The Impact of Missing Covariates on Individualized Cost-Effectiveness Analysis

• Main Authors: Tsai, Wan-Jung, 蔡宛蓉
• Other Authors: CHIEN-FU JEFF LIN
• Format: Others
• Language: zh-TW
• Published: 2014
• Online Access: http://ndltd.ncl.edu.tw/handle/29630013134526564693

碩士 === 國立臺北大學 === 統計學系 === 102 === Cost effectiveness analysis (CEA) is one of the important tools in health economics research. Cost and effectiveness are both affected by individual's characteristics; therefore, individualized CEA that includes individual's covariates in regression models become popular. At present some studies used multiple imputation only for missing value of cost or effectiveness. However, few studies addressed the effect of missing covariates on individualized CEA. This study aimed to explore the effects of missing covariates on individualized CEA from a population database. This study used a population database from National Health Insurance Research Database and compare individualized CEA between two treatments: internal fixation and hemiarthroplasty for femoral neck fracture. We generated 1,000 full data sets using bootstrap sampling. Then, we constructed missing data sets from 3 different missing mechanisms. We generated imputed data sets by using multiple imputation. We assessed the effects of missing mechanisms and multiple imputations on individualized CEA. We found that imputed data sets had better performance than missing data sets on estimation and variation. However, when data had observed outliers, the outliers had significant impact in imputation resulted in higher biases. Because multiple imputation used these outliers to imputed missing values, the effects of outliers on the ratios of ICER are unpredictable. We suggest that a complete sensitivity analysis for missing data is necessary in individualized CEA.