Electronic Thesis and Dissertation Repository


Master of Science


Epidemiology and Biostatistics


Dr. William Hodge


Background and objective: For cost-effectiveness analyses (CEA) of glaucoma interventions to be of use they require valid and accurate cost and effectiveness data. Costs remain understudied relative to effectiveness. The impact of cost estimation methods on resultant estimates is unknown in glaucoma. Direct measurement of costs is labour-intensive and expensive. Decision-analytic modelling of costs using literature sources, expert opinion, institutional experience and assumptions provides a quicker, less laborious alternative to empirical costing. A lack of long-term effectiveness data in chronic diseases like glaucoma means that modelling is widespread and inevitable, both for CEAs and budget impact projections. The same problem precludes validation of models and there are concerns about their validity and possible arbitrariness given the discretionary nature of their construction. In this thesis we investigate whether costs from a decision-analytic model of repeat laser trabeculoplasty among glaucoma patients provide a valid alternative to direct measurement of costs alongside an effectiveness trial. Secondary aims were to compare the ministry and societal perspective and to identify main cost drivers for repeat laser trabeculoplasty.

Methods: Trial-based costing was conducted as part of an effectiveness trial comparing argon- and selective-laser trabeculoplasty (ALT and SLT) after previous SLT among glaucoma patients at an ophthalmologic clinic in Ontario. For model-based costing a decision tree was formulated and populated with parameter estimates based on previous literature supplemented with assumptions. Mean trial and model cost were compared for ALT and SLT.

Results: Model and trial cost estimates differed minimally for the ministry perspective (4% and 8% for SLT and ALT respectively) – this in spite of large differences in modelled and observed parameter values. Labour accounted for almost 90% of total cost. Model and trial costs were also similar for the societal perspective (8% and 1% for ALT and SLT), although there was more sensitivity to assumptions about patient time loss. Indirect and patient costs were at least as large as direct medical costs. Our results indicate that modelled costs are an acceptable substitute for directly measured costs for some clinical scenarios – glaucoma interventions in Ontario possibly being such a case.