KSCREENER Patient Pre-Screening

    AI patient pre-screening - with the reasoning to prove it.

    KScreener reads de-identified EHR at the site, applies your protocol's eligibility criteria with chain-of-thought reasoning, and ships a coordinator-ready cohort each morning - every match annotated with a rationale a CRA can defend at audit.

    KScreener KT-204 · INC/EXC v0.3 HIPAA Screening Matches ready
    2,418 charts reviewed 84 high-confidence
    De-identified matches12 of 84
    PT-9012
    67F · Stage IV · L858R · ECOG 1
    0.96
    Likely eligible Coordinator-ready
    PT-4418
    58M · Stage IIIB · Ex19del · ECOG 0
    0.94
    Likely eligible
    PT-7750
    72F · Stage IV · L858R · ECOG 1
    0.91
    Likely eligible
    PT-3309
    61M · Stage IV · Ex19del · ECOG 1
    0.78
    Needs review
    Eligibility reasoning
    PT-9012 · KT-204
    EGFR-mutant NSCLC · 1L
    0.96
    Confidence
    ✓Inclusion matched · histology + stage
    ✓EGFR L858R mutation confirmed
    ✓ECOG within range (≤ 1)
    ✓Prior therapy check complete
    –Exclusion cleared · no active CNS mets
    Evidence: Path note 2025-03-12 confirms adenocarcinoma; NGS report flags EGFR L858R; MRI brain 2025-04-02 negative for CNS involvement. Coordinator-ready rationale attached.
    12k+
    Charts scanned per site / day
    3.4×
    Match rate vs. manual chart review
    -62%
    Screen-fail rate, post-deployment
    How KScreener works

    A screening engine, not a chart hunt.

    KScreener reads de-identified EHR data, applies your protocol's eligibility logic, reasons through each match, and delivers a coordinator-ready patient list with rationale attached.

    01 Source data
    Source data
    • De-identified EHR
    • Lab values & diagnoses
    • Medication history
    • Prior treatment data
    EHR feed live
    02 Protocol logic
    Protocol logic
    • Inclusion criteria
    • Exclusion criteria
    • Biomarker requirements
    • Line-of-therapy rules
    03 Screen
    Screen
    • Patient-by-patient review
    • Structured + unstructured signals
    • Criteria applied in sequence
    • Match candidates surfaced
    04 Reason
    Reason
    • Eligibility rationale
    • Criteria-level explanation
    • Evidence snippets
    • Coordinator-defensible logic
    05 Prioritize
    Prioritize
    • High-confidence matches
    • Needs review
    • Confidence ranking
    0.96
    0.82
    06 Route to site
    Route to site
    • Coordinator-ready cohort
    • Daily patient list
    • Review-ready output
    • Audit trail attached
    Delivered to coordinator
    One source of truth

    Every matched patient is backed by the same screening logic.

    Hover any stage to see how KScreener connects source data, protocol eligibility, reasoning, and coordinator-ready output into one defensible screening workflow.

    01The problem

    Eligible patients are already in the chart.

    They get missed because nobody has the time to read every record. Manual chart review is the bottleneck - and it's why a quarter of activated sites in any given trial enroll zero patients despite having the right population walking through the door.

    67%
    Of trials miss enrollment targets

    The single most common cause of trial delay is the patient who never gets identified - sitting in a chart on the third floor of a hospital that already runs the trial.

    40 min
    Per chart, manually

    A coordinator reviewing eligibility against a 30-criterion protocol averages 40 minutes per candidate. Most sites can't afford that across thousands of charts.

    46%
    Median screen-fail rate

    Half of patients who make it to a screening visit don't pass. The signal that would have stopped them was already in the chart.

    A different way to screen patients.

    Manual chart review depends on fragmented notes, inconsistent interpretation, and coordinator time. KScreener applies protocol eligibility logic consistently and returns patient matches with evidence and rationale attached.

    Traditional patient screening
    KScreener
    Screening basis
    Traditional patient screeningManual chart review across fragmented EHR notes
    KScreenerProtocol-specific screening across de-identified clinical records
    Eligibility criteria
    Traditional patient screeningApplied manually and inconsistently across reviewers
    KScreenerInclusion and exclusion logic applied consistently to every patient record
    Evidence and rationale
    Traditional patient screeningReasoning often lives in notes, spreadsheets, or coordinator memory
    KScreenerEach match includes criteria-level rationale and supporting evidence
    Review workflow
    Traditional patient screeningCoordinators spend time finding candidates before clinical review begins
    KScreenerLikely eligible patients are surfaced as a coordinator-ready worklist
    Screen-fail risk
    Traditional patient screeningWeak matches may move forward before key criteria are checked
    KScreenerPatients can be triaged by match confidence, review status, and exclusion signals
    Audit trail
    Traditional patient screeningScreening logic is reconstructed after the fact
    KScreenerEligibility reasoning, evidence, and review status stay attached to the match
    02How it works

    From raw chart
    to coordinator-ready.

    KScreener is built for the regulated, de-identified, on-premise reality of working with EHR data. The reasoning happens at the site. The PHI doesn't move. The output ships every morning at 7am local.

    IINGEST
    Ingest at the source

    Connects to the site's EHR (Epic, Cerner/Oracle, Athena) via FHIR or HL7. Raw PHI never leaves the site's tenant - de-identification happens in place.

    IISTRUCTURE
    Normalize the chart

    Free-text notes are parsed into structured concepts (SNOMED, RxNorm, LOINC) - pathology reports, oncology histories, lab trajectories.

    IIIREASON
    Reason over criteria

    Applies the protocol's inclusion and exclusion criteria with chain-of-thought reasoning - pulling supporting evidence from across the longitudinal chart.

    IVSHIP
    Ship a worklist

    Coordinators wake up to a ranked worklist with rationale-per-criterion attached - green to schedule, yellow to review, red to skip.

    03Explainability

    Every match comes with its work shown.

    A coordinator can trust an AI match exactly as far as they can defend it. Every KScreener candidate is shipped with a per-criterion rationale, a citation back to the source record, and a CRA-ready audit trail.

    9012
    Patient PT-9012 · 67F
    SITE 14021 · MOFFITT · LAST VISIT 2025-01-18
    ELIGIBLE · 0.96 confidence
    Inclusion 6.2.1METNSCLC adenocarcinoma - pathology report 2024-11-14, central node biopsy.PATH/2024-11/3409
    Inclusion 6.2.2METEGFR L858R confirmed - local NGS panel 2024-12-02; central retest pending.LAB/EGFR-NGS-7712
    Inclusion 6.2.3METECOG performance status 1 - clinic note 2025-01-18 (28d ago).NOTE/MD/2025-01-18
    Inclusion 6.2.4REVIEWAdequate organ function - labs 19d old, ALT slightly above bound (re-check needed).LAB/CMP/2025-01-09
    Exclusion 6.3.1CLEAREDNo prior EGFR TKI therapy in last 6 months - med history reviewed.MED/HIST
    Exclusion 6.3.2CLEAREDNo active CNS metastases - MRI brain 2024-11-09 negative.IMG/MRI/3001
    Reasoning trace · 14 source records · 1 follow-up neededSchedule screening
    04Data & privacy

    Built for the way real EHRs actually work.

    The patient privacy posture isn't a slide deck. It's an architectural commitment - KScreener is engineered so that the most sensitive data stays at the site, even from us.

    PHI never leaves the site

    KScreener runs in the site's tenant. The only thing that crosses the boundary is the de-identified, tokenized worklist - never raw PHI.

    HIPAA · 21 CFR Part 11

    Audit-ready by design. Every reasoning trace is hashed, every model version stamped, every coordinator action logged.

    Re-identification only by the site

    Tokens map back to MRNs only inside the site's EHR session. The platform itself never sees a name, a date of birth, or an address.

    IRB-approved workflows

    Standard IRB packet, BAA, and DPIA available out-of-the-box. Used in 40+ IRBs to date - your IRB has probably already seen us.

    05The economics

    Real money, both sides of the table.

    A faster trial is worth tens of millions to a sponsor and tens of thousands per month to a site. KScreener moves both numbers - and the second one is what makes the site network sustainable.

    For sponsors
    $8–14M

    Median net cost avoidance per Phase II oncology trial - from compressed enrollment, reduced screen failures, and fewer rescue site activations.

    • 5–7 month enrollment compression
    • 60%+ reduction in screen failures
    • Better representation across catchment
    For sites
    $1–4M

    Reported range of incremental annual revenue across Kitsa network sites - from higher match rates, more trials placed, and reduced coordinator overhead.

    • Coordinator hours redirected to consenting
    • Auto-matched to incoming sponsor RFPs
    • Performance score travels across studies
    “We used to spend three weeks finding ten patients. Now we get the worklist at 7am. The screen-fail rate went from 51% to 14%, and the coordinators actually have time to talk to families again.”
    Director of Clinical Research
    Mid-Atlantic academic medical center
    The rest of the platform
    Protocol authoring
    KScribe
    Regulatory-ready drafts in 48 hours.
    Site selection
    KScout
    400+ sites scored against your protocol.

    Questions, answered.

    Bring faster patient screening forward.

    See how KScreener helps teams match eligible patients to the right clinical trials with secure, AI-assisted screening workflows.

    Request a demo →