Simerjit Gill

Agent-Based Simulation of Healthcare for Type II Diabetes

Healthcare delivery systems around the world are faced with growing demand to provide a better-quality standard of care and at the same time keep the cost and resource utilization minimal. Statistics reveal that diabetes is a major issue in the healthcare industry and there is a strong need to address this problem in order to improve and reduce the cost of its treatment. Furthermore, evidence of the effectiveness of self-monitoring blood glucose levels is unclear, particularly in patients with Type 2 diabetes. In this study, interaction between the diabetic patient and the physician in current healthcare settings is modeled and simulated using software agent technology to demonstrate the long term clinical-effectiveness and cost-effectiveness of various diabetic interventions such as self-monitoring and lifestyle adjustments.

An agent-based simulation system for modelling a diabetic patient

This paper presents a new paradigm for modelling illness in the human population. In this work, we propose the development of a patient model using Software Agent technology. The Patient Agent is developed in accordance with the general parameters used in describing diabetes mellitus in terms of inputs, patient factors and outputs. Wus (2005) mathematical model has been adapted to transform input variables such as food, exercise and medications, as well as other risk factors like age, ethnicity and gender, into the primary output variable of blood glucose. Blood pressure and heart rate signal are additionally generated using typical physical data and linear scaling. The primary contribution of this work is in the demonstration of the ability of an agent mediated system to model the indicators of illness, specifically diabetes mellitus, in a short term interval with high accuracy, and in the interactions between a patient/physician in order to produce complex, dynamic system behaviours.

Agent-Based Modeling and Simulation

Computer simulation, or just simulation, is a decision support technique that enables stakeholders to conduct experiments with models that represent real-world systems of interest.

Patient-Physician Interaction Model

This work integrates Wu’s (Sycara AI magazine 19:79, 1998 [1]) model with the agent technology to develop the diabetic Patient Agent model (Martens and Benedicenti EEMA TRLabs execution environment for mobile agents, 2001 [2].

The Diabetic Patient Agent

This chapter presents and discusses the results obtained from the actual simulation of the proposed model. Various simulations have been performed to evaluate and examine the behaviour of the diabetic patient and the characteristics of the healthcare system under different scenarios. All scenarios in this section are simulated for 1 year with a fixed cost of physician and hospital visits. 5.1 Manipulating the Frequency and the Period of Self-monitoring This simulation examines the characteristics of a patient’s blood sugar and the financial cost associated with the healthcare system by modifying the frequency and the period of self-monitoring of blood glucose by the patient with an extremely unhealthy diet regimen. The parameter settings for this simulation are as follows: • Food: The patient consumes a high to very high amount of carbohydrates at breakfast, and a consistently very high amount of carbohydrates at lunch and dinner. • Physical Activity: The level of the patient’s physical activity is low to medium during the morning, afternoon and evening. • Mealtime: The patient eats breakfast around 6:00 am, lunch between 11 am and 12:40 pm, and dinner between 5:10 pm and 6:20 pm. • Medication: The patient is initially not on any medication. • Willingness to adopt healthier lifestyle: The patient’s behaviour for this parameter is set to 0%.

Self-aware Patient Agent Model

The CPA incorporates the influential parameters in diabetes management and it can represent typical diabetic patients. However, human behaviour is far more complicated than the proposed model. For instance, people may frequently sample their BG and respond to knowledge of their own condition, which is called SMBG and self-management. To incorporate such issues, the SPA is constructed as an enhancement of the CPA. This model also incorporates risk factors of age and health status.

Blood Glucose Monitoring Frequency Evaluation

As mentioned in Sect. 7.1, SMBG for people with Type 2 diabetes is commonly recommended and the efficacy of SMBG has been demonstrated in the SPA model. The hypothesis is based on the fact that self-management or lifestyle changes are facilitated by SMBG. The other hypothesis is that the samples can properly represent the CBG, and obviously, the more samples taken, the better the representation. However, the frequency of monitoring is an issue due to the related cost, discomfort and potential infection. There should be an optimal frequency of SMBG which supports an accurate representation of the CBG without excessive or oversampling.

Control Patient Agent Model

After observation of the patient-physician interaction model, we return to improve the individual patient agent model of developing the CPA and SPA in the second part of this work, since we believe patients are the most important components in the healthcare system. The improvement is mainly on the enhancements of the Ackerman model which has been discussed in Sect. 3.2. The CPA integrates several essential attributes in diabetes management that may affect diabetes, such as eating habits, physical activity levels and medication. The name of the CPA does not have a special meaning, but only comes from comparison with the SPA, which is characterized by self-awareness.

The Ackerman Mathematical Model

A variety of mathematical models have been developed in the past decades to address different aspects of diabetes (Wooldridge, An introduction to multiagent systems. Wiley, [1]; Oliveira et al., Robot Auton Syst 27:91–106, 1).