Digital-twin models compared for reliable short-term blood sugar risk prediction
Reliability-Aware Short-Term Glycemic Risk Prediction Using an Anchor-Based Comparative Framework of Sequence, Multimodal, and State-Updating Digital-Twin Models
Machine LearningArtificial Intelligence
Summary
Predicting low or high blood sugar shortly ahead is important for people with type 1 diabetes to avoid health problems. The authors developed a way to fairly compare different computer models that use daily blood sugar data and other personal info. They found no big differences in accuracy but saw that each model was better at certain reliability aspects like confidence and explanations. This shows it’s important to check these models carefully before using them in medical settings.
What this means in practice
- •For diabetes device developers: Build more trustworthy glucose prediction features that include evaluation of confidence and explanation robustness for type 1 diabetes patients.$Commercial implications: Enables creation of improved glucose monitors offering reliable risk alerts, enhancing device marketability and patient trust.
- •For hospital diabetes care teams: Use calibrated prediction models to identify patients at short-term risk of hypo- or hyperglycemia and tailor interventions more safely.
Tested on one dataset.
Authors
Pir Bakhsh Khokhar, Carmine Gravino, Fabio Palomba
Abstract
Anticipating short-term hypoglycemic and hyperglycemic risk is essential for proactive diabetes management. However, existing machine learning approaches often use inconsistent evaluation protocols and emphasize predictive performance over reliability. This study introduces an anchor-based benchmarking framework integrating longitudinal glucose sequences and multimodal patient information. Using the T1DiabetesGranada dataset, 5,503,441 standardized prediction anchors were constructed from 736 individuals with type 1 diabetes to evaluate hypoglycemia and hyperglycemia prediction at 30-, 60-, and 120-minute horizons. Three neural architectures, a Hybrid LSTM-Transformer, a Multimodal Digital Twin, and a State-Updating Digital Twin with gap-aware longitudinal state progression, were compared under identical patient-level validation protocols. Beyond predictive performance, the evaluation examined calibration, classification thresholds, explanation reliability, cross-model disagreement, and robustness to controlled perturbations. No statistically significant predictive differences were observed after Holm correction, although the models exhibited complementary reliability characteristics. The Multimodal Digital Twin achieved the lowest mean Brier score, while the State-Updating Digital Twin demonstrated the lowest mean expected calibration error (ECE-10). The Hybrid LSTM-Transformer showed strong explanation faithfulness and stability under its temporal attribution protocol. These findings highlight the importance of evaluating digital-twin models beyond predictive accuracy and emphasize the need for reliability assessment and external validation before clinical deployment.