Category: eClinical Systems
Coding adverse events and medications to standard dictionaries is repetitive expert work that piles up ahead of every interim and final lock.
Every adverse event and concomitant medication a site reports in free text has to be coded to standard dictionaries — MedDRA for events, WHODrug for medications — before analysis can happen. The volume is large, the work is repetitive, and it concentrates right before interim analyses and database lock, exactly when time is scarcest. AI-assisted coding proposes dictionary terms for incoming verbatims with a confidence signal, so coders spend their expertise on the ambiguous cases instead of re-keying the obvious ones.
Auto-coding tools take the verbatim text a site entered — with its abbreviations, misspellings, and local phrasing — and propose the matching dictionary term, along with a signal of how confident the match is. Clear, high-confidence matches can be routed for streamlined review, while genuinely ambiguous verbatims — vague descriptions, multiple candidate terms, clinical context that changes the coding — are queued for a coder's full attention. Modern language models handle the messy input better than exact-match synonym lists, which is where most of the manual effort used to go. The coder remains the decision-maker, and coding conventions remain the organization's: which term granularity to prefer, how to handle combination products, when to query the site instead of guessing. A proposed code is a draft judgment, and for safety-relevant terms especially, the review step is the point — miscoded events distort the safety profile the whole trial exists to measure.
Accuracy on your own verbatims is the only benchmark that matters: therapeutic areas differ, site phrasing differs, and a tool's published performance may not survive contact with your data. Run it against a study your team has already coded and compare its proposals to the final human-approved codes — overall, and specifically on the ambiguous cases where errors cost the most. Check multilingual handling if your sites report in more than one language. Operationally, examine dictionary version management — MedDRA and WHODrug update on fixed cycles, and version discipline is auditable — plus the audit trail on every proposal, acceptance, and override, and how the tool fits your existing coding workflow and conventions rather than imposing its own.
The retrospective test is nearly free: run the tool over a completed study's verbatims and score its proposals against the codes your team actually approved. That yields an accuracy figure on your own data and shows where the tool is trustworthy versus where it guesses. Teams typically go live in suggestion-only mode first — every proposal still passes a coder — and consider streamlined review for high-confidence matches only after the accuracy record on live data supports it.
It reads the free-text verbatim a site entered and proposes the matching MedDRA or WHODrug term with a confidence signal. Coders review the proposals — quickly for clear matches, carefully for ambiguous ones — and their decision is what enters the database. The tool drafts; the coder codes.
Not responsibly. A meaningful share of verbatims are ambiguous — vague phrasing, several plausible terms, context that changes the answer — and miscoding adverse events distorts the trial's safety picture. The defensible model is automation of the clear cases with expert review retained, especially for safety-relevant terms.
It depends on your data — therapeutic area, site phrasing, and languages all move the number, so treat published figures as a starting claim, not an answer. The reliable test is retrospective: run the tool on a study you've already coded and score it against your approved codes, paying particular attention to the ambiguous cases.
How the tool performs on our own coded studies, how it handles misspellings, abbreviations, and non-English verbatims, how dictionary version upgrades are managed and documented, what the audit trail captures for each proposal and override, and how it fits our existing coding conventions and review workflow.
Syntegra partnered with Datavant to connect its synthetic data engine with Datavant's de-identification and record-linking network, enabling privacy-preserving exchange of synthetic versions of linked real-world health datasets across the healthcare ecosystem.
Vendor: Syntegra · press release (Syntegra (GlobeNewswire)) · 2022-12-15 — Partner: Datavant
Abridge was named the first member of Epic's 'Pals' partnership program, integrating its generative-AI medical conversation summaries directly into Epic EHR workflows so health systems can adopt ambient documentation inside the tools clinicians already use.
Vendor: Abridge · press release (Abridge) · 2023-08-16 — Partner: Epic
Saama's AI-driven clinical analytics supported Pfizer's review of clinical trial data during its COVID-19 vaccine program; the companies subsequently expanded their relationship across Pfizer's broader R&D data operations.
Vendor: Saama Technologies · press release (Saama (Business Wire)) · 2024-02-12 — Partner: Pfizer
Kaiser Permanente rolled out Abridge's ambient AI clinical documentation technology across its integrated health system — 40 hospitals and more than 600 medical offices — letting clinicians generate draft visit notes from patient conversations instead of typing during encounters.
Vendor: Abridge · press release (Kaiser Permanente) · 2024-08-14 — Partner: Kaiser Permanente
A quality-improvement survey at the University of Kansas Medical Center, published in JAMIA Open, reported that clinicians using Abridge's ambient AI documentation self-reported lower perceived work burden and burnout and higher job satisfaction after implementation. Vendor-affiliated authors participated in the study.
Vendor: Abridge · peer reviewed (JAMIA Open (Oxford University Press)) · 2025-02-21