OpenFDA Drug Safety Finder

What does OpenFDA Drug Safety Finder do?

OpenFDA Drug Safety Finder queries the official openFDA JSON API and saves one flat row per record for drug safety, labeling, and recall work. It reads three public datasets on https://api.fda.gov:

• FAERS adverse events (mode=event, /drug/event.json)
• Drug labels / Structured Product Labeling (mode=label, /drug/label.json)
• Drug recalls / enforcement reports (mode=recall, /drug/enforcement.json)

The Actor does not download or parse FDA website HTML, and it does not need a login. An openFDA API key is optional. Every row includes mode and source set to openFDA. Empty fields are left out.

Why use OpenFDA Drug Safety Finder?

Pharmacovigilance, medical information, and quality teams use openFDA when they need a spreadsheet of public safety records instead of a browser session:

• Pull recent FAERS adverse event reports for a brand, generic, or substance
• Export drug label sections such as indications, warnings, dosage, and adverse reactions
• List drug recall classification, reason, recalling firm, and distribution
• Schedule the run, call it from the Apify API, or connect it to Make, Zapier, or n8n

Because the source is the documented API, the run does not depend on the FDA website layout, a headless browser, or proxy rotation. That is a different approach from Store Actors that scrape Drugs.com, DailyMed HTML, or fda.gov pages and charge more for the same public records.

What data can OpenFDA Drug Safety Finder extract?

Dates stored as YYYYMMDD are rewritten as YYYY-MM-DD. Text fields longer than 4,000 characters are cut and end with … [truncated]. Lists are de-duplicated.

Adverse events (FAERS)

Field	What it contains
safetyreportid	FAERS safety report id
receivedate	Date FDA received the report
serious	true when openFDA marks the report serious (1), false when it is not (2)
seriousnessdeath, seriousnesshospitalization	true when that outcome flag is 1. The same rule applies to seriousnesslifethreatening, seriousnessdisabling, seriousnesscongenitalanomali (openFDA's field name), and seriousnessother
patientSex, patientAge	Mapped from FAERS codes (Female, Male, Unknown; age such as 65 years)
reactions	MedDRA preferred terms
drugs, indications	Product names as reported, and drug indications. Suspect products are listed before concomitant ones
manufacturer	openFDA manufacturer names, or the sender organization when it is not the public-use sender
brand_name, generic_name, substance_name, rxcui	openFDA harmonized names and RxNorm CUIs, when FDA attached them
reportUrl	Official API URL for that safety report id

A FAERS report can list several drugs and several reactions. openFDA does not link one drug to one reaction. Presence on a row is not evidence that the drug caused the reaction.

Drug labels

Field	What it contains
set_id, id	SPL set id and document id
brand_name, generic_name, manufacturer_name, substance_name	openFDA names
indications_and_usage, warnings, dosage_and_administration, adverse_reactions	Label text, truncated
product_ndc, package_ndc, rxcui, spl_id, unii, application_number	openFDA identifiers
effective_time	Label effective time
spl_url	DailyMed SPL page built from set_id when that id is a UUID

Recalls

Field	What it contains
recall_number, status, classification	Enforcement report id, status, and class
product_description, reason_for_recall, code_info	What was recalled and why
recalling_firm, report_date, distribution_pattern, product_quantity	Firm, date, where it was distributed, and how much
brand_name, generic_name, substance_name, manufacturer_name	openFDA names when the report has them

How to search FAERS adverse events, drug labels, and recalls

1. Open the Actor and choose Try it or Start.
2. Leave Search as a plain name such as aspirin, or paste an openFDA search expression.
3. Choose Mode: adverse events, drug labels, or recalls. The default is adverse events.
4. If Search is empty, fill Brand name, Generic name, or Substance name instead. Those three are combined with OR.
5. Set Maximum results (1–500, default 50). The Actor sorts by the dataset date, newest first, and stops at this cap.
6. Optionally paste an openFDA API key. Without it the public limit is 240 requests per minute and 1,000 requests per day per IP. With a free key the daily limit is 120,000. The Actor still waits between calls. The key is sent only to api.fda.gov.
7. Run the Actor and open the dataset. Download JSON, CSV, Excel, or XML, or call the dataset API.

How a plain name becomes a search

If Search contains field:term syntax, it is sent as written. Examples:

• patient.drug.medicinalproduct:aspirin
• openfda.brand_name:"tylenol extra strength"
• openfda.generic_name:"acetaminophen" AND effective_time:[20200101 TO 20241231]

If Search has no field colon, the Actor quotes the whole phrase and ORs it across the name fields for that mode.

Adverse events:

(patient.drug.medicinalproduct:"aspirin" OR patient.drug.openfda.brand_name:"aspirin" OR patient.drug.openfda.generic_name:"aspirin" OR patient.drug.openfda.substance_name:"aspirin")

Drug labels:

(openfda.brand_name:"aspirin" OR openfda.generic_name:"aspirin" OR openfda.substance_name:"aspirin")

Recalls use those three openFDA fields plus product_description.

A lowercase word such as “cold and flu” stays one phrase. The uppercase operators AND, OR, and NOT are only special when you write a field expression yourself. Parentheses in an expression must balance. Do not paste an API key into Search.

How much will it cost to get openFDA drug safety records?

On the Apify Store this Actor is priced as pay per result: about $0.003 per result written to the dataset. A run that saves 50 rows is about $0.15 in result charges, before any platform plan discounts. Rows that are not saved are not result rows. Raising maxResults raises the maximum number of charged rows.

The price is not hard-coded in the Actor. Store billing uses the normal dataset-item event. An openFDA API key does not change the per-result price. It only raises the daily request cap. You can create a key from the openFDA authentication page.

There is no browser time and no residential proxy in a normal run, so platform compute stays small next to Actors that render FDA or pharmacy HTML.

Input

See the Input tab for every field. storage/key_value_stores/default/INPUT.json is a small sample (aspirin, adverse events, 10 rows) for a local apify run.

Field	Required	Notes
search	One of the search fields	Plain text or an openFDA expression. Example prefill: aspirin
mode	Yes	event (default), label, or recall
brandName, genericName, substanceName	One of the search fields	Used only when search is empty
maxResults	No	Integer from 1 to 500. Default 50
apiKey	No	Secret. Never stored in the dataset or the run overview
skip	No	Advanced. Integer from 0 to 25000, for resuming a result set

The run fails with a clear status message when the effective search is empty, mode is not one of the three values, or maxResults / skip / apiKey is not valid.

Output

You can download the dataset as JSON, CSV, Excel, XML, or RSS. The key-value record OUTPUT stores the row count, the openFDA total when the API sent one, the effective search, and the endpoint. It does not contain the API key. Adverse-event runs also store the FAERS disclaimer there.

{
  "mode": "event",
  "source": "openFDA",
  "safetyreportid": "10003304",
  "receivedate": "2014-03-12",
  "serious": true,
  "seriousnessdeath": true,
  "patientSex": "Female",
  "patientAge": "65 years",
  "reactions": ["Nausea", "Headache"],
  "drugs": ["ASPIRIN 81 MG", "ACETAMINOPHEN"],
  "indications": ["Pain"],
  "manufacturer": "Example Pharma",
  "brand_name": ["ASPIRIN"],
  "generic_name": ["ASPIRIN"],
  "rxcui": ["243670"],
  "reportUrl": "https://api.fda.gov/drug/event.json?search=safetyreportid%3A%2210003304%22&limit=1"
}

{
  "mode": "label",
  "source": "openFDA",
  "set_id": "0058175f-3474-40c3-a046-6cfaec86d84b",
  "brand_name": ["ASPIRIN"],
  "generic_name": ["ASPIRIN"],
  "manufacturer_name": ["Example Labeler"],
  "indications_and_usage": "Temporarily relieves minor aches and pains.",
  "effective_time": "2024-04-16",
  "spl_url": "https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=0058175f-3474-40c3-a046-6cfaec86d84b"
}

{
  "mode": "recall",
  "source": "openFDA",
  "recall_number": "D-0170-2024",
  "status": "Terminated",
  "classification": "Class II",
  "product_description": "Low Dose ASPIRIN, 81 mg, enteric coated tablets.",
  "reason_for_recall": "Products were stored outside labeled conditions.",
  "recalling_firm": "EXAMPLE DISTRIBUTION",
  "report_date": "2024-01-03",
  "brand_name": ["ASPIRIN"],
  "generic_name": ["ASPIRIN"]
}

The output tab has separate tables for adverse events, labels, and recalls. A run uses one mode, so the other tables are empty.

Tips

• Start with maxResults at 10 while you confirm the drug name, then raise it.
• Prefer a field expression when you need one column only, for example patient.drug.openfda.generic_name:"ibuprofen".
• Use skip to continue a previous run. openFDA will not page past a skip of 25000; narrow the search instead of looping forever.
• The Actor requests at most 100 records at a time, waits between calls, and retries HTTP 429 and 5xx with backoff. A search openFDA cannot parse fails immediately instead of being retried.
• A run with no matches finishes successfully and writes resultCount: 0.
• This Actor does not count reactions, compute disproportionality, or join a drug to a reaction. Those steps need a method you choose after export.

How this differs from an FDA HTML scraper

HTML scrapers read fda.gov, DailyMed, or pharmacy pages and break when the markup changes. This Actor uses the JSON API documented at open.fda.gov/apis. Drug adverse events, labels, and recalls are three mode values on that API, not three website crawls. openFDA annotates records with brand, generic, substance, and RxCUI when it can. The data are U.S. government public-domain data. API use follows the openFDA terms.

FAQ, disclaimers, and support

Is this medical advice?

No. This Actor is for research, pharmacovigilance review, and education. It is not a diagnosis, a treatment recommendation, or a substitute for the product label or a clinician. Do not make a medical decision from a row in this dataset.

Do FAERS reports prove the drug caused the reaction?

No. FDA Adverse Event Reporting System reports are submitted by patients, clinicians, and manufacturers. FDA does not require a proven causal link. Reports are not verified scientific studies, they are not a count of how often an event happens, and a drug can appear often simply because many people take it. Submission of a report is not an admission that the drug caused the event. The same limit is stored on the OUTPUT record for every adverse-event run: a report does not prove that the drug caused the reaction, and these rows are not medical advice or incidence estimates.

Does the Actor collect private patient data?

It reads public openFDA records. Those records can include patient sex and age as FDA published them. They are not a medical record from your system. Do not put patient names, emails, phone numbers, or medical record numbers in the search box. Personal data is protected by the GDPR in the European Union and by other laws elsewhere. Use the output only for a legitimate purpose.

What are the API limits?

Without a key: 240 requests per minute and 1,000 requests per day per IP. With a free key: 240 requests per minute and 120,000 requests per day per key. See openFDA authentication. The Actor stays under the per-minute cap and does not log the key.

What if openFDA is busy?

The Actor retries rate-limit and server errors, then fails with a short status message if openFDA keeps failing. It does not scrape a backup HTML page. A search with no matches is a successful empty run, not an error.

Where do I report a problem?

Use the Issues tab on the Actor page. For programmatic runs, use the API tab and the dataset items endpoint from the run.

Local development

pip install -r requirements-dev.txt
pytest

Tests use fixtures and a mock HTTP client. They do not call openFDA. apify validate-schema checks the input, dataset, and key-value store schemas.