Aims: Diabetes self-management involves a difficult balancing act between insulin, food and exercise. The challenge is to develop innovative, validated algorithms to aid patient decision-making and optimize glycaemic control. 'Librae' is a computerized diabetes simulator in diary format, developed as an educational predictive tool for patients, reducing 'trial and error' by allowing patients to simulate and experiment with dietary or insulin adjustments on a 'body double'. We have evaluated the predictive ability of Librae using continuous blood glucose monitoring (CGMS).
Methods: Patients with Type 1 Diabetes attending the Paediatric Clinic were invited to use 'Librae' for 1 week and were then fitted with a CGMS for 72 h. The predictive ability of 'Librae' was compared with concurrent data obtained from the CGMS.
Results: Seven thousand nine hundred and sixty paired blood glucose values were obtained from the 11 patients who completed the study. 'Librae' underestimated the measured CGMS values, the error having a positive mean of 0.35 mmol/l (95% confidence interval 0.22-0.48 mmol/l). However, Librae tended to overestimate at low levels of blood glucose readings, and underestimate at high levels of blood glucose readings.
Conclusion: The modelled values of 'Librae' correlated well with the CGMS data, but clinically unacceptable errors occurred at extremes of blood glucose levels. Concurrent CGMS recordings have provided a large data set to modify and improve the existing Librae model and patient feedback has led to improvements in its usability. Librae may provide a useful tool to improve diabetes self-management education and optimize glycaemic control.