Step 2: Choose the Method of Economic Evaluation

The next step is to decide which economic method best fits the chosen perspective and phase of intervention development. There are three main methods of economic evaluation: prospective, retrospective, and modeling approaches as described in Table 18.2.

Here we focus on approaches for establishing the costs of an intervention. Costs of an intervention should be considered early on as one is developing the delivery characteristics of a behavioral intervention and then more formally as one evaluates the intervention. That is, cost data should be collected prospectively, which is typically accomplished by integrating economic evaluations within efficacy and effectiveness studies. This is referred to as “piggybacking” economic evaluations alongside prospective studies. While the main goal of such trials is to establish the efficacy or effectiveness of the behavioral intervention, secondary goals should include an economic evaluation. Piggybacking allows the investigator to collect all relevant cost data. One important piece of cost data is the actual cost of delivering the intervention. As there is often little to no data available on the cost of delivering novel behavioral programs, this is an important endeavor. Piggybacking also has some disadvantages; it can increase respondent burden, and with small sample sizes, there may be insufficient power to analyze the cost data.

A retrospective economic evaluation is performed using historical data (e.g., claims data) with the goal of determining what the cost or cost-effectiveness of the treatment is compared to usual care or a relevant comparator. Retrospective economic evaluations are generally less expensive to conduct and can include large sample sizes. However, many retrospective data (e.g., health care claims or medical records) lack data on the cost of delivering behavioral interventions.

TABLE 18.2 Summary of Three Methods of Economic Evaluation

Method of Economic Evaluation

Question Intended to Answer





What is the financial impact of the intervention, when actually measured in a defined population?

Economic measures are included within study variables and actually measured during the course of the study Sometimes referred to as "piggybacking" when economic measures are added to studies that are primarily being conducted to test efficacy or effectiveness




randomized clinical trial



What was the financial impact of the intervention, based on costs and effectiveness actually observed, and considering investments made in the intervention?

An examination of health care investments made, in relation to financial savings (in other words, calculating the net financial benefit of the intervention, or return on investment) and/or in relation to the effectiveness achieved

Health care claims analysis; chart or electronic medical record review



What might be the financial impact of the intervention, given reasonable assumptions about its costs and effects?

Retrospective data from credible sources are scientifically analyzed to estimate costs with versus without the treatment, or costs per unit of effectiveness, with versus without the treatment

Decision analysis; Markov model; budget impact model

Modeling is performed using data from multiple sources (e.g., meta-analysis, published clinical trial results, registries, and databases). Modeling is often used when all the necessary components of an economic evaluation cannot be collected in a single study, hence necessitating simulation on the basis of available literature and existing data. Examples of modeling approaches commonly used in health care include decision analyses and Markov models. These models synthesize results from multiple sources to estimate cost and cost-effectiveness. Yet, in the case of behavioral interventions, data on costs and effectiveness are still emerging, so at the present time, models may be difficult to accurately inform.

For the purposes of this chapter, we discuss modeling as an approach suited for conducting sensitivity analysis of piggybacked economic evaluations. In this case, modeling serves as a vehicle for performing sensitivity analysis on trial-based cost studies to account for potential real-world uncertainty resulting from restrictive inclusion criteria and/or inadequate sample size.

< Prev   CONTENTS   Source   Next >