Failure Mode and Effects Analysis for Treatment Workflow of a Novel Ring Gantry Linac

Abstract

A new ring gantry Linac machine combined with a PET ring is available in the market. It has 3 separate centers, one for laser, one for CT imaging and a 3rd for the Linac. As a first-generation machine of its kind, the workflow heavily depends on user-input. As example, it uses two separate coordinate systems, IEC for lasers and CT localization offsets and DICOM for treatment planning. The planner must manually convert between the two systems. We hypothesize that the unique design of the machine that is heavily dependent on users' input increases the potential of failure of treatment. The present work investigates failure modes for treatment delivery using the methodology of failure modes and effects analysis (FMEA) and proposes solutions to mitigate some of the failure modes (FMs). A group of two radiation oncologists, two radiation therapists and three medical physicists was assembled. The process map for treatment delivery on the X1 was created and FMs were identified. Members independently graded each FM on 3 parameters, likelihood of occurrence, detectability of FM and level of severity on patient treatment. A grading scale of 1-5 was used with five representing the worst outcome in each parameter. Each member also identified the origin of each FM to be human, machine or clinical process. Mitigation solutions were proposed. The process map of treatment delivery on X1 consists of six major processes and 24 sub-processes. A total of 27 FMs were identified, with many 19/27 (70%) caused by human errors and 7/27 (27%) caused by machine. From all responses, we tallied a median of 11 FMs (40%) that have both S≥ 4 and O or D ≥ 4. To further focus our analysis, we looked at the highest PRN scores from each member and found 7 FMs that were common. 1 FM was in "initiation of Treatment" sub-process and 6 FMs in "Treatment delivery" sub-process. Proposed solutions to these FMs were concerned with software upgrades. Examples are, allow changes in fractionation, allow dose tracking, auto calculation of couch position for various sub processes, allow DICOM image transfer and many more. The current clinical workflow that we adopted for these FMs either puts the burden on the user to confirm/verify parameters or the use of third-party software. In addition to these high-scoring FMs, we adopted changes in clinical workflow to mitigate other FMs, most of them through third-party software. This study confirmed that the unique design and user-dependent machine environment, human failures are high accounting for 70% of FMs in treatment delivery alone. Our current workflow of the machine depends on direct user input to calculate/confirm certain parameters or the use of third-party software, which also depends on the user for proper completion of the task. Suggested solutions also included proposed improvement to the machine's software and user interface.