Public health data was never built to be read together.
Unicasys reconciles payments, device events, trials, and publications into one searchable entity graph — and watches ClinicalTrials.gov and PubMed so that every version change and new paper finds you automatically.
Five registries. One record of truth.
Each source publishes on its own schedule, in its own format, with its own quirks. Unicasys normalizes all five so a physician, a device, a facility, or a sponsor reads the same everywhere.
OpenPayments
Track industry payments to providers and teaching hospitals — consulting fees, research funding, ownership stakes — patterned across specialties and institutions.
MAUDE
Surface device malfunction and adverse event reports as they're filed — spotting the pattern of a recall long before the recall notice exists.
ClinicalTrials.gov
Follow trials from registration to results — flagging enrollment gaps, missed deadlines, and sponsors who never post outcomes.
Publications
Link peer-reviewed research on PubMed back to the physicians, sponsors, and trials that produced it — citation by citation.
Facilities
Track hospitals, surgical centers, and care facilities — ownership, quality ratings, and how they tie back to the payments and devices above.
From raw filing to flagged signal.
Pulled at the source
Filings are ingested directly from CMS, FDA, NIH, and NLM as they're published — no scraping delay, no stale exports.
Matched across registries
The same physician, device, facility, or sponsor is resolved across five inconsistent naming conventions into a single entity.
Outliers surfaced
Statistical thresholds and cross-source correlation flag what's worth a closer look — before it's news.
Know the moment a trial changes.
ClinicalTrials.gov edits studies silently — amended protocols, changed endpoints, moved enrollments. Trials Watcher polls the trials you care about and flags the moment a version changes, so a quietly altered study never slips past.
- Track up to 10 clinical trials for free.
- Monitor version changes on ClinicalTrials.gov.
- Get notified when a tracked trial is updated.
Stay current on new publications — automatically.
Publications Watcher monitors PubMed for articles that match your search criteria, generates concise AI summaries, and delivers them straight to your inbox. No manual searching required.
- Automated PubMed tracking with email alerts for new articles.
- AI-generated text summaries delivered to your inbox.
- Exclude journals you're not interested in.
Summaries are text-only. We do not process or analyze graphs, figures, or images.
How records become one entity.
Matching is deterministic and reproducible end to end: every record carries a stable, content-addressed ID, and the same real-world person, device, facility, or sponsor always resolves to the same entity — no matter which registry it came from.
Canonical names
Names are lowercased, legal suffixes stripped (Inc, LLC, Corp), and punctuation removed — so "Company X Corporation" and "company x corp." collapse to the same key.
Stable content-addressed IDs
A hash of the normalized name plus a type prefix (pro_, co_, fac_, dev_) yields the same entity ID every time the same record appears. No sequential keys, no drift between loads.
NPI, NCT, CCN anchors
Providers, trials, facilities, and devices anchor on their official identifiers, so payments, enrollments, and event reports attach to the right record instead of a text guess.
Bridge tables
Manufacturer and taxonomy cross-walks reconcile company aliases and provider specialties across OpenPayments, MAUDE, and trial sponsors — the step that connects a device maker to a trial it funds.
Blocking & collapse
Records are grouped by blocking keys and deduplicated, so alias sets collapse into one canonical entity instead of a pile of near-duplicates.
Bloom-filter narrowing
A compact bloom index prunes candidate sets in milliseconds, so searches and link queries run across millions of resolved entities without scanning everything.
A single-person research project.
Unicasys is an independent research project built and maintained by one person E. Hakim. The data comes from public U.S. government registries; the reconciliation, analysis, and tooling are independently developed.