Analysis: toward algorithms in diabetes self-management.

Abstract

371 IN HEALTH, fuel homeostasis is a robust process. Central to this mechanism, the endocrine pancreas controls the storage and release of all the vital fuels, including fat, carbohydrate, and protein precursors. It is easily within the physiologic range to control all metabolic needs from feast to famine. In diabetes, the exquisite control of protein, fat, as well as carbohydrate metabolism is lost because the release of insulin from the beta cells of the endocrine pancreas is either lacking (Type 1 diabetes) or inappropriate (Type 2 diabetes). Blood glucose concentrations, which are easily measured, partially reflect this loss of metabolic control. Thus for survival, persons with diabetes must substitute for the loss. This is achieved primarily by replacing the lost (closed-loop) endocrine function with subcutaneous injections of depots of insulin. In diabetes, self-management therefore involves strategic (open-loop) decisions about the timing and amounts of these injections needed to correct for abnormal blood glucose levels and anticipated carbohydrate intakes. Because a depot exists, the patient must ideally anticipate the effects of subsequent physical exercise. The rules used to govern these decisions are called algorithms, and their purpose is to control subsequent blood glucose levels. Algorithms in general can be thought of as a formal set of rules, procedures, or computer-executable routines that safely translate outcomes into interventions. In diabetes self-management, one outcome is self-measured blood glucose level, another is carbohydrate intake, and a third is planned exercise. Each modifies the intervention, which in this case is insulin dosing. For practical reasons, clinical algorithms have to be simple and easily self-administered by the patient at each injection. Clinicians teach the patients how to do this based on the blood glucose level they measure at each meal, the desired blood glucose level, the anticipated carbohydrate intake, the individual’s sensitivity to insulin at that time, and the anticipated level of exercise. If the parameters are correct, then the amount injected will control the following blood glucose levels. Although rules for the self-control of diabetes are not new, their formalization as mathematical algorithms is relatively recent. Albisser1 and Albisser et al.2 introduced the concept of programmable, hand-held computers into diabetic self-management and outlined closed-loop algorithms for the adjustment of shortand intermediate-acting insulins. The Albisser Insulin Dosage Computer executes algorithms that are responsive to just four blood glucose levels measured by the patient each day, with target levels and insulin sensitivity factors set by the provider. Patients can enter changes in anticipated carbohydrates. They can also enter changes in anticipated exercise that