FDA MAUDE · openFDA · Post-market intelligence
Recall risk, ranked months before the announcement.
MedRecall reads every adverse-event report the FDA publishes and ranks your device portfolio for post-market review — so your QA team works the drift, not the deadline.
| Device family | AE velocity · 90d | Severity mix | Mfr history | Risk score | Est. horizon |
|---|---|---|---|---|---|
Infusion pump · IP-400 series Class II · 510(k) K183221 |
+212% | 14 inj · 2 death | 3 prior recalls | 87/100 | ~5 mo |
Surgical stapler · SS-2 Class II · 510(k) K201104 |
+64% | 31 inj · 0 death | 1 prior recall | 61/100 | ~9 mo |
Patient monitor · PM-7 Class II · 510(k) K192288 |
+11% | 3 mal · 0 inj | 0 prior recalls | 24/100 | — |
Monitoring tells you a recall happened.
Ranking tells you where to look now.
Every commercial tool in this category monitors: it matches FDA notices to your inventory after the agency acts. The category's own market leader states plainly that it does not predict — signals are aggregated, humans adjudicate. The step from after to ahead is unclaimed.
Recall monitoring tools
- Match published FDA recall notices to your device inventory
- Alert after the enforcement announcement
- Aggregate signals; a human reads everything
- Workflow: track the response, document the close-out
MedRecall
- Ranks your portfolio by recall-risk trajectory, continuously
- Works from adverse-event velocity — the input that moves before notices do
- Quantified scoring: 90-day velocity, severity mix, manufacturer history
- Decision support: a prioritized review queue with estimated lead time
Category floor for monitoring-only intelligence: $449/mo (published competitor pricing, verified 2026-10).
Public data. Disciplined engineering.
Published science.
The methods behind adverse-event-based recall prediction are peer-reviewed and public — the 2025 West J Emerg Med work showed machine learning can flag recall-bound devices months out; nothing commercial has built on it. Our pipeline is the engineering layer that paper named as future work.
Every report, deduplicated
Continuous ingestion of MAUDE adverse events and recall enforcement records from openFDA, with linkage and deduplication — the unglamorous layer naive dashboards skip.
Trajectory, not counts
30/60/90/180/365-day adverse-event velocity and acceleration, severity language scoring, and manufacturer recall history — features that describe a drift, not a snapshot.
Tuned to stay quiet
A Random Forest classifier tuned for high specificity — a flag should mean something. Every score exposes the features that drove it.
A ranked queue
Monthly signal sheet, continuous API, and estimated review horizon per device family. Your team's judgment does the rest — on the right devices first.
“MAUDE is garbage” is the objection.
Methodology is the answer.
FDA's own MAUDE documentation warns against naive use: underreporting, unverified causation, no cross-device rate comparisons. We agree — which is why the methodology, not the data access, is the product.
The hidden-reports era is handled
From ~1997–2019 the FDA's ASR program kept 1.1M+ injury reports out of MAUDE. Any model trained across that boundary sits on a false baseline. Ours treats pre- and post-2019 as structurally different regimes.
Volume ≠ danger, and we never say it is
Over 90% of MAUDE reports come from manufacturers; high volume can reflect pharmacovigilance diligence. Scores rank trajectories for review — they do not assert causation or compare device safety rates.
Linkage before inference
Duplicate and follow-up reports inflate naive event counts. Record linkage and deduplication run before any feature is computed — the standard methods literature, applied.
Tracking AEMS from day one
The FDA's AEMS platform is consolidating legacy adverse-event systems. Our ingestors are being migrated to the successor structures so the pipeline survives the data-platform change.
Primary sources: FDA openFDA (MAUDE, recall enforcement); GAO-09-190; KHN investigation of the ASR program (2019); Ensign & Sandoval, eGEMs (2017).
Priced as intelligence,
not as workflow.
Monitoring tools charge hundreds a month to tell you what already happened. Pilot pricing starts where monitoring ends.
- One device portfolio, continuously ranked
- Monthly portfolio signal sheet (PDF)
- Read API — scores, features, horizons
- Methodology walkthrough on onboarding
- Every portfolio under one roof
- Alerting on score and velocity thresholds
- Full API access + historical series
- Named onboarding & quarterly methodology review
Pilot onboarding is by application while we stay small enough to walk every customer through the method.
What MedRecall is not.
Clear scope is a feature. It keeps the tool honest, your team oriented, and the regulatory line clean.
Not clinical decision support. Outputs inform manufacturer QA/RA post-market review — never patient-care decisions.
Not a regulatory system of record. Your QMS and your FDA reporting obligations remain yours.
Not a verdict. Scores prioritize human review; the judgment stays with your people — the category's own liability-safe voice.
Not a data dump. We sell ranked judgment, not API access to public data you already have.
Put a number on the drift before the notice does.
Tell us your device families. We'll run the pilot ranking and walk your QA lead through every feature that moved.
Include your device classifications and portfolio size. Applications reviewed within two business days.