How Vaccine Safety Is Monitored and Why VAERS Numbers Are Misread

A number pulled from the VAERS database tells you how many people filed a report after a vaccine, and nothing more. It cannot tell you the vaccine caused any of those events, because VAERS was built to raise questions, not to answer them. The answers come from a separate tier of systems, the Vaccine Safety Datalink and the FDA's BEST system, which compare vaccinated people against the rate of the same event expected in a similar unvaccinated group. Once you see that United States vaccine surveillance is two tools doing two different jobs, the usual misreadings fall apart.

Key points#

• VAERS is a passive, open-submission system: a report logs a coincidence in time, not a proven cause.
• Any large population has a constant background rate of heart attacks, strokes, seizures, and deaths that continue after a shot by chance alone.
• A raw report count has a shaky numerator and no denominator, so it cannot express how often an event truly happens.
• The Vaccine Safety Datalink (VSD) and FDA BEST use linked medical records and comparison groups to test whether an event happens more often than expected.
• Real safety signals, including rare clotting and myocarditis events, were found and confirmed through this comparison, not by counting reports.

Two systems, two very different jobs#

Vaccine surveillance in the United States is layered on purpose. The first layer is designed to be sensitive and fast. The second is designed to be slow, quantitative, and hard to fool. Treating a first-layer tool as if it were a second-layer answer is the single most common source of confusion about vaccine safety data.

Think of a smoke detector next to a fire investigator. The detector should be touchy: a missed fire is far worse than a false alarm, so it goes off for burnt toast as readily as for a real blaze. The investigator arrives afterward, measures, and decides whether there was ever a fire at all. VAERS is the detector. The VSD and BEST are the investigators. No one treats every beep as a house burning down, and no one should treat every VAERS entry as a proven injury.

Why a tally cannot prove harm#

Start with the arithmetic, because that is where the misreading lives. Health emergencies happen every day in any population, vaccinated or not. Give a vaccine to tens of millions of people and, purely by chance, an enormous number of ordinary heart attacks, strokes, miscarriages, and deaths will land in the days right after the shot. Those events were going to happen regardless. A report captures when something occurred, not why.

This is why the CDC warns that VAERS data alone cannot establish that a vaccine caused a reported event, and that it is usually not even possible to use the raw reports to calculate how often an event occurs. The database offers a numerator of loosely defined, unverified reports, and no trustworthy denominator, no comparison group, and no guarantee that each event is real or vaccine related. Counting deaths in that file and presenting the figure as a casualty list skips the only question that matters: how many of those same deaths would have occurred in the same people over the same window with no vaccine involved at all.

The honest question is always comparative. Not whether events happened after vaccination, but whether they happened more often than the background rate expected in a similar group that was not vaccinated. Passive reporting cannot answer that, and was never meant to.

What VAERS is genuinely good at#

Its openness is a feature, not a flaw, as long as you use it for what it does. Because patients, families, clinicians, and manufacturers can all submit, VAERS can surface something new, unusual, or very rare quickly, sometimes fast enough for regulators to act before a product reaches large numbers of people. The CDC describes it as a passive early warning system that monitors the safety of authorized vaccines, and passive is the operative word: nothing is collected automatically, and reports can lack details or contain errors.

So a VAERS entry is best read as a tip, not a verdict. Some tips reflect true reactions. Others are pure coincidence. The volume can also swing with news coverage rather than with any change in a vaccine, which is why a sudden spike in reports says more about attention than about biology.

The systems that actually measure risk#

To answer the comparative question you need defined populations, linked medical records, and a control group. That is the job of the Vaccine Safety Datalink, a partnership between the CDC and a set of integrated healthcare organizations. The VSD draws on electronic health records from its member sites to know who was vaccinated, with what, and when, then follows those same people for specific health outcomes. Analysts rerun the data on a near weekly cycle to test whether the rate of a predefined event of interest after a given vaccine is higher than in a comparison group. When a rate crosses a threshold set in advance, a formal investigation opens.

The introduction of newer vaccines shows the method in action. As pneumococcal conjugate products came into use, VSD investigators used a group sequential monitoring approach to track predefined events prospectively, checking week by week whether their rates climbed past a statistical boundary relative to a comparison group. That is a controlled analysis with a real denominator, which is precisely what a raw report count lacks.

The FDA runs a parallel effort through its BEST system, part of the wider Sentinel Initiative. BEST reaches into large insurance claims, electronic health records, and linked claims-and-record databases covering many millions of people, giving it the scale and structure to support rate comparisons. Its role is to confirm signals raised by passive reporting and to detect new ones. Passive reporting nominates a candidate; active surveillance decides whether the candidate is real.

This layering is why the genuine safety findings of recent years hold up. Rare clotting events and vaccine-associated myocarditis were identified and characterized through comparison against expected rates in defined populations, not by adding up entries in a public file. A concern earns credibility when it moves from the first tier to the second and survives that comparison.

How to read a VAERS number without being misled#

A few habits keep the figures in proportion. Treat any raw count as a list of reports, not a list of confirmed harms. Ask whether the claim compares vaccinated people to an expected background rate, and distrust any figure that skips that step. Notice whether a spike tracks a real biological signal or simply a wave of publicity. Check whether the controlled systems, the VSD or FDA BEST, have looked at the same question, because their answers, not the report tally, are what regulators act on.

None of this makes VAERS worthless. It is a sensitive front door that has repeatedly done its job. The failure is one of interpretation, and it happens whenever a number built to raise a question is presented as though it had already settled it.