City of Hope Uses Oncology AI Model to Cut Clinician Workload and Speed Patient Matching
City of Hope has deployed Hope LLM, a large language model built specifically for oncology, to handle tasks that historically consumed hours of clinician time. The system synthesizes fragmented patient records, reduces after-hours administrative work, and matches patients to clinical trials in real time.
Nasim Eftekhari, Chief AI and Analytics Officer at City of Hope, said the model's core value lies in its ability to reason across unstructured clinical data. Cancer patients often arrive with decades of medical records: thousands of pages of PDFs, handwritten notes, radiology reports, and fragmented treatment histories spread across multiple institutions.
Hope LLM now creates patient summaries in minutes-work that previously required hours of manual review. Clinicians can quickly understand prior treatments, adverse events, disease progression, and key decision points in a patient's journey.
Reducing unpaid administrative work
One measurable outcome: reduced "pajama time," the personal hours clinicians spend preparing for the next day instead of with family.
Eftekhari said this matters because AI adoption in healthcare isn't driven by novelty. It's driven by relief from real operational pain. When clinicians see immediate value, they use the tool.
Clinical trial matching at the point of care
Hope LLM continuously evaluates hundreds of active trials against patient-specific data and inclusion-exclusion criteria. When a doctor sees a patient, the system displays matching trials with justification for each recommendation.
This shifts clinical trials from a separate administrative process into integrated care delivery. It improves enrollment rates and patient access to new therapies.
City of Hope also runs the matching in reverse: Hope LLM identifies potential patient populations before a trial opens. This lets the organization assess enrollment feasibility and select sites based on real data, reducing the risk of under-enrollment.
Eftekhari said many trials fail to recruit enough patients and close. "We are trying to close that gap."
Governance as the limiting factor
As clinician demand for AI tools increased, Eftekhari's role shifted. "I've gone from promoting the use of AI to now trying to put governance in place," she said.
Large language models still hallucinate. Over-reliance on a system that sounds confident but produces false information creates clinical risk.
City of Hope established a multi-layered governance model with specific roles and responsibilities:
- Executive governance committee for strategy and high-risk decisions
- Multidisciplinary AI work group reviewing every use case
- Members include clinical experts, legal, compliance, ethics, IT, security, and HR
The organization designed governance to move at the speed of deployment. "We can review and approve or reject use cases in a matter of days," Eftekhari said.
Eftekhari emphasized that governance and AI education go hand-in-hand. Without both, organizations risk deploying systems that clinicians don't fully understand.
For executives building AI strategies, the City of Hope model demonstrates how domain-specific models, clear governance structures, and clinician-focused use cases create adoption without sacrificing safety. The shift from push to pull-where clinicians request AI tools rather than resist them-signals genuine operational value.
AI for Executives & Strategy and AI for Healthcare resources cover governance frameworks and enterprise deployment patterns relevant to this type of implementation.
Your membership also unlocks:
