Improving the management of hyperglycemia in type 1 diabetes therapy: an algorithm for the suggestion of corrective insulin boluses based on future glucose prediction

Type 1 diabetes (T1D) is a chronic disease characterized by elevated blood glucose (BG) levels, due to the absence of insulin secretion by pancreatic β-cells. People affected by T1D, must adhere with a life-long therapy which aims to maintain BG concentration within the safe range of [70−180] mg/dL. The standard therapy, which consists of multiple burdensome daily actions, can be inefficient and difficult to implement in the long-term, thus negatively affecting patient’s quality of life. In this context, decision support systems (DSSs) based on minimally-invasive continuous glucose monitoring (CGM) sensors, which allow measuring BG almost continuously for several days, play a key role in improving T1D management. In partic- ular, DSSs are software tools which can assist individuals with T1D by suggesting appropriate therapeutic actions as carbohydrates intake and insulin injections. The availability of algorithms that can accurately predict future BG levels would facilitate proactive therapeutic interventions, further improving glucose control and patients well-being. In this paper, we present a novel algorithm for DSSs that suggests the administration of corrective insulin boluses (CIBs) lever- aging a CGM-based predictive model aimed to forecast the future CGM concentration. The effectiveness of the proposed algorithm has been assessed retrospectively on a dataset consist- ing of 30 T1D patients recorded in real-life conditions and the performance has been compared with that of a recent state-of-art heuristic-based strategy for hyperglycemia correction. Results indicate that the proposed algorithm for CIB suggestion decreases the time in hyperglycemia (34.77% vs 39.42%) and increases the percentage of time spent in euglycemia (64.33% vs 59.16%) with statistical significance, additionally lowering the overall risk for patients.