AI Identifies Severe Symptoms in Childhood Survivors
St. Jude Children's Research Hospital researchers report in Communications Medicine that large language models (ChatGPT and Llama) can analyze interviews from 30 childhood cancer survivors (ages 8–17) and caregivers, producing over 800 analyzable items to detect symptom severity. The study found complex prompts—chain-of-thought and generated-knowledge—outperformed zero- and few-shot prompts, suggesting richer prompting may aid clinical workflows.
Key Points
- 1Shows LLMs (ChatGPT, Llama) detect symptom severity from 30 interviews and 800+ data items
- 2Finds complex prompts (chain-of-thought, generated knowledge) substantially improve accuracy versus zero/few-shot prompting
- 3Suggests using sophisticated prompting to extract conversational symptom data for clinical decision support
Scoring Rationale
Peer-reviewed, practice-relevant demonstration boosting prompt strategies, but small sample and single-center data limit generalizability and adoption.
Sources
Public references used for this report.
Practice with real Health & Insurance data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Health & Insurance problems
