Optimizing prompting strategies improves large language model classification of pain- and fatigue-related functional impact in childhood cancer survivors

Prompt engineering improves how large language models interpret survivor-reported pain and fatigue. These findings support the use of carefully designed prompts to enable automated, context-aware analysis of symptom narratives, providing a scalable approach to support symptom monitoring and survivor-centered care.