Expected Value of Information and Synthesis Methods for Research Prioritisation and Study Design
Lead Research Organisation:
University of Bristol
Department Name: Community-Based Medicine
Abstract
At present there is a gap between the way decisions are made about whether to adopt new treatments for use in the NHS, and the way decisions are made about whether more research is needed on these treatments. As a result, resources may be wasted researching treatments that were never likely to be cost-effective, or by adopting treatments which, if more evidence were collected, would not be cost-effective. It is possible to calculate the probability that a decision to adopt a treatment on the basis of current evidence would turn out to be the wrong decision if more evidence was collected, and how much resource waste and loss of quality of life this is likely to represent. These calculations provide an objective and fair method for informing the decision of which studies to fund, and how best to design those studies. The methods are, however, complex. This project aims to find ways of making the calculations easier so that they can be used more widely to help research funding bodies commission research that will help make better treatment decisions for patients in the NHS.
Technical Summary
The National Institute for Health and Clinical Excellence (NICE) and similar bodies worldwide use formal modelling techniques to answer policy questions about whether to adopt a new technology for routine use. But a series of further policy issues then arise that are relevant to research commissioning bodies such as the National Institute for Health Research (NIHR) Health Technology Assessment Programme (NCCHTA) and the MRC: (a) is more evidence required to support the adoption of this technology (b) which model inputs in particular should we collect evidence on? (c) what research designs deliver the best societal pay-off from research? (d) what is the optimal sequence of multiple new studies? These questions can be addressed using Expected Value of Information (EVI) methods.
While the principles of EVI methods are well understood, applications in health and medicine face technical challenges before they can be routinely implemented. These include: correlated model inputs; non-linear decision models; complex relationships between evidence and model inputs; heterogeneity in baseline characteristics, subgroups, and model uncertainty; biomarkers and biased evidence; optimisation over large design-spaces and optimal sequences of research. Changes in the policy environment, leading to coverage with evidence and risk-sharing schemes raise further technical challenges.
The aim of this proposal is to develop methods that address these technical challenges for EVI computation that are practical within the time constraints faced by NICE, NIHR NCCHTA, and those designing trials, and sensitive to the developing policy context. To achieve this, a series of carefully chosen worked illustrations will be assembled, covering the most important and commonly encountered evidence structures and cost-effectiveness models. The examples will include the design of randomised controlled trials, epidemiological studies, drug-development studies, use of biomarkers, and design of research portfolios.
EVI analysis depends on the validity of underlying systematic review, evidence synthesis, and cost-effectiveness analysis, and has implications for all of these disciplines. The training element is designed to broaden my knowledge of these disciplines, and will expose me to policy considerations at National level surrounding the use of these methods in Health Technology Assessment. Thus ensuring the policy-relevance of the research, and facilitating future research across the whole EVI agenda.
Research findings will be disseminated through publications and presentations at international conferences. Wherever possible generic methods to prioritise and design new trials/studies will be made available for use in standard software. The worked exemplars and methods will form the basis of short courses and a book.
400 words
While the principles of EVI methods are well understood, applications in health and medicine face technical challenges before they can be routinely implemented. These include: correlated model inputs; non-linear decision models; complex relationships between evidence and model inputs; heterogeneity in baseline characteristics, subgroups, and model uncertainty; biomarkers and biased evidence; optimisation over large design-spaces and optimal sequences of research. Changes in the policy environment, leading to coverage with evidence and risk-sharing schemes raise further technical challenges.
The aim of this proposal is to develop methods that address these technical challenges for EVI computation that are practical within the time constraints faced by NICE, NIHR NCCHTA, and those designing trials, and sensitive to the developing policy context. To achieve this, a series of carefully chosen worked illustrations will be assembled, covering the most important and commonly encountered evidence structures and cost-effectiveness models. The examples will include the design of randomised controlled trials, epidemiological studies, drug-development studies, use of biomarkers, and design of research portfolios.
EVI analysis depends on the validity of underlying systematic review, evidence synthesis, and cost-effectiveness analysis, and has implications for all of these disciplines. The training element is designed to broaden my knowledge of these disciplines, and will expose me to policy considerations at National level surrounding the use of these methods in Health Technology Assessment. Thus ensuring the policy-relevance of the research, and facilitating future research across the whole EVI agenda.
Research findings will be disseminated through publications and presentations at international conferences. Wherever possible generic methods to prioritise and design new trials/studies will be made available for use in standard software. The worked exemplars and methods will form the basis of short courses and a book.
400 words