The shape of severe workplace injuries: A topological analysis of ten years of OSHA severe-injury reports

Severe workplace injuries result from physical energy-transfer processes, yet surveillance systems reduce these events to discrete classification codes that may be assigned inconsistently. This study examines the structure revealed by the U.S. severe-injury record without imposing predefined categories and evaluates its implications for prevention. Using topological data analysis, we map 103,750 OSHA severe-injury reports from January 2015 through August 2025 from two complementary perspectives. Injury narratives form a continuous mechanism-outcome axis linking a machinery-and-amputation regime with a fall-and-fracture regime, connected by an interpretable hand-and-arm transition zone. The statistically significant, monotonic axis combines distinct event codes that share mechanism signatures, providing a consistency check for identifying records whose assigned codes differ from those dominant among similar cases. The axis reflects injury mechanism rather than harm magnitude: amputation and hospitalization rates vary smoothly along it (row-standardized Moran’s I = 0.86 and 0.84) but in opposite directions. Industry-by-state risk profiles are organized primarily by industry rather than geography, distinguishing amputation-dominant manufacturing from hospitalization-dominant service regimes across states. These patterns persist under federal-only reporting and replicate across the full dataset and all analytic choices tested. The findings provide a reproducible method for auditing narrative-code consistency, identify the machinery transition population as a hypothesis-generating target for machine-guarding attention, and support pooling sparse state-level data by industry.