Predictions about healthcare have a poor track record. Every year brings confident forecasts that some technology will transform care within twelve months, and most of those forecasts quietly expire. So a word of warning before we start: everything below is opinion. These are our bets, not findings. Each one is anchored in a trend we can document today, and each comes with the reason it might not happen.
We would rather be specific and wrong than vague and safe. Hold us to these.
1. Cheap semaglutide will become a public health story, not just a pharma story
The trend: In early 2026, core semaglutide patents expired in several large markets. In India, generic versions launched in the first quarter of the year, and generics have been approved in Canada. In China, where the core patent also expired in March, more than a dozen domestic candidates are in development or under review, but approvals have been slower; Bloomberg reported in May that Novo Nordisk expects generic competition there to be delayed until next year.
The prediction: In 2027, the debate about GLP-1 drugs in these markets will shift from access for the wealthy to questions health ministries actually have to answer. Should public systems fund them? For whom? Who monitors long-term use, side effects and discontinuation? We expect at least one large middle-income country to set out a formal public policy on subsidised GLP-1 use for obesity or diabetes.
Why we might be wrong: Supply, quality regulation and litigation could slow generic uptake. And cheap drugs do not automatically become public policy; they can simply become a private market.
2. The price of a QALY will become a trade issue elsewhere
The trend: From April 2026, England’s NICE raised its cost-effectiveness threshold from £20,000 to £30,000 per QALY to £25,000 to £35,000, the first change in more than two decades. The announcement came in the context of a US-UK agreement on pharmaceuticals.
The prediction: We think other countries with health technology assessment bodies will face similar pressure in 2027 to change how they value medicines, with drug pricing appearing more explicitly in trade negotiations. The age of treating cost-effectiveness thresholds as purely technical decisions is ending.
Why we might be wrong: Many systems have strong political and legal protections around their pricing processes, and may resist pressure that the UK did not.
3. Global health funding gaps will turn into visible service cuts
The trend: The United States’ withdrawal from the WHO took effect in January 2026. Meanwhile, the UN’s child mortality estimates show that progress on under-five mortality has slowed markedly since 2015, and the latest report explicitly warns about uncertainty around stable funding.
The prediction: In 2027, the funding shifts of the past two years will become more visible on the ground: in stock-outs, reduced surveillance, cancelled programmes and delayed campaigns. We also expect more countries to announce plans to increase domestic health financing, though fewer will deliver it.
Why we might be wrong: Other donors, philanthropies and domestic budgets may fill more of the gap than expected. And some effects will take longer than a year to show up in data.
4. Local validation of clinical AI will become a procurement requirement
The trend: The US FDA’s list of AI-enabled medical devices has passed 1,000 authorisations. Yet evidence that tools perform as advertised in new settings remains thin; the widely used Epic Sepsis Model, for example, performed far worse in an independent validation at Michigan Medicine than its developer had reported.
The prediction: In 2027, more large health systems will make local validation and ongoing performance monitoring a condition of buying clinical AI, rather than trusting vendor figures. Expect procurement contracts to include requirements for local testing, bias audits and the right to switch tools off.
Why we might be wrong: Validation is expensive, and smaller hospitals may simply not have the capacity. Speed-to-deploy pressures could win.
5. AI documentation tools will be judged on workload, not wow
The trend: Ambient AI scribes, which draft clinical notes from recorded consultations, have spread rapidly through health systems over the last two years.
The prediction: 2027 is the year these tools face harder questions. Do they reduce time spent on documentation after hours? Do they introduce errors? Do they change what happens in the consultation? We expect the conversation to shift from pilot enthusiasm to hard evidence, and some systems to quietly scale back where gains do not materialise.
Why we might be wrong: If clinicians love the tools, adoption may continue regardless of what the evidence says, as has happened with many other technologies.
6. Newborn survival will become the headline child health issue
The trend: The UN Inter-agency Group for Child Mortality Estimation reports that of an estimated 4.8 million under-five deaths in 2023, 2.3 million were newborns, and that newborn mortality has fallen more slowly than mortality in older children.
The prediction: We expect global child health strategy in 2027 to shift further towards care around birth and for small and sick newborns, because that is where the remaining deaths are concentrated.
Why we might be wrong: Funding constraints may push donors back towards cheaper, campaign-style interventions with faster measurable results.
7. The workforce question will dominate everything else
The prediction: None of the above works without people. We think 2027 will be the year when health workforce shortages, migration and burnout become the constraint discussed first, not last, in almost every health policy debate, from AI to financing.
Why we might be wrong: It is possible the conversation is already there, and 2027 will simply repeat it without changing much.
How we will score ourselves
A prediction that cannot be checked is just a mood. So here is what would count as being right or wrong on each, when we look back at the end of 2027.
On semaglutide, we are right if at least one large middle-income government publishes a formal policy on public funding or subsidy of GLP-1 drugs for obesity or diabetes, with defined eligibility. We are wrong if the only story is private-market price competition. A pilot in one region or a statement of intent does not count.
On thresholds and trade, we are right if drug pricing or health technology assessment rules appear explicitly in at least one more bilateral trade agreement or negotiation involving a country with an HTA body. We are wrong if the UK case remains a one-off.
On funding cuts, we are right if major global health programmes report documented service reductions, such as stock-outs, suspended campaigns or reduced surveillance, attributed to funding shortfalls. This is the prediction we most hope to get wrong.
On clinical AI validation, we are right if large health systems or national purchasers publicly adopt local validation, bias auditing and ongoing monitoring as written procurement conditions. Guidance documents that are not tied to contracts do not count.
On AI scribes, we are right if the debate in 2027 is driven by published evaluations of time saved, errors and patient experience rather than by vendor case studies, and if at least some systems publicly narrow or pause deployments based on that evidence.
On newborns, we are right if major global child health strategies and funding announcements in 2027 give explicit, prominent priority to care around birth and for small and sick newborns.
On workforce, we are honest enough to admit this one is the hardest to score. We will count it as right if workforce is the headline constraint in the major national health plans and global reports published during the year.
What connects the seven
Read together, these predictions share a thread. Almost all of them are about the gap between a change that has already happened and the institutions that have to absorb it.
Cheap semaglutide is a change in price that forces a decision about public policy. A higher QALY threshold is a change in one country that creates pressure on others. Funding withdrawals are decisions already made whose effects have not yet reached clinics. Clinical AI and ambient scribes are technologies already deployed whose evidence base is catching up. Newborn mortality is a trend already visible in the data that strategy has not fully caught up with. And the workforce is the institution that has to absorb all of the above.
That is deliberate. The safest predictions are not about new inventions, which arrive on unpredictable timelines. They are about the consequences of things that have already happened. Institutions move slowly, but they do move, and the direction is often visible a year or two in advance.
The strongest case against our whole approach
There is a fair objection to this list. By anchoring each prediction in a trend that is already visible, we may be predicting the obvious and missing the surprises that will actually define 2027. Most years in health are shaped by events nobody forecast: an outbreak, a court ruling, a political shock, a trial result that changes practice overnight. A list of extrapolations will always miss those.
We accept the point. Our defence is that surprises are, by definition, not things anyone can usefully bet on, while the slow consequences of known changes are where health leaders actually have room to prepare. A ministry cannot plan for an unknown outbreak in detail. It can plan for what happens when a drug that millions of people want becomes cheap.
What each prediction asks of health leaders now
If even half of these prove right, several groups have work to do before 2027 arrives.
Payers and ministries in countries where GLP-1 generics are arriving need a position on eligibility, monitoring and budget impact before demand forces one on them.
HTA bodies and their governments need to decide in advance how they will respond to pressure on their methods, rather than improvising during a trade negotiation.
Hospital procurement teams buying clinical AI need contract language for local validation and switch-off rights now, because it is much harder to add after a tool is embedded.
Clinical leaders running scribe pilots should be measuring after-hours documentation time and error rates from the start, so that the 2027 conversation is about their data rather than someone else’s marketing.
And everyone planning services should assume that workforce, not money or technology, will be the binding constraint, and test their plans against it.
The prediction we almost made
We considered an eighth bet: that 2027 would bring a serious, public failure of a clinical AI tool in routine use, of the kind that changes regulation. We left it out because it is too easy to claim after the fact and too hard to define in advance. But it is worth naming, because several of our other predictions, especially those about validation and scribes, become much more likely if that kind of event happens. Health policy often moves after a visible failure rather than before one. If you think that is the real prediction for 2027, tell us, and tell us which kind of tool you think is most exposed.
Your predictions
Which of these do you think we got wrong? What is the one change you expect to see in your own system in 2027? Put your prediction in the comments. At the end of 2027, the comments will show who called it better.
