What we make
Two products, one engine.
Both work off the same thing: the eligibility criteria a trial publishes to
ClinicalTrials.gov. The Screener checks those criteria against one patient. The Board scores
how readable they are in the first place. They share all the same code, so they always agree.
For patients, caregivers and coordinators
The Screener
Type a condition. It checks one person against every study currently
recruiting, answers what it can answer, and shows you the exact wording of everything it
could not.
- PullEvery recruiting study for that condition, up to 400 per search,
with the full eligibility section.
ClinicalTrials.gov v2 API
- Rule out on the easy stuff firstSex, age bounds and site countries arrive as
structured data, so they decide exactly and instantly. Most of a result set is ruled
out here, before any prose is parsed.
- Read the written criteriaEach remaining criterion is segmented into clauses,
matched against 35 clinical concepts, negation scoped to the clause, numbers extracted
with their units. Every predicate keeps the byte offsets it came from.
- Answer, or admit it cannotAbove 0.70 confidence it decides. Below it, or
on a criterion naming investigator judgement, it abstains out loud and prints the
deductions. Anything it could not compile is shown verbatim.
- Ask you the one question that helps mostA solver scores every unanswered concept
by how many still open studies it would settle, asks that one, and reranks against what
survives.
- Give you a shortlistNCT numbers, the predicate behind every verdict,
and the lines still needing a human. Export as CSV or a printable brief.
100.0% precision, 95.7% recall on the compiler ·
0 bytes about the patient transmitted
For researchers, sites and sponsors
The Study Board
Post the study you are recruiting for. It goes on a board patients and
caregivers already read. A private dashboard shows aggregate listing views, screener clicks,
and contact reveals without identifying anyone.
- Write the listingTitle, condition, what the study involves, who you
need and where. Plain words, not the protocol title. Takes about two minutes and there
is no account.
- It goes on the boardPublished immediately, next to every other study
recruiting. Patients and caregivers browsing for their condition see it.
- You get a private linkThe only key to your dashboard. No password to
lose, and nothing to sign into.
keep it somewhere safe
- People read it and actThey can reveal your published study address and email
you from their own inbox, or run the Screener against every recruiting trial.
- You see the numbersListing views, screener clicks, and contact reveals. No
names, notes, patient list, or CSV exists on Inclusion Health.
- You screen repliesAnyone who emails you still has to be screened against the
full protocol. Inclusion Health does not claim they are eligible.
Free · no account · public study inbox required
Shared underneath
One engine, so the two can never disagree.
The same compiler
Both products import the identical rule engine and the identical criterion splitter.
A study's score on the Board is computed by the code that screens a patient in the
Screener, so a sponsor sees exactly what a patient's tool saw.
The same learned layers
A conformal multi label classifier trained by distant supervision on 32,628 criteria,
and optional transformer embeddings for phrasing the lexicon misses. Neither is allowed
to produce a verdict in either product.
The same place to run
All inference happens in the browser. The Screener sends the condition name to the public
registry but transmits no patient profile. The Board sends aggregate counters keyed to a
public study identifier, with nothing about who triggered them.
| The Screener | The Board |
|---|
| Answers |
Which trials is this person not ruled out of? |
Are people finding my study, and how can they contact the team? |
| Unit of work | One patient against many studies | One study and its aggregate activity |
| Who opens it | Patients, caregivers, research nurses | Researchers, study coordinators, sponsors |
| Input | A condition, then age, sex and country | What you write about your own study |
| Output | A shortlist with the reason behind every verdict | Views, screener clicks, and contact reveals |
| Transmits | Nothing. The condition name goes to the public registry, as any search box would |
Your public listing and three aggregate counters. No patient data |
| Costs | Free permanently | Free, no account |
A third thing, which is not a product
The collaboration brief.
Both products rest on an evaluation of 46 criteria labelled by one rater, which is a
regression harness rather than a validation. The brief states that plainly, sets out a study
design for fixing it, and asks for the one input that cannot be generated from the inside: a
clinician labelled corpus. If you screen patients or study accrual, that is the page worth
your time, not either product.
