Leyao Ma

The etiology of pediatric costochondritis, a common cause of musculoskeletal chest pain, remains poorly understood. This study aimed to identify nutritiona

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.

Objective: Gestational diabetes mellitus (GDM) is one of the most common pregnancy complications. Electronic health records (EHRs) promise GDM risk prediction, but missing data poses a challenge to developing reliable and generalizable risk prediction models. This study aims to address the problem of missing EHR data in GDM prediction before 12 weeks gestation. Methods: A total of 5066 women with singleton pregnancies, aged 18 to 50, were included in this retrospective study. This study evaluated 6 imputation methods, combined with 4 classification machine learning models. The evaluation encompassed downstream predictive performance, robustness to variable missingness, ability to restore original data distribution, and influence on feature selection based on 10-fold cross-validation. Results: Our findings revealed a significant improvement in model performance with imputation. When using the top 30 features, logistic regression (LR) with multivariate imputation by chained equations using classification and regression trees (mice) achieved the highest area under the receiver operating characteristic curve of 0.6899, compared to 0.6336 for the LR model without imputation. Mice also led to the best average performance across prediction models and yielded the most accurate restoration of the original data distribution. LR models trained on data imputed by mice remained the most robust across varying levels of missingness. The classification algorithm primarily accounted for differences in predictive performance. In addition, we identified 18 key features for early GDM prediction in the Chinese population. Conclusion: This study demonstrates the critical role of imputation in improving the performance and fairness of GDM prediction models. The findings provide practical guidance for integrating imputation into clinical machine learning pipelines.