Making decisions under uncertainty—the role of probabilistic decision modelling
Centre for Health Economics, University of York York YO10 5DD, UK
Correspondence to: Mark Sculpher, Email: mjs23{at}york.ac.uk
The phenomenon of explicit priority-setting in health care is international and cuts across very different types of health care systems and levels of health care spending. In the context of increasingly stretched resources, economic evaluation in health care is becoming a central component of formal decision-making procedures; in particular, in those countries with a longer tradition in health technology assessment (HTA) such as Australia, Canada, UK, The Netherlands and the Scandinavian countries. In reality, it is unlikely that explicit rationing would ever replace implicit rationing entirely, being as they are the two ends of a continuum, and a certain degree of professional discretion is inevitable at the point of service delivery, with doctors always likely to play a role in resource allocation decisions. However, it is the extent to which a society wishes to move along this continuum that is important, being a political and not just a technical decision that implies making value judgements on national objectives. To the extent that resource allocation is supported by the available evidence on cost-effectiveness produced by independent expert agencies following formal and transparent processes, citizens can feel reassured that the adoption of such decisions represent value for money for collective resources.
Although the use of randomised controlled trials as a source of evidence for decision-making is widely accepted, they are not sufficient to address the question of whether a technology is cost-effective. Only decision modelling can characterise and synthesise a specific problem where a decision is required, whatever the limitations of the evidence. This task inevitably demands a number of requirements that phase III trials are normally unable to satisfy: firstly, the inclusion of all relevant comparisons in the analysis, rather than the use of a single arbitrary option or placebo; secondly, trials tend to use surrogate outcome measures rather than measures of absolute gain in health, such as life years gained or quality-adjusted life years (QALYs); thirdly, trials also tend to be based on protocol, which restricts the generalisability of their results and thus their value in making decisions about resource allocation; finally, clinical trials normally have a short follow-up, whereas the aim of economic evaluation concerns estimates of the total length of time in which costs and health outcomes are expected to differ between interventions.
The last point, the appropriateness of the time-horizon, is particularly relevant when technologies such as screening and diagnostic tests, and chronic conditions in general, are the object of analysis; hence, an extrapolation of short-term effectiveness measures is needed. Screening has the potential to save lives or improve quality-of-life through early diagnosis of serious conditions, reducing the risk of developing a condition or its complications, but the negative long-term consequences for those cases not detected or incorrectly identified should also be accounted for in any economic analysis. The article published in this journal by J. Thompson et al.1 presents a cost-effectiveness analysis of case finding for hepatitis C in primary care, with a model structure that combines a short-term decision tree and a Markov model to extrapolate the long-term consequences of hepatitis C infection for those people who were not identified for testing through case finding.
Needless to say, establishing the cost-effectiveness of a health care intervention is a complex and challenging task, which requires a few fundamental issues to be tackled such as how should the inevitable uncertainty in the data be incorporated in the analysis, how to synthesise data from a variety of sources, how the resource allocation decision should be taken in the context of uncertainty and whether more research should be funded to support future decisions. This complexity means that models must operate within an integrated analytical framework for decision making, which can handle parameter and decision uncertainty appropriately; thus, it should be based around probabilistic decision modelling.2,3
Probabilistic sensitivity analysis (PSA) allows the uncertainty in the individual parameters, estimated from available evidence, to be fully characterised using probability distributions to reflect their imprecision, and propagated through the model using second-order Monte Carlo simulation. In this way, probabilistic models allow the joint effect of parameter uncertainty across all input parameters in the model to be translated into uncertainty in the mean cost-effectiveness results.4 Although estimated with uncertainty, mean cost-effectiveness represents the best estimate of cost-effectiveness and adopting this approach to handle uncertainty does not only offer a technical but also a conceptual advantage over deterministic univariate sensitivity analysis, allowing a more intuitive interpretation of probability.5 The use of cost-effectiveness acceptability curves, similar to the ones presented by J. Thompson et al.1 allows for the presentation of decision uncertainty in terms of the probability that a technology is considered cost-effective for a given maximum willingness to pay on the part of the decision maker.6 One minus this probability reflects the decision uncertainty around adoption, that is, the chance that in adopting the technology the ‘wrong’ decision would have been made. The authors show in their analysis that the error probability that case finding for hepatitis C in general practice would be cost-effective is 27% for a population based strategy approach, and 23% for the targeted strategy.
The latest methodological guidelines on technology appraisal published by the National Institute for Health and Clinical Excellence (NICE) recommend as an important feature of cost-effectiveness analyses the use of PSA.7 In the United Kingdom, NICE is an example of an independent agency which undertakes appraisals of mainly new technologies at the request of the Department of Health and the Welsh Assembly Government. Many other independent HTA organisations are currently operating in the OECD, evaluating new and/or existing health care technologies and contributing to the governmental task on evidence-based policy and priority setting.8 At the European level, some important collaborative projects have been undertaken so far with the aim of harmonising the methodology for assessment and promoting European co-operation in the HTA field, such as the Eur-Assess and the ECHTA/ECAHI project. Economic evaluation is here to stay, and its continuous methodological development and growing literature prove the strength of this young discipline. In this context, it is increasingly important for clinicians to become more familiar with economic evaluation methods and the interpretation of cost-effectiveness results. The involvement of clinical experts on multidisciplinary teams undertaking economic evaluations of health care technologies for independent HTA agencies and for technology manufacturers, and the increased interest in cost-effectiveness analysis in some clinical areas such as cardiology and oncology proves that a proportion of the medical profession has already taken the lead.
Notes
Bravo Y and Sculpher M. Making decisions under uncertainty—the role of probabilistic decision modelling. Family Practice 2006; 23: 391–392.
References
1 Thompson Coon J, Castelnuovo E, Pitt M, Cramp M, Siebert U, Stein K. (2006) Case finding for hepatitis C in primary care: a cost utility analysis. Family Practice 23:393–406.
2 Sculpher M and Claxton K. (2005) Establishing the cost-effectiveness of new pharmaceuticals under conditions of uncertainty—When is there sufficient evidence? Value Health 8:433–446.[CrossRef][ISI][Medline]
3 Claxton K, Sculpher M, Drummond M. (2002) A rational framework for decision making by the National Institute for Clinical Excellence (NICE). Lancet 360:711–715.[CrossRef][ISI][Medline]
4 Briggs A. (2000) Handling uncertainty in cost-effectiveness models. Pharmacoeconomics 17:479–500.[CrossRef][ISI][Medline]
5 Briggs A, Goeree R, Blackhouse G, O'Brien B. (2002) Probabilistic analysis of cost-effectiveness models: choosing between treatment strategies for gastroesophageal reflux disease. Med Decis Making 22:290–308.[Abstract]
6 Fenwick L, Claxton K, Sculpher M. (2002) Representing uncertainty: the role of cost-effectiveness acceptability curves. Health Econ 10:779–789.
7 National Institute for Clinical Excellence. (2004) Guide to the Methods of Technology Appraisal (NICE, London).
8 International Network of Agencies for Health Technology Assessment (INAHTA). Available at: www.inahta.org.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||