The EuroQol 5-dimension (EQ-5D) instrument is a 5-item questionnaire that gives a simple summary of treatment outcomes in terms of the patient’s perception of health-related quality of life. It is designed to measure the impact of treatment on the patient’s mobility, self-care, usual activities, pain and anxiety/depression. The original version of EQ-5D (known as EQ-5D-3L) has only three levels in each dimension (no difficulties; some difficulties; extreme difficulties). A redesigned version (known as EQ-5D-5L) now exists, with improved question wording and five levels of response (no difficulties; slight difficulties; moderate difficulties; severe difficulties; extreme difficulties).
Following the redesign, decision-making bodies like the National Institute for Health and Care Excellence (NICE) are now faced with evidence on the effectiveness of treatment in two different forms, based on EQ-5D-3L and EQ-5D-5L. It is not feasible to demand that drug companies revisit the very large number of past NICE decisions based on old (3L) measurement of treatment outcomes and re-run those clinical trials using the newer measure. Equally, it seems unwise to decline the improvements in measurement offered by the new (5L) version – which drug companies are, in any case, already starting to use.
If NICE is to be able to make consistent decisions taking account of all available evidence, it needs to have a way of converting evidence measured on a 3L basis to a 5L basis and vice versa. This process of conversion is known as statistical mapping from one type of evidence to another. It should be emphasised that this type of statistical mapping is used only to help NICE make decisions about which drugs (or other healthcare technologies) should be recognised as cost-effective overall. It is not an “algorithm” that determines which treatment is offered to which patient – that remains a decision for individual patients and their clinicians.
To carry out statistical mapping, we need so-called multi-instrument survey data in which individuals report the state of their health outcomes in both EQ-5D-3L and EQ-5D-5L terms. We can then use statistical modelling of data from the multi-instrument survey to predict the unmeasured EQ-5D-3L in clinical trials that only use EQ-5D-5L questionnaire, or to predict the unmeasured EQ-5D-5L in trials that only use the EQ-5D-3L questionnaire.
This project has three stages:
Stage 1 design and implementation of a randomised experiment carried in a large national household survey (known as the Understanding Society Innovation Panel and covering almost 3,000 adults), to investigate the best way to design a multi-instrument survey.
Stage 2 design and implementation of a large-scale (50,000 individuals) multi-instrument online survey.
Stage 3 statistical modelling of the multi-instrument data and associated computer software that can be used in practice by cost-effectiveness analysts for mapping between the old EQ-5D-3L and newer EQ-5D-5L outcome measures.
Stage 1 of the project was completed and written up by March 2020. Stage 2 was completed, with fieldwork carried out in April-May 2020 by the OnePoll survey organisation. An interim report on Stage 3 was submitted to NICE and the Department of Health and Social Care June 2020, but statistical work remains in progress.