Combining statistical, rule-based, and physiologic model-based methods to assist in the management of diabetes mellitus.

Self-monitoring of capillary blood glucose is used by most patients with insulin-dependent diabetes mellitus as a means of assessing metabolic control. Therapeutic interventions are based on retrospective analysis of glycemic response to various factors, with insulin and diet playing the key roles. We describe a computer system being developed for intelligent automated analysis and interpretation of data relevant to glycemic control. CADMO (Computer-Assisted Diabetes Monitor) is intended to assist health care professionals with the management of patients with insulin-dependent diabetes. It takes as input glucose values and insulin doses collected via a memory meter by the patient over a period of several weeks. Rule-based logic, statistical methods, and a physiologic model of insulin pharmacokinetics and glucose dynamics are used to help detect meaningful patterns and trends in glucose and insulin data and to suggest approaches for optimizing insulin regimens.