The diabetic patient software agent.

Abstract

The purpose of this thesis is to develop an agent-based simulation model of a diabetic patient‟s blood glucose levels in which the efficiency of various treatment strategies can be evaluated in a micro scope. A further aim is to establish a multi-agent system of evaluating the healthcare system response under various scenarios of healthcare policies in a macro scope. A framework is endeavoured to be constructed in order to extend its applications into other diseases. The incidence of Type 2 diabetes mellitus is reaching epidemic proportions in the world in recent times. On one hand, the disease can result in various serious complications such as limb loss, blindness, ischemic heart disease and end-stage renal disease. On the other hand, people with diabetes can expect to live active, independent and vital lives if they try to keep their blood glucose in a target range through diabetes management strategies such as education, medication and lifestyle control. To quantitatively asses the efficiency of various treatment strategies, several cost-effective experiments in different simulation scenarios of treatment strategies are implemented in the proposed models in silico otherwise it is often not possible or too difficult, dangerous or unethical to do them in vivo. Two individual patient agent models of a Controlled Patient Agent and a Self- Aware Patient Agent are presented. The author extends the original seminal work of Ackerman et al. of a mathematical model of the human glucose regulatory system and incorporates the enhanced model in the Controlled Patient Agent. The Self-Aware Patient Agent is enhanced based on the Controlled Patient Agent by introducing a blood glucose sensor in silico and a reasoning model of responding to the measures. Furthermore, the signal technique of calculating a cross-correlation function and average blood glucose deviation between the continuous blood glucose and the interpolation of samples is proposed to evaluate blood glucose monitoring frequency in the Self-aware Patient Agent model. A design of a multi-agent system is finally presented by introducing other healthcare components so that more interesting insights such as the healthcare quality, cost and performance can be observed in a macro scope. The major observations from the experiments illustrated that the two agent models could represent typical diabetic patients. The agent models evaluate the efficiency of various lifestyles, the self-awareness and the treatment of self monitoring blood glucose in a quantitative way. This can help patients explore their prognosis if they are not meticulous in controlling their blood glucose levels and can assist patients to determine an optimal frequency for monitoring their blood glucose in order to reduce discomfort, cost and potential infection caused by intensity of measurement. By analyzing the simulation results, a base line of treatment strategies in silico is built, which may aid people to discover a personalized management regimen based on their situation.