Pathology
Its own export
Clinical Data Infrastructure
Connect clinical data. Accelerate trials.
Trialayer links the spreadsheets a site already exports from services that do not talk to each other, pseudonymizes them in the browser, and counts how many patients fit a protocol — so a feasibility question that used to be answered from memory is answered with an auditable number.
- Local pseudonymization
- No EHR integration
- Patient records never leave the browser
The problem
The whole industry decides where to place a trial using a number nobody counted.
The sponsor asks how many patients the site has with a given condition. The site answers from memory, and optimistically, because the high number wins the study. Sites get activated on that answer.
1 in 3
activated sites never enroll a single patient
20–50%
of trials need a rescue mission
US$600k
to 8 million is the cost of each day of delay
Industry reference figures, not Trialayer results. We do not have customers yet, and we would rather say so.
Who it's for
Two sides of the same table
Sites and sponsors negotiate over a number nobody measures today. Trialayer measures it once, inside the site, and both sides look at the same figure.
Research sites
Hospitals, clinics, dedicated sites and academic teams.
- Answer a feasibility questionnaire with evidence, not from memory.
- Organize what you already have without integrating your EHR.
- Keep the screening log you are already required to keep.
- One account per person: whoever uploads is not whoever signs off.
- See the number before anyone else, and decide whether to send it.
Sponsors and CROs
Feasibility and site selection teams.
- Get measured figures, not numbers declared on a form.
- With methodology in the header: how many, over what period, counted how.
- With the undetermined rate declared, not hidden.
- Find the sites that already see your population, before you activate any.
- Avoid activating a site that will not enroll.
How it works
Four steps, and the order between them matters
It runs in your browser. There is nothing to install and no EHR integration: the spreadsheet is read from your own disk, and no request leaves the tab while it works.
Drop in what you already export
Excel or CSV from pathology, pharmacy, discharges. Exactly as they come out of the system, with no renaming of columns and no format conversion. The file is read from your disk — it is not uploaded anywhere.
Tell it which column is which
The system proposes and you confirm. It never applies a guess on its own: if the identifier is wrong, different patients get linked together, and that does not show up when you look at the result.
Records are linked, then pseudonymized
Linking happens inside the site, using the identifier, because linking requires one. It is discarded immediately after: everything downstream works on subject codes.
A number comes out, and nothing else
With its period declared, its methodology in the header and the cell threshold already applied. And only when you press send.
Your exportsxlsx · csv
Linkedinside your site
Pseudonymizedfail-closed
One countnothing else
Our position
A meter, not a data warehouse
The difference is not cosmetic. It decides the entire architecture, and it decides what happens the day someone makes a mistake.
A warehouse centralizes and keeps the contents. A meter measures while you work, and hands over a reading.
The records stay insideOne reading leaves
TrialayermeasuresA data warehousecollects
| A data warehouse | Trialayer | |
|---|---|---|
| Where the data lives | Centralized | Patient records never leave the browser. Counts live on our server |
| Who is accountable | Owns the contents | Processor. The site is the controller |
| What is sold | Access to records | A number derived from records we never held |
| Who sees the number first | Whoever pays | The site. And it decides whether to send it |
What shows up when you count
The patient who was never tested
When a criterion cannot be evaluated because nobody ordered the test, that patient is not ruled out: they are unknown. The difference decides whether the site orders the test or does nothing.
That number exists nowhere today. Trialayer separates it from genuine exclusions instead of mixing the two, because mixing them deflates the figure and loses studies the site could have won.
ScreenedRuled outclinical reasonEligibleUndeterminednobody ordered the test
What is guaranteed
Your data team's questions, answered up front
These are not brochure promises. Each one is built as a rule of the system, and there are automated tests that fail if any of them breaks.
No EHR integration
We work on what the site already exports. It does not connect to your
system and needs no IT project.
Data stays at the site
The file never leaves your browser. An aggregate number goes out, and
only when you send it.
Fail-closed pseudonymization
If it cannot guarantee a record came out free of identifiers, it blocks
it. It does not continue with a warning.
Nothing is deleted
Append-only from day one. A correction inserts a new version and keeps
the previous one, with who and when.
Chile Law 21.719
Designed against the Chilean sensitive-data regime, in full force on
1 December 2026. The DPA clauses and the DPIA scope are drafted;
neither is signed, and no real record is touched until they are.
Cell threshold
No small-count category is ever released on its own. Applied at count
time, not at display time.
An account per person, not per site
Whoever uploads the spreadsheet is not whoever answers the sponsor.
Roles are checked on every request, not drawn on a screen.
Sending is a separate permission
Only the principal investigator and study coordination can release a
number. Recruitment staff and monitors can work; they cannot send.
Two-factor sign-in
A six-digit code from a phone, on top of the password. Sessions die
after 20 idle minutes — nursing stations are shared.
Every attempt is logged
Who signed in, when, and every failure. Never the password that was
typed: a wrong one here is usually the right one somewhere else.
What it does not do
Saying this is part of the product
It does not diagnose or recommend treatment. It counts, and shows the basis for every claim with a citation to the document and the date of the data.
It does not claim population prevalence. The system has no way to say "this site has N patients". It can only say "of the N screened in this period, M qualify", because that is what the data supports.
It never pays or rewards per referred patient, in either direction. It is prohibited by Article 46 of the Chilean Medical Association code of ethics, and no function exists that would allow it.
It does not read charts to hunt for candidates. That requires an ethics committee waiver and is out of scope for the product.
FAQ
What comes up in the first meeting
Do I have to integrate my EHR?
No, and it is not on the roadmap. Trialayer works on spreadsheets your site already exports — pathology, pharmacy, discharges — in Excel or CSV, exactly as they come out of the system.
No renaming of columns and no format conversion: when you drop the file in, you declare which column is which. It never leaves the browser.
Where does my patient data live?
Where it already lives: on your own disk, in the files your services already export. We never receive a copy, because the spreadsheet is read in the browser and the identifiers stop there.
You do not have to take our word for it. Press F12, open the Network tab, and drop your file: the request list does not move.
Do I need ethics committee approval to use it?
To count your population from administrative data the site already has the right to cross-reference, no. That is the provider's own administrative processing.
What does require a committee waiver is retrospective chart screening without the patient present, and Trialayer does not do that.
Is it a medical device? Does it need regulatory clearance?
It does not diagnose or recommend treatment. It counts, and shows the basis for every claim with a citation to the document and the date of the data, so the clinical decision stays with the clinician.
This is a position with no precedent in Chile yet, and we say so rather than claiming otherwise.
How long does it take to get running?
There is nothing to install, so there is no IT, security or procurement queue to wait in — which was the expensive part. What takes time is the first conversation about which data sources exist and what columns they carry, and that is one meeting.
What does it cost?
There is no price list yet, and we would rather say so. What is decided: nothing is ever paid or charged per referred patient, in either direction, per Article 46 of the Chilean Medical Association code of ethics.
In other markets site software is passed through in the study budget. We are checking whether that applies in Chile.
What if the count makes my site look bad to the sponsor?
You see the number first, and it does not leave until you press send. Without that rule the product would be an informant and no site would use it.
Besides, a low number you can back up protects you more than a high optimistic one: the site that promises and does not enroll is the one that drops in the scorecard.
Status
Under construction, and looking for our first site
The demo runs on a synthetic corpus: no real patient data, not even de-identified. What we are looking for right now is not a customer, it is a site willing to look at the column headers of one of its spreadsheets and say whether this would fit what they already have.
That conversation takes twenty minutes and commits you to nothing.