Glossary
Every term below appears somewhere on this site, and each one is the kind of word that changes what a number means. Where a term is used in a page's prose, the first mention is marked with a dotted underline you can expand in place.
Definitions describe what the words mean, not what any particular result shows. A page's own caveats and source lines still apply.
-
AEFI
Adverse Event Following Immunisation: any health problem reported after a vaccination, whether or not the vaccine caused it.
The name describes timing only, which is why causality has to be assessed separately.
Also written: adverse event following immunisation, adverse event following immunization.
-
Background rate
How often a condition occurs anyway in a population, with no exposure involved.
Without it, a raw count of reports cannot be judged high or low.
Also written: expected rate.
-
Bradford Hill viewpoints
Nine considerations — strength, consistency, specificity, temporality, dose-response, plausibility, coherence, experiment and analogy — used to weigh whether an association is causal.
They are viewpoints to reason through, not a checklist that proves causation.
Also written: Bradford Hill criteria.
-
Causality assessment
A structured judgement of whether an exposure plausibly caused an event, weighing timing, dose, biology, alternative explanations and the wider literature.
The judgement is about evidence in populations, not about any one person's case.
-
Confounding
When a third factor is tied to both the exposure and the outcome, creating an association that is not cause and effect.
Example: sicker patients get a drug more often, so the drug looks harmful.
Also written: confounder.
-
Data-quality grade
This site's A-to-D rating of how complete and current a country's reported vaccination data is.
A low grade describes the reporting, not the country's vaccination programme.
Also written: data quality grade.
-
Disproportionality analysis
Comparing how often an event is reported for one product against the background of every other product in the database.
It is a screening tool for finding things to study, not a measure of risk.
Also written: disproportionality.
-
Drug repurposing
Using a medicine already approved for one condition to treat a different one.
The safety record carries over; the evidence of benefit does not.
Also written: repurposing.
-
Druggable target
A protein or pathway in the body whose behaviour a medicine could realistically change.
Being druggable says nothing about whether altering it helps patients.
-
EudraVigilance
The European Medicines Agency's database of suspected side-effect reports for medicines used in the EU.
Like VAERS and FAERS, it collects suspicions, not confirmed causes.
-
FAERS
The US FDA Adverse Event Reporting System: the same idea as VAERS, but for drugs rather than vaccines.
Reporting is voluntary, so counts reflect who reported, not how often an effect happens.
Also written: FDA Adverse Event Reporting System.
-
Lot (batch)
One manufacturing run of a vaccine, identified by a code printed on the vial.
Report counts per lot are not comparable unless you know how many doses that lot contained.
Also written: batch number, lot/batch.
-
Observational study
A study that watches what happens in real-world care without assigning who gets what.
Useful for rare or long-term effects, but open to confounding.
-
Passive surveillance
Safety monitoring that waits for people to send in reports, rather than actively following a defined group of patients.
It can reveal a signal worth investigating, but it cannot measure how common something is.
Also written: spontaneous reporting.
-
Preprint
A research paper published by its authors before independent peer review.
Findings may change or fail review entirely.
-
PRR
Proportional Reporting Ratio: how much more often an event is reported for one product than for all other products in the same database.
A high PRR can come from real risk, from publicity, or from how reports are coded.
Also written: proportional reporting ratio.
-
Randomised controlled trial
A study that assigns participants to treatment or control by chance, so the groups differ only in the treatment.
Randomisation is what lets a trial support a causal claim.
Also written: randomized controlled trial, RCT.
-
ROR
Reporting Odds Ratio: a second way of measuring the same disproportionate reporting that the PRR describes.
Reported alongside the PRR because the two can disagree on sparse data.
Also written: reporting odds ratio.
-
Safety signal
A pattern in reported data suggesting a possible link between a product and an event, strong enough to justify a closer look.
A signal is a question, not an answer: most do not survive investigation.
Also written: disproportionality signal.
-
VAERS
The US Vaccine Adverse Event Reporting System: a database anyone can file a report to after a vaccination.
Reports are unverified and do not establish that the vaccine caused the event.
Also written: Vaccine Adverse Event Reporting System.
Using these definitions elsewhere
The definitions live in one file, /data/glossary.json, shared with opensourcemed.info. It is published under CC BY 4.0: reuse it with attribution. Corrections are welcome through the corrections form.
Last updated 2026-10-31.