Introduction
Health state values can be elicited through different methods, framed under certainty or uncertainty and based on scaling or choice [
1‐
3]. The most widely used methods are Standard Gamble (SG), Time Trade-Off (TTO), rating scale, including the visual analogue scale (VAS) as one of its variants, and Discrete Choice Experiment (DCE) [
1‐
3]. Different valuation methods have yielded different values for the same health state, and the relationship has been shown to be affected by the severity of the health state [
4,
5]. Transformation of VAS valuations to SG and TTO values have also been discussed in different studies, demonstrating the possibility of mapping of VAS scores to SG and TTO scores [
5‐
7].
Generally, choice-based methods, such as SG and TTO, are preferred by many health economists over rating scales, since choosing is considered to be a natural task for humans that is observable and verifiable [
1‐
3]. Rating scales are also claimed to be subject to measurement bias and to not possess interval scale properties [
1,
3]. However, other practical aspects weigh in favour of scaling methods, such as reduced amount of time required than for other methods as well as high response and completion rates [
1,
3].
The VAS has been used in the context of health and healthcare for various purposes; to measure symptoms (e.g. pain) or different domains of health (e.g. mobility), and to provide a single-index measure of health-related quality of life (HRQoL) [
1]. It is also used in economic evaluation as a valuation method, by directly asking individuals about their own health, or as a means of valuing health state classifications, including Quality of Well-being Scale (QWB), Health Utility Index (HUI), 15-D and EQ-5D [
1].
There are shortcomings with all valuation methods. While, as an example, challenges with using the SG and TTO for mild health states, temporary health states and for children’s health states have been shown in previous literature [
8‐
10], little is known about related concerns with VAS. There is a lack of a comprehensive overview of how the VAS is used in health state valuation, its purpose, design, benefits and drawbacks. This scoping review aims to describe how the VAS has been used for health state valuation in the published literature. Specifically, we address the following questions:
-
What are the overall purposes of using a VAS for health state valuation?
-
How have the health states valued using a VAS been defined?
-
What designs or forms of the VAS have been used for health state valuation?
-
What are the advantages and disadvantages of using a VAS for health state valuation mentioned in the identified articles?
Methods
Prior to data extraction the protocol was registered at the International prospective register of systematic reviews (PROSPERO) (number CRD42020210041). The review is presented following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews (PRISMA-ScR) checklist [
11].
The search for relevant studies was carried out in Medline (OVID), Web of Science (Clarivate) and PsycInfo (OVID), June 16th, 2020; with an updated search January 27th, 2022, no date limits were applied.
Two blocks of search terms were used: ‘visual analogue scale’ and ‘health state valuation’. A search strategy was developed by the authors and a librarian, were limited to the English language and studies published in peer reviewed journals. The index terms were Medical Subject Headings (MeSH). The full search strategy for each database is presented in Online Resource Table S1.
Selection criteria and screening process
Inclusion criteria were studies where a VAS was used for health state valuation, and studies where a specific dimension or disease stage were valued using VAS. It was not sufficient for a study to solely report on current health using a VAS. Exclusion criteria were studies where the endpoints for the VAS were specified for a specific condition or disease, if a comparison of only secondary data was performed (e.g. comparisons of value sets), if the publication was a study protocol or a mapping study (e.g. mapping from a condition-specific instrument to a generic instrument).
Titles and abstracts of all identified studies were independently screened for relevance by two reviewers (M.Å. and F.S.T.) using the software Rayyan [
12]. In the case of disagreement, the titles and abstracts were read again and discussed among the two reviewers. If not reaching consensus, the studies were included for full-text screening. Authors K.B. and J.B. were engaged in the discussion. All studies meeting the inclusion criteria were used for data extraction, and there was no assessment of data quality in the included studies.
Data extraction for study characteristics and the four research questions was performed using a self-designed data extraction form in MS Excel. The extracted background information included authors, year of publication, country, the overarching objective of the study, sample size for VAS valuation, setting (university, clinics, etc.), information on study population (general population or specific disease groups), age, mode of administration, and if additional valuation methods had been used. Information relating to the four research questions was extracted, including purpose, health state definitions, perspective (own vs. described health state), design, and advantages and disadvantages of using a VAS identified by the authors of the articles. The latter were reported in the extraction form using quotes from the individual articles.
A subset of the articles deemed eligible for analysis was initially read independently by two reviewers (M.Å. and F.S.T.) who compared the findings to test the tabulation procedure. The majority of the articles were then read and tabulated by a third reviewer (Z.M.T.L.). Data extraction was continuously discussed among these reviewers and any disagreement was discussed and resolved among all authors.
The findings of the included studies were summarized and presented descriptively using frequencies and proportions of different categories in relation to the research questions. Articles where the VAS has been used in the context of economic evaluation were described separately, as this use has been subject to previous discussion [
1,
3].
Discussion
This study set out to review how the VAS has been used for health state valuation to date. We included a total of 308 articles published between 1991 and 2021 from 40 countries in six continents. A rapid rise in relevant publications was observed in 1998 after the introduction of EuroQol health state valuation by Dolan [
54], and in 2005 after the EQ-5D was included by the National Institute for Health and Care Excellence (NICE) as a preferred instrument for technology appraisal [
55].
Most of the reviewed studies were community-based or healthcare facility-based with varying sample sizes of participants aged 18 years and above. Participants under 18 years were included in some studies among which the youngest participants were aged 5 years [
56,
57]. A previous study has shown VAS to be reliable from 5 years [
58].
In most studies, a VAS was used together with one or more other valuation methods. One of the main reasons not to use the VAS alone is that it can be prone to measurement bias [
5]. Despite this, the VAS has been shown to have its own advantages as well [
1,
3,
59]. Taking the generic EQ-5D instrument as an example, the EQ VAS provides complementary information to the EQ-5D profile, as it reflects the overall health status and probably captures other aspects of health not included in the EQ-5D descriptive system [
60].
Regarding modes of administration, telephone, web-based or postal studies have become popular since 2001. This trend can be expected to grow, particularly in the post-COVID era in which unnecessary physical interactions may be reduced. The use of VAS also requires minimum explanation to respondents, and self-administration is generally easier when valuing health states with the VAS compared to SG or TTO [
5].
Despite the ongoing discussions on whether VAS valuations are appropriate for use in cost-utility analysis, we identified 13 studies which applied a VAS for this purpose, mostly as a stand-alone method. In fact, the use of a VAS is accepted as the simplest approach to measuring preferences for both chronic and temporary health states [
3]. It is also seen as a direct measure of self-rated health and as able to provide additional information by measuring health on a single dimension [
1].
The main criticism regarding the use of VAS in economic evaluation is that it does not measure utility under uncertainty [
1,
3]. However, Parkin and Devlin [
59] argue that QALYs do not need to be based on utility theory as the use of QALYs in economic evaluation is primarily to inform the allocation of limited resources for improvements of health rather than utility [
59]. They argue that using the VAS involves both choice and trade-off across sets of health states, and that other methods also suffer from biases or concerns regarding generating reliable preferences [
59]. Although our scoping review did not yield any further theoretical arguments, it shows that the VAS has been used empirically in a number of economic evaluation studies since 2000. In these studies, many in a clinical context, an important attribute of VAS as a method easier to use than other valuation methods has been demonstrated. This was particularly exemplified by two of the reviewed economic evaluations which employed power functions to transform VAS to TTO values which were comparable to TTO elicited values [
61,
62], illustrating the practicality of VAS valuation and the possibility to generate values equivalent to TTO.
For the health states valued using VAS, the most common ones were disease-specific or related to the current health status of respondents, indicating the usefulness of a VAS for decisions in everyday clinical practice [
63]. In addition, as the EQ VAS is part of the EQ-5D instrument it was not surprising that the VAS was used to value EQ-5D health states, alone or in combination with other valuation methods.
There can be variations in the design when applying the VAS method. The line can be vertical or horizonal, vary in length, and be with or without intervals marked for different numbers [
1]. The response on the VAS is also indicated differently, such as drawing a line to indicate the position of a health state [
1,
2,
60] or placing cards to describe different levels of health states [
64]. Confirming earlier observations by Brazier et al. [
1], our review showed a large variation in the designs of a VAS. Such heterogeneity in VAS designs could have implications for the field by affecting comparability and uniformity of studies. An early study comparing horizontal and vertical VAS reported very high correlation [
65]. In contrast, a relatively recent study comparing a patient global assessment VAS (10 cm, horizontal) and the EQ VAS reported moderate correlation with poor concordance [
66]. A systematic review on design differences in paper-based and electronic VAS reported equivalence of results despite differences in scale length and format [
67]. Use of standardized and clear descriptions of VAS designs used in specific studies could improve uniformity across studies and in turn facilitate comparisons of results. The diversity in design can also be seen in other valuation methods. For example, regarding the TTO, both open-ended and iteration based TTO tasks have been used [
68]. There has also been a development of the TTO task over time, for example regarding health states worse than dead [
69]. Implications of these variations, as well as interview effects and how cognitively challenging the task is, need to be addressed further for all valuation methods.
There are several ways of anchoring the VAS, for example at 1 (‘full health’) and 0 (‘dead’). Health states valued as worse than ‘dead’ are consequently assigned negative values. However, it has been argued that anchoring 0 at ‘dead’ is not a theoretical requirement for health status measurement or for cost-utility analysis [
70]. While the terms ‘dead’ and ‘death’ are sometimes used interchangeably, the choice of term may matter in valuation studies, as ‘dead’ always refers to a state, but ‘death’ can mean either a state or an event [
70]. Despite the variations of the anchors used, the most common ones were 0 for the worst imaginable health state and 100 for the best imaginable health state. Some studies also used the lower anchor of 0 for dead/death and negative values for the states worse than dead/death. Amongst these, a majority used the term ‘death’ for the lower anchor while only a few used the term ‘dead’. This seems to be opposed to what was suggested by OHE, seeing ‘dead’ as a more representative term for a state than ‘death’ which could be ambiguous [
70]. However, the phrasing is not always internally consistent, as some authors used the terms ‘dead’ and ‘death’ interchangeably in the same manuscript, and the same authors sometimes used different terms in other publications. This could be due to different journals’ requirements or changes in the field over time. Harmonization of the endpoints of the VAS would enable the comparison of results. Furthermore, conducting studies using qualitative methods to explore how the phrasing of different endpoints are perceived among participants is recommended.
Among the articles that reported advantages and disadvantages of using a VAS, more advantages were mentioned, which could also be due to our selection criteria (focusing on articles that used a VAS). Common advantages stated were its simplicity, reliability, validity, and practicality in health state valuation, which consequently lead to feasibility and acceptability in its application. The applicability of VAS in economic evaluations was also acknowledged by some authors.
There were also criticisms regarding using a VAS as a utility measure, based on methodological considerations. These include poor theoretical foundations and no risk or trade-off property, although arguments against these criticisms have been put forward by Parkin and Devlin [
59] as discussed above. Some authors also questioned its validity in valuing certain health conditions and the explanatory power of resulting models.
Strengths and limitations
The main strength of our study is its comprehensive and systematic coverage of the literature. At least two reviewers were engaged in screening and selection of the articles and extraction of data in order to minimize bias throughout the process.
A limitation is that no quality assessment of the reviewed articles was conducted. However, the aim was to describe how the VAS had been used for health state valuations in the published literature and not to judge the obtained values from these studies. Our review focused on applications of the VAS, which means that many of the included articles did not systematically report observed advantages or disadvantages of using a VAS, and more theoretical arguments are likely to be found in other types of literature. Another shortcoming is that there may be other articles where the VAS has been used in economic evaluations which we did not include as our search strategies and selection criteria were primarily targeted towards valuation studies, following from our aim to describe how VAS has been used for health state valuation.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.