The Future of Healthspan and AI: From Sick Care to Predictive Medicine

I have spent my career at the end of the pipeline.

Emergency medicine is where decades of small decisions arrive all at once, usually at an inconvenient hour, usually past the point where anything upstream would have been easier. The 58-year-old with crushing chest pain did not become a cardiac patient that evening. He became one over twenty years, and I met him at minute four.

That is the actual argument for predictive medicine. Not that technology is exciting. That the current system meets people too late.

The number that should reframe the debate

Americans are not just dying of chronic disease. They are living with it, for a long time, at the end.

A study of all 183 World Health Organization member states found that the global gap between lifespan and healthspan has widened to 9.6 years. The United States has the largest gap on earth at 12.4 years, driven by the burden of noncommunicable disease. Women worldwide carry a gap 2.4 years wider than men. According to PubMed, that is Garmany A, Terzic A, JAMA Network Open, 2024 (DOI).

Sit with that. The average American spends over a decade at the end of life in a health state the data classifies as impaired.

Longevity is not the problem anymore. The problem is the shape of the last chapter.

Where AI genuinely moves this

I want to separate what has evidence from what has a podcast.

Earlier detection, with trial data behind it

The strongest published example is breast cancer screening. The MASAI trial randomized 105,934 women in the Swedish national screening program to AI-supported screening or standard double reading. AI-supported screening detected 6.4 cancers per 1,000 participants against 5.0 per 1,000, a 29 percent increase, and the additional cancers were mainly small, lymph-node negative invasive cancers. False positives did not rise significantly. Screen-reading workload dropped 44.2 percent. According to PubMed, that is Hernström V, Josefsson V, Sartor H, et al., Lancet Digital Health, 2025 (DOI).

Catching more small node-negative cancers is a healthspan intervention, not only a mortality one. Stage at detection determines how much of a person's remaining life is spent in treatment.

Risk stratification that reaches the people who never come in

The population that most needs preventive attention is the population least likely to schedule it. Predictive models built on existing records can find high-risk patients who have not presented, which converts prevention from something patients must initiate into something a health system can go out and offer.

That only works if somebody funds the outreach. A risk score with no team attached to it is a report, not a program.

Giving clinicians the time to actually do prevention

Prevention takes minutes that primary care does not have. Reducing documentation burden creates some of them. In a multicenter study of 263 ambulatory clinicians, an ambient AI scribe was associated with a mean reduction of 0.90 hours per day of after-hours documentation and a large drop in note-related cognitive task load, alongside burnout falling from 51.9 percent to 38.8 percent. According to PubMed, that is Olson KD, Meeker D, Troup M, et al., JAMA Network Open, 2025 (DOI).

Unglamorous, and probably the highest-yield healthspan intervention available to most health systems this year.

Where the field is overselling

I say this as someone who speaks about the future of medicine for a living, which is exactly why I should say it.

Biological age scores. Interesting research, weak clinical actionability today. Different algorithms disagree with each other on the same person, and almost none have been tested against outcomes in a way that would change what I do for a patient.

Consumer wearable diagnostics. Genuinely useful for behavior and for atrial fibrillation detection in defined populations. Not a substitute for a diagnostic workup, and prone to producing anxiety and downstream testing in healthy people.

Whole-body imaging marketed on healthspan. The incidental finding rate is real, the follow-up burden is real, and the evidence that routine screening of asymptomatic adults improves outcomes remains thin.

Supplement and protocol stacks with AI branding attached. A recommendation engine wrapped around weak evidence is still weak evidence.

The honest position is that the biggest healthspan levers remain sleep, movement, diet, blood pressure, tobacco, alcohol, social connection, and access to primary care. AI helps most by making those cheaper to deliver and easier to sustain, not by replacing them with something novel.

Why the emergency department is the wrong place to judge this

There is a bias I have to name in my own thinking.

Emergency physicians see prevention only when it fails. Nobody comes to my department to tell me their blood pressure has been controlled for fifteen years. That selection effect makes clinicians like me systematically pessimistic about upstream care, and it is worth correcting for.

It also makes the department an early warning system that nobody reads. Patterns show up in the emergency department years before they show up in population statistics: which neighborhoods are losing primary care access, which medications people have stopped filling, which conditions are arriving later and sicker than they did two years ago. Most of that signal is sitting in records that no one analyzes for public health purposes.

If a health system wanted one underused source of predictive information, it already owns it. The people who are using the emergency department as their primary care are telling the organization exactly where its prevention strategy has holes.

What predictive medicine requires that we do not yet have

Evidence standards that match the claim. Prediction is not prevention. A model that identifies risk has to be paired with an intervention that changes outcomes, and the pair has to be tested together.

Equity built in, not audited later. The healthspan gap is not distributed evenly, and neither is technology adoption. Ambient AI adoption ran at 64.7 percent in metropolitan hospitals against 54.3 percent outside them, and 67.6 percent among hospitals in the highest operating-margin quartile against 58.0 percent in the lowest (Yang F, Graetz I, American Journal of Managed Care, 2026). A predictive medicine era that follows that same curve widens the gap it claims to close.

Payment that rewards the years, not the visits. This is the real obstacle, and it is not technical. Almost nothing in the US payment system pays for a decade of health that never turned into an admission.

Governance, because prediction touches people who are not yet patients. Telling a healthy 45-year-old that a model flagged her carries consent, privacy, and psychological weight that a diagnostic tool applied to a symptomatic patient does not.

What I would build first

If I ran a health system and wanted a healthspan strategy that survives contact with reality, I would start with three things.

Deploy the tools that give clinicians time back, and spend that time on prevention rather than on throughput. Adopt AI-supported screening where randomized evidence exists, starting with breast imaging. And pick one high-prevalence chronic condition, build a risk model, and staff the outreach team before turning the model on.

Then measure something honest. Not engagement. Years of healthy life in a defined population, tracked over a decade.

The last chapter

Medicine got very good at extending the end of life. It has been much worse at improving it.

Twelve point four years. That is what the United States currently hands its citizens at the close, and no algorithm fixes it alone. But the tools now exist to find disease earlier, to reach the people who never make the appointment, and to give clinicians back the minutes that prevention requires.

I would rather meet that 58-year-old at year twenty than at minute four.

Harvey Castro, MD, MBA is a board-certified emergency physician, 5x TEDx speaker, and author of more than 30 books on AI and healthcare, including AI in Emergency Medicine (Wiley). He serves on Singapore's Ministry of Health Regulatory Advisory Panel and advises the Texas Medical Association's Committee on Health Information Technology.

Related DR GPT™ reading

• How AI Can Help the Healthcare Workforce Crisis Without Replacing Clinicians
• The Physician's Role in Responsible AI Adoption

More from DR GPT™: who DR GPT™ is, healthcare AI keynotes, board advisory, speaking, TEDx talks, books, and the media kit.

Book Harvey Castro, MD, MBA, known as DR GPT™, for a healthcare AI keynote, board retreat, or executive session. Start the conversation.