Clinical trial patient matching that shows the line in the chart for every criterion
Eligible patients found in minutes, each criterion backed by the record.
See one case, screen by screen ↓Software that checks a site's patient records against a protocol's inclusion and exclusion criteria to find who could be screened. Clinical Trial Patient Matching uses six agents to read every de-identified record at the site, mark each criterion met, not met or unknown with the line that proves it, and turn every unknown into a question for the patient's next visit. The principal investigator decides who is referred.
How to choose AI solutions for life sciences →
Updated
A protocol arrives with 20 to 30 inclusion and exclusion criteria. The research coordinator reads charts against them by hand: the tumour-board list, new-patient notes, labs, the CT report, the ECG. The facts that decide eligibility are scattered — the EGFR result in a scanned outside pathology report, the ECOG score in a free-text note from May, the QTcF in an ECG PDF. A structured query sees none of it.
When a fact is missing, the chart is set aside as a no and the patient is never asked. Hand pre-screening covers the main-campus list, so patients seen at satellite clinics — often the groups a Diversity Action Plan counts — are rarely reviewed. The trial misses its enrolment timeline while eligible patients sit in records nobody had time to read.
Estimated split of a research coordinator's working days spent pre-screening for one open trial in a month, by hand and with the solution.
Six agents read every record against every criterion and turn each gap into a question; the investigator decides.
Lakemont Cancer Institute runs VLT-411 for Veltara Therapeutics — a Phase 2 study of velrastinib after a third-generation EGFR inhibitor in EGFR-mutant advanced NSCLC, with 24 criteria. Last quarter, hand pre-screening found 5 eligible patients. Here is Tuesday 13 October, screen by screen, in the working solution.
Nadia runs pre-screening for four lung trials. Overnight, 1,212 de-identified records from the main campus and three satellite clinics were read against VLT-411's 24 criteria. Her grid shows 9 patients eligible on all 24, 47 one question away, 86 two or three away and 1,070 not eligible. Each dot is a criterion: green met, red not met, hollow amber unknown.
“For exclusions, met means the exclusion is absent · click any dot for its source line.”
The criteria parser loads 24 checks from protocol v2.0, confirmed by the investigator on 6 Oct. The record reader reads 1,212 records — 48,216 documents — at the site. The criterion checker runs 29,088 checks, each with its line. The question writer turns 244 unknowns into questions, 49 of them askable this week, and the ranker lifts the 9 eligible and 47 one question away to the top. Nadia can re-run it any time.
Every step is shown as it runs, with what it found.
P-0412, 58, main campus, is met on every criterion but one. I6 asks for ECOG 0 or 1 within 28 days before screening, and the last ECOG in his record is from 19 May. The checker will not call it met on old evidence; it marks it unknown and writes the question.
“ECOG not recorded since May; ask at Thursday’s visit.”
Add hepatitis B and C serology to the next blood draw. Order a 12-lead ECG at the next visit. Ask pathology whether the Nov 2024 block is still available. Ask the investigator whether an echocardiogram is needed. Each question names the criterion it settles and when to ask it. Nine eligible now and 7 projected from this week's answers make 16, against 5 found by hand last quarter.
“Answer it and the grid fills in.”
P-0883, 64, seen at the Eastgate clinic, is met on 23 of 24 criteria. Only I3, EGFR mutation, is unknown: the clinic note says “EGFR+ per outside records” but no report was on file. Then a new document arrives — an outside pathology report, scanned in by Eastgate on 12 Oct, after last night's run.
“Outside pathology report scanned in 12 Oct: read it.”
The record reader opens the three-page scanned report and finds the line: “EGFR exon 19 deletion (E746_A750del)”, path report dated 14 Aug 2026. The criterion checker marks I3 met under protocol 5.1.3, which accepts an exon 19 deletion by tissue test. P-0883 moves into the eligible rows: eligible on all 24 rises to 10, one question away falls to 46.
“P-0883 is eligible on all 24 criteria — every line cited.”
Her record opens with the 24 criteria on the left and the lines that decide them highlighted in place: the 30 Sep oncology clinic note (ECOG 1, post-menopausal, no brain metastases on the 24 Sep MRI), the outside report (T790M not detected, no MET amplification), and the 7 Oct labs — ANC 3.8, platelets 241, CrCl 72 mL/min. The record was read in place, identifiers removed.
“Pre-screening only. Dr. Ferreira decides who is referred; eligibility is confirmed at the screening visit after consent.”
The screening-log writer drafts the pre-screening worksheet: each criterion as written in the protocol, its verdict, the line in the record and the source with its date. Nadia sends it to Dr. Ferreira, who reviews it on the clinic tablet and chooses: refer to screening, hold or not eligible. He signs with his PIN and a note — “Approach at Thursday visit; consent then screening labs”.
The worksheet now reads “Referred to screening · e-signed”, and the decision goes into the screening log with his name. She is to be approached at her visit on Thursday 15 Oct; eligibility is confirmed at screening, after consent, under protocol v2.0.
“Only Lakemont staff see this worksheet. Veltara Therapeutics receives counts by site, never the patient.”
The funnel runs from 1,212 records read to 142 with no criterion failed, 56 eligible or one question away, 10 eligible on all 24 and 4 referred in 30 days. With this week's answers the site projects 17 eligible — 3.4 times the 5 found by hand last quarter. Of the 56, 21 are from groups under-represented in lung-cancer trials; the main-campus list alone would have shown 16 of them.
“Reported to the sponsor as counts for the study’s diversity action plan.”
The same morning, seen by the four people who carry pre-screening — what their week looked like, and what it looks like now.
Six agents read every record against every criterion and turn each gap into a question; the investigator decides.
Reads section 5 of the protocol and every amendment and turns each inclusion and exclusion criterion into a check: field, rule, threshold, look-back window and the record types to read, with its section number.
Reads clinic notes, labs, pathology and molecular reports, imaging and ECG reports, medication lists and scanned outside records, and keeps each fact with its quoted line, document and date.
Marks every patient against every criterion met, not met or unknown, citing the line that decides it.
Turns every unknown into one specific question — what to ask, of whom, and at which visit or lab draw — from the clinic schedule, routing tissue to pathology and judgement calls to the investigator.
Sorts patients by how close they are to eligible and projects how many will be eligible once this week's questions are answered, from how similar questions closed over the last 90 days.
Drafts the pre-screening worksheet for the investigator, keeps the screening log of every decision and shares counts by site and week with the sponsor.
Decides who is referred. The agents propose; a named person decides.
Ask why a patient is eligible, what to ask at Thursday's visits, or who is one question away at a clinic — every answer points to its source.
Why is P-0883 eligible now?
P-0883 was met on 23 of 24 criteria; only I3 (EGFR mutation) was unknown — the 30 Sep clinic note says “EGFR+ per outside records” but no report was on file. The outside pathology report scanned in on 12 Oct reads “EGFR exon 19 deletion (E746_A750del)”, dated 14 Aug 2026. Protocol 5.1.3 accepts an exon 19 deletion by tissue test.
Which questions can I close at Thursday’s visits?
10 questions can be closed at Thursday’s visits (15 Oct), on 9 patients. 3 of them are one answer from eligible: P-0412 (I6: ECOG not recorded since May), P-0761 (E6: add hepatitis B and C serology to Thursday’s blood draw) and P-0594 (E4: order a 12-lead ECG at the next visit).
Who is one question away at Eastgate?
10 patients seen at Eastgate are one question away, among them P-0883 (I3), P-0761 (E6), P-0026 (I5) and P-0914 (E11). Eastgate was not part of hand pre-screening, which covered the main-campus tumour-board list.
Draft the worksheet for P-0883 for Dr. Ferreira
Drafted: all 24 criteria with the line and date that proves each, the EGFR result from the outside report, and her next visit (Thu 15 Oct). It is a pre-screen only: Dr. Ferreira decides, and eligibility is confirmed at the screening visit after consent.
13 screens from the working solution, on its sample data. Pick one to see it large.
What the agents read, how long evidence counts, who is asked and what leaves the site are settings, saved for the next run.
| Setting | Default | Choose from |
|---|---|---|
| Records the agents read | All seven sources on | Clinic notes · laboratory results · pathology and molecular reports · imaging reports · ECG and echocardiogram reports · scanned outside records · medication lists — each on or off |
| ECOG performance status counts for | 28 days | 7 days · 28 days · 60 days |
| Blood count and chemistry count for | 14 days | 7 days · 14 days · 60 days |
| ECG counts for | 90 days | 7 days · 60 days · 90 days |
| No evidence in the window | Unknown — never met | Locked |
| Who is asked | Visit questions to Nadia Okoye and Priya Das | Lab questions to the coordinators, ordered by the investigator · tissue and outside reports to pathology · judgement calls to Dr. Samuel Ferreira |
| Re-match | Nightly at 02:00 | and whenever a new document is filed |
| Privacy and sign-off | Locked by the research office | Records stay at the site · sponsor sees counts only · investigator e-signs every referral · agents never contact patients |
Every figure is labelled: a target the solution is built to, an estimate, a typical published result, or a proven one.
“demo” = seen in the working solution, on its sample site data · “target” = the design goal, measured in the live solution · “estimated” = our estimate · “typical” = published figures (enrolment-barriers review via DataAlly — about 80% of trials miss their enrolment timeline; a retina-trial study at ASRS 2026, via Ophthalmology Times — 30–60 minutes a patient by hand; FDORA 2022 on Diversity Action Plans). People, sites, sponsors, trials and patients named on this page are fictional — characters and sample data in the working solution.
Software that checks a site's patient records against a protocol's inclusion and exclusion criteria to find who could be screened. Clinical Trial Patient Matching uses six agents to read every de-identified record at the site, mark each criterion met, not met or unknown with the line that proves it, and turn every unknown into a question for the patient's next visit. The principal investigator decides who is referred.
A missing fact is never a no and never a yes: it is unknown. The question writer turns it into one specific question — order a 12-lead ECG at the next visit, ask pathology whether the archival block is still available — timed to the clinic schedule. When the answer is recorded, the cell fills in and the patient moves up the grid.
Yes. The record reader reads clinic notes, labs, pathology and molecular reports, imaging and ECG reports, medication lists and scanned outside records, and reads a new document as soon as it is filed. In the working solution it found an EGFR exon 19 deletion in a scanned outside report and the patient became eligible on all 24 criteria.
Records are read in place with identifiers removed, and nothing is copied off site. The sponsor sees counts by site and week — records read, eligible, one question away, referred — never a patient. Those privacy rules are locked by the clinical research office.
The criteria parser reads the amendment, compares it with the confirmed rules and sends any changed criterion to the investigator. In the working solution, HLV-207 amendment 2 changed two criteria, and its rules wait for the investigator before they are used.
It reads records from every clinic, not just the main-campus list. On the sample data, 21 of the 56 patients eligible or one question away are from groups under-represented in lung-cancer trials, against 16 from the main-campus list alone. They are reported to the sponsor as counts, with each group as it is recorded in the record.
Yes. The agents never decide eligibility and never contact a patient. The investigator confirms the rules built from the protocol, decides from the worksheet and e-signs every referral, the coordinator asks the questions under the site's own process, and eligibility is confirmed at the screening visit after consent.
The Agentic Solution Engine builds and deploys it from your requirements and documents — your protocols, pre-screening SOP, record sources and clinic schedule — and it goes live once every quality gate has passed. We will walk you through it on one of your own open protocols first.
A checked first-draft protocol in days, in your approved language.
1-min demoeTMF Quality CheckAn inspection-ready trial master file every day, each site scored on what is due today.
1-min demoClinical Study Report AuthoringResults drafted within days of database lock, linked to tables.
1-min demoWe'll run Clinical Trial Patient Matching on one of your open protocols and a sample of de-identified records.