PODCAST | Rethinking AI Success in Oncology: Lessons from City of Hope
April 8, 2026·9 min
The episode opens with host Eric Pupo of GuideHouse introducing Nassim Eftikari, who leads the Department of Applied AI and Data Science at City of Hope, an NCI-designated national academic cancer center with 40-plus locations across the US. Eftikari describes City of Hope's research infrastructure, including the Beckman Research Institute and TGen genomics testing company, and explains why this breadth makes it a strong environment for AI beyond operations and patient care. On the question of what metrics matter as Chief AI Officer, Eftikari places performance metrics above ROI, stating that the first question for any oncology AI tool is whether it actually works for their specific patient population. She notes that FDA approval or documented success elsewhere does not transfer automatically, making local clinical validation a non-negotiable prerequisite. On where AI is generating tangible return, Eftikari identifies two areas: AI-assisted diagnostics in radiology and pathology, with many tools FDA-approved and embedded in clinical devices; and documentation tools such as ambient transcription, patient chart summarization, and clinical trial matching that replace manual review of large volumes of free text. She extends this to research, noting that real-world evidence extraction for expediting oncology research is already happening, not a future-state ambition. On prioritization, Eftikari describes a six-dimensional framework developed over nine years at City of Hope that covers patient care impact, patient experience, provider experience, direct financial impact, indirect financial impact including capacity and downstream revenue, and depth-versus-breadth of patient impact. She explains that the framework applies consistently across internally built and externally sourced tools. The episode closes with Eftikari noting that the framework's depth dimension allows a small-reach tool affecting life-and-death decisions to rank above a broad-reach tool with marginal per-patient benefit.
