EVIDENCE QUALITY IN URBAN SEARCH AND RESCUE READINESS ASSESSMENT: REGISTRY AMBIGUITY, OPERATIONAL INFLUENCE, AND COMPLIANCE-RECORD ERROR

Abstract

Readiness judgments can be misleading when classification labels, operational endpoints, and requirement records are treated as interchangeable evidence. We assess these boundaries using 61 first INSARAG classifications in 48 countries, 43 BASARNAS offices, descriptive national indicators, and controlled simulations. Exact Brier-loss envelopes and exhaustive refitting cover four disputed labels. Under equal team weights, ridge improvement over prevalence ranges from −0.02841 to +0.00348; under equal country evaluation weights, the same predictions remain favorable across all 16 assignments. Robustness is therefore estimand-dependent. The office training coefficient is imprecise and reverses sign after removing one office. Source checks confirm that its influential annual means agree with published totals and that the endpoint measures readiness to depart, not incident arrival. A descriptive category decomposition localizes the recorded increase without identifying its cause. In 1600 paired synthetic experiments, exact gate features reduce hard-gate ridge RMSE at n = 27 from 0.2223 to 0.0688. An independent 800-experiment recording-error sweep yields RMSE 0.0511 with exact gates and 0.1526 with 10% per-gate error at n = 240. The contribution is a reproducible evidence audit with explicit interpretation boundaries, not a validated certification model or a causal estimate of training effectiveness.