An Educational Review About Using Cost Data for the Purpose of Cost-Effectiveness Analysis

Abstract

This paper provides an educational review covering the consideration of costs for cost-effectiveness analysis (CEA), summarising relevant methods and research from the published literature. Cost data are typically generated by applying appropriate unit costs to healthcare resource-use data for patients. Trial-based evaluations and decision analytic modelling represent the two main vehicles for CEA. The costs to consider will depend on the perspective taken, with conflicting recommendations ranging from focusing solely on healthcare to the broader 'societal' perspective. Alternative sources of resource-use are available, including medical records and forms completed by researchers or patients. Different methods are available for the statistical analysis of cost data, although consideration needs to be given to the appropriate methods, given cost data are typically non-normal with a mass point at zero and a long right-hand tail. The choice of covariates for inclusion in econometric models also needs careful consideration, focusing on those that are influential and that will improve balance and precision. Where data are missing, it is important to consider the type of missingness and then apply appropriate analytical methods, such as imputation. Uncertainty around costs should also be reflected to allow for consideration on the impacts of the CEA results on decision uncertainty. Costs should be discounted to account for differential timing, and are typically inflated to a common cost year. The choice of methods and sources of information used when accounting for cost information within CEA will have an effect on the subsequent cost-effectiveness results and how information is presented to decision makers. It is important that the most appropriate methods are used as overlooking the complicated nature of cost data could lead to inaccurate information being given to decision makers.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 58 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.