ABSTRACT
Consequences of Decisions . . . . . . . . . . . . . . . . . . . . . . . . . 411 13.2 Basic Steps in Developing a Decision Model . . . . . . . . . . . . . . . . . . 412
13.2.1 Conceptualize the Problem . . . . . . . . . . . . . . . . . . . . . . . . . 413 13.2.2 Conceptualizing the Model . . . . . . . . . . . . . . . . . . . . . . . . . 414 13.2.3 Specifying Type of Model to Be Used . . . . . . . . . . . . . . . . . . 414
13.2.3.1 Decision Trees . . . . . . . . . . . . . . . . . . . . . . . . . . . . 415 13.2.3.2 State Transition (Markov) Models . . . . . . . . . . . . . . 415 13.2.3.3 Individual Microsimulation . . . . . . . . . . . . . . . . . . 415 13.2.3.4 Dynamic Transmission (Compartment) Models . . . . 415 13.2.3.5 Discrete-Event, Agent-Based Simulation . . . . . . . . . 415
13.2.4 Structure the Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 415 13.2.5 Analyze the Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 416 13.2.6 Conduct Sensitivity Analysis . . . . . . . . . . . . . . . . . . . . . . . . 416 13.2.7 Validate the Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 418
13.3 Types of Modeling Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 418 13.3.1 Basic Branch and Node Decision Trees . . . . . . . . . . . . . . . . . 419 13.3.2 State Transition (Markov) Models . . . . . . . . . . . . . . . . . . . . 420
13.3.2.1 Adding Clinical Detail . . . . . . . . . . . . . . . . . . . . . . 424 13.3.3 Microsimulation Models . . . . . . . . . . . . . . . . . . . . . . . . . . . 426 13.3.4 Discrete-Event and Agent-Based Models . . . . . . . . . . . . . . . 428 13.3.5 Infectious Disease (Compartment) Models . . . . . . . . . . . . . . 429 13.3.6 Hybrids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 429
13.4 Examples of Decision Models Used in Comparative Effectiveness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 430 13.4.1 Use of Virtual Colonoscopy . . . . . . . . . . . . . . . . . . . . . . . . . 430 13.4.2 First-Line Therapy for Follicular Lymphoma . . . . . . . . . . . . 430
13.5 Decision Analysis Worked Example . . . . . . . . . . . . . . . . . . . . . . . . 431 13.5.1 Basic Structure and Definitions of a Decision Analysis
(Branches/Nodes/Outcomes) . . . . . . . . . . . . . . . . . . . . . . . 431 13.5.2 Branches and Nodes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 432
13.5.3 Probabilities and Outcomes . . . . . . . . . . . . . . . . . . . . . . . . . 433 13.5.4 Mechanics: Expected Value of the Outcomes . . . . . . . . . . . . 435 13.5.5 Sensitivity Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 436
13.6 Value of Information Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 437 13.6.1 Expected Value of Perfect Information . . . . . . . . . . . . . . . . . 439 13.6.2 Expected Value of Partial Perfect Information . . . . . . . . . . . 441 13.6.3 Research Prioritization Using VOI Analysis . . . . . . . . . . . . . 442
13.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 442 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 443
ABSTRACT Mathematical models are being increasingly used to evaluate the future impacts of various therapies for particular problems, and to inform many types of clinical and policy decisions facing clinicians, patients, and policy makers in healthcare today. One advantage of mathematical models is that they can incorporate evidence from many different sources and provide a framework to integrate this knowledge and estimate an optimal decision under various different conditions. The purpose of this chapter is to provide an overview of the use of mathematical modeling in comparative effectiveness research, with a concentration on the commonly used methods, including simple decision trees, time-varying state transition (often called Markov) models, and microsimulation. Several examples are described, and references to more complete expositions on the methodology are provided.
