Short-Term Continuous Glucose Forecasting with Large-Language-Model-Derived Nutrient Estimates from Real-World Chinese Dietary Records.
- 1Institute of Medical Information, Chinese Academy of Medical Sciences & Peking Union Medical College, Beijing, China.
Abstract
Background: Accurate short-term glucose forecasting is important for diabetes self-management, yet dietary information remains difficult to quantify and integrate into forecasting models in free-living settings, especially for complex Chinese meals. Methods: We investigated whether large language models (LLMs) can convert free-text Chinese dietary records into structured nutritional features for short-term continuous glucose forecasting. Five Chinese-capable LLMs were evaluated as nutrient estimators, and the resulting nutritional features were integrated with continuous glucose monitoring histories in a seasonal-patch transformer for glucose forecasting. Performance was assessed in 2 real-world cohorts (ShanghaiT1DM and ShanghaiT2DM) for both full-period and postprandial glucose forecasting. Results: LLM-derived nutrient estimates significantly outperformed semantic embeddings as dietary representations for glucose forecasting (P < 0.001). In the primary comparison setting, the best-performing LLM-based model achieved a postprandial 3-h root mean square error (RMSE) of 34.668 mg/dl, reducing the error by 1.967 mg/dl compared to non-LLM settings. In the full-period setting, the optimal LLM-based model achieved an RMSE of 27.035 mg/dl, reducing the error by 0.42 mg/dl relative to non-LLM settings. Across both forecasting tasks, direct nutrient estimation by DeepSeek-v3.1 achieved the best average RMSE. Forecasting gains were positively associated with nutrient-estimation quality, particularly for carbohydrate, dietary fiber, and fat. Architecture ablations further showed that patch-based encoding was the most influential component, with additional contributions from temporal decomposition and side branch design. Conclusion: These findings show that LLM-based nutrient estimation provides a practical and physiologically interpretable way to convert free-text dietary records into structured meal-related inputs for glucose forecasting.
Related
- Digital Twin Workflow for Continuous Glucose Forecasting: Real-Time Data Assimilation Using Reinforcement Learning
- Nutrient Calculation Accuracy, Expert-Rated Clinical Appropriateness and Safety Indicators, and Reproducibility of Large Language Model-Generated Diet Plans for Polyendocrine Metabolic Ovarian Syndrome: A Real-Case-Based Evaluation
- Knowledge, attitude and practice on continuous glucose monitoring and self-monitoring of blood glucose among healthcare professionals across India: A cross-sectional study
- Changes in CGM-Derived Glycemic Metrics and Triglyceride-Glucose Index During Ramadan in Adults With Type 2 Diabetes Using Real-Time Continuous Glucose Monitoring
- Continuous Glucose Monitoring for Metabolic Phenotyping and Precision Nutrition: Current Evidence, Clinical Applications and Future Perspectives
BibTeX
@article{ma2026shortterm,
title = {{Short-Term Continuous Glucose Forecasting with Large-Language-Model-Derived Nutrient Estimates from Real-World Chinese Dietary Records.}},
author = {Leyao Ma and Jie Hao and Lin Yang and Xiaowei Xu and Jiao Li},
journal = {Health data science},
year = {2026},
month = oct,
doi = {10.34133/hds.0471},
url = {https://europepmc.org/article/MED/42953316},
note = {CC BY},
}
RIS
TY - JOUR
TI - Short-Term Continuous Glucose Forecasting with Large-Language-Model-Derived Nutrient Estimates from Real-World Chinese Dietary Records.
AU - Leyao Ma
AU - Jie Hao
AU - Lin Yang
AU - Xiaowei Xu
AU - Jiao Li
PY - 2026
DA - 2026/10/28
JO - Health data science
DO - 10.34133/hds.0471
UR - https://europepmc.org/article/MED/42953316
AB - Background: Accurate short-term glucose forecasting is important for diabetes self-management, yet dietary information remains difficult to quantify and integrate into forecasting models in free-living settings, especially for complex Chinese meals. Methods: We investigated whether large language models (LLMs) can convert free-text Chinese dietary records into structured nutritional features for short-term continuous glucose forecasting. Five Chinese-capable LLMs were evaluated as nutrient estimators, and the resulting nutritional features were integrated with continuous glucose monitoring histories in a seasonal-patch transformer for glucose forecasting. Performance was assessed in 2 real-world cohorts (ShanghaiT1DM and ShanghaiT2DM) for both full-period and postprandial glucose forecasting. Results: LLM-derived nutrient estimates significantly outperformed semantic embeddings as dietary representations for glucose forecasting (P < 0.001). In the primary comparison setting, the best-performing LLM-based model achieved a postprandial 3-h root mean square error (RMSE) of 34.668 mg/dl, reducing the error by 1.967 mg/dl compared to non-LLM settings. In the full-period setting, the optimal LLM-based model achieved an RMSE of 27.035 mg/dl, reducing the error by 0.42 mg/dl relative to non-LLM settings. Across both forecasting tasks, direct nutrient estimation by DeepSeek-v3.1 achieved the best average RMSE. Forecasting gains were positively associated with nutrient-estimation quality, particularly for carbohydrate, dietary fiber, and fat. Architecture ablations further showed that patch-based encoding was the most influential component, with additional contributions from temporal decomposition and side branch design. Conclusion: These findings show that LLM-based nutrient estimation provides a practical and physiologically interpretable way to convert free-text dietary records into structured meal-related inputs for glucose forecasting.
SN - 2097-1095
SN - 2765-8783
ER -