AI Helps Identify Childhood Cancer Survivors' Health Needs

St. Jude study finds sophisticated AI prompting strategies outperform simple approaches in detecting symptom severity and impact

Mar. 28, 2026 at 2:54am

Got story updates? Submit your updates here. ›

Artificial intelligence (AI) could help physicians determine if survivors of childhood cancer need extra support, according to a new study from St. Jude Children's Research Hospital. The researchers found that more complex AI prompting strategies, which provide additional information to the language models, performed significantly better than simpler approaches in analyzing interview transcripts to detect symptoms causing severe disruptions in survivors' daily lives.

Why it matters

Identifying which childhood cancer survivors have symptoms severe enough to need targeted support is difficult for physicians, as much of the relevant data exists in hard-to-review conversational transcripts. This study demonstrates how AI could unlock that underutilized data to assist physician decision-making and improve care for this growing population of survivors.

The details

The researchers interviewed 30 childhood cancer survivors aged 8-17 and their caregivers, then had human experts analyze the transcripts to categorize symptoms by severity and physical, cognitive or social impact. They then gave the same transcripts to two large language models, ChatGPT and Llama, using four prompting strategies - two simple approaches (zero-shot and few-shot) and two complex ones (chain-of-thought and generated knowledge). The complex prompting methods performed significantly better, distinguishing physical/cognitive impacts well and having moderate ability to detect social impacts.

  • The study was published on March 26, 2026.

The players

I-Chan Huang

Corresponding author, Department of Epidemiology & Cancer Control at St. Jude Children's Research Hospital.

St. Jude Children's Research Hospital

A leading pediatric treatment and research facility located in Memphis, Tennessee.

ChatGPT

A large language model developed by OpenAI.

Llama

A large language model developed by Meta AI.

Got photos? Submit your photos here. ›

What they’re saying

“About 40%-60% of a clinical encounter is a patient talking to their physician about symptoms and related health experiences. We have provided a proof of concept that large language models could help analyze that underutilized conversational data to detect symptom severity and its functional impact and assist physician decision-making to provide better care to survivors.”

— I-Chan Huang, Corresponding author, Department of Epidemiology & Cancer Control at St. Jude Children's Research Hospital

What’s next

The researchers say much more testing will be required before clinical use, but the findings provide an early example of how AI could improve survivorship care by making it easier to capture and analyze complex symptom information from patient-physician conversations.

The takeaway

This study demonstrates the potential for sophisticated AI prompting strategies to unlock valuable data from patient interviews and assist physicians in identifying childhood cancer survivors who need additional targeted support, improving care for this vulnerable population.