Canonical text (the exact bytes fingerprinted)

TRUVACE RECORD VERSION
record: TRV-2026-0666
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-08-06T06:26:42.301733Z
status: published
lens: trace
sector: health
headline: TrialTriage, a Semiautonomous Prescreening Workflow for Resolving Ambiguity in Phase I Oncology Trial Eligibility: Development and Proof-of-Concept Study Using Synthetic Cases
dek: Enrollment in phase I oncology trials remains low largely because potentially eligible patients are not identified and evaluated quickly enough. Current clinical trial matching systems can identify candidate patients from the electronic health record, but cases with missing or uncertain eligibility data are often routed for offline manual review. This delay impedes clarification and prolongs the final eligibility determination. This study evaluated TrialTriage, a semiautonomous system built on the n8n platform a…
gain_title: TrialTriage achieved perfect concordance with ground truth on 90 synthetic phase I oncology cases and reclassified ambiguous cases to definitive eligibility after capturing investigator email replies, processing cases in seconds compared to slower manual review.
problem_title: Some ambiguous cases remained unresolved when investigator replies lacked actionable information, and cases with no reply after 48 hours still required deferral to offline manual review.
trace_subject: semiautonomous prescreening and email-based ambiguity resolution for phase I oncology trial eligibility
gain_reading: TrialTriage achieved perfect concordance with ground truth on 90 synthetic phase I oncology cases and reclassified ambiguous cases to definitive eligibility after capturing investigator email replies, processing cases in seconds compared to slower manual review.
gain_evidence: TrialTriage's classifications were 100% concordant with the author-confirmed ground truth in all 90 synthetic cases (95% CI 96.0%-100.0%). | 4 of 6 were reclassified definitively after investigator response
problem_reading: Some ambiguous cases remained unresolved when investigator replies lacked actionable information, and cases with no reply after 48 hours still required deferral to offline manual review.
problem_evidence: 2 remained ambiguous because the replies lacked actionable information. | after 48 hours without a reply, the case was referred for manual review.
quick_read: Researchers developed TrialTriage, a semiautonomous prescreening workflow on the n8n platform that uses large language model extraction from clinical narratives and investigator email replies plus a 7-criterion deterministic rule engine to classify phase I oncology trial eligibility, automatically emailing investigators when information is missing and reclassifying after reply capture.

In a proof-of-concept test on 90 synthetic cases, the system matched ground truth in all cases and processed 30-case batches in about 2.3 minutes versus 9.8 minutes for human reviewers, with 4 of 6 initially ambiguous cases resolved after email response, but the authors note the synthetic, protocol-aligned evaluation does not demonstrate real-world EHR performance and requires prospective validation.
limitation: Evaluation used only synthetic cases with label definitions aligned to the same protocol rules used to design the rule engine, so results reflect implementation fidelity not real-world clinical performance.
tag: Automated dual reading
key_points: System built on n8n platform combining large language model-based variable extraction from free-text clinical narratives and investigator email replies with deterministic 7-criterion rule engine. | Ambiguous cases triggered structured email query to investigator with two requests at 24-hour intervals; after 48 hours without reply referred for manual review. | Tested on 90 synthetic cases generated by Claude Sonnet 4.6, Gemini 3.1, and Grok 4, balanced across eligible, not eligible, and ambiguous. | Five independent reviewers achieved mean accuracy 96.7% with Fleiss ba 0.910 and required mean 9.8 minutes per 30 cases versus 2.3 minutes for TrialTriage.
rundown: TrialTriage was implemented on n8n to extract variables from clinical narratives and email replies using large language models, then apply a prespecified 7-criterion protocol via a deterministic rule engine to classify each case as eligible, not eligible, or ambiguous.

In testing, 90 synthetic cases were generated independently by three different models with balanced eligibility distributions and author-confirmed answer keys, and five reviewers classified the Claude subset using a uniform survey form for comparison.

When first-pass classification was ambiguous, the workflow sent structured email queries at 24-hour intervals and reran classification after reply capture; in a 6-case subset, four converted to definitive status while two stayed ambiguous due to non-actionable replies.
sources:
- peer_reviewed | JMIR Formative Research | https://doi.org/10.2196/100779 | 2026-08-05
prev: 0000000000000000000000000000000000000000000000000000000000000000

sha256

244046a80329f8345f3acbd1396b88c0da5b8cf816bb619563d8d8ccdbcc39a2

previous

0000000000000000000000000000000000000000000000000000000000000000