Capturing human-machine interplay in automated insulin delivery systems through day-to-day variability of the automated correction index (ACI-V%).

BACKGROUND

The Automated Correction Index (ACI) reflects the proportion of algorithm-driven correction boluses relative to total bolus insulin in advanced hybrid closed-loop (AHCL) systems. While mean ACI has been associated with glycemic outcomes in patients with type 1 diabetes mellitus (T1DM), the clinical significance of its day-to-day variability (ACI-V%) remains unknown.

METHODS

In this multicenter real-world study, 330 pediatric and adult patients with T1DM using the MiniMed™ 780G system were studied. ACI-V% was defined as the coefficient of variation of daily ACI values over a 30-day period. Associations between ACI-V% and continuous glucose monitoring (CGM) metrics were assessed.

RESULTS

Higher ACI-V% was associated with greater time in range (TIR70-180 mg/dL) and lower time above range, while no significant association with time below range was observed. ACI-V% showed an inverse correlation with mean ACI, suggesting greater day-to-day variability in settings of lower overall automated correction insulin delivery. Moreover, ACI-V% independently predicted achievement of TIR70-180 mg/dL > 70%, with a threshold of ≥ 32.6% identifying patients more likely to reach this target (sensitivity 52.4%, specificity 71.4%).

CONCLUSIONS

ACI-V% provides an exploratory descriptor of day-to-day variability in automated correction insulin delivery that may complement mean ACI for clinical interpretation of AHCL system behavior.