Multilevel variation in clinical trial enrollment in community oncology settings: A real-world EHR-derived data analysis.

123 Background: Clinical trials are essential for advancing cancer care, yet enrollment rates remain low. While patient characteristics are known to influence participation, the impact of physician- and practice-level factors is less understood, particularly in community oncology settings where most cancer patients receive care. This study examined enrollment trends and quantified the influence of physician- and practice-level variation in trial participation across community practices. Methods: We retrospectively analyzed data from the Flatiron Health Research Database from 205,743 adult patients with metastatic/advanced cancers (melanoma, head/neck, breast, thoracic, GI, GU, or GYN cancers) treated by 1,664 unique physicians across 107 trial-available community practices (2010–2024). Trial participation was defined as receipt of clinical study, with each patient attributed to an oncologist using a validated visit-based algorithm. Mixed-effects logistic regression with random intercepts for physicians and practices assessed enrollment variation, adjusting for patient demographics and clinical characteristics. Variance decomposition quantified the proportion of explainable variability attributed to patient-, physician-, and practice-level factors. Stratified analyses were performed across cancer groups and diagnosis periods (2010–2014, 2015–2019, 2020–2024). Results: Of 205,743 patients, 10,869 (5.3%) were enrolled in trials; annual participation rate rose from 4.0% (2011) to 6.9% (2024). Enrollment varied widely among physicians (mean: 6.53; median [IQR]: 2[0–5]), with 72.3% enrolling fewer than 5 patient. Notably, 80% of enrollments concentrated among just 20% of physicians. Practice-level enrollment rates ranged from 0.1% to 59.1%, with nearly half enrolling < 20 patients. Patient characteristics explained less variability (11.4%) compared to physician- (12.3%) and practice-level (14.9%) factors. Specifically, practice-level variation rose from 12.9% in 2010–2014 to 22.1% in 2020–2024, while physician-level variation rose from 12.1% to 14.6% over the same period. Variation attributable to patient characteristics was higher in breast and head/neck cancer groups, while other groups were more subjected to physician or practice-level influence (Table). Conclusions: Despite modest increase in trial enrollment in community oncology, overall participation remains suboptimal with growing variation across physicians and practices. These findings highlight the need for targeted interventions to reduce inequities by enhancing provider referral behaviors and increasing trial capacity in community oncology. Variance Decomposition (%) Patient Physician Practice Breast 14.82 8.77 14.49 GI 14.19 14.66 16.94 GU 11.79 15.89 25.49 Melanoma 11.52 14.8 18.56 Head & Neck 17.8 15.61 11.35 GYN 9.56 14.46 17.45 Thoracic 11.62 12.82 16.79