Christian E H Boehler1, Joanne Lord2. 1. Institute for Prospective Technological Studies, Joint Research Centre-European Commission, Seville, Spain (CEHB) 2. Health Economics Research Group, Brunel University, Uxbridge, UK (JL).
Abstract
BACKGROUND: Published cost-effectiveness estimates can vary considerably, both within and between countries. Despite extensive discussion, little is known empirically about factors relating to these variations. OBJECTIVES: To use multilevel statistical modeling to integrate cost-effectiveness estimates from published economic evaluations to investigate potential causes of variation. METHODS: Cost-effectiveness studies of statins for cardiovascular disease prevention were identified by systematic review. Estimates of incremental costs and effects were extracted from reported base case, sensitivity, and subgroup analyses, with estimates grouped in studies and in countries. Three bivariate models were developed: a cross-classified model to accommodate data from multinational studies, a hierarchical model with multinational data allocated to a single category at country level, and a hierarchical model excluding multinational data. Covariates at different levels were drawn from a long list of factors suggested in the literature. RESULTS: We found 67 studies reporting 2094 cost-effectiveness estimates relating to 23 countries (6 studies reporting for more than 1 country). Data and study-level covariates included patient characteristics, intervention and comparator cost, and some study methods (e.g., discount rates and time horizon). After adjusting for these factors, the proportion of variation attributable to countries was negligible in the cross-classified model but moderate in the hierarchical models (14%-19% of total variance). Country-level variables that improved the fit of the hierarchical models included measures of income and health care finance, health care resources, and population risks. CONCLUSIONS: Our analysis suggested that variability in published cost-effectiveness estimates is related more to differences in study methods than to differences in national context. Multinational studies were associated with much lower country-level variation than single-country studies. These findings are for a single clinical question and may be atypical.
BACKGROUND: Published cost-effectiveness estimates can vary considerably, both within and between countries. Despite extensive discussion, little is known empirically about factors relating to these variations. OBJECTIVES: To use multilevel statistical modeling to integrate cost-effectiveness estimates from published economic evaluations to investigate potential causes of variation. METHODS: Cost-effectiveness studies of statins for cardiovascular disease prevention were identified by systematic review. Estimates of incremental costs and effects were extracted from reported base case, sensitivity, and subgroup analyses, with estimates grouped in studies and in countries. Three bivariate models were developed: a cross-classified model to accommodate data from multinational studies, a hierarchical model with multinational data allocated to a single category at country level, and a hierarchical model excluding multinational data. Covariates at different levels were drawn from a long list of factors suggested in the literature. RESULTS: We found 67 studies reporting 2094 cost-effectiveness estimates relating to 23 countries (6 studies reporting for more than 1 country). Data and study-level covariates included patient characteristics, intervention and comparator cost, and some study methods (e.g., discount rates and time horizon). After adjusting for these factors, the proportion of variation attributable to countries was negligible in the cross-classified model but moderate in the hierarchical models (14%-19% of total variance). Country-level variables that improved the fit of the hierarchical models included measures of income and health care finance, health care resources, and population risks. CONCLUSIONS: Our analysis suggested that variability in published cost-effectiveness estimates is related more to differences in study methods than to differences in national context. Multinational studies were associated with much lower country-level variation than single-country studies. These findings are for a single clinical question and may be atypical.
Authors: Scott Ramsey; Richard Willke; Andrew Briggs; Ruth Brown; Martin Buxton; Anita Chawla; John Cook; Henry Glick; Bengt Liljas; Diana Petitti; Shelby Reed Journal: Value Health Date: 2005 Sep-Oct Impact factor: 5.725
Authors: Veerle M H Coupé; Cindy Veenhof; Maurits W van Tulder; Joost Dekker; Johannes W J Bijlsma; Cornelia H M Van den Ende Journal: Ann Rheum Dis Date: 2006-07-31 Impact factor: 19.103
Authors: Massimo Brunetti; Ian Shemilt; Silvia Pregno; Luke Vale; Andrew D Oxman; Joanne Lord; Jane Sisk; Francis Ruiz; Suzanne Hill; Gordon H Guyatt; Roman Jaeschke; Mark Helfand; Robin Harbour; Marina Davoli; Laura Amato; Alessandro Liberati; Holger J Schünemann Journal: J Clin Epidemiol Date: 2012-08-03 Impact factor: 6.437
Authors: Anna Joy Rogers; Nathaniel G Rogers; Meredith L Kilgore; Akila Subramaniam; Lorie M Harper Journal: Value Health Date: 2016-11-11 Impact factor: 5.725
Authors: Christian E H Boehler; Gimon de Graaf; Lotte Steuten; Yaling Yang; Fabienne Abadie Journal: BMC Med Inform Decis Mak Date: 2015-09-04 Impact factor: 2.796
Authors: T I Armina Padmasawitri; Gerardus W Frederix; Bachti Alisjahbana; Olaf Klungel; Anke M Hövels Journal: PLoS One Date: 2018-05-09 Impact factor: 3.240