as the Galleri test (see here for my evaluation), and a full-body MRI, which for most people introduces a distinct risk of overdiagnosis. (What’s the value of knowing that you have some asymptomatic health condition that’s best left untreated?)

Some of the testing is downright sketchy. Viome’s “Gut Intelligence Test”, for instance, is mostly hokum, because it relies on questionable data about links between gut microbiome and health.

Worst of all, perhaps, are the autism tests that analyze blood, skin cells, or a strand of hair. This is irresponsible, not just because the causes of ASD are poorly understood, but because what we do know about the genetic predispositions would only sustain predictions for a small percentage of people that would be better than chance, statistically speaking, but not by much.

For those who want real-time data, there are wearable or otherwise portable devices, which, like so much other personalized health technology, straddle the divide between valuable medical uses (credit card-sized devices that monitor abnormal heart rhythms – see this Skeptical Cardiologist post) and wellness hacks.

For instance, if I wanted to, I could ask my Coros watch to feed performance analytics into Strava, which tracks physical activity and allows it to be shared on a social network. (The title of a recent blog post from Das Z illustrates where this technology has led: “Nobody cares about your Strava year”. Subtitle: “Comparison culture disguised as celebration.” Um, no thank you.)

Health-related wearables are the fastest-growing type of wearable, according to industry analyses, and we can expect the market to expand in 2026 as the FDA eases restrictions on what manufacturers can claim.

Here again, the need for personalized biotracking should be a “personalized” decision. I ignore most of it – I use my running watch to detect distance, pace, and altitude ­– though I am seriously considering a coronary artery calcium test, as it’s a more direct measure of arterial health. In my case, even though my blood pressure and lipid profile are normal, the CAC test might be useful because all my immediately family on both sides have experienced cardiovascular disease.

Finally, I should mention the growing synergy between biotracking and AI. For instance, last year researchers reported that the GPT-based Delphi-2M predicts the rates of more than 1,000 diseases based on medical records, with accuracy comparable to that of single-disease predictions.

A thousand diseases. You can imagine how a technology like this will be marketed. Instead of going to a cardiologist and being told that you’re at risk of, at most, a handful of problems specific to heart health, you can just give Delphi-2M access to your medical records and it will calculate your risk of just about everything, including diseases you may not have heard of.

Unfortunately, personal risk can’t be inferred very accurately from group data. Even if AI could somehow nail the estimates, how would you change your behavior if you knew that 10% of the population develops disease X, and your chances turned out to be, say, 14%? Would you worry more if your chances were 38% as opposed to 31%? (I would, but whatever behavioral changes I made would be the same in each case.) These are tough questions with no clear answers. I doubt AI can help much.

Gene editing

2025 has been described as a breakthrough year for gene editing. The first gene therapy for Huntington’s disease was found to slow the rate of cognitive decline by 75% over a 3-year period. Last month, researchers reported that gene-edited immune cells successfully treated T-cell leukemia. These are small studies, but the data are extremely promising. In the 20th century vaccines eliminated or greatly reduced many diseases. Perhaps the same will be said about gene editing in the 21st century.

Studies like these illustrate what’s more properly called “precision medicine” as opposed to personalized medicine, but in 2025, the first genuinely personalized gene therapy was successfully carried out, saving the life of a baby born with a rare liver disorder that impaired his body’s ability to process protein.

With gene editing we have yet another promising trend on the horizon – accompanied, once again, by concerns that the market is promising too much. For $599, you can swab your cheek, send it to 10X Health, and get all sorts of dubious lifestyle advice, including guidance on supplements that have no established benefits. Buyer beware.

Final thoughts

The growth of personalized medicine – my pick for the #1 public health trend of 2026 – is driven by new technologies (including AI) designed to improve health care, by the feeling that health care is not personalized enough, and by the seductive notion that a person’s own physiological data holds some of the secrets to better health. Surely it does. Surely there’s value in many aspects of personalized medicine. But not everything can be quantified – at least not reliably – and not every biomarker has a simple relationship to health. I believe that personalized medicine, with or without AI, is only part of what’s needed to be as healthy as possible.

What else do we need? Healthy diets, good sleep, sufficient activity, minimal stress, and many other practices I hope to cover in this newsletter in the coming year.

Thanks for reading!