Related Experiment Videos
Constructing confidence intervals for cost-effectiveness ratios: an evaluation of parametric and non-parametric
A H Briggs1, C Z Mooney, D E Wonderling
1Health Economics Research Centre, Oxford Institute of Health Sciences and Nuffield College, University of Oxford, U.K. andrew.briggs@ihs.ox.ac.uk
Statistics in Medicine
|December 22, 1999
Summary
Estimating confidence intervals for cost-effectiveness ratios is crucial in health economics. Monte Carlo simulations reveal that methods assuming normal distributions perform poorly, especially with skewed data.
Area of Science:
- Health Economics
- Biostatistics
- Statistical Modeling
Background:
- The incremental cost-effectiveness ratio (ICER) is a key statistic in health economic evaluations.
- Estimating confidence intervals for ICERs is challenging due to the intractable variance of ratio estimators.
- Various methods have been proposed in health economics literature to address this challenge.
Purpose of the Study:
- To evaluate and compare the appropriateness of different confidence interval estimation methods for cost-effectiveness ratios.
- To determine which methods provide accurate coverage properties under various conditions.
Main Methods:
- Monte Carlo simulation techniques were employed to repeatedly sample from known distributions.
- Different confidence interval estimation methods were applied to simulated data.
- Response surface analysis was used to analyze the performance of methods across varying sample sizes, coefficients of variation, and cost-effect covariances.
Main Results:
- Substantial differences in the performance of various confidence interval estimation methods were identified.
- Parametric and non-parametric methods assuming a normal sampling distribution performed poorly.
- The approach of combining separate cost and effect intervals also yielded inadequate results.
Conclusions:
- The choice of method for estimating confidence intervals significantly impacts the estimated limits for cost-effectiveness ratios.
- Normal approximation methods are particularly unreliable and should be avoided when skewed sampling distributions are suspected.
- Accurate confidence interval estimation is vital for reliable economic appraisal in healthcare.