Where climate hits hardest: Geographic patterns in climate education across 266 U.S. oncology fellowship programs.

Abstract

e23155 Background: Climate change disproportionately affects cancer patients through care disruptions and treatment delays with 65% of US cancer patients living in climate-vulnerable counties and documenting worse survival following disasters. Oncology trainees will practice for decades in a climate-altered reality, yet it remains unknown whether US training programs are positioned to prepare fellows for climate-resilient care. We mapped fellowship programs against climate hazard and social vulnerability indices to identify where climate curricula are most urgently needed. Methods: We identified all ACGME-accredited hematology-oncology and radiation oncology programs (January 2025), linking locations to the CDC Social Vulnerability Index (SVI) and FEMA National Risk Index for heat, flooding, wildfire, and hurricane hazards. High vulnerability was defined as top-quartile scores. Double vulnerability was defined as high SVI plus high climate risk. We performed chi-square tests and Random Forest predictive modeling. Results: We identified 266 programs (188 hematology-oncology, 78 radiation oncology) across 46 states. 72 programs (27.1%) were in high-SVI states and 158 (59.4%) in high climate-risk states. 50 programs (18.8%) faced double vulnerability with marked geographic clustering (coefficient = 0.70, p < 0.001), located in seven states (AZ, CA, FL, LA, NM, OK, SC) that collectively diagnose over 438,000 new cancer cases annually. Regional variation was highly significant (χ² = 120.09, p < 0.0001). Predictive modeling achieved near-perfect discrimination (ROC-AUC≥0.998, Table 1). SVI correlated strongly with wildfire risk (r = 0.73, p < 0.0001) but not hurricane risk (r = 0.10, p = 0.11). Conclusions: Nearly 1 in 5 US oncology programs (18.8%) face both high social vulnerability and climate hazards, serving populations where disaster exposure is associated with worse cancer survival. Despite the critical need, only 1 of 9 existing climate education initiatives is located in a hotspot region, leaving most programs without targeted curricula. We propose targeted implementation of our proposed HEAT-ONC model (Hazards, Effects, Adaptation, Teaching) with regional customization. This evidence-based approach addresses a critical training gap where vulnerability is highest. Regional analysis of double vulnerability: climate risk, social vulnerability, and predictive modeling. Region Programs (n) High SVI n (%) Climate risk n (%) Double Vulnerability n (%) Test Statistic P-value West 39 5 (12.8) 35 (89.7) 30 (76.9) Clustering coef=0.70 <0.001 South 86 8 (9.3) 41 (47.7) 20 (23.3) χ²=16.44 0.0009 Northeast 78 0 (0.0) 77 (98.7) 0 (0.0) χ²=4.64 0.031 Midwest 63 0 (0.0) 5 (7.9) 0 (0.0) χ²=120.09 <0.0001 Overall 266 13 (4.9) 158 (59.4) 50 (18.8) Predictive model: SVI (28.7%),Wildfire(21.3%),Region (13.2%)ROC-AUC≥0.998