tooluniverse-adverse-event-detection

Provide the generic drug name and the skill queries FAERS for top adverse events, then calculates PRR, ROR, and IC with 95% confidence intervals for each event. A signal is detected when PRR is at least 2.0, the lower confidence interval exceeds 1.0, and there are at least 3 cases.

Disproportionality analysis compares the reporting rate of an adverse event for a specific drug against all other drugs in FAERS using metrics like PRR, ROR, and IC. It identifies events reported more often than expected, flagging potential safety signals for investigation.

Yes, the skill supports head-to-head comparison using FAERS_compare_drugs, which returns PRR, ROR, and IC side-by-side for both drugs on a given adverse event. It can also aggregate adverse events across an entire drug class to distinguish drug-specific from class-wide signals.

The Safety Signal Score is a 0-100 metric combining FAERS signal strength (0-35 points), serious adverse events (0-30), FDA label warnings (0-25), and literature evidence (0-10). Scores of 75-100 indicate high concern, 50-74 moderate concern, and below 25 low concern.

If no FAERS reports exist, the skill skips FAERS phases and relies on FDA labels, mechanism-based predictions, and literature. A NOT_FOUND error from FDA label tools is normal and simply means that section, such as a boxed warning, does not exist for the drug.

FAERS data is subject to reporting biases and confounding by indication, since patients taking a drug often have the underlying disease. The skill mitigates this through drug class comparison, demographic stratification, and Naranjo causality reasoning, but signals indicate association rather than proven causation.