Cohortbridge connects to your EHR via FHIR R4, runs a two-pass eligibility screening model against your trial protocol, and delivers a ranked candidate list directly into your CTMS — reducing coordinator pre-screening time from 45 minutes to approximately 8 minutes per candidate.
Mid-size CROs managing Phase II and Phase III oncology and rare disease trials for emerging biotech sponsors face a familiar bottleneck: manual patient screening against complex eligibility criteria causes enrollment timelines to slip 4–8 months per trial. The root cause isn’t a shortage of potentially eligible patients. It’s that site coordinators spend 60–70% of pre-screening time on patients who fail a single secondary exclusion criterion that could have been detected from their EHR record before consent was ever sought.
The numbers are hard to ignore. Mid-size CRO studies screen 8–25 patients per enrolled participant — a screen-failure rate that adds $3,000–$8,000 per failure in coordinator labor and site fees. A coordinator spending 45 minutes per manual chart review, working through 15 patients to enroll one, is spending 10+ hours per enrollment. That’s not a people problem. It’s an information retrieval problem that should have been solved at the EHR layer.
Input: Trial protocol inclusion and exclusion criteria structured by the Cohortbridge onboarding team, patient demographic and clinical data from the site’s EHR via HL7 FHIR R4 or a direct Epic MyChart Bedrock API connection, and site coordinator referral lists provided by participating clinical sites.
Processing: Cohortbridge maps the trial’s eligibility criteria to FHIR data elements and runs a two-pass screening model. The first pass applies deterministic rule-based filters for hard exclusion criteria retrievable from structured EHR fields (diagnosis codes, lab values, prior medications, procedure history), eliminating the majority of ineligible patients before any coordinator time is spent. The second pass runs a probabilistic NLP model against clinical notes to surface potentially qualifying patients whose eligibility depends on narrative documentation — scoring each on likelihood of meeting each inclusion criterion and flagging the criteria requiring coordinator verification.
Output: A ranked patient short-list delivered to the site coordinator’s existing trial management system showing each candidate’s eligibility score, the specific criteria met and flagged, the data fields driving each determination, and a consent-ready summary of the patient’s relevant clinical history — reducing coordinator pre-screening time from 45 minutes per patient to approximately 8 minutes per prioritized candidate.
Integrations: Epic MyChart Bedrock API, Cerner Millennium FHIR, MEDITECH Expanse FHIR, HL7 FHIR R4, Medidata Rave, and Veeva Vault CTMS.
Converts complex protocol inclusion and exclusion criteria into structured FHIR-queryable logic within 5 business days
Clinical trial eligibility criteria are written in protocol language that does not map directly to EHR data fields without expert translation. Cohortbridge's onboarding team structures each criterion as a combination of FHIR query parameters, diagnosis code sets, lab value thresholds, medication class definitions, and temporal relationship rules during a five-day protocol intake process.
The resulting eligibility rule set is reviewed with the sponsor's medical monitor before deployment to confirm clinical intent is preserved. Rule sets are versioned, so protocol amendments can be incorporated mid-enrollment without restarting the screening model from scratch.
Hard exclusion filtering eliminates 80-90% of ineligible patients before any coordinator time is spent
The first screening pass applies deterministic rule-based filters against structured EHR data — ICD-10 diagnosis codes, LOINC lab values, RxNorm medication records, and CPT procedure history — to immediately eliminate patients who fail hard exclusion criteria: prior cancer diagnosis, recent conflicting medication, age or organ-function thresholds.
This pass runs in under two seconds per patient record and typically eliminates 80-90% of the screened population, leaving a much smaller set for the second-pass NLP model that reviews clinical notes for softer criteria.
Surfaces potentially eligible patients whose qualifying conditions are documented in narrative clinical notes, not structured fields
Rare disease and oncology trials often require eligibility evidence that lives in clinical notes — pathology report language, physician assessment of disease severity, prior treatment response descriptions — rather than in ICD codes or lab values alone.
Cohortbridge's NLP model processes discharge summaries, progress notes, and pathology reports using a transformer-based extraction pipeline fine-tuned on oncology and rare disease clinical text. For each patient who passes the deterministic first pass, the NLP model scores the likelihood of meeting each note-dependent inclusion criterion and highlights the specific note passages that drove the score.
Ranked candidate list with eligibility breakdowns and consent-ready patient summaries in the coordinator's existing CTMS
Cohortbridge delivers its output as a ranked candidate list integrated into the site's existing trial management system — Medidata Rave or Veeva Vault CTMS — so coordinators never leave their primary workflow. Each candidate card shows the overall eligibility score, a criterion-by-criterion pass/fail breakdown with supporting data, and a one-page consent-ready clinical summary.
Coordinators can override any AI determination with a note, which feeds back into model improvement for subsequent screening cycles.
Tracks screen failure patterns across sites to identify protocol criteria causing disproportionate enrollment friction
Mid-size CROs managing multi-site trials need visibility into which eligibility criteria are driving screen failures at which sites so they can flag protocol amendment opportunities to sponsors and optimize site selection for future trials.
Cohortbridge's analytics dashboard aggregates screen failure data across all participating sites, showing the proportion of failures attributed to each eligibility criterion, the site-level variation in failure rate, and the estimated enrollment delay caused by each high-friction criterion.
Connects to Epic, Cerner, and MEDITECH via FHIR R4 with HIPAA Business Associate Agreement coverage and no PHI retention after screening
Cohortbridge connects to site EHRs via FHIR R4 APIs under a HIPAA Business Associate Agreement executed with each participating site or health system. The platform processes patient records ephemerally — PHI is used to compute eligibility scores and generate candidate summaries but is not stored in Cohortbridge's systems after the screening session completes.
All data in transit is encrypted via TLS 1.3; at-rest storage for intermediate processing uses AES-256 encryption in an AWS GovCloud environment. Audit logs of every query and API call are available to site compliance officers on request.
Cohortbridge is designed to reach first candidate output quickly. Protocol intake, EHR connection, and CTMS integration happen in parallel — most CROs see the first ranked candidate list within the first week of a new trial.
Your trial team shares the protocol and eligibility criteria. The Cohortbridge onboarding team maps each inclusion and exclusion criterion to FHIR data elements — ICD-10 codes, LOINC lab values, RxNorm medications, CPT procedures, and temporal relationship rules — within five business days. The rule set is reviewed with your sponsor’s medical monitor before any screening run begins.
We connect to each participating clinical site via HL7 FHIR R4, Epic MyChart Bedrock API, Cerner Millennium FHIR, or MEDITECH Expanse FHIR. A HIPAA Business Associate Agreement is executed with the site or health system before any patient data is accessed. Setup for a standard FHIR connection typically completes within two to three business days per site.
The deterministic first pass runs structured EHR data against hard exclusion criteria, eliminating 80–90% of ineligible patients in under two seconds per record. The second pass applies a transformer-based NLP model to discharge summaries, progress notes, and pathology reports for candidates who passed the first gate, scoring each on likelihood of meeting note-dependent inclusion criteria and flagging the passages that drove the score.
A ranked candidate list is pushed into Medidata Rave or Veeva Vault CTMS within your existing coordinator workflow. Each record includes an overall eligibility score, a criterion-by-criterion breakdown, the specific data fields or note passages supporting each determination, and a one-page consent-ready clinical summary. Coordinators can override any determination with a note; those overrides feed back into model improvement for subsequent screening cycles.
Cohortbridge is designed for mid-size CROs managing multi-site Phase II and Phase III trials — teams of 15–200 FTE with 3–25 active trials and annual CRO services revenue of $8M–$120M. The sweet spot is CROs running oncology and rare disease trials for emerging biotech sponsors, where eligibility criteria are complex, screen-failure rates are high, and coordinators are stretched across multiple active studies.
Cohortbridge is not built for large global CROs that already have proprietary patient-matching platforms deployed (IQVIA, PPD/Thermo Fisher, and Labcorp Drug Development-tier organizations). It’s also not the right fit for academic medical centers running investigator-initiated trials without a professional CRO partner, or for Phase I first-in-human trials with fewer than 20 enrollment targets, where the manual screening overhead is manageable and the regulatory constraints around automated pre-screening are more stringent.
If you’re a mid-size CRO and your coordinators are routinely spending 40+ minutes per manual chart review, Cohortbridge is built specifically for that workflow.
We’ll walk through what eligibility screening looks like for your specific trial criteria and EHR environment. No prepared slide deck — just a working demo against a real protocol structure.