The 90% syndrome in construction safety: distributional concentration of construction accidents in the final project phase

Abstract

This study examines the distribution of reported construction accidents across project lifecycle stages, moving beyond micro-level causation toward phase-dependent concentration patterns. A dataset of 33,963 accident records from the Korean Construction Safety Management Integrated Information (CSI) system was analyzed using chi-square goodness-of-fit, odds ratio, and Cochran-Mantel-Haenszel stratified analyses, alongside interviews with eight senior practitioners within an explanatory sequential mixed-methods design. Reported accident records were not uniformly distributed across project progress intervals (χ2 = 1,143.91, df = 9, p < 0.001). The ≥90% interval accounted for 14.7% of reported accidents (OR = 1.552, 95% CI: 1.482–1.626) and 16.6% of fatalities (OR = 1.793, 95% CI: 1.441–2.230). Associations between institutional safety mechanism absence and late-phase concentration persisted after controlling for client type (Safety Management Plan absence: MH-OR = 1.999; DfS absence: MH-OR = 1.295). Interviews identified schedule compression, multi-trade spatial convergence, and accumulated fatigue as key field-level conditions. Findings inform recommendations across three stakeholder levels: regulatory bodies (formalizing the ≥90% phase as a heightened oversight period and extending DfS requirements to private-sector projects), project owners (contract design for schedule contingency), and contractors (enhanced Permit-to-Work and logistics coordination in the final phase). This study offers a large-scale empirical analysis of non-uniform accident concentration across construction lifecycle stages using national administrative data, extends the “90% Syndrome” into the construction safety domain, and introduces a bidirectional safety-schedule feedback model as a construction-specific theoretical contribution.