Use this URL to cite or link to this record in EThOS:
Title: Temporal uncertainty in cost-effectiveness decision models : methods to address the uncertainties that arise when the appropriate analysis time horizon exceeds the evidence time horizon in cost-effectiveness decision models as applied to healthcare interventions
Author: Mahon, Ronan
ISNI:       0000 0004 5347 3493
Awarding Body: University of York
Current Institution: University of York
Date of Award: 2014
Availability of Full Text:
Access from EThOS:
Access from Institution:
The problem of predicting outcomes over time and expressing uncertainty about the future is one common to many scientific disciplines. For cost-effectiveness analysis used to aid resource allocation decisions in healthcare, this problem presents itself in the form of a disparity between the evidence time horizon (which is typically short-term) and the appropriate analysis time horizon (which is often long-term). To date, this problem has been primarily characterised as one of a need to extrapolate, i.e. an imperative to interpret the available short-term evidence and project this into the long-term in order to plug the evidence gap. Furthermore, the issue has been strongly associated with estimations of survival, but less so with other measures of disease progression, with estimates of cost, or with estimates of health-related quality of life. This thesis strives to take a broad and thoughtful approach to examining the general problem of a dearth of evidence pertaining to the long-term. It is argued that this problem is most accurately and most usefully thought of as one of uncertainty. As such, in this thesis, the term ‘temporal uncertainty’ is employed. Consideration is given to the nature of temporal uncertainty and when it is of significance in the context of decision making with evidence development. Where a full expression of temporal uncertainty is necessary in order to make an informed decision, a number of approaches are described and appraised. Caution is advised in relation to extrapolating evidence over time due to the implicit assumption that outcomes in the short-term are good predictors of outcomes in the long-term. It is recommended that temporal uncertainty be characterised by a single uncertain ‘temporal’ parameter and incorporated into a probabilistic analysis in order to provide a true estimate of expected cost-effectiveness and to estimate the value of obtaining information that would lessen temporal uncertainty. In the context of these principles, a review of the health technology assessment (HTA) literature reveals that approaches to addressing temporal uncertainty to date have been inconsistent and largely inadequate. The review also makes apparent the full range of model parameters that are regularly exposed to temporal uncertainty and the specific analytical challenges that must be overcome. A motivating example (the RITA-3 decision model) is employed in order to develop and apply methods that appropriately quantify temporal uncertainty for a range of model parameters given the available evidence. The motivating example also facilitates an examination of the effects of expressing temporal uncertainty throughout a decision model. It is found that the replacement of ‘conservative’ temporal assumptions with expressions of temporal uncertainty alters the adoption recommendation for several of the risk groups under examination, that overall uncertainty around costs and health benefits is greatly inflated, that there is likely to be value in obtaining further information specifically in relation to the long-term temporal nature of certain model parameters and that there may also be value in ‘waiting’ for further evidence to be revealed if there is the potential for significant irrecoverable costs to be incurred. In summary, this thesis represents a contribution to the development of methods to aid decision making in healthcare. In particular, the significant issue of temporal uncertainty is expounded and methods to appropriately address temporal uncertainty are developed and demonstrated.
Supervisor: Manca, Andrea Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available