Purpose: Automated treatment planning may reduce the occurrence of errors but may also reduce the detectability of these errors before plan approval, due to reduced human interaction. Thus, plan checks should be designed to catch failure modes unique to auto-generated plans.
Methods: A custom checklist was developed using guidance from AAPM TGs 275 and 315 and the results of an FMEA of the Radiation Planning Assistant, an automated treatment planning tool. In the first study, 8 physicists were recruited to review 10 automatically generated plans, 5 of which contained errors (incorrect isocenter identification, wrong prescription, wrong laterality, incorrect target contour, and incorrect field boundary). 4 head and neck VMAT plans, 3 cervix 4-field-box plans, and 3 chest wall plans were reviewed. Physicists performed plan checks, recorded errors, and evaluated for clinical acceptability without the use of our checklist. Two weeks later, the physicists reviewed 10 additional plans using our initial checklist that included 90 distinct items. The checklist was then modified based on a usability survey. The revised checklist was assessed by 14 physics residents who completed 10 plan checks each, 5 of which contained errors. Eight and six participants completed their reviews without and with the checklist respectively.
Results: For plan reviews performed by physicists without the checklist, 67.5% of errors were detected; with the initial checklist, 87.5% errors were detected. Modifications were then made to the checklist to address concerns identified, limiting checks to only 18 items that are of additional concern for automated plans. For plan reviews performed by residents without the checklist, 52.5% of errors were detected; with the revised checklist, 70% of errors were detected.
Conclusion: Our results indicate that the use of a customized checklist when reviewing automated treatment plans will result in a higher rate of error detection and thus improved patient safety.
Funding Support, Disclosures, and Conflict of Interest: This work was partially funded by Varian Medical Systems