The Trump administration gives US states 6 more months before Medicaid's work rules can end anyone's cover

The stories

22 on this beat today.

Every story told the same way: what happened, who is involved, why it matters, how it works, the truth and the good news.

1. 01

   The Trump administration gives US states 6 more months before Medicaid's work rules can end anyone's cover

   6 months

   more time before states must enforce Medicaid's work rules

   announced by the Trump administration on Friday [1]

   20 million

   adults in 44 states covered by the work rule

   they must show 80 hours a month from 1 January 2027 [3]

   69%

   of people who lost Medicaid in the last recheck lost it over paperwork

   over 25 million lost cover in all [6]

   $326 billion

   saved by the US government over 10 years, budget officials estimate

   with 5.3 million more people uninsured [5]

   What happened

   • The Trump administration said on Friday it will give US states 6 more months to enforce Medicaid's new work rules. [1]
   • Medicaid is the US government's health cover for people on low incomes. [3]
   • States can now wait until July before ending cover for people who do not meet the rule. [2]
   • KFF, a US health research group, counts about 20 million adults covered by the rule, in 44 states. [3]
   • From 1 January 2027, those adults must show 80 hours a month of work, study or volunteering. [3]
   • They must also prove it again every 6 months, instead of once a year. [3]
   • People who are disabled, caring for young children or 'medically frail' are let off. [1]
   • Lawyers for 24 states sued over the rules in June 2026. [4]
   • On 29 October a US judge ruled that the government's test for 'medically frail' went beyond the law. [4]
   • The judge also found states had too little time or money to rebuild their systems. [1]
   • The delay came in a letter to state Medicaid directors, after that loss in court. [2]
   • Dan Brillman oversees Medicaid at the Centers for Medicare and Medicaid Services, or CMS. [2]
   • Brillman wrote: "We have heard directly from states about their implementation timelines." [2]
   • States must still begin checking in January whether people on Medicaid work. [2]
   • Hospitals have prepared big campaigns to help patients with the extra paperwork. [2]
   • After pandemic rules ended, US states rechecked everyone on Medicaid, and over 25 million people lost it. [6]
   • 69% of them lost it for paperwork reasons. [6] Only 31% were found no longer eligible. [6]
   • Arkansas tried a work rule in 2018, and more than 18,000 people lost cover that year. [7]
   • Researchers found no significant rise in the number of people working. [7]
   • The US Congressional Budget Office says the rule saves the US government $326 billion over 10 years. [5]
   • It also expects 5.3 million more people to end up uninsured. [5]

   Who is involved

   • Disabled people let off
   • Carers of young children let off
   • The medically frail let off
   • Dan Brillman runs Medicaid at CMS, wrote the letter
   • The Trump administration the US government; it wrote Medicaid's work rules and gave states 6 more months
   • US states they run Medicaid day to day; 24 of them sued over the rules and won part of the case
   • Jacob Wallace a Yale public health expert on Medicaid; he says the rules' paperwork drives people off
   • KFF a US health research group that counts who is covered and who loses cover

   How it unfolded

   1. June 2026 24 states sue over the rules [4]
   2. 29 Oct a judge rules the frailty test went beyond the law [4]
   3. January 2027 checks begin
   4. July 2027 cover can end
   5. 1 January 2027 work rules start
   6. 6 months on cover can end
   7. 2018 Arkansas tries a work rule; more than 18,000 lose cover [7]
   8. 2023-24 states recheck everyone; over 25 million lose Medicaid [6]
   9. 13 Nov states get 6 more months to enforce [1]

   Why it matters

   States that wait would end no one's cover under this rule before July 2027. [2]

   Iowa begins its own checks on 1 December, and Montana began in July. [10]

   Think it through

   1. 1

      Do most people on Medicaid not work?

      Most adults the rule covers already work or could qualify to be let off, KFF finds. [5]

      Already workingmost

      ++

      Could be let offmany more
   2. 2

      So how do people who qualify lose cover?

      Jacob Wallace of Yale says people drop off over a wrong address, a wrong phone number or a confusing form. [7]

      A form is sent→Wrong address or confusing→Cover ends
   3. 3

      How big was that last time?

      In the last big recheck, 69% of the people who lost cover lost it over paperwork. [6]

      Paperwork69%

      Found not eligible31%
   4. 4

      Who gains if people drop off?

      The US government spends $326 billion less over 10 years, the budget office says. [5] It also counts 5.3 million more uninsured people. [5]

      saves $326 billionUS government

      vsvs

      5.3 million uninsuredPeople
   5. 5

      Does the rule get more people into work?

      Budget officials expect the rule to add almost no jobs, KFF reports. [9]

      Work rule→Almost no new jobs→Millions uninsured
   6. 6

      Who has to do all the checking?

      Each state must build new systems, hire and train staff and reach enrollees, in a short time. [8]

      New rules→New computer systems→Staff to do the checks
   7. 7

      Is anyone let off but still at risk?

      Native Americans are let off the rule, but states plan to ask them for proof in different ways. [10]

   8. 8

      So what is really going on?

      Each state can now wait 6 months before ending anyone's cover. [2] The more checks it asks for, the more people drop off, Wallace says. [7]

   How it works

   Public administration researchers call this administrative burden. It is the time, forms and proof it takes to get help.

   The rule adds a check every 6 months, not once a year. [3] Each extra hoop loses more people, Wallace says. [7]

   Seen it before

   A gym that makes members sign up again every month loses some who simply forget, not because they quit.

   The truth

   • What is really going on The Trump administration gave US states 6 more months before Medicaid's work rules can end anyone's cover. [1][2]In the last big recheck, 69% of the 25 million people who lost Medicaid lost it over paperwork, not because they stopped qualifying. [6]
   • Why it works on people The name 'work requirement' points at jobs, while budget officials expect the rule to add almost no jobs and to leave millions uninsured. [9]
   • The 5.3 million is a forecast KFF cited in June, before October's ruling eased the frailty test. [5][4]
   • Nobody knows yet

     How many people will lose Medicaid once states enforce the work rules.

     The 5.3 million forecast KFF cited in June came before the October court ruling, and states may now wait 6 months. [5][4][2]

   What comes next

   The rules still start on 1 January 2027. [8] States may then wait 6 months before ending cover. [2]

1. 02

   Native Americans' Medicaid proof differs by state

   What happened

   • Native Americans are let off Medicaid's new work rules under the 2025 US law. [10]
   • But the federal rules set no common way for states to check who is Native American. [10]
   • KFF Health News asked 14 states, and found they plan to ask for proof in different ways. [10]
   • Arizona and 5 other states let people tick a box on their form. [10] North Carolina asks for a tribal card or record. [10]
   • About 87% of Native people live outside tribal areas such as reservations, 2020 data shows. [10]
   • In the last Medicaid recheck, about 850,000 Native Americans lost cover, often over paperwork. [10]

   Who is involved

   • Native Americans let off the work rule

   Why it matters

   A Navajo citizen living in Connecticut may struggle to prove their status, a tribal policy director says. [10]

   Think it through

   1. 1

      Why can't states just check a list?

      Several states say they do not track tribal ties, because the question on forms is optional. [10]

      Question is optional→No record kept→Proof is harder
   2. 2

      Who could fix this before January?

      CMS, the US agency that runs Medicaid, did not answer whether it will issue more guidance. [10]

   3. 3

      So what is really going on?

      The law lets Native Americans off, but each state's proof decides who stays covered. [10]

      Law lets them off→State asks for proof→Proof decides cover

   How it works

   Public administration calls the time and forms needed to prove you qualify compliance costs, paid by the person.

   Native leaders fear people who qualify will fall through the cracks, as in the last recheck. [10]

   Seen it before

   A free school lunch still needs a form, and a child whose form gets lost goes without.

   Why people who qualify for health cover can still lose it to paperwork

   Public administration researchers call it administrative burden: every form and proof a programme asks for makes some people who qualify drop off.

   The twist

   A programme can shrink without anyone becoming less eligible. The extra forms do the shrinking.

   The picture

   Lost it over paperwork69%

   Found not eligible31%

   Step by step

   1. 1 A programme sets who qualifies for help
   2. 2 It asks people to prove it, again and again
   3. 3 Each form costs time, know-how and nerve
   4. 4 Some who qualify miss a step
   5. 5 Their help stops, though nothing about them changed

   The same force, elsewhere today

   Where this chain is also running, in today's other stories.

   • Native Americans' Medicaid proof

     The law lets them off the work rule, but each state sets its own proof, so a missing tribal card can still end cover.

   • The blind woman told to seek care in India

     She meets the rules, yet must prove her case again every 30 months for about 20 years.

   • States' rules for AI in health care

     Eight states now stop insurers denying cover by computer alone, at the sign-off patients need before treatment.

   Where you've seen this

   Free school meals

   families who qualify miss out when the form is long or in a language they do not read

   Tax refunds

   people owed money never claim it because the claim takes hours

   Disability benefits

   repeat assessments stop payments to people whose condition has not changed

   The catch

   Some checks do stop help going to people who do not qualify. The hard part is counting how many people who do qualify each check also loses.

   And the whole of it

   Most of us meet forms like these and think of them as small. For someone ill, poor or far from an office, the same form can decide whether they see a doctor next year.

   The truth

   • Nobody knows yet

     How each state will check that a Native American is let off the rule.

     CMS did not answer whether it will issue more guidance, and the 14 states asked plan different proof. [10]

   Good news

   • In Montana, Native people who declare their status need no more proof. [10]

1. 03

   New governors could widen Medicaid

   What happened

   • Democrats won the governor's race in Georgia, Kansas, Texas and Wisconsin in the US midterms. [11]
   • Those 4 are among the 10 US states that never widened Medicaid to more low-income adults. [11]
   • KFF counts 2.6 million uninsured adults in those 10 states who would qualify if they did. [11]
   • Nearly half of them live in Texas. [11]
   • The new governors in Georgia and Texas promised to widen Medicaid. [11]

   Who is involved

   • Georgia new governor
   • Kansas new governor
   • Texas new governor
   • Wisconsin new governor

   Why it matters

   In Georgia, about 185,000 people could fall into a gap when a state programme ends this year. [11]

   Think it through

   1. 1

      Why would work rules make widening easier?

      In Mississippi, plans to widen Medicaid failed because work rules could not be added. [11]

      No work rule allowed→Plan fails→Work rules now national
   2. 2

      Can the governors decide it alone?

      No: past attempts in these states failed in their state lawmakers' votes. [11]

   How it works

   Political scientists call it divided government when one party does not hold every branch.

   In Georgia and Texas, Democrats won the lower house, so control of state lawmaking is now split. [11]

   Seen it before

   A class president can promise a later bell, but the teachers still have to agree.

   What comes next

   The new governors take office in January 2027. [11]

2. 04

   Rural health money splits evenly, not by need

   What happened

   • On 9 November CMS announced the second year of awards from a $50 billion US rural health fund. [12]
   • The 2025 law set up the fund to soften its own cuts to rural areas. [12]
   • Awards average $200 million a state, from $142 million in New Jersey to $282 million in Texas. [12]
   • Texas has about 30 times as many rural residents as New Jersey. [12] It gets only about twice as much. [12]

   Why it matters

   The law is expected to cut Medicaid spending in rural areas by about $137 billion. [12]

   Think it through

   1. 1

      Why does a state with few rural people get so much?

      By law, half the fund is split equally, so each state gets $100 million a year from it. [12]

      $100 milliona year to every state, whatever its need

   2. 2

      Does the money match the cut?

      The fund is $50 billion over 5 years. [12] The rural cuts are about $137 billion over 10 years. [12]

      $50 billion, 5 yearsFund

      vsvs

      $137 billion, 10 yearsRural cuts

   How it works

   Public finance calls money sent out by a fixed rule, not by request, formula funding.

   Because 75% of the fund is locked by fixed rules, every state's second award was within 4% of its first. [12]

   Seen it before

   Splitting a pizza into equal slices gives the small eater as much as the hungry one.

3. 05

   Hundreds of children in England get weight-loss jabs

   What happened

   • Children in England got 419 prescriptions for weight-loss jabs in 2025-26, the Guardian found. [13]
   • The drugs, called GLP-1s, were first made for type 2 diabetes and include Wegovy and Mounjaro. [13]
   • From April to July this year there were 341 more, 15 of them for children under 12. [13]
   • The figures leave out private, prison and hospital prescriptions, so the real number is likely higher. [13]
   • Last month the WHO said these drugs should not be given to children under 10. [13]

   Why it matters

   In England, 22.8% of children aged 10 and 11 are obese, national figures show. [13]

   Think it through

   1. 1

      Why give a child a slimming drug?

      NHS clinic doctors say they give it to very heavy children with illnesses such as type 2 diabetes. [13]

   2. 2

      What happens when the jabs stop?

      Weight tends to return when the drugs stop, says Martin Whyte, a professor of metabolic medicine. [13]

      Jab taken→Jab stopped→Weight returns
   3. 3

      Is use about to grow?

      A clinic doctor says they may widen use from age 12 to 8 once the licence allows. [13]

      from 12Now

      ↓→

      from 8Planned

   How it works

   Pharmacologists say these drugs copy a gut hormone that tells the brain the body has eaten.

   Whyte says children who start need a long-term plan, partly because weight tends to return. [13]

   Seen it before

   Glasses help you see only while you are wearing them.

   The truth

   • The count came from freedom of information requests, and it is unclear how many were for weight loss. [13]
4. 06

   US organ system writes new donor rules

   What happened

   • The US organ transplant network approved its first national rules for NRP on 6 November. [14]
   • In NRP, doctors pump blood back through a donor's organs after they are declared dead. [14]
   • The rules start on 6 December and tell surgeons to cut, not only clamp, the vessels to the brain. [14]
   • Some teams saw signs of life when blood reached the brain during NRP, a transplant surgeon says. [14]
   • A separate rule from 18 November lets donation teams talk to families before they decide to stop treatment. [15]

   How it unfolded

   1. 18 November families approached earlier
   2. 6 December NRP rules begin

   Why it matters

   More than 109,000 Americans wait for a transplant, and about 12 die each day. [14]

   Think it through

   1. 1

      Why restart blood flow after death at all?

      It stops organs spoiling, so more and better organs reach patients, a US transplant official says. [14]

      Blood flow restarted→Organs stay healthy→More transplants
   2. 2

      Then why do critics object?

      Bioethicist Lainie Ross says: "It's taking organs from people who are no longer dead." [14]

   3. 3

      Do the new rules settle it?

      Critics say most surgeons already clamp and cut, so the rules add little protection. [14]

   How it works

   Medical ethicists call it the dead donor rule: organs may be taken only after death. [14]

   NRP restarts circulation after death was declared by circulation stopping, so critics say it bends that rule. [14]

   Seen it before

   A referee who restarts the clock after the final whistle makes people ask whether the game really ended.

   The truth

   • The first part started as an emergency, before the public could comment, a surgeon on a network committee objects. [14]
   • Nobody knows yet

     Whether the new US rules make organ donors any safer.

     Critics say most surgeons already clamp and cut the vessels, and the first part began before the public could comment. [14]

   What comes next

   The early approach to families starts on 18 November. [15] The NRP rules start on 6 December. [14]

5. 07

   US moves to limit strong kratom products

   What happened

   • A US federal limit on some strong 7-OH products could start as early as 13 November, NPR reported. [16]
   • 7-OH is one of the main mind-altering chemicals in kratom, which comes from a tree in the coffee family. [16]
   • Concentrated 7-OH is sold as tablets, gummies and drinks in petrol stations and smoke shops. [16]
   • In August the DEA, the US drug police, restricted 3 related opioids made in labs. [16]
   • After US health officials advised on 9 October, the DEA narrowed its plan to higher-strength products. [16]
   • Ordinary kratom leaf is not expected to be banned. [16]

   How it unfolded

   1. August 3 related opioids restricted
   2. 9 October plan narrowed to strong products

   Why it matters

   Calls about kratom to US poison centres rose about 1,200% between 2015 and 2025. [16]

   The same report counted 233 deaths, and 79% of them involved other drugs too. [16]

   Think it through

   1. 1

      Why do people take it at all?

      Some people use it to ease opioid withdrawal or long-term pain, and say it helped them. [16]

   2. 2

      Then where is the harm?

      7-OH acts more like an opioid in the brain, and labels often do not show how much is inside. [16]

      Strong 7-OH→Unclear labels→Hard to know the dose
   3. 3

      Who decides what stays legal?

      The DEA and health officials do, after a kratom group sued over the August rule. [16]

   How it works

   Pharmacologists call how strongly a drug acts at a given dose its potency.

   The leaf holds only small amounts of 7-OH, so concentrated 7-OH products are far stronger. [16]

   Seen it before

   Lemon squash is a drink, but the concentrate straight from the bottle is far stronger.

   The truth

   • Estimates of US kratom users run from 5 million to 15 million. [16] Most do not separate leaf from 7-OH. [16]
   • Nobody knows yet

     Whether the US limit on strong 7-OH kratom products has taken effect.

     NPR reported it could start as early as 13 November, and no source here confirms that it did. [16]

6. 08

   Little regret after starting hormones young

   What happened

   • Of 372 people who began hormone therapy as young people, 2.7% said they regretted it, a US study finds. [17]
   • The study in JAMA Pediatrics surveyed former patients of one US gender clinic. [17]
   • On average they had started 4.9 years earlier. [17]
   • 96.8% said starting had been the right decision. [17]
   • 17.3% had stopped taking hormones. [17]

   Why it matters

   Many US states have restricted this care for young people, citing fears of regret. [17]

   Think it through

   1. 1

      Does stopping mean regret?

      Not usually: 83% of those who stopped still said starting was right. [17]

      83%of those who stopped still said it was right

   2. 2

      What does the main number hide?

      An outside psychologist notes the 17.3% who stopped is far above a Dutch study's 2%. [17]

      % who stopped

      This US clinic17.3

      Dutch study2
   3. 3

      Who did not answer?

      People assigned male at birth were less likely to answer the follow-up survey. [17]

   How it works

   Survey researchers call it non-response bias when the people who answer differ from those who do not.

   Nearly 300 of the 698 people invited did not reply, so their views are unknown. [17]

   Seen it before

   A cafe that asks only returning diners about the food hears mostly from people who liked it.

7. 09

   North Dakota caps insulin at $25 a month

   What happened

   • North Dakota caps what patients pay for insulin at $25 a month. [18]
   • North Dakota is one of 29 US states with such a cap, NPR and KFF Health News report. [18]
   • About a fifth of the 40 million Americans with diabetes depend on insulin. [18]
   • Insurers warned the cap would raise premiums for everyone else. [18]
   • Supporters showed a similar limit for state workers had little effect on premiums. [18]

   Why it matters

   The Kritzberger family pays at least $8,000 a year for drugs, supplies and doctor visits, despite insurance. [18]

   Think it through

   1. 1

      Why do people with insurance pay so much?

      High deductibles make some pay thousands before cover starts, but the insulin cap applies every month anyway. [18]

      High deductible→Pay thousands first→Cap applies anyway
   2. 2

      Haven't insulin prices come down?

      Yes: average yearly costs for privately insured people fell from $470 in 2018 to $220 in 2025. [18]

      2018$470 a year

      2025$220 a year
   3. 3

      So what is the cap really for?

      It guards families against future price rises, NPR reports. [18]

   How it works

   Health economists call a fixed limit on what a patient pays a cost-sharing cap.

   The cap limits only what the patient pays, so insurers said everyone else's premiums would rise. [18]

   Seen it before

   A school trip with a fixed low price means the school covers whatever it really costs.

   Good news

   • In North Dakota, Republicans and Democrats joined to pass the cap. [18]
8. 10

   The kind of fat changes keto's effect in mice

   What happened

   • Keto diets high in long-chain saturated fat raised inflammation in mice, a Science Advances study finds. [19]
   • A keto diet cuts carbohydrates so the body uses fat as its main fuel. [19]
   • Mice on a Crisco-based keto diet built up inflammatory immune cells in heart, liver, kidneys and brain. [19]
   • Crisco shortening has about 26% long-chain saturated fat, against about 12% in coconut oil. [19]
   • Mice on the coconut oil keto diet did not show the effect. [19]

   Why it matters

   Keto diets were first made to treat epilepsy, and are now widely used to lose weight. [19]

   Think it through

   1. 1

      So is keto bad for you?

      The University of Cincinnati team says not all keto diets are the same. [19]

      Crisco ketomore inflammation

      vsvs

      Coconut oil ketolittle change
   2. 2

      Why would one fat inflame and another not?

      The authors link long-chain saturated fat to stress in mitochondria, the parts of a cell that make its energy. [19]

      Long-chain saturated fat→Stressed mitochondria→Inflammation

   How it works

   Immunologists call an immune reaction set off by damaged cells, with no germ present, sterile inflammation.

   Here signs of mitochondrial stress drew inflammatory white blood cells into the mice's organs. [19]

   Seen it before

   A smoke alarm can go off from burnt toast, not only from a real fire.

   The truth

   • The results come from mice fed for 21 days. [19]
9. 11

   Coffee drinkers' guts work differently

   What happened

   • Regular coffee drinkers had more varied gut bacteria, a University of Copenhagen study of 61 adults finds. [20]
   • Volunteers swallowed a small electronic capsule that measured their gut as it passed through. [20]
   • Those drinking 1 to 5 cups a day had much less acid in the small intestine. [20]
   • They also had more squeezing movements in the colon, which push waste along. [20]
   • The study was published in the journal Gut Microbiology. [20]

   Why it matters

   Many people notice that coffee affects their digestion, says Henrik Munch Roager, who leads the group. [20]

   Think it through

   1. 1

      Does coffee cause these changes?

      The researchers say the study cannot prove it, because it only compared people's habits. [20]

      Compare habits→Find a link→Cause not shown
   2. 2

      Is less gut acid good?

      The team says it cannot yet tell whether that helps or harms. [20]

   How it works

   Epidemiologists call a study that watches habits without changing them observational.

   Coffee drinkers may differ in other ways. [20] So the team plans a study that sets how much each person drinks. [20]

   Seen it before

   Children who eat breakfast may do better at school, but they may also simply sleep more.

10. 12

    UK tells a blind Nigerian woman to seek care in India

    What happened

    • The UK Home Office told a blind Nigerian woman, 67, that she could get treatment in India. [21]
    • She has lived in the UK for 26 years and has never been to India. [21]
    • She went blind from acute glaucoma, an eye disease, soon after arriving in 2000. [21]
    • Under new settlement rules, she may have to renew her visa every 30 months for about 20 more years. [21]
    • Home Office sources say a corrected letter will be sent. [21]

    Why it matters

    Some older applicants may die before they qualify: one, a woman from Pakistan, is 103. [21]

    Think it through

    1. 1

       Why can't she stay for good now?

       A fast track exists for exceptional cases, but the Home Office refused her. [21]

       Asks to settle now→Exceptional case route→Refused
    2. 2

       Who decides what counts as exceptional?

       The Home Office says it judges each case on its merits. [21] Her letter named a country she has no link to. [21]

    How it works

    Public administration calls the time spent proving a case again and again compliance costs, paid by the person.

    Each 30-month renewal is a new application, so she must keep proving the same facts into her 80s. [21]

    Seen it before

    A library card that must be proved again every few months, after years of use, wears people out.

    She meets the rules, yet must prove her case again every 30 months for about 20 years.

    The idea, in story 02: Why people who qualify for health cover can still lose it to paperwork →

    The truth

    • A Home Office spokesperson said: "We do not routinely comment on individual cases." [21]
11. 13

    US states write rules for AI in health care

    What happened

    • Lawmakers in 30 US states passed 115 measures on AI this year, an advocacy group counts. [22]
    • 8 states now stop insurers from using only AI or automated systems to deny or cut cover. [22]
    • That includes prior authorisation, the insurer's sign-off before a treatment. [22]
    • At least 5 states restrict AI in mental health care, mostly requiring a licensed human therapist. [22]
    • The American Medical Association, the biggest US doctors' group, released voluntary guidance in October. [22]

    Why it matters

    The Trump administration wanted one national framework that would override state AI laws. [22]

    Think it through

    1. 1

       Will a national rule override the states?

       KFF says it is less likely now, with Democrats controlling the US Congress from January. [22]

       Trump administrationone national rule

       vsvs

       New Congressless likely to agree
    2. 2

       Does it matter how an AI tool is set up?

       A study out on 14 November found that the instructions given to AI models changed their medical answers. [23]

       Instructions given→Answers shift→Different care

    How it works

    Regulators call keeping a person responsible for an automated decision human oversight. [22]

    Under these laws, a person, not software alone, must stand behind a denial of cover. [22]

    Seen it before

    A self-checkout still needs a staff member to approve some items.

    Eight states now stop insurers denying cover by computer alone, at the sign-off patients need before treatment.

    The idea, in story 02: Why people who qualify for health cover can still lose it to paperwork →
12. 14

    Instructions sway AI medical answers

    What happened

    • Telling AI models to play a certain kind of doctor changed their medical decisions, a study finds. [23]
    • Researchers tested 20 open AI models on 5,000 patient cases, with the facts held the same. [23]
    • With no persona, the models said yes to 42.8% of decisions. [23]
    • With 9 doctor personas, the yes rate swung from 36.9% to 46.4%. [23]
    • Bigger models and models trained on medicine were no safer from the effect. [23]

    Why it matters

    Hospitals using these tools could get different decisions just from how a tool is set up. [23]

    Think it through

    1. 1

       Which decisions moved most?

       Choices about patients' own wishes and about treatment moved the most. [23]

       Patients' own wishesmoved most

       ++

       Treatmentmoved most
    2. 2

       Who sets these instructions?

       Whoever runs the tool writes them, and the authors say that makes it a policy choice. [23]

    How it works

    Decision scientists call the point where a choice flips from no to yes a decision threshold.

    The persona moved that point, so one patient could get a yes from one set-up and a no from another. [23]

    Seen it before

    Two teachers marking one essay give different grades if one is told to be strict.

13. 15

    Leaked data made Alzheimer's AI look perfect

    What happened

    • Two popular brain-scan datasets made AI models look far better at spotting Alzheimer's than they are, a study finds. [24]
    • In one set, every test image came from a patient also in the training images. [24]
    • Fixing that cut the models' accuracy from 100% to about 46%. [24]
    • In the second set, about a fifth of the images were near-copies of others. [24]
    • A model trained on one set scored 25% on the other, no better than chance. [24]

    Why it matters

    Much published accuracy on these sets reflects memorised patients, not real diagnosis, the authors say. [24]

    Think it through

    1. 1

       How can a model score 100% and know nothing?

       It learned to recognise the same patients' scans, not the signs of disease. [24]

       Patient in training→Same patient in test→Recognised, not diagnosed
    2. 2

       What fixes it?

       The authors released new splits that keep each patient's scans on one side only. [24]

    How it works

    Machine-learning researchers call this data leakage: test answers slip into the training. [24]

    With one patient on both sides, the model only had to recognise each brain, so its score said little. [24]

    Seen it before

    A pupil who has seen the exam paper can get full marks without understanding the subject.

    Good news

    • Researchers found that two popular Alzheimer's scan datasets let AI models cheat, and released fixed versions that keep each patient's scans on one side.
14. 16

    The brain's seeing area helps fear memories last

    What happened

    • Lasting fear memories in people depend partly on the brain's visual area, a Nature Communications study finds. [25]
    • Volunteers learned to fear an image, then were tested at 15 minutes, 24 hours and 7 days. [25]
    • Traces of the memory showed 40 to 60 milliseconds after seeing the image. [25]
    • Electrical stimulation through the scalp over the visual area disrupted storing the memory, but not first learning it. [25]

    How it unfolded

    1. 15 minutes first test
    2. 24 hours second test
    3. 7 days third test

    Why it matters

    The authors say the visual area could be a target for treating fear that will not fade. [25]

    Think it through

    1. 1

       Isn't fear stored deep in the brain?

       Limbic regions take part, but this study shows early visual areas do too. [25]

       Limbic regionstake part

       ++

       Early visual areastake part too
    2. 2

       Could this treat phobias soon?

       No: this was a lab study of how memories form, not a treatment trial. [25]

       This studyhow memory forms

       vsvs

       A treatment trialnot done

    How it works

    Neuroscientists call the step that turns a new memory into a lasting one consolidation.

    Learning the fear made the visual area less held back. [25] Stimulation stopped that, and the fear memory weakened. [25]

    Seen it before

    Wet cement takes a footprint easily, but once it sets, the print stays.

15. 17

    Pregnancy danger videos rated below average

    What happened

    • Short videos about preeclampsia on two Chinese platforms are below average in quality, a study finds. [26]
    • Preeclampsia is a serious complication of pregnancy, affecting 2% to 8% of pregnancies worldwide. [26]
    • Researchers rated 164 videos on Douyin and Bilibili with 3 standard quality tools. [26]
    • Videos by doctors in the right specialty made up 64% of the sample. [26]
    • Causes and how common it is made up less than 10% of what the videos covered. [26]

    Why it matters

    Short-video apps are now an important way people find health information, the authors note. [26]

    Think it through

    1. 1

       Are popular videos better?

       No consistent link was found between likes or shares and quality. [26]

       Likes and sharespopularity

       vsvs

       Quality scoreno steady link
    2. 2

       Who could fix it?

       The authors say platforms should check the credentials of those uploading. [26]

    How it works

    Health communication researchers score videos with tools like DISCERN, which rates how reliable health information is.

    Here specialist doctors' videos scored higher on 2 of the 3 tools. [26]

    Seen it before

    A recipe from a trained chef is more likely to work than one from a stranger, but not always.

16. 18

    In rural Somalia, a birth emergency waits on others

    What happened

    • In rural Somalia, a woman's pregnancy emergency often waits on her husband's decision and money, a study finds. [27]
    • Researchers interviewed 32 women, husbands and traditional birth attendants in Jowhar and Burhakaba districts. [27]
    • Cost and transport delayed care further. [27] Fear of poor hospital care slowed referrals. [27]
    • Traditional birth attendants were trusted, but often limited in sending women on. [27]

    Why it matters

    Deaths of mothers remain very high in fragile places such as Somalia, the authors note. [27]

    Think it through

    1. 1

       Is the problem just the distance to a hospital?

       No: the delays fed each other, and expected bad care slowed the first decision. [27]

       Expect poor care→Decide later→Arrive later
    2. 2

       Would teaching danger signs fix it?

       The authors say awareness or more clinics alone are unlikely to work. [27]

    How it works

    Public health uses the three delays model: deciding to seek care, reaching it, and receiving it. [27]

    Here the first delay came because men held the decision and the money, even when women saw danger. [27]

    Seen it before

    A pupil who feels ill must wait for a teacher's note before going home.

17. 19

    Vegetables on prescription in Cornwall

    What happened

    • Patients with type 2 diabetes or pre-diabetes in Cornwall get free weekly vegetable boxes, from a scheme called WellFed. [28]
    • More than 100 people have joined, sent by GPs at more than 20 surgeries. [28]
    • They also cook and garden together at 20 sites across Cornwall. [28]
    • An early report says the group's average blood sugar fell from the diabetic to the pre-diabetic range. [28]
    • One man's HbA1c, a blood sugar test, fell from 50 to 42, below the risk level of 48. [28]

    Why it matters

    Cornwall has a high rate of pre-diabetes compared with the rest of England. [28]

    Think it through

    1. 1

       Does this show vegetables treat diabetes?

       No: it is an early report, and one man also had other medical treatment. [28]

    2. 2

       Who pays for it?

       Cornwall council's public health team and the local NHS fund it. [28]

    How it works

    Doctors track long-term blood sugar with a test that averages it over about three months.

    Michael Waller's HbA1c fell from 64 to 47 after he swapped chocolate and ice cream for vegetables and pulses. [28]

    Seen it before

    A school report averages a whole term, so one bad test barely moves it.

    Good news

    • In Cornwall, the average blood sugar of people on a free vegetable-box scheme fell from the diabetic range to the pre-diabetic range, an early report says.

    What comes next

    If the final results match, the scheme may be run more widely, the project hopes. [28]

18. 20

    Breathing training adds to weight training

    What happened

    • Adding breathing training to weight training improved lung and heart measures more than weights alone, a small trial finds. [29]
    • 28 overweight young women in China trained 3 times a week for 8 weeks. [29]
    • Both groups saw lower blood fats and blood pressure. [29]
    • The breathing group gained more in lung capacity and fitness. [29]

    Why it matters

    Inactivity and overweight are big health problems for young women in China, the authors write. [29]

    Think it through

    1. 1

       How sure is this?

       The trial had only 28 people, split into 2 groups. [29]

       28people in the whole trial

    2. 2

       Why would breathing help fitness?

       Breathing training raised how much air the lungs could move at their peak. [29]

       Breathing muscles trained→More air moved→Better fitness

    How it works

    Exercise physiologists call the most oxygen a body can use in hard effort its peak oxygen uptake.

    Both groups raised it, but breathing training also raised how much air the lungs could move. [29]

    Seen it before

    A wider straw lets you drink a milkshake faster.

    The truth

    • The trial had no group that did no exercise at all. [29]
19. 21

    UK tops a first index of disease prevention

    What happened

    • The UK ranks first of 114 countries in a new index of how ready they are to prevent disease. [30]
    • FTI Consulting, an advisory firm, built the Preventive Health Readiness Index from 27 measures. [30]
    • Finland, Australia, Norway and Sweden came next. [30]
    • Even the UK scored only 77.8 points. [30]
    • The UAE came 23rd, with a perfect 100 for its taxes on sugary drinks. [30]
    • It scored just 45 on tobacco control, against 82.5 for the UK. [30]

    Who is involved

    • UK first
    • Finland second
    • Australia third

    Why it matters

    Some middle-income countries, including Thailand, did better than richer ones. [30]

    Think it through

    1. 1

       How can a tax change what is in a drink?

       The UAE's tax now rises with the sugar in each 100ml, which led makers to change their drinks. [30]

       Tax set by sugar→Makers cut sugar→Less sugar on sale
    2. 2

       Why is the UAE weak on smoking despite a big tax?

       A pack of 20 cigarettes costs about Dh20 in the UAE, against about Dh90 in the UK. [30]

       UAE20 Dh a pack

       UK90 Dh a pack
    3. 3

       What does the index leave out?

       It scores policies and readiness, not how healthy people are today, its co-author Karim El Msan says. [30]

    How it works

    Public health calls stopping an illness before it starts primary prevention.

    The UAE's sugar tax led makers to reformulate their drinks, so what is on sale carries less sugar. [30]

    Seen it before

    A school that takes out the sweet machine cuts sugar for every pupil, without a lecture.

20. 22

    Russia's health budget falls behind prices

    What happened

    • Russia's health spending has fallen 728 billion rubles behind inflation since its full-scale war began. [31]
    • Novaya Gazeta Europe, an independent Russian outlet, made the calculation. [31]
    • Education fell 322 billion rubles behind. [31]
    • Its reporters found more than 800 staff cuts and closures in health and education since 2021. [31]
    • More than a quarter of them came in 2026. [31]

    Why it matters

    The cuts reach rich industrial regions such as Kemerovo and Sverdlovsk, not only poorer ones such as Karelia. [31]

    Think it through

    1. 1

       Did Russia actually cut its health budget?

       Not exactly: the gap is against inflation, so spending fell behind rising prices. [31]

       Budget rises a little→Prices rise more→Less care bought
    2. 2

       Why now, and not when the war began?

       Sharp cuts began in 2025. [31] The authors link them to the end of Russia's wartime boom of 2023 and 2024. [31]

       Wartime boom→Boom ends→Social spending cut

    How it works

    Economists call spending measured after rising prices real spending.

    Prices rose faster than Russia's health budget, so each year's budget paid for less care. [31]

    Seen it before

    If pocket money stays the same while sweets get dearer, it buys fewer sweets.

    The truth

    • The outlet did not publish exact counts of this year's cuts. [31]