Bayesian Variable Selection in Cost-Effectiveness Analysis

Abstract

Linear regression models are often used to represent the cost and effectiveness of medical treatment. The covariates used may include sociodemographic variables, such as age, gender or race; clinical variables, such as initial health status, years of treatment or the existence of concomitant illnesses; and a binary variable indicating the treatment received. However, most studies estimate only one model, which usually includes all the covariates. This procedure ignores the question of uncertainty in model selection. In this paper, we examine four alternative Bayesian variable selection methods that have been proposed. In this analysis, we estimate the inclusion probability of each covariate in the real model conditional on the data. Variable selection can be useful for estimating incremental effectiveness and incremental cost, through Bayesian model averaging, as well as for subgroup analysis.

Cost-effectiveness analysis of domiciliary topical sevoflurane for painful leg ulcers

“…Thus, we planned to perform a linear regression analysis to ascertain if the main results could have been influenced by the assignation to the treatment groups. Specifically, a multiple regression analysis was conducted to identify the part of the difference in costs or effectiveness between groups that was not attributable to the treatment but instead to other covariates, specifically, baseline characteristics of the patients [ 44 – 51 ]. Therefore, covariates allow to reduce the bias and uncertainty of the estimation of the parameters, even if the treatment groups have similar characteristics [ 48 – 51 ].…”

“…Specifically, a multiple regression analysis was conducted to identify the part of the difference in costs or effectiveness between groups that was not attributable to the treatment but instead to other covariates, specifically, baseline characteristics of the patients [ 44 – 51 ]. Therefore, covariates allow to reduce the bias and uncertainty of the estimation of the parameters, even if the treatment groups have similar characteristics [ 48 – 51 ]. The specification of this regression model was properly evaluated as it fitted the checklist developed to assess statistical methods for addressing selection bias in CEAs based on observational data [ 52 ].…”

Bayesian Regression Model for a Cost-Utility and Cost-Effectiveness Analysis Comparing Punch Grafting Versus Usual Care for the Treatment of Chronic Wounds

“…Our CUA and CEA were carried out in a Bayesian framework [35][36][37][38][39][40][41][42]. A multiple regression analysis was conducted to verify if the results in costs, utility, and effectiveness could be explained by the treatment performed or by other covariates [43][44][45][46][47][48][49][50]. To do that, Costs, QALYs, and Wound-free period were selected as dependent variables.…”

“…They also allowed us to identify the part of the difference in costs, utility, and effectiveness that was not attributable to the treatment but to the differences in the characteristics of the patients. In addition, covariates allowed us to reduce the bias and uncertainty of the estimation of the coefficients [47][48][49][50]. Our regression model specification was properly evaluated as it fitted the checklist developed to assess statistical methods for addressing selection bias in CEAs based on observational data [61].…”

Educational Data Mining

“…The researcher explores various alternatives, trading off theoretical plausibility, computational tractability, model fit, statistical reliability, interpretability, and informativeness with respect to the research questions of interest. This problem falls in the domain of model selection but differs from prior work on selecting structure [e.g., Madigan and Raftery, 1994] or variables [e.g., Negrin et al, 2010] in that we focus on selecting a specific parameterization of the given variables.…”