README.md

---
license: cc-by-4.0
task_categories:
  - tabular-classification
tags:
  - fda
  - pharmaceutical
  - recalls
  - medical-devices
  - drug-safety
  - compliance
  - regulatory
pretty_name: FDA Recall Intelligence
size_categories:
  - 10K<n<100K
---

# FDA Recall Intelligence

**74,000+ FDA drug and device recalls** enriched with facility inspection history, compliance action context, and AI-classified failure categories. Answers the question: what led to this recall, and what's the facility's track record?

This is a 10% sample. Get the full dataset (74,000+ recalls with AI classifications) on Gumroad → https://mandiasdata.gumroad.com/l/fda-recall-intelligence

## Why This Dataset Exists

FDA publishes recall data separately from inspection and enforcement data. To understand why a recall happened and whether the facility has a pattern of problems, analysts must manually cross-reference 3–4 databases. This dataset pre-joins that context into a single row per recall.

## Schema

| Column | Type | Description |
|--------|------|-------------|
| `recall_number` | str | FDA recall identifier |
| `recall_type` | str | "drug" or "device" |
| `firm_name` | str | Recalling firm |
| `product_description` | str | Product being recalled |
| `reason_for_recall` | str | FDA-stated reason (free text) |
| `status` | str | Ongoing, Completed, Terminated |
| `recall_class` | str | Class I / II / III (device recalls) |
| `initiation_date` | date | When the recall was initiated |
| `report_date` | date | When FDA was notified |
| `distribution_pattern` | str | Geographic distribution of recalled product |
| `product_quantity` | str | Amount of product recalled |
| `product_ndc` | str | National Drug Code (drug recalls) |
| `fei_number` | str | Matched FDA Establishment Identifier |
| `facility_total_inspections` | int | Inspection count for this facility |
| `facility_oai_count` | int | Serious inspection failures at this facility |
| `facility_last_classification` | str | Most recent inspection outcome |
| `facility_last_inspection` | date | When facility was last inspected |
| `facility_compliance_actions` | int | Warning Letters / Injunctions / Seizures |
| `facility_warning_letters` | int | Warning Letters issued to this facility |
| `preceded_by_oai` | bool | Was the most recent inspection before this recall an OAI? |
| `days_since_last_inspection` | int | Days between last inspection and recall initiation |

## Key Insights

- **74,000+** total recalls (23,000+ drug, 50,000+ device)
- **79%** matched to an FDA-inspected facility
- **4,100+** recalls preceded by an OAI inspection
- **Date range**: 2005 – present

## AI-Enriched Failure Categories (Drug Recalls)

17,000+ drug recalls have been classified by failure mode using Claude AI:

| Category | Count | Share |
|----------|-------|-------|
| Sterility | 6,469 | 37% |
| cGMP Deviations | 3,536 | 20% |
| Contamination | 1,922 | 11% |
| Potency | 1,573 | 9% |
| Mislabeling | 1,416 | 8% |
| Stability | 890 | 5% |
| Foreign Material | 694 | 4% |

Each classification includes a subcategory (e.g., "penicillin cross-contamination", "subpotent API", "label mixup") for granular analysis. Device recall classifications are in progress.

## Sample

This is a 10% random sample. The full dataset is available for purchase — details below.

## Sources

- OpenFDA Drug Enforcement (drug recalls)
- OpenFDA Device Enforcement (device recalls)
- DDAPI Inspections & Classifications (facility context)
- DDAPI Compliance Actions (enforcement context)

## License

Derived from U.S. government public data. AI classifications generated by Claude (Anthropic). Provided as-is.

## Purchase

For the full dataset: https://mandiasdata.gumroad.com/l/fda-recall-intelligence