Error in the delivery of radiation therapy: results of a quality assurance review.

PURPOSE To examine error rates in the delivery of radiation therapy (RT), technical factors associated with RT errors, and the influence of a quality improvement intervention on the RT error rate. METHODS AND MATERIALS We undertook a review of all RT errors that occurred at the Princess Margaret Hospital (Toronto) from January 1, 1997, to December 31, 2002. Errors were identified according to incident report forms that were completed at the time the error occurred. Error rates were calculated per patient, per treated volume (>/=1 volume per patient), and per fraction delivered. The association between tumor site and error was analyzed. Logistic regression was used to examine the association between technical factors and the risk of error. RESULTS Over the study interval, there were 555 errors among 28,136 patient treatments delivered (error rate per patient = 1.97%, 95% confidence interval [CI], 1.81-2.14%) and among 43,302 treated volumes (error rate per volume = 1.28%, 95% CI, 1.18-1.39%). The proportion of fractions with errors from July 1, 2000, to December 31, 2002, was 0.29% (95% CI, 0.27-0.32%). Patients with sarcoma or head-and-neck tumors experienced error rates significantly higher than average (5.54% and 4.58%, respectively); however, when the number of treated volumes was taken into account, the head-and-neck error rate was no longer higher than average (1.43%). The use of accessories was associated with an increased risk of error, and internal wedges were more likely to be associated with an error than external wedges (relative risk = 2.04; 95% CI, 1.11-3.77%). Eighty-seven errors (15.6%) were directly attributed to incorrect programming of the "record and verify" system. Changes to planning and treatment processes aimed at reducing errors within the head-and-neck site group produced a substantial reduction in the error rate. CONCLUSIONS Errors in the delivery of RT are uncommon and usually of little clinical significance. Patient subgroups and technical factors associated with errors can be identified. The introduction of new technology can produce new ways for errors to occur, necessitating ongoing evaluation of RT errors for quality assurance. Modifications to processes of care can produce important reductions in error rates.