A Multiple Hypothesis Approach to Estimating Meal Times in Individuals with Type 1 Diabetes

In order to properly titrate insulin dosages for individuals with type 1 diabetes (T1D), it is essential to have accurate information regarding when they have eaten and taken insulin to reconcile those events with their blood glucose levels throughout the day. A verifiable record of when insulin was taken can be obtained by downloading data from the patient's insulin pump. While this record shows exactly when insulin was injected, it remains unclear when that person actually ate. Although information about consumed carbohydrates is often logged at the time of an insulin bolus, it has been shown that individuals with T1D often dose insulin long after they have eaten. This practice is not advised and has been linked to an increased risk of developing complications. This project demonstrates a method to estimate the times of meals using a multiple hypothesis approach. When an insulin dose is recorded multiple hypotheses are spawned describing different variations of when the meal in question occurred. As postprandial glucose values further inform the model, the posterior probability of the truth of each hypothesis is evaluated, and from these posterior probabilities an expected meal time is found. This technique could be used to help advise physicians about the mealtime insulin dosing behaviors of their patients and potentially influence changes in their treatment strategy.