Atmospheric-Hazard Enclosed-Space Accidents on Ships: A Barrier-Based Analysis of Accident Investigation Reports

Despite successive regulatory reforms, entry into enclosed spaces remains one of the deadliest operations in ship management. The IMO GISIS database was screened, and every case involving actual entry, atmospheric exposure and a death or injury was retained, yielding 39 official reports published between 2007 and 2025 by the authorities of 15 flag states. Each case was coded on seven barriers (B1–B7) defined from IMO A.1050(27) and ISGOTT on a three-level ordinal scale (absent/inadequate/adequate) adapted from Kjellén’s (2023) barrier index; the barrier concept rests on Hollnagel, and the joint failure of the barriers is interpreted through Reason’s defences-in-depth model. The cases resulted in 53 deaths and 33 injuries (86 victims), of which 64.1% were multiple-victim events. Oxygen deficiency was present in 79.5% of cases and associated with 83.0% of deaths. The preventive barriers were systematically weak: permit-to-work (B2) was absent in 89.7% of cases and respiratory protection (B5) in 87.2%; all six failed in 30.8%, and the mean score was 3.00 out of 14. Emergency response (B7), by contrast, never collapsed entirely and was uncorrelated with prevention quality, whereas the preventive barriers formed a coherent scale (Cronbach α = 0.72). A rescuer-victim pattern was observed in half of the determinable cases and was the principal driver of multiple fatalities (U = 55.5; p < 0.001). Neither severity nor barrier performance improved over time. Adopting the revised recommendations is necessary but not sufficient; what is decisive is applying the rules to every entry.