City Research Online

Abstract

This thesis is concerned with two specific cost measurement issues commonly raised within the context of economic evaluation of health care interventions. Both these issues arise due to limited availability of cost information from medical studies. The first relates to the process of cost data collection while the second relates to the statistical analysis of cost data. Investigation of this subject matter is undertaken with reference to clinical trials although this setting does not restrict the generality of the findings. The typical pattern of cost collection records volumes of resource use at the patient level but not resource unit cost in the treatment centre where the resource was utilised. The calculation of treatment cost then is normally based on some average unit cost estimate obtained from a variety of sources as opposed to centre specific unit cost information. The question arises as to whether the source of unit cost information has an impact on the calculated total treatment costs. This is addressed using a simulation setting and assuming specific underlying production and cost relations which determine the behaviour of treatment centres in delivering a health outcome. The results show that assuming the treatment centres operate in a manner consistent with economic theory, using average instead of centre specific unit cost information will lead to biased estimates of the total cost of treatment. The issue of primary concern in the thesis relates to the incompleteness of cost infonnation for analysis due to censoring. Censoring occurs whenever patients are not observed for the full time to event and affects both effectiveness and cost data. Any analysis of such data that fails to account for the presence of censoring will result in biased estimates of the statistics of interest. This issue has only recently been addressed in the literature within the context of cost analysis and a well established methodology for dealing with this problem is lacking. There are a limited number of parametric and non- parametric estimators which have been proposed in an attempt to adjust cost estimates for censoring all of which are considered here. A subset of those lack theoretical justification and as such lead to erroneous inferences, while those whose use is justified on theoretical grounds have not been empirically assessed under conditions of heavy censoring using real medical data. This is undertaken in the present analysis using a clinical trial dataset which displays extreme levels of censoring. Although the theoretical investigation shows that under specific assumptions the approaches provide consistent estimators of mean cost while accounting for the loss of information due to censoring, the analysis reveals various performance patterns ranging from generally stable estimators under the conditions considered to estimators which become increasingly unstable with increasing levels of censoring.

Publication Type:	Thesis (Doctoral)
Subjects:	H Social Sciences > HB Economic Theory H Social Sciences > HC Economic History and Conditions
Departments:	Bayes Business School Bayes Business School > Bayes Business School Doctoral Theses Doctoral Theses

Export

Downloads