— records across 6 emergency specialties, scored against physician-designed criteria.
AI clinical models drown in search engine noise and risk hallucinating on edge-case scenarios. Universal Document filters out the noise, providing — records, machine-scored against hardcoded, physician-designed clinical scorecards across 6 emergency specialties.
| Specialty | Records | Share |
|---|
| Status | Records | Share |
|---|
Every record is graded the same way, by the same ten rules, whether it was scraped an hour ago or a year ago. Passing records ship without further manual review. Records that fail — low evidence grade, low ER-applicability — are held, visibly, until a reviewer clears them. That queue is the second chart above, not a number we'd rather not show you.
Ingestion of raw clinical studies from ClinicalTrials.gov, PubMed, and OpenFDA.
Automated scorecards verify design type, sample size, evidence grade, and ER relevance.
Records that fail the automated scorecard are held for human review; the rest are released on the strength of the scorecard alone.
Curated datasets delivered instantly in flat CSV, JSON, and self-rendering UDS formats.
Customize your dataset properties before checkout. If no filters are selected, you will receive the full un-truncated dataset.
Hardcoded in the validator, not a marketing checklist — every exported record carries its full rule breakdown.
| No. | Rule | What it checks |
|---|---|---|
| 01 | ER Applicability Score | Every record scored 0-10 for real-world emergency department applicability. |
| 02 | Guideline Alignment | Flagged if it contradicts current standard-of-care / ACLS-ATLS-aligned guidelines. |
| 03 | Statistical Integrity | Sample size >=30 and p<0.05 required to pass; underpowered studies are flagged. |
| 04 | Outcome Relevance | Differentiates surrogate endpoints (lab values) from patient-centered outcomes. |
| 05 | Bias Detection | Flags industry-sponsored, single-center, and unblinded studies. |
| 06 | Clinical Plausibility | Compares reported effect sizes to plausible ranges; flags outliers for review. |
| 07 | Actionability | Rates how immediately actionable the finding is in an ER setting (STAT/Routine/N-A). |
| 08 | Evidence Grade | A-F grading, A = meta-analysis down to F = case report / adverse-event report. |
| 09 | Population Fit | Matches study population to typical adult ER demographics. |
| 10 | Recency Weight | Higher weight for studies under 5 years old. |
Every record ships with its full 10-rule scorecard. Records that fail the scorecard are held for human review before release.
A: Every record is scored against a 10-rule checklist designed by a physician. Records that fail the checklist are held for human review before release; passing records are machine-scored and released without further manual review.
A: ClinicalTrials.gov, PubMed, and OpenFDA. We verify and annotate high-yield records across multiple ER specialties.
A: Yes. Our datasets are designed to reduce AI hallucinations by providing high-quality training data, machine-scored against physician-designed criteria, that filters out noise and low-evidence studies.
A: CSV and JSON, ready for any AI pipeline. We also offer UDS (Universal Document) format for customers requiring cryptographic verification.
A: — records across multiple ER specialties (Oncology, Cardiovascular, Sepsis, Trauma, Stroke, Toxicology).
We typically respond within 4 business hours.