Per-code design-input scores, audit materials, and analysis code for injury classifications (ICD-10-CM Chapter 20, ICD-11 Chapter 23 and extension chapter, OIICS 2.01/3.03, ESAW, Korean injury-type)

Supporting information for a manuscript under review. The archive contains per-code design-input field scores for 32,622 version-specific classification records across seven components: ICD-10-CM Chapter 20 (7,461 codes), the ICD-11 Chapter 23 pre-coordinated stem titles (909) and the ICD-11 extension chapter used for postcoordination (17,159), the US Occupational Injury and Illness Classification System versions 3.03 (3,231) and 2.01 (3,035), the five ESAW case variables (383), and the Korean injury-type classification (444). Every label was scored against four minimum design-input fields derived from published protective-clothing performance standards - hazard agent identity, exposure pathway, body region and exposure magnitude - under a unified three-reading scheme (primary agent-identity, permissive hazard-cue, liberal agent-axis bound), with all-published-code denominators, and with inhalation or ingestion recorded under a routing flag outside the four-field total. Exposure magnitude is reported as an adjudicated result rather than an assumption: all 115 labels the lexicon flagged in that field were adjudicated against a published rule, and none expresses a delivered quantity. The archive also contains the audit materials - the seeded false-negative audit (seed 20260807), the four-expert content review of the provisional framework (I-CVI/S-CVI), and the blinded double coding (seed 20260808) with unified kappa confidence intervals - the Korean and English scoring lexicons, and the deterministic Python pipeline. run_himj.py in Scripts_analysis_code regenerates every reported value from a clean extraction of this archive and checks 22 headline values against the values the manuscript prints; MANIFEST.txt describes each file. Version history: v2 superseded the v1 values computed under non-residual denominators. v3 added ICD-10-CM Chapter 20 and the ICD-11 extension chapter, taking the corpus to 32,622 records. v4 adds the corpus-wide exposure-magnitude adjudication table, aligns the primary-reading exclusion list across the fatality-weighting scripts, completes the body-region adjudication counts, and adds four modules the package was missing, so every shipped script runs from a clean extraction.