What harms are most missed by passive systems like VAERS?

The short version: passive systems like VAERS miss harms that are delayed, rare, chronic, or require a denominator to detect, and the evidence shows exactly which categories fall through.

The 2026 HPV VAERS study found that 90% of reported adverse events occurred within 30 days of vaccination [1]. That means anything taking longer to appear—autoimmune conditions, neurological decline, cancer—is structurally undercounted. The same study found POTS had the strongest signal strength but only 680 reports [1]; a rare condition with a plausible biological link can be flagged by VAERS, but the system cannot tell you whether that is 680 out of 680 real cases or 680 out of 68,000. Without a denominator of total doses administered, you cannot calculate incidence.

The 2006 Pediatrics analysis showed that by 2002, litigation-related reports made up the majority of VAERS reports for overdose, neuropathy, and thimerosal, and nearly half of mental retardation reports [3]. That means the database is shaped by who is suing, not by who is injured. Harms that are not being litigated—low-grade chronic inflammation, metabolic disruption, cognitive decline in children without a formal diagnosis—are invisible.

The Alberta HPV study, which linked vaccination records to hospital and ED data, found that among 195,270 vaccine recipients, 958 were hospitalized within 42 days but only 4 of those had a reported adverse event [6]. That is a 0.4% capture rate for hospitalizations. The system missed 99.6% of events serious enough to require admission.

My call: passive systems preferentially miss delayed harms, rare conditions without a denominator, and any injury that does not generate a lawsuit or an acute medical visit. The Alberta data suggests the capture rate for serious events requiring hospitalization is below 1%. Confidence: moderate, because the Alberta study is the only direct linkage in the retrieved evidence, and it covers only one vaccine in one province.