Health policy-makers use cost-effectiveness analyses (CEA) to decide which interventions should be provided. For CEA to deliver accurate results, the use of appropriate statistical methods is essential. Much recent progress has been made in developing appropriate statistical approaches for CEA. This theme aims to encourage development, and dissemination of statistical methods for CEA, and to stimulate debate. Within the overall theme, the following areas are of specific interest to researchers at LSHTM.
CEA that use patient-level data from randomised controlled trials (RCTs)
Significant advances have been made in statistical methods for CEA that use data from RCTs where individual patients are randomised.1 These include Bayesian regression models that have the flexibility required to recognise the correlation between costs and outcomes, imbalances in baseline covariates, and skewed costs.2 Methods have been proposed for multicentre studies (national and international) where it can be important to recognise the data hierarchy, and subgroup analyses.2-4 Other areas of general interest include extending methods for handling missing5 and censored data to the CEA context.6
CEA that use data from cluster randomised trials (CRT)
CRTs provide important evidence for evaluating the relative cost-effectiveness of alternative health service and public health interventions as well as different clinical guidelines. CRTs raise analytical issues for CEA; costs in particular may be more similar within, rather than between, clusters. CEA should recognise both that costs and outcomes are correlated and that individuals are clustered within settings. However, almost all CEA based on CRTs currently use standard regression methods that make the implausible assumption that observations are independent.7 This is a concern whether the CEA are based solely on CRT data or uses the trial estimates in a decision model. Statistical methods are required that recognise the clustering of individuals within settings. As part of an MRC funded study researchers at LSHTM in collaboration with colleagues at the MRC Biostatistics Unit, are developing multi-level models8 and bootstrap methods, specifically for CEA that use CRT data.
CEA using observational data
CEA commonly use observational studies, either alongside or instead of data from RCTs. A major concern is that the results suffer from treatment selection bias due to confounding. Commonly used analytical methods for dealing with selection bias such as regression analysis or propensity score (Pscore) matching can be highly sensitive to model specification.9 Researchers at LSTHM are undertaking methodological research investigating an alternative approaches including a matching method, Genetic Matching (GenMatch).10 This multivariate matching method extends Pscore matching by using an automated search algorithm to optimise balance on baseline covariates, given the data. In the context of CEA where a major concern is balancing baseline covariates, GenMatch can reduce bias compared to propensity score matching.11 Ongoing work is investigating alternative methods for bias reduction such as inverse probability of treatment weighting. This work is being undertaken in collaboration with the University of California at Berkeley and with colleagues at LSHTM working under the causal inference theme. A two-day workshop for disseminating the methods is planned for September 2011.
CEA using decision models
Ideally CEA report cost-effectiveness over the patients’ lifetime for a full range of competing alternatives. Here, a single RCT or observational study is insufficient. Method guidance recommends the use of decision-models to allow for a broader range of comparators, and to facilitate lifetime analyses. For CEA of interventions to treat or prevent infectious disease, standard decision models are insufficient, and researchers at LSHTM have developed more appropriate modelling approaches.12 LSHTM researchers have also developed approaches for validating and calibrating decision models.13-14 Work on calibrating models is in collaboration with the LSHTM centre for mathematical modelling of infectious disease.
1. Willan A, Briggs A. Statistical analysis of cost-effectiveness data. Wiley, Chichester, 2006.
2. Nixon R, Thompson SG. Incorporating covariate adjustment, subgroup analysis and between-centre differences into cost-effectiveness evaluations. Health Economics 2005;14:1217-29.
3. Manca A, Lambert PC, Sculpher M, Rice N. Cost-effectiveness analysis using data from multinational trials: the use of bivariate hierarchical modelling. Med Decis Making 2007;27:471-90.
4. Grieve R, Thompson S, Nixon R, Cairns J. Multilevel models for estimating incremental net benefits in multinational studies. Health Econ 2007;16:815-826.
5. Grieve R, Cairns J, Thompson SG. Improving costing methods in multicentre economic evaluation: the use of multiple imputation for unit costs Health Economics 2010 19(8): 939-954.
6. Willan AR, Lin DY, Cook RJ and Chen EB. Using inverse weighting in cost-effectiveness analysis with censored data. Statistical Methods in Medical Research, 11: 539-551.
7. Gomes M, Grieve R, Nixon R and Edmunds J. Statistical methods for cost-effectiveness analyses that use data from cluster randomised trials: a systematic review and checklist for critical appraisal. Medical Decision Making 2011 (in press).
8. Grieve R, Nixon R, Thompson SG (2010). Bayesian hierarchical models for cost-effectiveness analyses that use data from cluster randomized trials Medical Decision Making 30(2):163-75.
9. Rubin, DB Estimating causal effects from large data sets using propensity scores. Annals of Internal Medicine 1997, 127:757–-763.
10. Sekhon, J. Multivariate and propensity score matching with automated balance search. Journal of Statistical Software 2011 (In Press).
11. Sekhon, J. and Grieve, R. A Matching Method for Improving Covariate Balance in Cost-Effectiveness Analyses. Health Economics 2011 (In Press).
12. Brisson M, Edmunds WJ. Economic evaluation of vaccination programs: the impact of herd-immunity. Med Decis Making. 2003 Jan-Feb;23(1):76-82.
13. Calibrating models in economic evaluation: a seven-step approach. Vanni T, Katnon J, Mada J, White RG, Edmunds WJ, Foss AM, Legood R; Pharmacoeconomics, 2011; 29(1):35-49.
14. Kim LG, Thompson SG. Uncertainty and validation of health economic decision models. Health Economics 2010 19(1): 43-55.