The diabetes system modeling project began in September 2003 primarily as a result of two concerns: first, that existing programmatic strategies focused on reducing the immediate burden of diabetes might soon become overwhelmed or lose effectiveness in the face of rapidly increasing disease prevalence, and, second, that diabetes leaders lacked effective, quantitatively-grounded decision-support tools that could improve diabetes strategies in light of the first concern. SD was determined to be an appropriate technique for looking formally at a broad spectrum of programmatic options and considering their relative effectiveness over the short and long term. Part of the appeal of SD was its flexibility and ability to deal in an integrative and transparent way with the diverse questions and rich information sources that characterize the current state of diabetes prevention and control. Thus, an SD approach held promise as a way of answering tough questions, diminishing the sense of information overload, pointing the way toward a more effective program mix, and thereby improving the ability of DDT to engage effectively with state and local colleagues, as well as other stakeholders who would ultimately be needed as partners to champion any change in programmatic direction.
The way that we have been able to start accomplishing these aims over the past two years is by developing a mathematical tool that allows one to do a wide range of experiments, not unlike the artificial conditions created in a prospective clinical trial. In an SD model, however, the conditions are defined by the formal structure of a simulation model along with its baseline assumptions, and the experiments allow decision-makers to compare outcomes under carefully specified alternative scenarios. The purpose of this experimental approach is not to forecast actual future values of the system, but rather to learn about the relative impacts of alternative assumptions and interventions. One may generally draw firm conclusions about those relative impacts, even if one is not absolutely certain about the baseline (or “status quo”) assumptions. Of course, the baseline assumptions should be plausible, and it is often instructive to experiment with different baseline assumptions to determine what effect they may have on the findings of relative impact.
The diabetes model is not a fixed entity but one that has evolved and continues to evolve as we learn more and engage others in an expanding dialogue about the system’s structure and behavior. The model has been developed in close collaboration with a team made up of DDT staff from both the program branch and ESB, with additional input from a steering committee of experienced individuals from throughout CDC, and also with detailed input from DNPA, DACH, and other related divisions. More recently, we have been working with diabetes analysts and policymakers in the Vermont Department of Health to explore conditions in their state, and we have made further changes to the model as a result of those interactions. Face-to-face meetings have been held throughout the process, and hundreds of pages of electronic messages and memoranda document the work we have done between meetings to improve the model’s logic, gather data, derive parameter estimates, and evaluate the model’s output and its implications.
CONTENTS
1. Background
2. Model Structure Overview
3. Population Stocks and Flows
- 3.1 Oveview, Population Inflow, and Deaths
3.2 Prediabetes Onset and Recovery
3.3 Diabetes Onset and Progression
3.4 Diagnosis of Complicated Diabetes
3.5 Diagnosis of Uncomplicated Diabetes
3.6 Diagnosis of Prediabetes
3.7 Simulation of Historical Diabetes and Prediabetes Prevalence
4. Control of Diabetes and Prediabetes
- 4.1 Factors Affecting the Controlled Fraction
4.2 Simulation of Historical Diabetes and Prediabetes Control
5. Obesity as Affected by Caloric Intake and Physical Activity
- 5.1 Modeling Obesity by Way of Average Weight and BMI
5.2 Caloric Intake and Physical Activity
5.3 Obese Fractions of Normoglycemic, Prediabetes, and Diabetes
Populations
5.4 Simulation of Historical BMI and Obese Fractions
6. Health Care Costs
- 6.1. Health Care Costs of Diabetes and Prediabetes
6.2. Simulation of Historical Health Care Costs of Diabetes and Prediabetes
7. Unhealthy Days
- 7.1. Unhealthy Days due to Diabetes
7.2. Simulation of Historical Unhealthy Days due to Diabetes
8.References
