Description and preliminary evaluation of a Multiagent Intelligent Dosing System (MAIDS) to manage combination...

Computer decision support systems are potentially effective methods for adjusting insulin, but current models do not take into account simultaneous changes of more than one agent. We describe the development of the Multiagent Intelligent Dosing System (MAIDS, Dimensional Dosing Systems, Wexford, PA) for predicting glycemic outcome in response to concurrent dose adjustments in oral hypoglycemic agents and insulin.

Retrospective data from a patient cohort with type 2 diabetes who had simultaneous changes in insulin and metformin were analyzed. Glycemic markers (fasting glucose, random glucose, or hemoglobin A1c) expected at the visit subsequent to dose changes were calculated using two methods: the previously reported Intelligent Dosing System (IDStrade mark, Dimensional Dosing Systems), which accounts for changes in only one agent, and the MAIDS. Expected results from both systems were correlated with levels actually observed.

We analyzed 32 patients with 40 paired visits. For fasting glucose (n = 8 paired visits), the correlation between expected and observed values was 0.07 when using the IDS but 0.78 when using the MAIDS. For random glucose (n = 16 paired visits) the correlation between expected and observed levels was 0.49 for the IDS but 0.79 for the MAIDS. With hemoglobin A1c as the marker (n = 16 paired visits), the correlation was 0.40 when using the IDS but 0.60 with the MAIDS.

The MAIDS allows better prediction of glycemic outcome in circumstances where both insulin and an oral hypoglycemic drug are changed concurrently. Application of the MAIDS to other clinical scenarios, such as simultaneous adjustment of insulin and carbohydrate intake, requires further study.