Fda 510K Five Year Recall
Will a medical device cleared by FDA through the 510(k) pathway appear in FDA's device recall enforcement reports, for a recall initiated within five years of the clearance decision?
Versioned research data. Read the suitability requirements before interpreting scores.
Suitability and evaluation
Evaluation setting: retrospective research. Read the prediction setting and limitations below. Public validation is development data; downloadable holdout labels must stay outside the evaluated agent workspace.
At a glance
| Question | Will a medical device cleared by FDA through the 510(k) pathway appear in FDA's device recall enforcement reports, for a recall initiated within five years of the clearance decision? |
|---|---|
| Field | U.S. Food and Drug Administration (openFDA, Center for Devices and Radiological Health) / binary-classification |
| Rows | 48,623 |
| Documented splits | 3 |
| This version | v-68e59a5ac91ad404 |
| Released | 2026-11-13 |
| Validation ROC AUC | 0.6798441538385072 |
| Validation Brier | 0.0794854534527989 |
| Artifact | 4402989d50d250e5d9eab6a8f93c00997c3be07ffecedd577a04ceddd1c15d15 |
Get this dataset
How to load it
from datasets import load_dataset
ds = load_dataset('earino/factory-fda-510k-five-year-recall', revision='v-68e59a5ac91ad404')
The version links and checksums identify this artifact. Exclude the labelled holdout from the agent's workspace when running an evaluation.
What is inside 
| Split | Rows | Positive labels |
|---|---|---|
| eval | 9,046 | — |
| holdout | 9,014 | — |
| train | 30,563 | — |
Files and checksums
| Download | Bytes | SHA-256 |
|---|---|---|
| LICENSE-EVIDENCE.json | 697 | 863cdacc293823d37d8480b06ea3a1dc0df4b081ad98acfbda7c50ebb7dd6008 |
| LICENSE.md | 1,172 | c451ca1e6eb9b0775760b04b23cfb825f3492c5aabc65ae0f30371c50a33e4f4 |
| README.md | 4,973 | 241266950dca7d7c833a12998c41a821f0c80cfd2572a8fcdc7ed8ceeb93d755 |
| SOURCES.json | 1,187 | 10cfe9b3d784fdc678c823fd3c1ff2a7f66d0838b850029c757b89e24db1e540 |
| evaluation.json | 2,290 | 54c5ee04d19b8136f2091bb7e6bda575a28ed8053301997e5ac69a4989945be6 |
| extract__meta.json | 1,748 | f1eb741be4fbcb543452b835cd6e33d1cdb0be59d635cd3d67e0e7f2514f0853 |
| extract__private__holdout.csv | 517,988 | b3e91aad60e68bddd48e7452d8d06cbb994766c799e6b988e606a5f11707407d |
| extract__public__eval.csv | 519,604 | 825221abdca3a5cddf6df1694107bc8951d1ec9bbed97f346ea75a2d9f7cfa09 |
| extract__public__train.csv | 1,742,647 | 3e449ebcacdbaf2919571aa443056b7ee78fd5805221bf68902a5edff6a1f25f |
| extract__quality.json | 17,027 | 435574c5e5e2757ff4dc699ccbeff9be89dcd73b71271072151933054aee1795 |
| get_dataset.py | 3,680 | fad89222d50f9d9709225c34909be74c591df409c08b8b5951eba9d3cadfdd48 |
| qualification.json | 16,632 | bde81d0a427f5cfcff7aadd311051801b8400e723d87f58fd744757728c95a77 |
| reproduction__LICENSE-CC0-1.0.txt | 7,048 | a2010f343487d3f7618affe54f789f5487602331c0a8d03f49e9a7c547cf0499 |
| reproduction__LICENSE-MIT.txt | 1,124 | 56b932ff064e5da113ae10175c8546325f64f26e1ece71683ffcbf8b2304fe1c |
| reproduction__artifact_scrub.py | 11,369 | 8c663253e691a7b8751f9c853b40a3752a833360caafb1ea8e8f534e64af4b68 |
| reproduction__build.py | 28,337 | 5b7198c03b84241760d2bacd707197a2d2b5b97e8285989bf52ecd5a5723f562 |
| reproduction__construction.json | 20,951 | c7ef917f65e5a389807ccd64fc6d4ceece359d0e0b0eae8cbcfb83c92575eafe |
| reproduction__dataset_package.py | 7,927 | c2364defe58dcc99322be59a1bb04894ce00e823ce37afeb37bc31012f744327 |
| reproduction__dataset_run.py | 12,381 | 8085fae14d6e93b852e692e18c1ef72113a3601f48f78901b6d1ae6db278de57 |
| reproduction__get_dataset.py | 3,680 | fad89222d50f9d9709225c34909be74c591df409c08b8b5951eba9d3cadfdd48 |
| reproduction__leakage_battery.py | 17,379 | adbf7ee0aca0012b78cc1f1e4aeb968a23717cae30f66a02e080215e75f830e3 |
| reproduction__provenance.json | 573 | 8880de985efd8e79e50bfe94346960358c3143b60c1ed423139fa2e3911da7bc |
| reproduction__qualify_dataset.py | 66,067 | bff6d43d0904182e2612c32bab702e41addbf389904a236c209e78be39e37a6b |
| reproduction__recovery_recipes.py | 17,609 | f64c542fcda232169c95ace17421b45275194b934938c0ea4c9b68e6f844ad0e |
| reproduction__requirements.txt | 160 | 5c72f3268174d20364624f76680e189c408a05dba2e7b8cfe12d077b3865e7f0 |
| reproduction__source.json | 16,989 | 5a6312f07712549bbf40169c6c84dcdac03e98f38441c267469026ca9dee63a0 |
| reproduction__source_io.py | 5,607 | 0c60eed4968cd342148efbaac7e4fa98f00708d8e6c14c0929928687b9af1ea3 |
| reproduction__train.py | 2,500 | a3c6bcf13735bc85c52129ded68f839090dffdc266ebc3810dc61b2e6ea5e7e8 |
| reproduction__validate.py | 1,277 | b597e7f84fed614e64b4a86fbecbd6ec0145916a24d0d221a796586976418e7b |
| reproduction__validate.sh | 482 | 3f06ca48f11c2e05d331b4d8394284925de660c1d4b086d2f8243959e83b7e6d |
| source-checks.json | 9,331 | 0104544e6bd5839225510077877ccb6ef6c9adea1a722916b624a2bcd32ac51c |
| source-snapshot.tar.gz | 528,983,417 | e9cf3993d15e54ddfbdbae0ab51b8870ec822493489ac6e59304fd1362d0bce0 |
| verify_dataset.py | 1,276 | f4a4bb7f565cdbd075c520621f881876b491eaed91a5e6e70afb31e4753f9834 |
Release notes · 2026-11-13
Will a medical device cleared by FDA through the 510(k) pathway appear in FDA's device recall enforcement reports, for a recall initiated within five years of the clearance decision?
Population: 510(k) clearance records that carry both a k_number and a decision_date, whose decision_date falls in 2004-01-01 to 2019-12-31, and whose decision_code is SESE (substantially equivalent). No restriction by device type, firm, size or geography. Records with a missing or unparsable decision_date are excluded, and any record whose 1826-day window is not fully covered by the recall snapshot is dropped rather than labelled 0.
Target: y = 1 if any openFDA device/recall product record lists the row's k_number in its k_numbers array and has event_date_initiated on or after the row's decision_date and on or before decision_date + 1826 days; y = 0 if that five-year window closes on or before the recall snapshot date (2026-11-12) and no such record exists. Rows whose window is not fully observed are dropped, not labelled 0.
Prediction time: The clearance decision date recorded on the 510(k) record (decision_date). Only fields on that record plus counts of strictly earlier clearance dates are used.
Splits: Fixed chronological boundaries on decision_date, no random sampling and no re-splitting: train 2004-01-01 to 2013-12-31, eval 2014-01-01 to 2016-12-31, holdout 2017-01-01 to 2019-12-31. All three windows end at least five years before the recall snapshot (last_updated 2026-11-12), so every row's horizon is observed. No embargo is required because each row is a single clearance event, no feature uses a recall outcome, and applicant_prior_clearances uses only strictly earlier decision dates. Splits, label rule and horizon are frozen before any performance measurement.
Outcome horizon: 60 months (5 years) from the clearance decision date.
Inputs: product_code: FDA three-character product code on the clearance record (for example MQB or DZE), the finest standard device-type bucket available at clearance.; advisory_committee: Two-letter review panel code and its description (for example RA Radiology, DE Dental, CV Cardiovascular).; device_class: Risk class 1, 2 or 3 attached to the record by openFDA's harmonised join against FDA's current classification database.; clearance_type: Review track: Traditional, Special, or Abbreviated 510(k).; decision_code: FDA decision code on the clearance record, for example SESE for substantially equivalent.; review_days: Calendar days from date_received to decision_date.; third_party_flag: Y or N flag for review by an FDA-accredited third-party reviewer.; applicant_prior_clearances: Count of 510(k) records with the same applicant string and a decision_date strictly earlier than this row's decision_date, a decision-time proxy for the firm's regulatory experience.; applicant_country_code: Two-letter country of the applicant's filed address, distinguishing domestic from foreign submitters.
Recorded structural checks and baselines
PASSED Gate 1.6.1.
Existing unmodified runner baseline on public validation only; no holdout scoring
These historical checks do not establish availability at each prediction time or protection against every shortcut. Scores below describe the recorded protocol.
Limits
The outcome is linked only through the 510(k) number: 20,792 of 59,705 device recall records (about 35%) carry no k_number and are invisible, so the label is a lower bound on true recall incidence for cleared devices.
One recall event frequently names several 510(k) numbers and one firm recurs across many rows, so positives are correlated within recall events and within firms; effective sample size for positives is smaller than the row count suggests.
event_date_initiated is the firm's action date, not FDA's classification date, so recalls initiated near the end of a five-year window but classified later can be miscounted at the boundary.
device_class comes from openFDA's harmonised join against FDA's current classification database, so it can differ from the class in force at clearance (current-vintage limitation, stated rather than hidden).
Applicant strings are free text, so the same firm appears under spelling variants and applicant_prior_clearances undercounts large filers.
The 510(k) record does not say whether the device was ever marketed; an unknown share of 'no recall' rows are devices that never reached patients, which is a definitional limitation of the label, not a data error.
The bulk files are a moving current vintage (FDA notes later updates can change old records), so the exact row set is reproducible only by archiving the two zips and their checksums.
510(k) records carry a named contact person and street address; those columns are dropped before release, and no individual-level field is published in the derived CSVs. Only the 510k and recall endpoints are used, so the GMDN-restricted content noted on the licence page is out of scope.
The outcome is linked only through the 510(k) number: 20,792 of 59,705 device recall product records carry no k_number and are invisible, so the label is a lower bound on true recall incidence for cleared devices.
event_date_initiated is the firm's action date, not FDA's classification date, so a recall initiated near the end of a window but classified later can sit at the boundary.
device_class comes from openFDA's harmonised join against the current classification database, so it can differ from the class in force at clearance.
The 510(k) record does not say whether the device was ever marketed; an unknown share of no-recall rows are devices that never reached patients.
When training labels became available is not documented by the publisher, so the label timing is recorded as unverified rather than as a forecasting claim.
No uncertainty interval; observed differences do not establish universal headroom
Source, licences, and citation
U.S. Food and Drug Administration (openFDA, Center for Devices and Radiological Health)
None required under CC0 1.0. Recommended credit: 'Source: U.S. Food and Drug Administration, openFDA device/510k and device/recall endpoints (public domain, CC0 1.0)', with the archive date.
Source: Public domain: CC0 1.0 Universal dedication by FDA, on the independent statutory basis of 17 U.S.C. 105 (no copyright in U.S. Government works). Compilation: CC0-1.0. Code and documentation: MIT.
Our additional compilation rights only; upstream terms remain in force.