Knapp et al. 2004 — EU verbal frequency labels cause dramatic overestimation of medical risk

Claim: Knapp, Raynor & Berry (2004, Quality & Safety in Health Care 13(3):176-180) found a side effect labeled "rare" (true ~0.01-0.1%) was estimated at ~18% by the verbal group vs ~2.1% by the numeric group; "common" constipation was estimated at 34.2% (verbal) vs 8.1% (numeric).

Büchter et al. (2014, BMC Medical Informatics & Decision Making 14:76, meta-analysis of 10 RCTs) found verbal descriptors led to systematic overestimation of absolute risk (range of means 3%-54%) vs smaller overestimation for numbers (2%-20%). Bal et al. (2021, JGIM, systematic review): average interpretations of "rare" ranged 7-21% and "common" ranged 34-71% across studies.

Source: Knapp et al. 2004; Büchter et al. 2014 meta-analysis; Bal et al. 2021 systematic review.

Confidence: Verified — peer-reviewed + meta-analysed + systematic-reviewed.

Why this matters for Candid: Independent corroboration of Budescu et al. 2009/2014 — IPCC verbal terms interpreted REGRESSIVELY; numeric pairing fixes from a different domain. The convergence (climate + medicine) is unusually strong. Anchor every widget tier word.