Mathematical model could help ICU doctors regulate blood glucose

Managing the blood glucose levels of critically ill patients is a challenging and burdensome process for healthcare providers. Accurate and robust forecasting of blood glucose levels could support the clinical decision-making process in intensive care units (ICUs).

Sirlanci et al. developed a new model for blood glucose forecasting. In theory, a high-fidelity nonlinear deterministic model for forecasting should be possible given the complex nonlinear behavior of blood glucose regulation. However, in most clinical settings, blood glucose is not monitored closely enough for such a model to be effective. Instead, the researchers developed a simpler mechanistic model to reduce uncertainty when used with a noisy data set.

“Instead of predicting the actual trajectory of blood glucose levels, which is very challenging due to real-world data limitations, we developed a physiology-based stochastic model that resolves the blood glucose dynamics to a level that results in blood glucose forecasts sufficiently accurate for glycemic management,” said author Melike Sirlanci.

The team’s model has two components: a deterministic component that models the average blood glucose level behavior, and a stochastic component to quantify oscillations. It can also be personalized based on patient-specific data.

Their model showed good accuracy in predicting future blood glucose levels when compared to experimental data collected from patients. The authors hope the model can be used in ICU settings. They also believe this technique, which sacrifices physiological fidelity to gain robustness and accuracy, could be applied to other problems that have similarly noisy or sparse data.

Source: “A simple modeling framework for prediction in the human glucose-insulin system,” by Melike Sirlanci, Matthew E. Levine, Cecilia C. Low Wang, David J. Albers, and Andrew M. Stuart, Chaos (2023). The article can be accessed at https://doi.org/10.1063/5.0146808 .