BACKGROUND. A widely reported ABO-mismatch accident in March of 2003 raised concerns about the reliability of the transplantation system. Because this type of failure is rare and significant, we performed a probabilistic risk assessment (PRA) of the donor-recipient matching processes for thoracic organ transplantation. METHODS. A probabilistic risk assessment was performed. RESULTS. The likelihood of accidental incompatible implantation was already low in 2003. The PRA model indicates that the likelihood of such an event was 1.38×10 per donated organ. This estimate correlates closely with the observed rate of these accidents. Based on this model, process changes put in place shortly after the accident reduced the probability to ∼3.08×10 and changes put in place in October 2004 further reduced the probability to ∼2.22×10 per organ donated. CONCLUSIONS. The observed and predicted likelihoods of accidental incompatible thoracic organ transplantation are comparable. These likelihoods are several orders of magnitude smaller than other hazards associated with solid organ transplantation. The PRA model indicates that changes that followed the March 2003 accident further reduced the likelihood of accidental incompatible implantation by roughly two orders of magnitude. Quantitative estimates from PRA can be used to assess risks in healthcare and to gauge the impact of system changes on these risks. © 2007 Lippincott Williams & Wilkins, Inc.