Introduction to statistics and modeling methods applied in health economics

Abstract

The increasing complexity of health economics methodology has raised the need for technical methods to systematically use patient-level data and characterize uncertainty around the decision problem for decision makers. This chapter provides an introduction to these methods, focusing on trial-based statistical techniques and economic modeling methods for the purpose of health economic analysis. This chapter describes some differences between the more commonly used frequentist approach for clinical analysis and the developing use of Bayesian methods for health economic analysis. Statistical methods described include the use of power calculations, hypothesis testing, and regression analysis, and their relevance for economic analysis. More advanced statistical methods are also introduced, such as the area under the curve method for assessing incremental benefit, controlling for missing data and baseline characteristics, and using mapping algorithms for eliciting preference-based tariff scores when a preference-based measure has not been collected within a study. The second part of the chapter focuses on modeling methods designed to synthesize data from multiple sources when the economic analysis needs to go beyond a single source of primary data or for a longer time horizon. Multiple types of economic models are described, including decision trees, state transition models (including Markov chain models), microsimulation, and discrete event simulation. The chapter breaks down key elements of model design and offers recommendations on possible sources of data that may be used to derive parameter estimates. The conclusion of the chapter includes recommendations for appropriately reporting results of the statistical and modeling analyses carried out as part of an economic evaluation.