Development and Assessment of Artificial Intelligence (AI)-Enhanced Pre-treatment Peer-review Process to Improve Patient Safety in Radiation Oncology

PROJECT SUMMARY Objectives: To develop and integrate an innovative artificial intelligence (AI) and machine learning (ML) based intervention focused on physician variability (Aim #1) to be used in the pre- treatment peer-review process to improve RT providers’ performance during these tasks. To implement and evaluate the impact of our integrated intervention on patient safety in the real clinical environment (Aim #2). Rationale: Radiation therapy (RT) plays an important role in the curative and palliative management of many cancers; ≈ 50% of people with cancer in the US receive RT (≈ 600,000 annually). While many advances over the last two decades have improved patient safety (e.g., image guidance, conformal beam shaping), the increasing complexity (e.g., intensity-modulated Presently, up to ≈10% of patients receiving RT are affected by treatment planning errors, with more harm present in minority and underserved populations, which is unacceptable. Using cluster randomized stepped wedge design, we will evaluate RT cases in the baseline period, and after our peer-review processes are adjusted, overtime to, + the providers’ variability data. RT) has created new error pathways. Methods: ‘Optimal’ algorithms and visual summaries of these data will be determined in Aim #1. Patient safety will be quantified via the number/severity of clinically relevant errors not detected during pre-treatment peer-review and detected by downstream QA processes; expressed as the rate per 1000 RT fractions delivered. We will use advanced regression models to evaluate patient safety over time. The significance level will be set at 0.05, two-tailed.