It is well-known that, to avoid diabetic complications, patients should keep their blood glucose levels as close to normal as possible.¹ However, this is no easy task for patients with Type 1 diabetes on insulin pump therapy. These patients must continually monitor their blood glucose levels, avoiding both hyperglycemia, or high blood glucose levels, and hypoglycemia, or low blood glucose levels.

Current monitoring devices can provide large volumes of blood glucose data for physicians to review. However, we do not yet have commercially available programs to interpret this data or to integrate it with data about lifestyle factors that can impact blood glucose levels. We have built a prototypical system to automatically analyze both patient blood glucose and lifestyle data, detect abnormal patterns in blood glucose control, and then recommend solutions to individual problems.²

Twenty patients with Type 1 diabetes on insulin pump therapy participated in a six-week pilot study. Each patient provided extensive electronic daily logs including self-glucose monitoring data, insulin dosages, work schedules, sleep patterns, exercise, meals, stress, illness, infusion set changes, pump problems and hypoglycemic episodes. Each patient also provided Medtronic MiniMed continuous glucose monitoring data for three separate 72-hour intervals. Physicians interpreted the data, identifying problems and recommending therapy adjustments to solve them. Patients then made the recommended adjustments, and physicians watched subsequent data to see how well the adjustments were working. Artificial intelligence researchers, known as knowledge engineers, worked with the physicians to record the problems, solutions, and outcomes. Each recorded problem was encoded as a case for the software prototype.

Fifty problem/solution/outcome cases were included in a prototypical decision support system. The software detects nocturnal hypoglycemia, morning hyper or hypoglycemia, over-correction for hyper or hypoglycemia, pre-meal or post-meal hyper or hypoglycemia, over-bolusing at meals, exercise-induced hypoglycemia, and some problems related to insulin pump or infusion set malfunction. Then it compares each newly detected problem to the fifty stored cases to find similar problems with known solutions. Similar past problems and solutions are displayed to the physician as an aid in deciding how to best handle the new problem.

This study demonstrates the feasibility of developing decision support software for patients with Type 1 diabetes on insulin pump therapy. The long-range goal is to enhance this software so that it could be used directly by patients, in non-critical situations, and would immediately alert physicians to critical situations. Additional research is needed to develop a practical tool that would be safe and effective for patients to use. A second research study is planned in which 28 patients will participate for three months each to further develop the intelligent decision support software.

^1. The Diabetes Control and Complications Trial Research Group (1993). The effect of intensive treatment of diabetes on the development and progression of long-term complications in insulin-dependent diabetes mellitus. New England Journal of Medicine, 329:977-986.

^2. C. R. Marling, J. H. Shubrook, W. A. Miller, A. J. Maimone and F. L. Schwartz (2007). Intelligent decision support software for Type 1 diabetics on insulin pump therapy. American Diabetes Association 67th Scientific Abstracts Book, abstract number 2087-PO.

We gratefully acknowledge support from Medtronic MiniMed, Ohio University's Russ College Biomedical Engineering Fund, and the Ohio University Osteopathic College of Medicine Research and Scholarly Affairs Committee. We would also like to thank the participating patients, research nurses, and graduate research assistants for their valuable contributions to this work.