Room: Track 3
Purpose: To improve the safety and efficiency of treatment planning and delivery by reporting, categorizing and then automating processes that lead to errors.
Methods: Incident learning has been present in our department since 2016. Errors relating to treatment planning were categorized and analyzed. The root cause of error was frequently manual data entry, often required by incompatibility of our treatment planning and management systems. Two approaches were required to counteract this: automatic check of data entry and automatic data transfer. From 2017 through 2019 error-prone processes were continuously identified, and subsequently automated.
Results: 507 treatment planning errors were reported. Changes to the department such as new planning system and reduction in planning timeline were shown to increase the error. The most common categories were “Prescription mismatch”, “Plan change”, “Setup Images” and “Field Parameters or Labeling”. A web-based plan check tool was developed to check 166 items, and reduced chart errors from 100 to 15 events per year. Automated tools for plan report generation, SBRT coordinate calculation and prescription import were developed. Automated prescription import required assistance from the vendors and reduced errors from 70 to 8 per year. After event reporting dropped to <5 per month in summer 2019, reporting was reemphasized, and reports increased to >15 per month. This enabled further planning errors to be identified and a pre-check API-script for dosimetrists has been developed to counter these.
Conclusion: The synergy of incident learning and automation effectively improved treatment planning quality and efficiency. System interoperability appears to be a gap leading to human failure and is a barrier to automation. Clinical operation is a continuous optimization process and encouragement of event reporting is the key to finding new areas of improvement to improve quality and safety.