An investigation of cost variation across health care settings and the implications for economic evaluation
This thesis is concerned with the estimation of costs in economic evaluation. The thesis reviews the theoretical and applied literature on costing and highlights that studies generally ignore cost variation across health care settings. The thesis aims to assess why costs vary across health care settings, and the implications for economic evaluations. The study uses microeconomic theory to pose hypotheses for cost variation across health care settings and uses a consistent methodology to collect costs across a range of health care settings. The analysis uses multilevel models (MLMs) to test hypotheses concerning cost variation. Statistical theory suggests that MLMs accommodate the hierarchical structure of the data and may therefore be more appropriate than ordinary least squares (OLS) models for identifying reasons for cost variation across settings. The use of MLMs and OLS models for analysing reasons for cost variation are compared. The OLS models find that both patient and higher-level covariates are associated with length of hospital stay (LOS) and total cost, but these models overestimate the precision of the higher-level variables. By contrast, the MLMs show that none of the higher-level variables are associated with LOS, and the national level of spending on health care is the only higher-level variable associated with total cost. The empirical investigation also illustrates that using OLS regression analysis to report cost-effectiveness can lead to inaccurate estimates. By contrast, the MLMs recognise the structure of the data and accurately quantify mean incremental cost- effectiveness and the associated levels of uncertainty. The thesis concludes that ignoring cost variation across health care settings can lead to inaccurate estimates of cost and cost-effectiveness. Basing decision-making on inaccurate information can move the allocation of health care resources away from the target of allocative efficiency. This thesis presents a methodology for improving the conduct of cost analyses that future economic evaluations can adopt.