Title: Uncertainty and value of information when allocating resources within and between different health care programmes: A mathematical programming approach
Speaker: Dr Claire McKenna, Research Fellow, Team for Economic Evaluation and Health Technology Assessment
Abstract: Standard cost-effectiveness decision rules compare the additional cost per extra unit of effect (ICER) of the more effective treatment option to some ‘notional’ threshold value that the decision maker is willing to pay for additional health benefits. Although this threshold should be based on some informal assessment of the shadow price of the budget constraint, its true value is unknown and it may not be consistent with an overall fixed budget for healthcare. Therefore, the adoption decision can lead to a suboptimal allocation of resources and the opportunity costs from displacing other unrelated programmes at the margin are not identified. Mathematical programming offers an alternative solution by solving the allocation problem as a whole. We present our proposed two-stage stochastic mathematical programming (SMP) formulation to optimally allocate resources within and between multiple healthcare programmes. The formulation incorporates uncertainty and variability within the allocation system and calculates the expected value of acquiring additional evidence to resolve all the parameter uncertainties. We compare the results from the SMP to traditional methods using standard decision rules.
Title: Expected value of sample information calculations for cluster randomised trials and mixed treatment comparisons
Speaker: Dr Nicky Welton (MRC Health Services Research Collaboration, Department of Social Medicine, University of Bristol)
Abstract: Expected value of sample information (EVSI) methods provide a framework under which to assess the expected health gains from new research given the study design (e.g. sample size). The optimal study design is that which maximises the Expected net benefit of sampling (ENBS). EVSI calculations present many technical challenges including: allowing for correlations between regression coefficients; non-linear relationship between the regression coefficients and the net benefit function; random effects models on the baseline; and the per-capita baseline effect for use in a decision analysis. We outline possible solutions to these problems, including Bayesian multivariate normal updating, Taylor approximation, and conditioning on the baseline. We illustrate the methods with two examples: a mixed treatment comparison for treatments for bipolar disorders and a cluster-randomised trial of interventions to increase attendance at breast cancer screening. We discuss the potential for and barriers to EVSI methods being used in practise.
Title: Do Italian cardiologists care about economic evidence? A discrete choice experiment
Speaker: Aleksandra Torbica. Research Fellow, Centre for Research in Healthcare and Social Management, CERGAS, Boconni University, Italy. Aleksandra was awarded an Alan Williams Health Economics Fellowship and is visiting the Centre for Health Economics, June-September 2007.
Abstract: The role of professionals is essential in evaluating novel treatment strategies in healthcare. In making their judgments professionals use different types of evidence and rely on various sources of information. It has been questioned whether the results of formal economic evaluation analyses influence their decision making process. The majority of studies investigating the impact of economic evidence on clinical decision making limited their findings to a description of clinicians’ views and perceptions, without providing any measure of the relative importance they attach to different types of evidence. The purpose of the study was to elicit Italian cardiologists’ preferences for different types of evidence in assessing new treatment strategies through discrete choice experiments (DCE) framework. Data were collected through a self-completed questionnaire administered during the Italian congress of hospital cardiologists (ANMCO). Results indicate that cardiologists consider economic evidence (i.e. cost-effectiveness) to be an important ingredient in their decision making, in addition to other attributes (quality of clinical evidence, size of health gain). Economic evidence is valued more highly among clinicians with higher level of self-assessed knowledge of economic evaluation techniques and among younger professionals (age< 45). Overall, a DCE framework appears to be a feasible methodology for eliciting clinicians’ decision making criteria.
Title: Exploring the social value of health care interventions
Speaker: Colin Green, Senior Lecturer in Health Economics, Peninsula Technology Assessment Group.
Abstract: When making decisions about the availability of health care interventions, in the UK NHS, what do decision makers take into account? It would not be surprising for decision makers to say they consider efficiency (or cost effectiveness), and it would not be surprising to have decision makers talk about equity arguments. But what would they mean by equity? And how would equity be taken into account. Equity trade-offs are rarely presented by decision makers. In this seminar I will talk about the findings from a discrete choice study that looked at social preferences, using interviews in a sample of the general public, to explore the relative importance of a range of competing social values. I will talk on the empirical literature reviewed to inform the study, and how findings from such studies might be used by decision makers. The research findings are used to consider what it is that might be meant by equity in a health technology appraisal setting.
Title: Economic evaluation of health technologies in Italy: current practices and future trends
Speaker1: Aleksandra Torbica (Economic Evaluation Analysis Department, Bocconi University, Italy. Aleksandra is visitor to CHE from June to September 2007)
Abstract: Over the past two decades the research in the field of economic evaluation analyses has exponentially grown in Italy, in terms of number of studies produced and initiatives promoted at national and regional level. The increasing research activity, however, still contrasts with the lack of interest by policy makers to use these studies within an institutional framework. In my seminar I will present the main research topics investigated by Unit for Economic Evaluation Analysis at CERGAS, Bocconi University. I will discuss some of the peculiarities of our approach both in methodological and applied research, including the contribution of management disciplines (eg. cost accounting, organization design) in economic evaluation analysis, and the necessity to acknowledge the complexity of decision making levels within a decentralized healthcare system. For illustrative purpose, I will present an example of our applied research: an observational, prospective, multi-centre study on costs and outcomes of alternative treatment strategies for cerebral aneurysm in Italy (endovascular-coiling vs surgical-clipping approach).
Title: Comparability of costing strategies to generalise multinational trial data across jurisdictions
Speaker 2: Marielle van der Burgt (Radboud University Medical Centre, Nijmegen, Netherlands. Marielle is a visitor to CHE from February to August 2007)
Abstract: When country-specific health economic data is not available, data from other settings may be used to inform local decision-making. This is accompanied by questions about their generalisibility, the extent to which cost results of a study, as they apply to a particular setting, hold true for another setting. This study explores factors responsible for between-setting differences in costs and resource use, and methods by which these differences can be quantified. As an illustration, the study utilises US based resource use data from an ongoing trial concerning management of end-stage HIV patients to approximate costs for the United Kingdom and the Netherlands.
Title: Pragmatic method for assessing the potential impact of a proposed clinical trial
Speaker: Dr Jonathan Michaels, Professor of Vascular Surgery, Academic Vascular Unit, University of Sheffield
Abstract: When large research trials are examined in technology appraisal or systematic literature reviews they often fail to answer the important issues in regard to clinical or cost-effectiveness. Although the protocols of proposed trials usually consider in detail the statistical analysis plan and power calculations, the justification for the choice of a relevant effect size is often sketchy. Whilst value of information analysis can provide an estimate of the expected value of clinical research, and assist in the estimation of relevant effect size, it has not been widely adopted in decision making regarding the commissioning or funding of such research. This may in part reflect the complexity of the modelling and a lack of understanding of the techniques that are used.
This paper proposes a simplified and transparent method for tabulating the plausible ranges of effect size and cost implications in order to provide an indication of the potential net benefit of a research study. The method requires that those proposing the research put forward suggestions for the expected treatment effect and incremental cost, with a 'best guess', likely values and a plausible range, and to estimate from these a range of net benefits. By also estimating the potential population, these figures can also be used to predict the overall potential benefit to the NHS in order to assist decision makers in assessing the potential value of the proposed study.
Title: Incorporating equity weights into cost-effectiveness analyses - opening Pandora's box?
Speakers: Dr Aki Tsuchiya (Sheffield Health Economics Group, University of Sheffield), Christopher McCabe (University of Warwick) Allan Wailoo (University of Sheffield) Jon Karnon (Univerity of Sheffield)
Abstract: Historically, health care decision makers only considered issues of safety and efficacy when deciding whether a therapy should be available or not. Over the last two decades, the increasing cost of health care has led decision makers to consider issues relating to the efficiency of alternative uses of their limited health care budgets. Current practice is to assign equal social value to a unit of health improvement ("a QALY is a QALY is a QALY"). However, there is now substantial evidence suggesting that members of the public do not hold this view. Against this background it has become increasingly clear that decision makers need to consider issues of equity alongside that of efficiency. Amendments to the QALY model have been proposed which would allow the incorporation of explicit differential societal preferences for health gain to different groups within society - so called 'equity weights'. To date, research on equity weights has focussed on candidate criteria for equity weights and methods for estimating such weights. It has implicitly assumed that should legitimate, valid, and reliable equity weights become available, they would be incorporated through a simple process of re-weighting the denominator in the ICER. In the first part of this paper we explain why such an approach is likely to be appropriate in only a handful of cases and in many more cases it would lead to a misrepresentation of society's preferences for health and equity. The second part of the paper examines the hidden equity weights in current practice, and highlights some implications adopting an explicit 'disaggregated equity' strategy in cost-effectiveness analyses.
Title: Synthesising evidence from summary statistics with Individual Patient Data
Speaker: Gerry Richardson (Senior Research Fellow, Team for Economic Evaluation and Health Technology Assessment, University of York)
Abstract: The 'Expert Patients Programme' (EPP) aims to provide self care support to any individual with a chronic condition in England, and a recent trial based cost-effectiveness analysis (CEA) demonstrated that it was likely to be cost-effective at commonly expressed threshold values for a QALY.
However, it has been argued that the use of single trial based CEA may be of limited use for decision makers and that other relevant information should be incorporated into the analysis. This workshop will describe the methods, results and implications of synthesising evidence from other sources with the data from the national EPP evaluation.
The workshop has three main aims:
Title: Reducing emotional distress in people caring for patients receiving specialist palliative care: A cost-effectiveness analysis.
Speaker: Gerhart Knerer (Senior Medical Statistician, MRC Clinical Trials Unit, London)
Abstract: Background: Caring for relatives with advanced cancer may cause psychological and physical ill health.
Aims: To assess the cost effectiveness of support from a carer advisor + specialist palliative care (CA) compared with conventional specialist palliative care (SPC) alone.
Methods: Cost-effectiveness analysis based on randomised controlled trial. A total of 271 informal carers were randomised to CA (n=138) or SPC (n=133). The intervention comprised six weekly visits by a trained advisor. Health outcomes expressed in terms of Quality Adjusted Life Weeks (QALWs) based on subject’s responses to the EQ-5D at baseline and at three points up to 12 weeks’ follow up. Twelve week costs estimated from NHS perspective. Cost-effectiveness was assessed using the net monetary benefit (NMB) regression approach and likelihood support intervals.Results: The mean QALWs per subject were lower for the CA group compared to SPC (coefficients from the effect regression [note: the NMB regression with lambda tends to infinity] are reported). Unadjusted mean Quality-Adjusted-Life-Weeks (QALWs) scores for the CA group were 3.09, approximately 0.076 less than the SPC group (95% CI, -0.246 to 0.398). Covariate adjustment (e.g. baseline costs, baseline EQ-5D scores and period of assessment [i.e. 4 week, 9 week or 12 week]) did not alter the substantive result (differential QALW: -0.191, 95% CI ; -0.058 to 0.440). Unadjusted mean costs (when lambda=0, NMB= -Cost) per CA case were £1287, approximately £190 more than the SPC group (95% CI, -65.7 to 448.5). Adjusting for covariates did not alter the substantive result (differential mean cost: £159.8 (95% CI, -148.3 to 467.9). Net benefit analysis and likelihood support intervals indicated that the intervention was not cost-effective. The probability that caregiver support plus specialist palliative care is cost effective was below 15% for a large range of values of willingness to pay for an additional QALW.Conclusions: Carer advisor + specialist palliative care is not cost-effective relative to specialist palliative care alone.
Title: Exchanging SF-12 for EQ-5D : caveat emptor
Speaker: Ling-Hsiang (PhD student, Outcomes Research Group, Centre for Health Economics, University of York) and Professor Paul Kind (Director of Outcomes Research Group, Centre for Health Economics, University of York)
Abstract: There is a relatively rich choice of alternative models that allow for the representation of SF-36/SF-12 data as a single index score. In such circumstances the user acts on the basis of a belief that the model can be generalised to any and all situations. This generous assumption needs to be confronted. Thus, the present study demonstrates possible modelling strategies for converting SF-12 responses into an index based on EQ-5D utility scores and examines the error of the estimation (the difference between observed and estimated EQ-5D index scores) generated by various mapping methods. The reliability of replacing missing EQ-5D index scores with estimated scores is also tested. Applying the data from the 2003 US Medical Expenditure Panel Survey, the results suggest that the choice of modelling option is largely a matter of indifference if mean EQ-5D index values are being estimated. There is a degree of variability in the results achieved in these different models but their capacity to represent a 'group' average seems to hold good. However, it must be pointed out that the degree of error increases in older and less healthy groups in every model. The range of error in estimating EQ-5D for a lone patient can be relatively large, sufficient to lead to the misclassification of treatment responses.Given the inherent danger of such misrepresentation, perhaps the best course of action would simply be to include EQ-5D in all clinical studies.