| Summary: | 碩士 === 逢甲大學 === 應用數學學系 === 103 === Cost-effectiveness analysis is often used in the biomedicine, it is mainly used in the comparison of the cost-effectiveness of new and standard treatments. Studies on the incremental cost-effectiveness ratio (ICER) as an index of evaluation has been frequently studied. However, ICER is vulnerable to numerical instability. The average cost-effectiveness ratio (ACER) is another index of evaluation. It avoids the problem of ICER. We proposed the probability which is modeled in the form of ACER whether or not the new treatment is more cost-effectiveness than standard treatment; it offers the possibility of obtaining cost-effectiveness when policy makers choose a new treatment. In practice, the cost and effect data are often from a skewed distribution, so this study will assume the cost and effect are from the binary logarithmic normal distribution. Using the maximum-likelihood estimation method and non-parameter method, we proposed three different kinds of estimation methods, which are (1) the estimated probability of cost-effectiveness of the maximum-likelihood estimation method, (2) standardization- differences estimation of the maximum-likelihood estimation method and, (3) non-parameter estimation. We made an attempt to provide a testing procedure by using a new index. We used the Monte Carlo simulation analysis to explore three different kinds of estimation methods and their performance by using their empirical size and power curve. The simulation results show that the standardization-differences estimation of the maximum-likelihood estimation method should be recommended.
|
|---|
