1. Introduction
With the increasing prevalence of mental health problems [
1], there is a need for a short patient reported outcome measure (PROM) to assess the quality of life outcomes for individuals living with mental health conditions. While it is important to treat symptoms, there is a growing recognition of the value of leading a meaningful life and the need to capture the impact of conditions on this, even in the presence of symptoms. It is also known that many people with mental health conditions are able to lead fully “functional lives” despite the presence of symptoms. Most importantly this “recovery” process “is best judged by the person living with the experience” [
2] (p. 3). However, it was agreed with the funder of this project that to capture the recovery process, it was important to develop some form of measurement with the following seven criteria. The first and most important criterion was that the measures were based on the outcomes service users identify as being most central to them in recovering their quality of life rather than symptoms [
3,
4]. The other six criteria were that they should be: available in a version that was short enough for initial assessment and repeated use in routine outcome measurement settings but with a longer version or item set for research purposes; suitable for use with a wide spectrum of mental health conditions and levels of severity; appropriate for individuals aged 16 and over; robust psychometric properties; suitable for self-completion; and free to publicly funded service delivery organisations.
The rationale for developing the new Recovering Quality of Life measures—ReQoL-10 and ReQoL-20—was two-fold. First, existing recovery measures did not meet the above criteria. A systematic review of recovery mental health outcomes assessed 11 instruments for their psychometric properties, ease of administration and service user involvement [
5]. None of the measures reviewed met the seven criteria above mainly because they contained too many items, were focused on processes and treatment options which are of course important but not outcomes, they were specific to one patient population or with inadequate psychometric properties (see
Supplementary Materials, Table S1). Boardman et al. [
3,
4] also identified the need for a new measure to contain the themes similar to those suggested by Leamy et al. [
6] around connectedness, hope, identity, meaning, and empowerment. Second, in line with the guidelines recommended by the National Institute for Care and Excellence (NICE), EQ-5D is used to calculate benefits to generate quality adjusted life years (QALYs) for use in economic evaluation [
7]. However, there is increasing evidence that EQ-5D may not be suitable for some conditions like anxiety [
8,
9], schizophrenia [
10], other psychotic disorders [
9,
11], and bipolar and personality disorders [
12]. Consequently, another preference-based measure may be more desirable for use in the economic evaluation of mental health interventions [
13]. Therefore, the ReQoL measures were developed to meet the seven criteria identified above as a routine outcome measure with the possibility of generating a set of preference weights.
The ReQoL measures were developed in four stages. The theoretical background for the measure which comprised a systematic review of the quality of life (QoL) literature and in-depth interviews with 19 service users identified the following six mental health themes (activity; belonging and relationship; choice and control; hope; self-perception and; well-being) and one physical health theme [
8,
14,
15]. In Stage I, items were generated under each theme using those from existing quality of life and recovery measures; phrases from the interview transcripts used to identify the themes [
15]; and items identified by the research team. These items, 1597 in all, were sifted using an adapted criteria list [
16,
17] to arrive at 87 items. In Stage II, these items were presented in turn to working age adult service users and younger service users to consider their appropriateness. Qualitative data on the items were also gathered on the 61-item set through a translatability assessment (
Table 1). In Stage III of the project, psychometric analyses were carried out in two separate studies recruiting 2262 and 4266 participants respectively. The qualitative evidence was integrated with the quantitative data to produce the final measures in Stage IV. In terms of governance of the project, the members of the stakeholders group consisting mainly of policy-makers, representatives from professional bodies, staff from various mental health charities and health care professionals (
n = 33); the advisory group (
n = 32) consisting mainly of academics and clinical academics nationally and internationally; and the expert users group (
n = 6) were asked to comment at each stage of the project. The members of the psychometrics group (
n = 6) provided specialist advice on the quantitative studies. In addition, the six expert service users were also members of the scientific group (
n = 18) which formed the decision-making group.
While it is quite common for both qualitative and quantitative evidence to be used in the development of PROMs, exact details of how the qualitative and quantitative data are combined are often not reported. A possible reason may be because in many cases, the qualitative and quantitative stages are separate stages and the data are used sequentially rather than combining the qualitative and quantitative evidence for the final item selection [
18]. The aim of this paper is to present the approach used to combine qualitative and quantitative evidence in the development of the ReQoL measures. While this approach is specific to ReQoL, there is scope for it to be applied more generally to measure development.
4. Discussion
This article set out to co-produce a measure of recovery of quality of life by service users and researchers using qualitative as well as quantitative approaches. Broadening the definition of recovery requires an equivalent broadening of the research methods used by combining qualitative and quantitative evidence when developing an outcome measure. We found that it was possible to present the multitude of often complex and technical evidence concisely so that service users, clinicians, and researchers with varying backgrounds in multi-disciplinary teams could equally contribute meaningfully to the final item selection.
The use of qualitative and quantitative methodologies is the hallmark of a mixed methods approach and has been widely adopted in the research literature [
23]. However, it has been less well used in pursuit of measure development. Indeed, virtually all measure development depends solely on quantitative methods with the evidence for validity derived from psychometric analyses. Although we did not follow any formal model for combining qualitative and quantitative methods, our approach has close parallels to the framework advocated by Luyt [
24] that, in turn, extended a model initially developed by Adcock and Collier [
25]. Luyt’s framework identifies three different levels (theory, domains, and items) that are informed and then refined through an iterative process of combining qualitative and quantitative data. Our informal model broadly followed these levels (termed stages in our approach) with each stage informed by feedback from qualitative analyses (as well as quantitative data), which derived from service users. An interesting component of Luyt’s framework is that the central concept of validity is viewed as being established across methodologies (i.e., both qualitative and quantitative research) rather than multiple aspects of validity (e.g., construct, concurrent, discriminative) being determined within only quantitative methods.
It is obvious that that the qualitative evidence could not have been generated by the researchers alone. The co-production of the ReQoL clearly shows that it is imperative to include service users with lived experience in the development of a measure that is to be relevant to them. The process of excluding service users from the construction of measures is an extension of their exclusion from other activities, and thereby increases their social isolation. By contrast, our view is that service users as well as clinicians, linguists, and researchers all need to be included in the process of production as they all have their own perspectives, life experiences, expertise, and biases. Therefore, co-production is not necessarily the most straightforward way of constructing an outcome measure, but is by far the best way to guarantee a more relevant one. As will have been clear, the ReQoL measures were co-produced in partnership by service users and researchers [
26]. However, co-production is only a first step towards social inclusivity whereby service users challenge services to both adopt and implement measures for which service users have a sense of joint ownership. In this sense, the development of measurement tools becomes one other aspect within the area of mental health in which the views of service users need to be central.
While it is paramount to recognise the importance of face and content validity as enhanced by multiple perspectives, it is crucial that an outcome measure assesses what it purports to measure—that is, ensuring that the measure retains the necessary psychometric standards. As shown in
Section 3.2, one misfitting item was selected based on qualitative evidence and the deliberation of the members at the Scientific Group meeting. Similarly, an item with a relatively poor information function was also chosen. It is therefore recognised that there has to be a trade-off between superior psychometric properties and the face and content validity. The key in constructing the best possible outcome measure lies in the ability to find the right balance between the two. We think that this has been achieved as the use of the ReQoL is increasing rapidly and the initial psychometric results are encouraging [
16].
It has been long recognised that qualitative and quantitative can inform each other and be very complementary [
27]. However, in health a lack of integration of the two was recognised [
28]. The value of this paper is that, instead of keeping the qualitative and the quantitative strands separately, it demonstrates in detail how the two have been successfully integrated. While it is not uncommon, for qualitative work to be carried out alongside psychometrics in measure development, to our knowledge, this is the first paper to use a diagrammatic approach and provide details on the process of integrating the two. The traffic light approach taken by Study 2 is a clear improvement on the tabular depiction in Study 1 and proved easier to understand by all the members of the scientific group.
There are various ways of using graphics [
29] but in this paper we have shown a simple and effective way of presenting complex information to individuals with a view of empowering everyone to participate fully in decision-making. Our evidence shows that it is possible to combine the results from qualitative experts, analysts (psychometricians), clinicians, and service users. However, it is important to be realistic as this approach involves greater planning, time and resources. This way of presenting information can be adopted in many areas of measure development and is reasonably generalisable.
One caveat regarding the study is that the various groups of adult service users, young service users, clinicians, and linguists assessed different item sets since it was part of an iterative process. For example, at the end of the interviews with the service users, the items that were slightly reworded were not reassessed by the linguists. However, care was taken to ensure that the revised items were in line with the comments received about them. Another caveat is that although the evidence is summarised based on criteria established, the aspect of colour-coding remains subjective. Although the qualitative data transcripts were analysed by three experienced researchers, they were summarised by one researcher only. Similarly, one researcher summarised the quantitative evidence.