Accounting for Insulin-on-board in Large Language Model Advice about Pediatric Hyperglycemia
- 1Boston Children's Hospital
Abstract
Large language models (LLMs) can be used by patients and caregivers to answer medical questions and the characterization of safety concerns in LLM outputs for patient-facing applications is an active area of research. In this work, we study LLM responses to caregiver questions about hyperglycemia in pediatric type 1 diabetes. We measure the frequency of a potential safety failure: advice to dose insulin in the absence of a clear warning to account for insulin-on-board. We found occurrences of the potential safety failure across four LLMs tested through programmatic endpoints and two LLMs tested through chat interfaces. The rates of the potential safety failure varied when different context was given about the patient's meal and insulin history and treatment regimen. These results highlight a need for guardrails for patients with diabetes who use LLMs to manage their care.
Related
- Navigating digital surgical advice: evaluating the reliability of large language models for pediatric phimosis
- Disruption of Anticipatory Insulin Secretion in Patients with Hyperglycemia and Obesity
- Sliding-scale Insulin Does Not Control Steroid-induced Hyperglycemia in a Non-diabetic Patient, No Matter How You Tune It
- Glargine-to-intravenous insulin dose ratio, rebound hyperglycemia and hypoglycemia after discontinuation of insulin infusion in critically ill adults: a retrospective cohort study
- Comparative Discriminative Performance of the HbA1c/HDL-C Ratio, Stress Hyperglycemia Ratio, and ZJU Index for HOMA-IR-Defined Insulin Resistance in Patients With Type 2 Diabetes.
BibTeX
@misc{healey2026accounting,
title = {{Accounting for Insulin-on-board in Large Language Model Advice about Pediatric Hyperglycemia}},
author = {E. Healey and S. Lee and C. M. Astley and K. D. Mandl},
howpublished = {medRxiv},
year = {2026},
month = nov,
doi = {10.64898/2026.11.06.26367316},
url = {https://www.medrxiv.org/content/10.64898/2026.11.06.26367316v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - Accounting for Insulin-on-board in Large Language Model Advice about Pediatric Hyperglycemia
AU - E. Healey
AU - S. Lee
AU - C. M. Astley
AU - K. D. Mandl
PY - 2026
DA - 2026/11/06
PB - medRxiv
DO - 10.64898/2026.11.06.26367316
UR - https://www.medrxiv.org/content/10.64898/2026.11.06.26367316v1
AB - Large language models (LLMs) can be used by patients and caregivers to answer medical questions and the characterization of safety concerns in LLM outputs for patient-facing applications is an active area of research. In this work, we study LLM responses to caregiver questions about hyperglycemia in pediatric type 1 diabetes. We measure the frequency of a potential safety failure: advice to dose insulin in the absence of a clear warning to account for insulin-on-board. We found occurrences of the potential safety failure across four LLMs tested through programmatic endpoints and two LLMs tested through chat interfaces. The rates of the potential safety failure varied when different context was given about the patient's meal and insulin history and treatment regimen. These results highlight a need for guardrails for patients with diabetes who use LLMs to manage their care.
ER -