About COIOS

Our mission is to reveal the drivers of future health through multi-disciplinary research. We believe the real insights come from combining perspectives from behavioural psychology, data science and emergent technology, and that those insights benefit us all.

Dr Fiona Garrett Ryan

CEO and Founder

Fiona founded COIOS to bring a clinician's perspective to research on population risk. She has been a general practitioner for more than twenty years, and has held the Diploma in Medical Acupuncture since 2003.

Her approach in the consulting room shapes how the company does its research. She looks at lifestyle and emotional factors behind symptoms rather than the symptoms alone, and holds that small, achievable changes have a large cumulative effect on long-term health.

Dan Ryan

Chief Scientist

Dan is a medical demographer. He has spent three decades working out what changes in health mean for life expectancy. He led global research teams at Swiss Re and at Willis Towers Watson, building forward-looking models of mortality and morbidity, and bringing behavioural science and emerging data sources into how risk is assessed.

At COIOS he is responsible for the research. He reads the clinical literature and national statistics across countries, decides which drivers of future health matter, and presents analysis that people can act on.

During the pandemic he contributed to the COVID-19 Actuaries Response Group, with published work on care home mortality and viral mutation. He sits on the Mortality Research Steering Committee of the Institute and Faculty of Actuaries. He is the Deputy Chair of the International Actuarial Association's Mortality Virtual Forum.

COIOS explores drivers of future health in four domains: meeting health needs today, learnings from behavioural science, the power of data to unlock insights and better outcomes, and finally, both disruption and innovation in tomorrow’s world. We report what the evidence says about health and the risks that change it. This page sets out where what you read here comes from, how it is chosen, and what we will and will not say about it.

What we publish, and what we don't

Everything on this site is there to inform. We report what a study found and how strong the evidence is. We give the official figures and when they were last revised. We say where authoritative sources disagree and why, and how the methods behind the numbers work.

We do not publish assumptions. Nothing here says what mortality improvement to expect, which projection to adopt, or what a finding means for a particular scheme, portfolio or policy. Turning evidence into assumptions is judgement applied to one organisation's own position. Where the site shows a projection, it is an official one. The point is to show how it fared against events, not to impose one of our own.

The line is deliberate. It keeps what is published here free, public and the same for every reader. And it keeps the advice where it belongs, with the person who is accountable for it.

Where the material comes from

We read three kinds of source, and only these.

Peer-reviewed journals. Standing searches run against Europe PMC, Crossref and the journals' own feeds, on the questions behind the twenty-three drivers on The Crucible: whether deaths from heart disease have stopped falling, what changes behaviour at scale, how well mortality can be forecast. A list of journals is read in full. Both lists are on the sources page, with what each source covers and when it last updated. They change as we learn what each contributes, and changes are noted in the corrections and changes log.

The medRxiv preprint server. Preprints posted there are read alongside the journal literature, and where the same work appears in both, the journal version is preferred to its preprint.

Research institutes. Three only: the Max Planck Institute for Demographic Research, INED and IZA, read through their own feeds.

Sources that publish in German, French or Japanese are read in translation. The full list, with what each source covers and when it last updated, is on the sources page.

We deliberately do not read industry or consultancy publications, including reinsurer bulletins and the insurance trade press, general news media, or anything behind a paywall. The first keeps this site from becoming an insurance publication by another route. Anything worth having in the second should have come from a primary source. The third keeps every claim here checkable by anyone.

How items are chosen

Each morning at 06:00 a scan runs across all the sources above. It turns up typically a few hundred new papers. Many do not advance our knowledge. A vocabulary of words for each driver then files each paper under the drivers whose words its title and abstract carry; papers that match no driver are set aside, duplicates are folded together, and a language model scores each filed paper's relevance to the questions we follow. That step is done by software because no person could read that volume daily.

A language model drafts each card that goes forward, in our house style: what was studied, what was found, and why it is interesting. A separate reader checks it makes sense to someone who has not read the paper, and a separate checker reads it against the paper's abstract. Every card is then read by a person on a private review screen, who approves, edits or rejects it. Nothing appears on this site without that approval.

We say this plainly because readers are entitled to know where automation begins and ends. The scanning software, the standing queries and the source list are described on this page and on the sources page. Any change to them is noted in the corrections and changes log.

Verdicts

Each item carries a one-word verdict on what it does to the evidence that came before it:

• new — a question not previously studied at this scale or in this population;
• confirms — consistent with the existing weight of evidence;
• refines — consistent, but changing the size of an effect, the population it applies to, or the mechanism;
• contradicts — inconsistent with the existing weight of evidence.

Data and trackers

The numbers on this site are official figures reproduced with attribution. Every chart states its source and the date of the data it shows. When a source revises its own past — a re-estimated baseline, a census rebasing, a change in how causes are coded — we record the revision and its size on the revisions page, and any driver it affects says so. We do not adjust, smooth or model official figures ourselves.

Every item, data-watch entry and revision, and every tracker's full series, can be downloaded as CSV and JSON from the downloads page.

Corrections

When we get something wrong we correct it in place, with a dated note on the item saying what changed, and we record it on the corrections and changes log. If you think an item is wrong, write to hello.central@coios.me with the source. We will look at it and reply.

Independence

The site is funded by COIOS Ltd and by nothing else. It carries no advertising, no sponsored content, and no material from clients or industry bodies. Reading the site is free and always will be.

Everything published here is written in a personal capacity, draws solely on publicly available sources, and represents the views of its authors alone. It is not the view of any employer, client or other organisation, and it is not advice of any kind.

Office

167–169 Great Portland Street
5th Floor
London W1W 5PF

hello.central@coios.me
+44 7990 805142

People

Dr Fiona Garrett Ryan, CEO and Founder
fiona.ryan@coios.me

Dan Ryan, Chief Scientist
dan.ryan@coios.me

Elsewhere

X
LinkedIn

This page records two things: corrections to what the site has published, and changes to how the site is made.

Corrections

Where we get something wrong, we correct it in place, with a dated note on the item saying what changed, and we record it here. If you think an item is wrong, write to hello.central@coios.me with the source.

• 26 September 2026 — An item on an mRNA vaccine tested in mice was withdrawn because it reported no outcome in people.
• 26 September 2026 — An item on genomic loci for ME/CFS was withdrawn because it falls outside the drivers the site follows.
• 26 September 2026 — An item on the cost-effectiveness of antiobesity drugs was withdrawn because it reported no new analysis of its own.
• 18 September 2026 — Three further items have been withdrawn: each reported no new analysis of its own — two scoping reviews and a review presenting no new data. The scan now holds any publication reporting no new analysis of its own, whatever it calls itself; systematic reviews and meta-analyses, which pool data and produce an estimate, are admitted.
• 16 September 2026 — Five items published on 15 and 16 September have been withdrawn. Three reported no finding about a human population: an animal study, a trial protocol with no results, and a record for which no abstract was available. Two were on subjects none of the site's drivers covers. The scan now holds all four kinds on the review screen rather than publishing them.
• 16 September 2026 — Two items have been refiled. A paper on e-cigarettes, snus and chemotherapy response no longer carries the smoking and vaping driver, which follows tobacco use and the mortality that follows from it; it remains under cancer. A paper on machine-scored autopsy descriptions now carries the AI in medicine driver, which it should have carried when published.

Changes to how the site is made

The scanning software, the standing queries and the list of sources are described on How we work and on the sources page. Any change to them is noted here, newest first.

• 26 September 2026 — Several journals were moved to their publishers' own records after we found the literature database was listing them weeks after publication; journals we cannot reach at all were removed from the source list.
• 16 September 2026 — From 16 September 2026 the twenty highest-scoring papers of each day's scan go to the editor for review, in place of a fixed score threshold. This is an interim arrangement while the scoring scale is reviewed.
• 14 September 2026 — The life-expectancy standing query lost an unfilled placeholder whose square brackets had stopped its date window applying, so it no longer returns papers published years earlier.
• 13 September 2026 — Two standing queries added, on alcohol consumption and physical activity; the vaccination-uptake query moved to What changes behaviour at scale, which now takes vaccine coverage from Infectious disease.