A systematic review, psychometric analysis and qualitative assessment of Generic Preference-Based Measures of Health in Mental Health Populations and the estimation of mapping functions from widely used specific measures

EQ-5D

The EQ-5D questionnaire comprises five dimensions: mobility, self-care, usual activities, pain and anxiety/depression ( Table 1 ). These can be seen as a combination of physical functioning (mobility, self-care), mental health (depression and anxiety), social functioning (which may be included in usual activities) and symptoms (pain and discomfort). Respondents are asked to report their level of problems (no problems, some/moderate problems or extreme problem/unable to do) on each dimension to provide a position on the EQ-5D health-state classification system. Responses can be converted into one of 243 different health states [ranging from no problems on any of the dimensions (11111) to severe problems on all five dimensions (33333)], each with its own preference-based score. Preference-based scores are determined by eliciting strength of preference data, that is, by establishing which health states are preferred from a sample of the general public. Preferences are elicited using the TTO technique, which involves asking participants to consider the relative lengths of time in full health (i.e. number of life-years) they would be willing to sacrifice to avoid poorer health states. The scoring algorithm, or social tariff, for the UK is based on TTO responses of a random sample (n = 2997) of non-institutionalised adults. Values are anchored by ‘1’ representing full health and ‘0’ representing the state ‘dead’, with states ‘worse than death’ bounded by ‘−1’. Utility values from the UK EQ-5D tariff range from −0.59 to 1. 15 The EQ-5D is often administered with the EQ visual analogue scale (EQ-VAS), which asks respondents to indicate their health state on a rating scale from worst health imaginable to best imaginable.

SF-36 and SF-12

The SF-36 is a generic health status profile measure consisting of eight dimensions of general health, bodily pain, physical functioning, role-physical, mental health, vitality, social functioning and role-emotional ( Table 2 ). 19,20 These eight dimensions produced separate scores by taking a simple summation of item responses and applying a linear transformation to place them onto a 0 to 100 scale. There is also an alternative normalised scoring system generating a mean normative score of 50, with 10 points representing a standard deviation (SD) in a normal population. Physical and mental health component summary scores (PCS and MCS) can also be generated from these dimension scores. 10 The SF-1211 is a shortened version of the SF-36, containing 12 SF-36 items, and also produces PCS and MCS scores. These measures are two of the most widely used generic measures of health and have been validated in many conditions and settings. However, the SF-36 and SF-12 cannot be used in their standard form in economic evaluation because the scoring method is not preference-based and, therefore, cannot be used to generate QALYs. Furthermore, they produce a profile of scores and cannot be combined with survival, so they provide no single measure of effectiveness required for assessing cost-effectiveness. For this reason, a preference-based measure called the SF-6D has been derived from the SF-36.

SF-6D

The SF-6D provides a means of translating the SF-36 or the SF-12 into a preference-based single index. 16,21 The SF-6D reduces the eight dimensions of the SF-36 to six: physical functioning, role limitations, social functioning, pain, mental health and vitality ( Table 3 ). Each dimension has four, five or six levels, giving a total of 18,000 possible health states. The values attached to each level and dimension generated by the classification system were derived from standard gamble valuations for a sample of 249 of these health states. Face-to-face interviews were conducted with a representative sample of 611 members of the UK population. 16 States were valued by asking respondents to choose between each of five SF-6D states (imagining they remained in those states for the rest of their lives), versus a gamble between the best and ‘pits’ health states. Respondents were then asked to value the worst state in relation to immediate death. The valuations for the SF-6D were derived from a linear random-effects model, and ranged from 0.29 to 1.0 (full health) (see www.shef.ac.uk/scharr/sections/heds/mvh/sf-6d). 16

Objective indicators

A tension exists in quality-of-life measurement over whether it should have a subjective or objective orientation. A more objective approach may place its emphasis on income, living conditions, access to resources, participation in occupational and social roles,23,24 or the presence or absence of a medical condition or symptom. While objective measures have an important place in the broader quality-of-life literature, within health there has been an increasing emphasis on the importance of the patient’s perspective and this has been assumed to imply a move away from objective measures. This has been partly because many of the commonly used objective indicators like income have been found to be only weakly related to well-being. It is also because objective indicators, by their nature, take a top-down, paternalistic approach, rather than reflecting what individuals might perceive to be important to their quality of life. There has been a similar movement in health care, with a move away from using clinical indicators to measure the outcome of health care towards a more user-focused approach using subjective measures of health and well-being.

Needs satisfaction

There has been a tradition of defining quality of life in terms of the extent to which human needs are met. This is based on Maslow’s25 hierarchy of needs necessary for human existence (e.g. food, drink and shelter, social and belonging, status and self-esteem and self-actualisation). Once the basic needs for food, drink and shelter are met, then human beings can look to fulfil higher needs such as control, autonomy, pleasure and self-realisation. 26,27 These have been found to correlate highly with life satisfaction. This approach continues to be influential in the development of measures, including recent work on an outcome measure for social care. 28

Subjective well-being

It can be argued that in the developed world, where basic needs have been met, a focus on objective or needs-based measures should be replaced by perceived well-being. 29 Economists in the 19th century saw utility as a cardinal indicator of happiness and this comes from the utilitarian Benthamite tradition concerned with maximising the happiness of the greatest number. 30 There has been something of a revival in using happiness and other measures of well-being in public policy in psychology and economics, such as through the work of Kahneman and others. 31 In its modern formulation, subjective well-being is seen to have two components. One is a hedonistic view based on how an individual feels in any given moment. It has typically been assessed using simple items asking people to rate their current level of happiness. 32 The second is a more evaluative concept based on a reflection on how satisfied we are with our lives, and this may include past and present happiness and future expectations, and embrace notions of self-fulfilment, realisation or actualisation. This has typically been assessed using items that ask people to rate how satisfied they are with their life and aspects of it. 33

The revival of subjective well-being in economics has in part been a response to criticisms of measures of benefit used by economists in economic evaluations such as the QALY. By the start of the 20th century, mainstream economics had moved away from an experiential view of utility to one based on preference satisfaction. 34 This is best seen in the marketplace where individuals make choices between goods and services on the basis of what they most desire. They are regarded as autonomous and perfectly informed individuals who will choose the mix of goods and services that maximises their well-being. Where such ‘revealed’ preferences are not available, as is often the case for health, respondents are asked to state strength of preference for different states of the world. This is a preference-based or decision-based notion of utility. However, there is a large body of evidence that consumers do not necessarily know what will maximise the well-being that they eventually experience. 31 While this debate has not been important in the general quality-of-life literature, it has been highly controversial in health economics since the EQ-5D and SF-6D are scored using the preferences of the general public. 35

Capabilities

The notion of capability was developed by Sen36 in response to what he regarded as the narrow perspective of economists in the way they assessed value. He argues that subjective well-being and preference-based utility both fail to consider all the factors that matter for informing public policy. Again, this is an economics-focused debate, but one that has important implications later in this report for the interpretation of qualitative data on what seems to matter to people with mental health problems (see Chapters 5 and 6 ).

Sen36 offers a framework for moving from the purchase of goods and services (including health care) to utility or well-being. He proposes that along the way individuals are transformed by the benefits of the goods or service; for example, a bicycle will confer improved functioning in terms of mobility. 36,37 Economics usually ignores the impact of what a person has on a person’s functioning, and focuses on utility as reflected in their choices. Sen takes it one step further for policy-making, as he argues that public policy is not concerned with functioning per se; so, in the case of the bike, public policy is concerned not with whether or not the person chooses to use the bike, but simply the fact that they are able to function in a particular way. This idea of capability offers someone the choice to be able to ‘do’ or to ‘be’. The object of public policy is to provide opportunities and not to make people do one thing or another. Sen accepts that for some basic capabilities this distinction between capability and functioning may not be important as individuals have little choice, but in large areas of public policy it will be important to separate out these concepts. 36

Sen did not prescribe a particular set of capabilities or a way to operationalise the concept. There have been attempts by others, with the most common being attempts by experts to construct a list of those functionings they think are important, covering basic capabilities through to higher-level issues around human rights and well-being. 38 These tend to be more like the objective lists described earlier and there has been little attempt to score them. In health economics, examples of capability measures include the recently developed ICECAP-O and ICECAP-A, though these use stated preferences to generate the scores. The items were developed from interviews with members of the general public, and attempts to capture the idea of capability by using terms such as ‘I am able to be . . .’ or ‘I can have . . .’. 39,40 Although there may be doubts about whether or not this is truly a measure of capability (including the way it uses preferences in the scoring), it is an important development in terms of going beyond health. This measure will be considered again in the light of the findings of this research in Chapter 6 .

Psychological well-being

This is a broad category of constructs that covers aspects of personality that include morale, self-esteem and self-concept, sense of coherence, self-efficacy, autonomy and control. This list often includes aspects of subjective well-being, but we have presented them separately as they can be seen as mediating variables rather than necessarily determining quality of life in their own right. Indeed, it has been proposed that there should be a model of quality of life based on these psychological factors and the way they explain variations in subjective well-being through cognitive mechanisms. 41,42 They clearly have an important role in the way people cope with and adapt to problems with their health. Knowledge of the precise mechanisms and the support for different models, however, are limited. It is interesting to note that these variables have also been found in measures of quality of life and capabilities as dimensions in their own right. 27,39,40

Health and health-related quality of life

There has been a separate tradition of measuring perceived health in health services research and health economics, and the EQ-5D and SF-36 have come out of this tradition. 43 Health can be seen as a part of quality of life, and is usually seen as narrower than quality of life. However, some definitions of health are very broad, such as the widely cited World Health Organization (WHO) statement, set out in its original constitution, that health is: ‘A state of complete physical, mental and social well-being, and not merely the absence of disease and infirmity’. 44 This too is a rather amorphous construct. The field is made more complicated by the idea of HRQoL, whereby the concern is not with health per se but those aspects of quality of life that are related to a person’s health. Given the central role of health in many people’s lives, this is likely to cover many different aspects of quality of life. A definition of HRQoL provided by researchers, representing what many people in the field mean by HRQoL, is: ‘A person’s subjective perception of the impact of health status, including disease and treatment, on physical, psychological and social functioning and well-being’. 45 This is also a broad definition, and as a result, the content of HRQoL measures varies widely and is no less wide-ranging in coverage than many measure purporting to measure quality of life; indeed, they are often wider.

Discussion of views of quality of life and their implications for this report

This section has demonstrated that there are many different and often competing views of quality of life: objective versus subjective accounts; what people do and feel versus capability; well-being versus psychological well-being (e.g. self-esteem, autonomy, choice and so forth); and narrow versus broader notions. No measure can capture all these elements and ultimately there is a value judgement to be made regarding what matters to policy-makers. It is also the case that these views are not entirely separable, so, for example, needs satisfaction covers objective aspects alongside those concerned with psychological well-being. In practice, measures often contain items from more than one of these views of quality of life. The EQ-5D, for example, covers negative aspects of well-being (i.e. depression and anxiety) and objective aspects such as mobility and self-care. Items of the SF-36 cover well-being (i.e. mental health and vitality), objective aspects (i.e. physical function) and psychological health (i.e. general health perception).

The lack of an agreed view or clear definition means that any attempt to measure quality of life and mental health-related quality of life is fraught with difficulty. This rules out the application of criterion validation in this field, whereby a measure is compared to some accepted gold standard. This has been a source of frustration for researchers and has led to some scepticism as to the value of trying to measure quality of life. 46 However, the primary aim of health services is to improve the quality (health related) and quantity of life, so it is imperative for policy-relevant research to have measures of what matters in the lives of people who use health services. The aim of this report is to examine the extent to which the EQ-5D and SF-36 achieve this through the application of qualitative and quantitative methods.

Assessing appropriateness

This report brings together important new evidence on the appropriateness of the two most widely used generic preference-based measures in mental health in the UK (the EQ-5D and SF-6D), and assesses whether or not it is necessary to adapt them in some way, or consider developing a new preference-based measure for mental health. It uses a combination of quantitative and qualitative evidence within the psychometric framework to assess the validity of the measures in mental health populations. 53,54

Selecting review studies

Data extraction

Data from all included trials were extracted using a form designed specifically for this review, and piloted on a sample of papers. Data extracted included: country of publication, type of disorder, study sample characteristics (numbers, age, sex), other measures used, mean scores on HRQoL measures, type and method of validity assessment, type and method of responsiveness assessment, and validity and responsiveness data. Extractions were performed by one reviewer (DP or TP). Where duplicate publications reported on similar data, the most complete and recent data were extracted.

Study characteristics

The initial search for studies for the wider review retrieved 4115 unique citations across the five mental health conditions ( Figure 1 ). Of these, 3849 were excluded at the title and abstract stage and 266 were examined in full text. From these, 154 studies were found that met the inclusion criteria. A further 12 studies were identified through reference list checking. Overall, the findings from 91 studies are discussed in this chapter for the five conditions. SF-36 and SF-12 studies were not ultimately included in the depression and anxiety review as a sufficiently large number of studies used the SF-6D to be able to extrapolate to these longer versions of the measure. Figure 1 shows a flow diagram of study identification and the characteristics of the studies reviewed by condition are presented in Appendix 2 .

FIGURE 1.

Flow diagram of study identification.

Schizophrenia

The majority of the evidence (25 studies) examined the validity and responsiveness of the SF-36. 66,67,69–85,88,92–95 Although there appears to be strong evidence that the SF-36 is able to distinguish between general population norms and scores of people with schizophrenia (known-group validity), the evidence for convergent validity and responsiveness is less certain ( Table 4 ). Similar findings exist for the EQ-5D, with mixed evidence for the properties of convergent validity and responsiveness. Indeed, when strong associations were found between individual EQ-5D health-state dimensions (e.g. anxiety/depression or self-care) and symptom or functioning measures, this did not necessarily translate into comparable changes in overall EQ-5D index scores, i.e. health-state utility values. 48,90 There was some evidence that associations with measures of depression were comparatively stronger than those with symptom measures of schizophrenia [e.g. the Positive and Negative Syndrome Scale (PANSS)]. 71,88–90

When testing associations between measures for convergent validity (or change scores in responsiveness), there are reasons to predict that stronger and more consistent correlations might exist between generic HRQoL measures and functioning [e.g. Global Assessment of Functioning (GAF), Social and Occupational Functioning Assessment Scale (SOFAS)] or mental health/schizophrenia-specific HRQoL [e.g. Quality of Life Scale for Schizophrenia (QLS)] measures than purely symptom-based measures such as the PANSS. By their very nature, symptom measures are measuring different concepts from HRQoL measures, so it might be reasonable to predict that it is less likely that a strong correlation might exist. A re-examination of the evidence, taking into account evidence for the type of measure used to assess convergent validity (symptom vs. functioning or HRQoL measures, subjective vs. objective measures), produced mixed results. Functioning and schizophrenia HRQoL measures did not fare much better than clinical and symptom-based measures, with four studies indicating strong evidence for convergent validity,82–84,86 and four indicating uncertain or no evidence of such a relationship. 69,81,88,89

Bipolar disorder

There was positive evidence that the SF-36 is a valid and responsive measure in bipolar disorder when individuals are in a depressed, euthymic or mixed state ( Table 5 ). There was little evidence available on the EQ-5D and SF-12 and none for the SF-6D. What evidence there is on the EQ-5D is mixed, with three tests supporting the validity of the EQ-5D in this group and three against it across four studies. 98–101

It is unclear if these generic measures are valid in manic or hypomanic individuals. Only 7 out of 22 SF-36 studies included individuals in a manic or hypomanic state,74,99,101,104,106,110,117 and these suggest it is not a valid and responsive measure within this population. However, where studies examined convergent validity with clinical measures of mania, the numbers of patients in the manic or hypomanic mood state were too small to be meaningful. More generally, there is some concern around how to obtain reliable HRQoL ratings within bipolar disorder individuals in manic or hypomanic states as this relies on self-report.

Depression and anxiety

The SF-6D and EQ-5D demonstrate good construct validity and responsiveness for patients with depression ( Table 6 ). They can both distinguish between groups that are known to vary according to severity of depression, and across differences in quality of life of depressed patients. Both measures respond to clinical and quality-of-life improvement and deterioration. Indeed, in many cases they are more responsive than depression-specific measures (this may be due to the integrated nature of mental and physical health problems and potential simultaneous improvement in comorbid conditions).

The performance of the EQ-5D for patients with anxiety is a little more mixed. The measures were found to be more highly correlated with depression scales than with clinical anxiety scales in patients with anxiety.

The relationship between the EQ-5D and the SF-6D reflects that found for other conditions. The EQ-5D shows a lower level of utility at the most severe end for depression, and the SF-6D shows equal or greater detriment at the milder end. The SF-6D identifies utility loss in patients that report full health on the EQ-5D, though patient averages for mild depression and anxiety are still able to show lower than normal population utility using the EQ-5D.

Personality disorder

The EQ-5D appears responsive in individuals with personality disorders ( Table 7 ). Data on other properties such as convergent and known-group validity were very limited. There was also little evidence on the SF-36 or SF-12 and none on the SF-6D. Nevertheless, the studies which did exist provided some positive evidence that the measures are valid for use in personality disorders. 100,102,107 An exception was Narud et al. ,119 who found that most dimensions on the SF-36 were not able to detect changes in patients in the same way as clinical measures. They concluded that this may be because some SF-36 dimensions are not relevant to HRQoL, so that, even if patients change clinically, this does not translate to a change in HRQoL.

Measures used in data sets

The EQ-5D and SF-6D were compared to one widely used and validated condition-specific measure within each data set. The measures used were restricted to those included in the data sets available. Table 8 displays the measures used in the analysis. Details of the two generic preference-based measures are presented in Chapter 1 . The condition-specific measures are described below.

Data sets

Analysis

Feasibility

Convergent validity

Common mental health conditions

The correlation between the EQ-5D and HADS-T, HADS-A and HADS-D indicated a moderate level of convergence ( Table 10 ). The SF-6D is correlated with the CORE-OM clinical score and the functioning, well-being and symptoms domain scores in the moderate-to-strong range across both samples (see Table 10 ). Correlation with the risk domain score is moderate for the IAPT sample and low for the PoNDER sample. All correlations are significant (p < 0.01).

TABLE 10 - Convergent validity of the EQ-5D and SF-6D in common mental health conditions and common mental health conditions and personality disorder samples

Significant at 0.01.

Condition-specific measure	Common mental health conditions			Mixed
	AHEAD	IAPT	PoNDER	POPMACT
	EQ-5D index	SF-6D index	SF-6D index	EQ-5D index
HADS-T	−0.38*	–	–	−0.49*
HADS-A	−0.25*	–	–	−0.39*
HADS-D	−0.35*	–	–	−0.45*
CORE-OM
Clinical score	–	−0.62*	−0.56*	–
Functioning score	–	−0.51*	−0.49*	–
Symptoms score	–	−0.65*	−0.59*	–
Well-being score	–	−0.56*	−0.46*	–
Risk score	–	−0.35*	−0.18*	–

Figure 2 displays scatterplots of the relationships between the generic and condition-specific measures and the LOWESS fit lines. The lines demonstrate that the relationship between the EQ-5D and HADS differs across the severity scale (the concordance between the measures is better at the less severe end of the scale). The relationship between the SF-6D and CORE-OM is more consistent across the severity scale, and is similar for both the IAPT and PoNDER samples.

FIGURE 2.

Scatterplots and LOWESS lines for the common mental health conditions and mixed diagnosis data sets. (a) AHEAD; (b) POPMACT; (c) IAPT; and (d) PoNDER.

Schizophrenia

The correlations between the EQ-5D and condition-specific measures vary across the two schizophrenia samples. Correlations with the BPRS-E in the QUATRO sample154 are moderate for the total score and the depression and positive symptom dimensions, whereas they are weak for the other dimensions ( Table 11 ). Correlations with the PANSS in the MATISSE sample155 are weak, indicating little convergence.

TABLE 11 - Convergent validity of the EQ-5D and SF-6D in schizophrenia and personality disorders

Significant at 0.01.

Condition-specific measure	QUATRO		MATISSE
	EQ-5D index	SF-6D index	EQ-5D index
BPRS-E total	−0.34*	−0.31*	–
BPRS-E disorganization	−0.22*	−0.17*	–
BPRS-E depression	−0.39*	−0.38*	–
BPRS-E negative symptoms	−0.14*	−0.13*	–
BPRS-E positive symptoms	−0.25*	−0.22*	–
PANSS total	–	–	−0.15*
PANSS positive	–	–	−0.14*
PANSS negative	–	–	−0.03
PANSS general symptoms	–	–	−0.21*

The correlations between the SF-6D and BPRS-E follow a similar pattern to those of the EQ-5D although the correlations are smaller in magnitude, with weak correlations across most of the dimensions apart from depression (see Table 11 ). There is, therefore, poor evidence of convergence for the SF-6D.

The LOWESS lines for the QUATRO sample (which completed both the EQ-5D and SF-6D) demonstrate a tendency for the generic preference-based measure scores to increase as scores on the BPRS decrease (equivalent to less severe problems on both measures). However, the EQ-5D displays a large ceiling effect, meaning that a score of 1 on the EQ-5D is associated with a wide range of BPRS scores. There is a trend towards a linear relationship between the EQ-5D and PANSS, and again there is a large EQ-5D ceiling effect whereby a score of 1 equates to a wide range of PANSS scores ( Figure 3 ).

FIGURE 3.

Scatterplots and LOWESS lines for the schizophrenia samples. (a) QUATRO (EQ-5D); (b) QUATRO (SF-6D); and (c) MATISSE.

Discriminant validity

Responsiveness

Analysis

Preliminary analysis was undertaken prior to mapping in order to identify which condition-specific dimensions/items to use (model specification) and which technique to use to model the relationship between these condition dimensions/items and the target generic measures (model type). Correlation analysis exploring the relationship between generic scores/dimensions and condition-specific scores/items was used to identify which dimensions and/or items would be expected to have a strong relationship with the EQ-5D or SF-6D scores/dimensions and hence the appropriate model specification. 180 Correlation analyses were also used to determine whether or not any independent variables were highly correlated and, therefore, not recommended for inclusion in the same regression model, at both the dimension and item levels. We assessed the distribution of the scores of the generic measures to help inform the choice of estimation technique.

Mapping functions can be generated using different specifications based on assumptions regarding the relationship between generic measures and condition-specific measures. The simplest function is the additive model, where our target measure, the EQ-5D or SF-6D score, is regressed on the total score of the condition-specific measure of interest, for example the HADS total score. This is a limiting specification as it assumes that the dimensions/items of the starting measure carry equal weight and that responses on all items lie on the same interval scale. 178 This assumption can be relaxed by using the dimensions or items from the starting measure instead. Dimension scores are treated as continuous variables whereas items are modelled as dummy variables, i.e. there is a variable for each level in the item (all of the time, some of the time, etc.), which has a value of 1 for those at that level and 0 for everyone else. The assumptions of the additive model can be relaxed by adding interaction terms and squared terms for dimensions. 178 This is often limited to dimension level, especially where there are a large number of items with several levels which can result in a large number of variables. Analysis may also be limited to statistically significant items. Other variables can also be included, for example clinical and sociodemographic information. In this study, we tested additive models with total scores and dimension scores and then tested the inclusion of interaction and squared terms at the dimension level. We also analysed item-level models but did not test for interactions or squared terms for these models. Sociodemographic information (age and sex) was added to the best-fitting models for each condition-specific measure. Comparisons were made between models that had all variables and those that only included variables that were significant at the 10% level.

We started by using ordinary least squares (OLS) regression to model the relationships, as this is the most common model technique. 178 However, OLS fails to take into account the bounded nature of generic preference-based measures which have a maximum value of 1. It also fails to take into account the distribution of some measures such as the EQ-5D, where there may be a large number of respondents at a score of 1 with non-normal (bimodal or trimodal) distributions. 178 The Tobit model can be used to take into account the bounded nature of the preference-based measure values. 180 The two-part model, which models the probability of having a score of 1 and the generic preference-based measure score for those not at 1 in two separate regressions, can also be used where there is a sufficient number of respondents at 1 (full health). 180 Assessment of the range and distribution of utility scores was used to determine whether to use the Tobit or two-part model in the analysis and the models were used where appropriate.

More complex models can be developed whereby, rather than the target measure being the generic measure score, separate regressions are undertaken on the dimensions of the generic preference-based measure, referred to as response-level mapping in the literature. 181 For example, for the EQ-5D, separate regressions are carried out for the five dimensions (mobility, self-care, usual activities, pain and anxiety/depression). Here, the condition-specific measure is used to model the probability of an individual being at a particular level of the generic measure dimension, for example level 2 for mobility, self-care, etc. This results in separate regressions for each dimension of the generic preference-based measure, which are then used to classify individuals in each dimension. Existing tariffs are then used to generate the utility scores. Multinomial or ordered logistic regressions are used to model these probabilities. We tested the application of response-level mapping.

Regression coefficients were assessed based on expected relationship with generic preference-based measure scores and dimensions as well as statistical significance using the Wald test (p ≤ 0.1). Model goodness of fit was measured using R ², adjusted R ², pseudo-R ² and Bayesian information criterion (BIC). With R ² values, the higher the value, the better the model, and with BIC, the lower the value, the better the model. The main concern when assessing mapping results is the accuracy of the predictions. 180 Model prediction properties were tested using the root-mean-square error (RMSE), a measure of the difference between the observed and predicted generic scores. Smaller RMSE values indicate better-performing models. We also compared RMSE alongside predicted means across the EQ-5D and SF-6D range as well as using severity measures where group differences were expected. As the aim of the study was to develop mapping functions that could be used in external studies, it was important to assess how well these functions would work in an external data set within a similar patient population. Follow-up samples were available for some of the data sets used. However, patient clinical characteristics may have changed at follow-up because of interventions, and so generic and condition-specific measure scores may not be comparable to baseline values. Validation was therefore undertaken for the best-fitting models by splitting the baseline samples (75%/25%). The larger sample (75%) was used to estimate the model and the smaller sample was used to validate this model.

We undertook standard tests for the modelling techniques that we used. For OLS, we assessed whether or not there was evidence of heteroscedasticity using White’s test. Multicollinearity was assessed using the collin user-generated command in Stata 11 (StataCorp LP, College Station, TX, USA) and variables exhibiting collinearity were excluded. Non-linearity was tested using fractional polynomials and squared terms were tested where there was evidence of non-linearity. The omodel user-generated Stata command was used to test for the proportional odds assumption when undertaking response-level mapping. Multinomial regressions were used where this assumption was violated in any one of the dimension regressions, and ordered logit regressions were used where this assumption held for all the regressions. Logit regressions were used where there were dependent variables that had been collapsed to two levels because of small proportions of individuals at the lowest levels; for example, very few individuals reporting level 3 in the mobility dimension of the EQ-5D. Model specification was assessed using the link test. We report the performance statistics for the best-fitting models for each condition-specific measure in the main report, and performance statistics for other models are reported in Appendix 2 . Stata 11 was used in all the analyses.

Measures

Data

We focused on two common mental health conditions data sets (AHEAD and IAPT) and one common mental health conditions and personality disorders data set (POPMACT), described in detail in Chapter 3 . The PoNDER sample was not used as it focused on women with postnatal depression and mapping functions developed from this sample may not be transferable to other, more general samples. We also included an additional sample (EMPIRIC), described below.

Preliminary analysis results

Range and distribution of scores

Descriptive statistics are reported in Table 15 . Across all four samples, most of the generic scores did not cover the full range of the measures, apart from in the POPMACT sample. None of the SF-6D scores were at 1, and there were only a few individuals (7/286) at 1 in the EQ-5D in the AHEAD sample. Most of the condition-specific measures cover the full range of scores.

TABLE 15 - EQ-5D, SF-6D and condition-specific measure summaries and descriptive statistics

CORE-OM F, CORE-OM functioning; CORE-OM R, CORE-OM risk; CORE-OM S, CORE-OM symptoms; CORE-OM T, CORE-OM total; CORE-OM W, CORE-OM well-being.

Variables	Mean	SD	Minimum	Maximum
AHEAD (n = 286)
EQ-5D	0.62	0.26	−0.13	1
SF-6D	0.62	0.09	0.30	0.80
HADS-D	10.48	3.95	0	20
HADS-A	13.06	3.47	2	20
HADS-T	23.54	5.89	6	38
Age	42.5	15.32	17	91
Female n (%)	192 (67.1)
POPMACT (n = 475)
EQ-5D	0.50	0.32	–0.59	1
SF-6D	–	–	–	–
HADS-D	11.22	4.55	0	21
HADS-A	14.15	3.95	0	21
HADS-T	25.37	7.45	0	42
Age	32.0	11.23	16	66
Female n (%)	325 (68.4)
IAPT (n = 394)
SF-6D	0.62	0.13	0.30	0.93
CORE-OM T	19.84	7.77	0.88	37.94
CORE-OM W	2.57	0.93	0.00	4.00
CORE-OM S	2.37	0.92	0.08	4.00
CORE-OM F	2.10	0.85	0.00	4.00
CORE-OM R	0.60	0.79	0.00	4.00
PHQ-9	15.59	7.05	0.00	27.00
GAD-7	13.42	5.59	0.00	21.00
Age	40.0	13.9	16	89
Female n (%)	283 (71.8)
EMPIRIC (n = 213)
SF-6D	0.59	0.11	0.35	0.88
GHQ-12	16.97	7.33	2	36
Age	39.0	13.97	16	73
Female n (%)	127 (59.6)

The EQ-5D scores showed evidence of bimodal distributions whereas SF-6D scores were approximately normally distributed across the samples ( Figure 4 ). Distribution of EQ-5D dimension levels was skewed, with < 5% of individuals in level 3 in all dimensions apart from the anxiety and depression dimension (25%) in the AHEAD sample. In the POPMACT sample, the EQ-5D dimensions had < 1% in level 3 for mobility and self-care. SF-6D dimension levels were evenly distributed in both IAPT and EMPIRIC samples.

FIGURE 4.

SF-6D and EQ-5D histograms. (a) SF-6D; and (b) EQ-5D.

Mapping results

In this section we report performance statistics of the best-fitting models for each source measure (HADS, PHQ-9, GAD-7, CORE-OM and GHQ-12) and the two target measures (EQ-5D and SF-6D). Performance statistics of the best-fitting models from the other models that were tested, and OLS regression results, are reported in Appendix 3 .

Hospital Anxiety and Depression Scale and EQ-5D

Ordinary least squares regressions were undertaken for the EQ-5D regressions with HADS dimensions and total scores in the AHEAD and POPMACT samples. The best-fitting models were the total score model in the AHEAD sample and the model with both anxiety and depression dimensions in the POPMACT sample. Coefficients were negative as expected, indicating that as depression and anxiety increase, EQ-5D scores fall. Figure 5 shows the predicted and observed scores for these two models; these have a weak association, particularly on the ends of the scale. Interaction terms were not statistically significant and fractional polynomials indicated that HADS dimensions and totals have a linear relationship with the EQ-5D. Summary goodness of fit and predictive measures are reported in Tables 17 (AHEAD) and 18 (POPMACT). As OLS is used, the predicted mean value for the best-fitting models is the same as the observed values, but the minimum and maximum scores indicate a smaller range of values. This reflects findings shown in Figure 5 that these models overpredicted at the bottom end of the EQ-5D and underpredicted at the top end. As expected, there was poor association between these predictors and the EQ-5D with small R ² values, particularly in the AHEAD sample. In the AHEAD sample, age was statistically significant and improved predictive performance ( Table 17 , model 2) based on BIC. However, the link test indicated that this model was misspecified and the White test showed evidence of heteroscedasticity. Predictive performance was poor, with a RMSE of 0.375 for EQ-5D scores < 0.3 and 0.193 for scores > 0.8 in the AHEAD sample. Although not directly comparable, RMSE scores in the POPMACT sample exhibited a similar pattern, with large RMSEs at the severe end (RMSE = 0.299 for EQ-5D values < 0). Predictions were slightly better at the top end of observed EQ-5D scores for the POMACT sample (RMSE = 0.082 to 0.088). Age and sex were not statistically significant in this sample. There was evidence that predicted EQ-5D scores could discriminate between individuals with and without depression in both samples but the RMSEs were still large.

FIGURE 5.

Predicted and observed EQ-5D – HADS. (a) HADS total score (AHEAD); and (b) HADS-A and HADS-D (POPMACT).

TABLE 17 - Summary performance statistics: OLS regressions, EQ-5D onto HADS (AHEAD)

Regression results are reported in Appendix 3 , Table 48 .

Independent variables	Observed EQ-5D		Predicted EQ-5D
			Model 1		Model 2
			HADS-T		HADS-T + age
Summary statistics
Mean	0.61		0.61		0.61
SD	0.262		0.104		0.115
Minimum	–0.126		0.3571		0.3332
Maximum	1		0.9208		0.9173
Regression performance statistics
R 2			0.156		0.193
Adjusted R 2			0.153		0.187
RMSE			0.194		0.188
BIC			8		1
EQ-5D range	n	Mean EQ-5D	Mean EQ-5D	RMSE	Mean EQ-5D	RMSE
< 0.3	59	0.1816	0.5565	0.375	0.5495	0.368
0.3 to 0.599	34	0.4186	0.5421	0.129	0.5254	0.124
0.6 to 0.699	52	0.6709	0.6142	0.098	0.6002	0.099
0.7 to 0.799	34	0.7402	0.655	0.098	0.655	0.105
0.8 to 1	107	0.8408	0.6495	0.193	0.6655	0.179
HADS-D	n	Mean EQ-5D	Mean EQ-5D	RMSE	Mean EQ-5D	RMSE
Non-case/possible ≤ 10	147	0.6931	0.6821	0.169	0.6816	0.160
Probable case > 10	139	0.5258	0.5374	0.221	0.5379	0.218

TABLE 18 - Summary performance statistics: OLS regressions (POPMACT – EQ-5D)

Regression results are reported in Appendix 3 , Table 49 .

Independent variables	Observed EQ-5D		Predicted EQ-5D
			Model 3
			HADS-D and HADS-A
Summary statistics
Mean		0.5039	0.5039
SD		0.3195	0.1573
Minimum		–0.594	0.1492
Maximum		1	1.016
Regression performance statistics
R 2			0.242
Adjusted R 2			0.239
RMSE			0.227
BIC			150
EQ-5D range	n	Mean EQ-5D	Mean EQ-5D	RMSE
≤ 0	33	–0.1245	0.4021	0.299
0.001 to 0.199	60	0.1295	0.4423	0.117
0.2 to 0.299	64	0.263	0.4324	0.045
0.3 to 0.599	90	0.4242	0.4449	0.017
0.6 to 0.699	57	0.6699	0.532	0.037
0.7 to 0.799	45	0.7374	0.5421	0.062
0.8 to 0.899	103	0.8382	0.586	0.082
0.9 to 1	23	1	0.7295	0.088
HADS-D	n	Mean EQ-5D	Mean EQ-5D	RMSE
Non-case/possible ≤ 10	210	0.622	0.6375	0.220
Probable case > 10	265	0.4104	0.3981	0.233

Tobit regressions were also undertaken. Summary statistics of the best-fitting model indicated that OLS performed better than Tobit in terms of BIC for both samples, with little difference in terms of overall RMSE (see Appendix 3 , Tables 44 and 45 ). However, RMSE over the EQ-5D range showed that the Tobit model performed slightly better at the severe end of the EQ-5D scale and worse at the top end. Response-level mapping models were also tested for the two best-fitting OLS models: those with anxiety and depression combined and with the total scores. As already noted, there were very few individuals at level 3 in the five EQ-5D dimensions and dimension levels had to be collapsed in order to run these models. This resulted in loss of information and poor predictive performance, especially at the severe end of the EQ-5D scale, with RMSE > 0.5. The best-fitting response-mapping model performance statistics are also reported in Appendix 3 , Tables 44 and 45 .

Hospital Anxiety and Depression Scale and SF-6D

Ordinary least squares regressions to predict the SF-6D from the HADS in the AHEAD sample indicated that the best-fitting model based on BIC was the one with both the HADS-D and HADS-A dimension scores. This model showed evidence of heteroscedasticity but no evidence of misspecification based on the link test. The coefficients were negative, as expected. Plots of the predicted and observed SF-6D scores indicated that the HADS was poor at predicting SF-6D scores, particularly at the severe end of the scale ( Figure 6 ). Minimum predicted SF-6D scores for Models 4 and 5 were much higher than the observed SF-6D scores ( Table 19 ), although RMSE for scores < 0.5 indicate that this might be an outlier as these range between 0.141 and 0.134. There was also evidence of underpredicting at the top end of the SF-6D scale. The R ² values were higher than those reported for the EQ-5D regressions, indicating better association. As with predicted EQ-5D scores, predicted SF-6D scores are able to discriminate between those with and without depression. Interaction and squared terms do not add to the model, although age was statistically significant and added to the model ( Table 19 ). Tobit models did not provide improvements over OLS and SF-6D dimensions were not available to test response-level mapping models.

FIGURE 6.

Predicted and observed SF-6D – HADS (AHEAD).

TABLE 19 - Summary performance statistics: OLS regressions (AHEAD – SF-6D)

Regression results are reported in Appendix 3 , Table 50 .

Independent variables	Observed SF-6D		Predicted SF-6D
			Model 4		Model 5
			HADS-A and HADS-D		HADS-A and HADS-D + age
Summary statistics
Mean	0.6118		0.6118		0.6118
SD	0.089		0.049		0.051
Minimum	0.2968		0.5003		0.4942
Maximum	0.8037		0.766		0.7645
Regression performance statistics
R 2			0.305		0.326
Adjusted R 2			0.301		0.319
RMSE			0.059		0.057
BIC			–660		–663
SF-6D range	n	Mean SF-6D	Mean SF-6D	RMSE	Mean SF-6D	RMSE
< 0.5	27	0.4478	0.5885	0.141	0.5816	0.134
0.5–0.599	82	0.5556	0.5963	0.050	0.5956	0.049
0.6–0.699	124	0.6496	0.6214	0.042	0.623	0.042
0.7–0.804	53	0.7379	0.6692	0.069	0.6702	0.068
HADS-D	n	Mean SF-6D	Mean SF-6D	RMSE	Mean SF-6D	RMSE
non-case/possible ≤ 10	147	0.6584	0.6575	0.059	0.6574	0.058
Probable case >10	139	0.5793	0.5803	0.058	0.5804	0.057

Patient Health Questionnaire and SF-6D

The best-fitting model from the PHQ-9 regressions based on the BIC was the item-level model with significant items only; however, the full item-level model performed better in terms of predictive ability and results from the latter model are therefore reported. The coefficients on most of the items were negative and statistically significant. Items related to suicidal thoughts and ‘feeling bad about yourself’ were positive, which is contrary to expectations as this indicates that reporting these symptoms is associated with higher SF-6D scores. Figure 7 shows the predicted and observed SF-6D scores indicating some association. R ² was high in the item-level models ( Table 20 , model 6). There was evidence of overpredicting severe health problems and underpredicting for those in better health. RMSEs are similar at both ends of the SF-6D scale. Predicted SF-6D scores are able to discriminate across patients with depression with RMSEs < 0.1. As expected, Tobit models did not offer any advantage over the best-fitting OLS model as there were no individuals at the top of the SF-6D scale (see Appendix 3 , Table 46 ). Response-level mapping models also performed poorly relative to OLS, with large RMSEs at the severe end of the scale (see Appendix 3 , Table 47 ). This was mainly because dimensions and items had to be collapsed into smaller numbers of levels in order for the models to run.

FIGURE 7.

Predicted and observed SF-6D – PHQ-9 and GAD-7 (IAPT). (a) PHQ-9 items; (b) GAD-7 items; and (c) PHQ-9 and GAD-7 items.

TABLE 20 - Summary performance statistics: OLS regressions (IAPT – SF-6D and PHQ-9, GAD-7)

Regression results are reported in Appendix 3 , Tables 51–53 .

Independent variables	Observed SF-6D (n = 394)		Predicted SF-6D
			Model 6		Model 7		Model 8
			PHQ-9 items		GAD-7 items		PHQ-9 and GAD-7 items
Summary statistics
Mean	0.6193		0.6193		0.6193		0.6193
SD	0.126		0.091		0.078		0.093
Minimum	0.301		0.4242		0.4834		0.4079
Maximum	0.929		0.8595		0.8336		0.8719
Regression performance statistics
R 2			0.516		0.383		0.538
Adjusted R 2			0.476		0.342		0.487
RMSE			0.070		0.079		0.067
BIC			–613		–554		–578
SF-6D range	n	Mean SF-6D	Mean SF-6D	RMSE	Mean SF-6D	RMSE	Mean SF-6D	RMSE
< 0.4	20	0.3610	0.4929	0.132	0.5319	0.171	0.4944	0.133
0.4–0.499	51	0.4690	0.5592	0.093	0.5699	0.101	0.5537	0.088
0.5–0.599	100	0.5587	0.5692	0.053	0.5871	0.056	0.5682	0.053
0.6–0.699	125	0.6386	0.6354	0.051	0.625	0.054	0.6344	0.048
0.7–0.799	60	0.7444	0.687	0.066	0.6747	0.079	0.6904	0.064
0.8–1	38	0.8552	0.738	0.119	0.7099	0.145	0.7453	0.111
PHQ-9 cut-offs	n	Mean SF-6D	Mean SF-6D	RMSE	Mean SF-6D	RMSE	Mean SF-6D	RMSE
None/mild	90	0.7272	0.7273	0.073	0.6934	0.085	0.726	0.069
Moderate	84	0.6501	0.6475	0.071	0.6481	0.079	0.6498	0.068
Severe	220	0.5634	0.5643	0.068	0.578	0.076	0.564	0.066

Generalised Anxiety Disorder Assessment and SF-6D

The GAD-7 OLS regression results indicated that there was poor association between this measure and the SF-6D. The best-fitting model was the full-item model, but most of the items were not statistically significant, and those that were are mainly negative, as expected. Figure 8 shows that there was poor association between observed and predicted SF-6D scores. R ² values were smaller than those of the PHQ-9. Maximum and minimum scores were lower and higher than those of the PHQ-9 (see Table 20 , model 7). RMSEs were therefore also larger. Despite evidence of poor predictive performance, predicted scores are able to discriminate across depression severity groups. Similar to the PHQ-9 results, Tobit did not offer any advantage over OLS and the response-level mapping performed poorly (see Appendix 3 , Tables 46 and 47 ).

FIGURE 8.

Predicted and observed SF-6D – CORE-OM (IAPT). (a) CORE-OM items; and (b) CORE-OM significant items.

General Health Questionnaire and SF-6D

The GHQ-12 item-level model performed better than the total score model. The coefficients were mainly negative, as expected, but the plots of observed and predicted SF-6D scores show that there is weak association in these values ( Figure 9 ). R ² and adjusted R ² values were low as expected. Exclusion of correlated items did not improve the GHQ-12 item-level models. There was no evidence of misspecification, but the White’s test indicated the presence of heteroscedasticity. There was evidence of over- and underpredicting of the SF-6D scores, with large differences between observed and predicted SF-6D scores at the minimum and maximum values ( Table 22 ). RMSE values ranged from 0.116 to 0.166 at the top and bottom of the SF-6D scale; these values were not as large as those reported by some of the models for other measures. Small RMSEs might be driven by small numbers at the top and bottom of the SF-6D range. The predicted SF-6D scores did not vary significantly based on CIS-R severity levels. Tobit and response-level models were not estimated as they were unlikely to improve on the poorly performing OLS models.

FIGURE 9.

Predicted and observed SF-6D – GHQ-12 (EMPIRIC).

TABLE 22 - Summary performance statistics: OLS regressions (EMPIRIC – SF-6D)

CIS, Clinical Interview Schedule.

Independent variables	Observed SF-6D (n = 213)		Predicted SF-6D
			Model 11
			GHQ items
Summary statistics
Mean	0.5895		0.5895
SD	0.111		0.051
Minimum	0.345		0.4329
Maximum	0.878		0.7231
Regression performance statistics
R 2			0.209
Adjusted R 2			0.047
RMSE			0.077
BIC			–185
SF-6D range	n	Mean SF-6D	GHQ items	RMSE
< 0.5	38	0.4341	0.5469	0.116
0.5 to 0.599	82	0.5515	0.5862	0.050
0.6 to 0.699	65	0.6421	0.6039	0.051
0.7 to 0.878	28	0.7897	0.6237	0.166
CIS	n	Mean SF-6D	GHQ items	RMSE
12 to 17	43	0.6306	0.6061	0.087
18 to 23	58	0.6231	0.6057	0.082
24 to 29	51	0.5858	0.576	0.072
30 to 35	45	0.5374	0.5779	0.073
36+	16	0.5159	0.562	0.062

Validation of best-performing models

Models 1 to 8 (HADS, PHQ-9, GAD-7) and model 11 (GHQ-12) were poor predictors of EQ-5D and SF-6D scores and these were therefore not validated. The CORE-OM items models (models 9 and 10) performed reasonably in estimating SF-6D scores within the sample. Validation of these models was therefore undertaken using 75% of the sample to estimate and 25% of the sample to validate the models. Results are reported in Table 23 . Mean predicted SF-6D values were lower than observed values. As with in-sample predictions, the RMSEs showed evidence of over- and underpredicting severe and better health states. The RMSEs were larger across the EQ-5D range and severity measures, although focusing on significant items improves the model’s predictive abilities. This is evident in Figure 10 , which shows observed and predicted scores with evidence of poor association, particularly at both ends of the SF-6D scale. However, a larger sample may be required to validate this model, as although the sample matches the estimation sample in terms of distribution of scores, there are very few individuals at the extremes of the scale.

FIGURE 10.

Predicted and observed scores: SF-6D and CORE-OM (IAPT) out-of-sample validation. (a) CORE-OM items (n = 99); and (b) CORE-OM significant items (n = 99).

Search methods

Systematic reviews of evidence of clinical effectiveness require extensive searching based on a clearly focused search question. Defining a focused question was not possible nor appropriate here, given the exploratory and inductive nature of the review process. An iterative approach to searching was used, in order to accommodate within the search process new themes emerging throughout the review. Extensive searching was undertaken, using a number of search techniques. Database searches were undertaken between October 2009 and April 2010 and included MEDLINE, Applied Social Sciences Index and Abstracts (ASSIA), Cumulative Index to Nursing and Allied Health Literature (CINAHL), PsycINFO and Web of Science. Search techniques included keyword searching, taking advice from experts, hand searching, citation searching of relevant references and internet searching. Four iterations of searching were undertaken (see Appendix 4 , Table 56 ). The searches were not restricted by date, language or country. Key search terms used were mental health, mental illness, mental disorder, quality of life, well-being, life satisfaction, life functioning, life change, recovery, subjective experience, lived experience, lifestyle, coping, adaptation, qualitative and qualitative research. For a full list of search terms and details of the search iterations see Appendix 4 .

Inclusion and exclusion criteria

Language

Although the searches were restricted to English-language articles, non-English-language articles were excluded because of the potential for misinterpretation. Five potentially relevant articles were excluded on the grounds of language. Figure 11 shows a PRISMA flow diagram of the searches undertaken.

FIGURE 11.

PRISMA flow diagram of searched articles.

Data extraction and analysis

The following details of the studies were extracted: mental health problem studied, author affiliation, time and location of study, number and demographic details of participants, research aims and questions, recruitment and sampling methods, method of data collection and analysis, and themes within the findings and discussion sections of the papers.

Framework analysis198 was used to allow the identification of common and variable patterns of themes within and across different studies. The first stage of framework analysis – familiarisation – was undertaken by reading all included papers. The second stage involved examining the findings from these papers to identify initial themes for a thematic framework. These 10 initial descriptive themes were either identified as main themes from more than one study or arose consistently across studies, and were as follows: activity, relationships, the self, the future/aspirations, symptoms/well-being/emotions, spirituality, control/coping, insight/education, health-care services/interventions, and resources/basic needs. The third stage – data organisation – involved charting data from the findings and discussion sections of papers that corresponded to each theme. Text was transferred verbatim to ensure contextual accuracy. It was common for text to be identified as supporting more than one theme, for example a quote describing how work was good for the individual's self-esteem would be placed in the thematic categories ‘activity’ and ‘self’. At the next stage each initial theme was examined, and further subthemes identified and documented within the framework chart. To assist with the final stage of framework analysis – mapping – the subthemes were listed and examined for their conceptual similarities and differences. To aid this process, we searched the wider literature to find papers which would help us to understand the data and make connections between subthemes, and would assist in the development of our final themes. For example, ‘belonging’ was an emerging theme, and we identified Hagerty et al. ’s201 research which explored and defined this concept. We then returned to our thematic framework to re-examine our data in the light of the wider literature. Other influential literature at this stage was on the theory of ‘doing, being, becoming’,202 ill-being compared with well-being and intrinsic and extrinsic quality of life,203,204 and demoralisation. 205

Validation and trustworthiness

Validation procedures were incorporated into the review at all stages. Two researchers (JC and MLJ) independently identified articles from the first search iteration, and compared results to clarify the inclusion and exclusion criteria. Potential full articles were identified from further searches by the primary researcher and independently checked by the second researcher. The included articles were examined independently by both researchers to identify the main themes for the initial framework. Disagreements at all stages were resolved by discussion. Additionally, a multidisciplinary team of researchers met regularly in addition to meetings with clinicians and a user representative to discuss and challenge the inclusion and exclusion criteria, thematic framework, and conceptual interpretations and conclusions.

Description of included studies

Thirteen studies were identified from 16 articles,206–220 one of which was unpublished (E Chambers and S Cook, Sheffield Hallam University, 2012; hereafter cited as Chambers 2012). Two of these journal articles208,209 had fuller reports available. ^{Chambers 2012},206 The fuller reports only have been referenced in the findings. A further study indicated that not all emerging themes were presented in the paper and had a supplementary paper dedicated to the impact of bipolar disorder on work functioning, which was included in our analysis. 207 The studies were published between 1994 and 2010 in a number of countries: Canada209,210,212,215–218 UK,208,211,217 Sweden,214,219 USA220 Austrailia216 and New Zealand. 213 The professional affiliations of the first author were occupational therapy,208,209,212,215,217 nursing,210,213,214,219 psychology,211,216 psychiatry218 and social work. 219

The mental health disorder most frequently represented was schizophrenia (or other psychotic disorder); this was the only population researched in three studies211,212,215 and the majority population in a further two studies. ^{Chambers 2012},220 Three studies included individuals with bipolar disorder only216,218,219 and one included those with panic disorder only. 213 Other studies206,210 had a mixed population including the above disorders plus persons with personality disorder, severe depression and anxiety disorders. Two studies did not specify the disorder; they included persons described as having ‘enduring mental health problems’217 and ‘psychiatric disability’. 214

Two studies had a primarily positive orientation in that they asked ‘what is required for a good quality of life?’,218,219 and four showed a negative orientation through asking ‘how has your mental health affected your quality of life?’. 210,212,213,216 The remainder considered both ‘what had helped and hindered quality of life’. Four studies had a conceptual orientation;210,212,213,219 the remainder presented their findings descriptively. Further details of the studies^{Chambers 2012},206,207,210–220 can be found in Appendix 4 (see Table 57 ).

We identified six major themes from the initial descriptive themes: well-being and ill-being; control, autonomy and choice; self-perception; belonging; activity; and hope and hopelessness. These themes are now explored in detail.

Well-being and ill-being

Well-being has long been regarded as an important dimension of HRQL scales (see Chapter 1 ). The emotional component of subjective well-being consists of high levels of positive affect (experiencing pleasant emotions and moods), and a lack of low levels of negative affect (experiencing few unpleasant emotions and moods). 204 Within our papers, symptoms of mental illness and aspects of emotional well-being were intertwined, with an emphasis on the negative rather than the positive. This suggested that ill-being, which is more akin to distress and the symptoms of mental illness, is an important aspect of quality of life for those with severe mental health problems. The most evident ‘ill-being’ themes were general feelings of distress, experience of psychosis/mania, depressed mood, fear and anxiety, and problems with energy and motivation and fear and anxiety.

Control, autonomy and choice

The importance of aspects of choice and control to quality of life was identified in eight studies^{Chambers 2012},206,211,212,215,216,218,219 and was a main theme in three of these. 211,215,216 It was often discussed in the context of the availability of external resources which enabled choice and control, including medication and treatment, support, information and finances.

Self-perception

A number of aspects of self associated with quality of life were identified: ‘self-efficacy’, having a belief and confidence in your own abilities; ‘self-identity’, having a perception of self and knowing who you are; and ‘self-esteem’, having a sense of self-worth and self-respect. These were linked to a further theme of self-acceptance. These ‘self’ concepts were closely associated and used interchangeably within the studies reviewed. Aspects of the self and self-perception were a major theme in three studies206,213,216 and were present in some form within all of the other studies except one,210 which had an abstract analytical style and only had undertones suggesting low self-esteem/image.

Belonging

The concept of belonging has been defined as the experience of integration and personal involvement in a system or environment at differing interpersonal levels. It can have two dimensions: ‘valued involvement’ – the experience of feeling valued, needed and accepted; and ‘fit’ – the person’s perception that his or her characteristics articulate with, or complement, the system or environment. 201 Of the primary research studies included in the review, one identified ‘connecting and belonging’ as being important to quality of life. 215 Others identified closely related main themes: being part of a social context,214 rejection and isolation from the community,216 a need for acceptance by others,220 social support,218 relationships,206 barriers placed on relationships,211 labelling and attitudes from others,211 stigma,206,211,218,220 alienation220 and detachment and isolation. 211

Activity

By ‘doing’, a person achieves a sense of self and mastery, and successfully participates in the external world. 202,222 The importance of activity in some form to quality of life was expressed in all of the studies except the one which examined panic disorder from a nursing perspective. 213 There was a difference in emphasis between studies: some focused specifically on the benefit of employment207,211,216,220 and others on activity or occupation in its broader sense, including both employment and leisure activity. ^{Chambers 2012},206,212,214,215,217,219 Whatever the type of activity, it was stressed that it should be meaningful or fulfilling,^{Chambers 2012},206,212,214,215,217,219 enjoyable,212,217,219 and suited to need and capabilities. 212,214

The benefit of activity is that it can provide the means for many of the factors important to quality of life discussed above. It is through activity that the opportunity arises to interact with others and hence develop a sense of belonging. 215 Activity can also improve mood,^{Chambers 2012},206,212,214,215 increase energy and/or motivation,206,215,219 relieve stress^{Chambers 2012} and boredom215,217 and provide a distraction from problems. ^{Chambers 2012},214,215 It also helps self-esteem and self-confidence, engenders a positive self-identity^{Chambers 2012},206,207,211,215,217,220 and enables people to take control of their lives. 215

One further factor not already discussed is how activity provides order, routine, and structure. 211,214,215,218,219 Routine and structure can be achieved through employment,211,218 childcare211,218 or activity in general, be it work or leisure. 214,215,218,219 However, one study highlighted how too much structure could be problematic and that what was important was flexibility and choice. 218 Having a physiological routine – particularly regular sleep, meals and exercise – was considered important for general well-being. 218,219 Using activity to structure time increased motivation, provided a diversion from problems and avoidance of negative moods:215

The actual work, whatever it is, is good for the mind and soul . . . you forget yourself. You forget your own problems when you are working[.]

In the morning I have to do something. Some job or something I should do. Otherwise, I become bored and then become depressed because I don’t have anything to do . . .

Laliberte-Rudman et al. 215

Although activity was almost universally considered to be beneficial, taking part could be difficult if the activity was too demanding and not suited to needs. 207,214,219 The symptoms of mental illness could make difficult even the most rudimentary of activities, such as self-care, cooking and shopping,^{Chambers 2012},217 and taking up employment was especially problematic. 212 Even potentially enjoyable leisure activities were avoided because of concern regarding other people’s reactions,^{Chambers 2012},211,217 problems relating to people,217 and the associated fear and stress resulting in a deterioration in health. ^{Chambers 2012},211 Lack of money also put a restriction on enjoyable pastimes. 211 For those in employment, interpersonal relationships at work were particularly affected due to social withdrawal and irritability, or interfering, inappropriate or volatile behaviour during hypomania, although work productivity could increase during hypomania. 207

Hope and hopelessness

The importance to quality of life of personal achievement or having dreams and goals was evident in six of the studies;206,210,212,214,215,218 the importance of activity and/or life in general being fulfilling and having some meaning and purpose was also evident. 206,214,215,218,220 Both were necessary to instigate change, make plans, and to move forward. Again, the difficulty of achieving this was stressed. 206,212,215

Losses experienced in the past affected the view of the future with a perception of reduced opportunities and choices216 and diminished hopes and dreams,210,212 particularly in the fields of employment207,211,220 and relationships. 210,220 Loss and the effect of past experiences was a theme in eight of the studies,^{Chambers 2012},206,207,210–212,216,220 and a major theme in three of these. 210,212,216 These losses included the loss of life roles generally and, more specifically, the loss of work and career opportunities, relationships and the parental role, skills and ability, time, financial losses, and, ultimately, the loss of a sense of self and identity. Losses which had occurred in the past were perceived as a burden,206 with a pervasive sense of ‘something missing’220 which had long-lasting effects and made life a constant struggle. 210,216,220 Participants compared their own lives negatively with those of others,210,216,220 or with their own lives before illness struck,210,212 and all of this brought about feelings of failure and of being cheated, and a sense of unfairness. 207,216,220

Past losses, including the loss of meaning and purpose in life, and a sense of helplessness and inability to cope, all brought about a feeling of hopelessness, necessitating a renegotiation and a lowering of aspirations and priorities. 206,210,212 The concepts of ‘hope’ and ‘hopelessness’ permeated the review studies206,210,211,216,219,220 and formed a major theme for two210,216 Hopelessness was an expression of the view that life would never change for the better, and brought about a pervasive feeling of distress. 220 Conversely, hope provided a catalyst for change and a better life. 219

Well, my whole life feels problematic, I feel as if I’m not going anywhere . . . I know it sounds negative and I’m not really negative like this all the time, but you know, I find it hard, projecting myself into the future, and leading a happy life . . .

I don’t have hope that I’ll ever have a nice boyfriend, I don’t have any hope that I’ll get married, I don’t have any hope that I’ll work a full week – week after week after week. I don’t really have hope for stability . . .

Lim et al. 216

Strengths and limitations

Complexities arose when setting boundaries around the concept of quality of life. It was evident that there was a considerable overlap in findings, with studies examining ‘recovery’, ‘lived/subjective experience’, ‘psychosocial issues’, ‘health needs’ and ‘strategies for living’. After much discussion and deliberation, these studies were excluded from the review. Since completion of our analysis, a systematic review of the concept of ‘personal recovery’ has been undertaken,233 a concept previously defined as ‘a way of living a satisfying, hopeful, and contributing life even with limitations caused by illness’. 234 Interestingly, the authors identified five recovery processes comprising ‘connectedness’, ‘hope and optimism about the future’, ‘identity’, ‘meaning in life’ and ‘empowerment’, which are very similar to our own final themes. They do not include ‘well-being’ and this may be due to the rejection of an emphasis on symptoms within the recovery movement and hence a separation of ‘clinical’ recovery and ‘personal’ recovery. This suggests that the concepts of ‘personal recovery’ and ‘quality of life’ are very closely related. This is important to understand as the concept of ‘recovery’ is gaining prominence as a guiding principle for mental health services. 235

There were also difficulties setting boundaries around themes because of the strong inter-relationship between the different domains which make up quality of life. To avoid repetition, subthemes have been placed within the main theme with which they were considered to be most strongly associated, but aspects of these themes could be placed in other themes. For example, ‘feeling normal’ has been included under the main theme of ‘belonging’ but could also be regarded as an element of ‘ill-being/well-being’ and ‘the self’.

Reliance on secondary data can limit the ability to make connections between subthemes, as can a reliance on that which researchers choose to report within a paper from their primary data. This is another reason for conducting our own primary research reported in the next chapter.

The primary studies included those with severe mental health problems only, with a majority covering schizophrenia or psychotic disorders. Where there was a mixed population, studies rarely indicated any differences between people with different diagnoses. The findings may therefore have biases towards those with psychotic rather than affective disorders. The interviews reported in the next chapter recruited more people with affective disorders to counteract this bias.

Implications for the wider project

This review has identified six potential domains of quality of life. These formed the basis of a topic guide for the interviews of people with mental health problems reported in the next chapter. They also provided an initial set of themes for the framework used to analyse the data collected from the interviews.

Measuring quality of life for people with mental health problems has been of growing interest because of concerns about the emphasis placed by mental health services on reducing symptoms. Yet this review has highlighted that the distress caused by symptoms is integral to the quality of life of people with mental health problems, and in some instances seeing beyond this distress is difficult. However, the review offers a partial picture because of the focus of studies on people with psychotic rather than affective disorders. The qualitative interview study reported in the next chapter addresses this limitation by including people with affective disorders. There is also concern that the review may be partial due to reliance on secondary data. The qualitative interview study in the next chapter addresses this possible limitation by gathering primary data on quality of life from the same samples as those included in the review.

Recruitment

Participants were recruited from three mental health service providers in a city in the north of England. One service (IAPT) provided psychological therapies for those suffering from mild to moderate depression and anxiety. The other two were community mental health teams (CMHTs) providing services for those with more severe problems, one working with individuals with severe and complex non-psychotic disorders (e.g. severe depression, PTSD, personality disorder) and the other with those with psychotic disorders (e.g. schizophrenia, bipolar disorder). Recruitment was undertaken by service providers who applied wide inclusion criteria in order to capture as broad a range of mental health problems as possible. Exclusions included those individuals facing acute episodes of their mental health condition, those not well enough to take part, and those who could not speak English or give consent, as well as cases where there was a known recent forensic history or any doubt relating to the safety of the researcher undertaking the interview in the person’s own home. After a brief explanation of the research by the mental health provider, those persons who expressed an interest in taking part were given a recruitment pack consisting of a letter from the university-based researcher and a flyer giving concise information about the research, plus an ‘expression of interest’ form to be returned to the researcher in an included SAE. The researcher subsequently contacted the person to give further details, answer any questions and arrange the interview if the person was still happy to go ahead (none refused at this stage). A confirmation letter was sent, together with detailed information about the research (including confidentiality, anonymity and right to withdraw) and a consent form for completion prior to the interview. A choice of venue was offered, which could be the person's own home, a room at a university in the north of England or a centre frequently used by mental health services in an easily accessible part of the city. A £10 shopping voucher was offered for participation in the research. Approval for the above procedures was given by the local Research Ethics Committee.

The services recruited 21 people to take part in the research and 17 were subsequently interviewed, nine from the IAPT service and eight from the CMHTs. Four potential participants were recruited by the service providers but did not attend interview, two could not be contacted and did not respond to messages left on their telephones (one IAPT, one CMHT), one arranged an interview but cancelled because they felt too ill to attend (CMHT) and another did not attend an arranged interview (CMHT). A further two participants with psychotic disorders were recruited by one of the participants subsequent to her own interview.

Interviews

All 19 participants were interviewed by a member of the project team (JC), a mental health researcher with a background in behavioural sciences and outcome measure development. The interviews were semistructured with the use of a topic guide (see Appendix 5 ) to ensure that a common set of questions was asked. The topic guide was based on the synthesis of a qualitative review presented in Chapter 2 and, as the ultimate aim of the research was to establish the suitability or otherwise of the most commonly used generic quality of life measures, it included the item content of the EQ-5D and the SF-36. Three pilot interviews had taken place with people with different diagnoses (schizophrenia, bipolar disorder and depression) to test the topic guide. The guide was not changed substantially apart from a decision being made to include questions relating to the items of the EQ-5D and SF-36 as part of the interview rather than administering the actual measures. The first part of the interview aimed to elicit what was important to quality of life from the perspective of the individual, without any prompts. The interview started by asking for background information, and information on current activities and the participant's health, using responsive questions to probe these aspects of his or her life and to determine their relative importance. Participants were asked general open-ended questions about what affected their quality of life both from a positive and negative perspective, what they enjoyed and why, what they would most like to change, what helped, and what was stopping them doing what they wanted to do. Once their own perceptions had been exhausted, the interviewer introduced concepts that the research review had raised or that were included in the EQ-5D or SF-36. The interviews took place when the researchers were part way through analysing the data from the review, when the themes were descriptive in nature rather than conceptual. These were raised only if they had not already been discussed in the interview, and included questions about the relative importance or effect on their quality of life of relationships, support, stigma, work, leisure activities, mental health symptoms and their relative effects, medication and side effects, physical health/pain, energy/motivation, self-esteem/confidence, mental health services/workers, and finances.

All the interviews were tape recorded apart from one, where notes were taken at the request of the interviewee. One further interview was recorded but accidently deleted; notes for this interview were made 3 days after the interview took place. The interviews lasted between 25 minutes and 1 hour 50 minutes, averaging 1 hour 16 minutes. Seven of the interviews took place in the participant’s own home. These tended to be people with more severe problems, five of whom were recruited by the CMHTs. The remaining two, from the IAPT service, had difficulties leaving their homes because of anxiety. The remainder of the interviews were conducted in a room at the university.

Participants

Interviewees included 11 men and eight women. These comprised one participant in his 20s, six in their 30s, eight in their 40s, three in their 50s and one in his 60s. One participant was of African Asian descent and another African Caribbean; the remainder were white European. A broad range of mental health problems were represented including schizophrenia, schizoaffective disorder, personality disorder, PTSD, mild to severe depression, anxiety, agoraphobia, eating disorder and anger. All, with the exception of one, were not in paid employment at the time of the interview although most had worked at some time in the past or were currently employed in a voluntary capacity. Five were married or with a partner; the remainder were not currently in a relationship. Further contextual information about the participants can be found in Table 24 .

Analysis

The interview recordings were transcribed verbatim. Notes were used for two interviews. The interview data were analysed thematically using framework analysis. 198 Framework analysis was used to allow the identification of common and variable patterns of themes within and across different groups, in particular relating to problem severity and type. This analytical method also allowed for comparisons to be made with thematic findings from the review of qualitative research previously undertaken. The themes identified in the review made up the initial themes of the framework. Next, each transcript was read for refamiliarisation. Interview text was then charted onto the framework for the most part using the predefined themes of the review. This was either charted verbatim or paraphrased depending upon the relevance and focus of the response to the research question. Nothing was considered irrelevant and all data were coded in some form. All charted text was indexed with the interview timings that had been noted in the transcript so that actual dialogue could be returned to ensure contextual accuracy. It was not unusual for text to be coded under more than one theme, often one descriptive and another conceptual. Cross-references between themes were noted.

Subjective well-being and ill-being

Within the review (see Chapters 5 and Connell et al. 199) the focus of subjective well-being was on the negative rather than the positive aspects of well-being; that is, what takes quality away from life rather than what adds quality to life. Thus the focus was on overall feelings of distress related to the symptoms of mental illness. These included the feelings of depression, a lack of energy, motivation and enthusiasm, fear and anxiety (particularly in social situations, and fear of relapse) and the distressing symptoms of psychosis including hearing voices, hallucinations and paranoia. These aspects of ill-being were also described in our interviews, particularly by those with severe and enduring mental health problems. Thus, when asked what would improve their lives, participants tended to cite the absence of these negative issues rather than the presence of the positive aspects of subjective well-being such as feeling happy. When asked what made them happy the response was often ‘nothing’. The expectation of ever feeling happy for this population was low.

I’m wanting to change – the biggest issue with my life, but there is a waiting list. There’s a new therapy coming out called EMDR [eye movement desensitisation and reprocessing]. I got put down for it once but I was having suicidal thoughts, erm I’ve not attempted this year so I have been accepted . . . I will still get flash backs but I won’t get the trauma afterwards and that’s what I want to get rid of, I don’t want to be traumatised every day, my coping strategies don’t work for me . . . I would just want limited life back, instead of the full quality of life[.]

Participant 8 (CMHT, PTSD)

Can you think of anything that might make you happy, or happier?

No! Perhaps if I was a different person that would make me happy, I don’t know what else to say . . .

Participant 4 (CMHT, anxiety/depression)

In contrast, interviewees with less severe mental health problems did speak of the experience of happiness and enjoyment, although it was often tempered. They spoke of enjoying current hobbies and pastimes, or happiness as a feeling they wished to regain.

What makes you happy?

Making sure that my wife is comfortable and happy, my son, my garden is just – simple things really, there is no major thing that I can think of . . . my grand-daughter makes me happy[.]

Participant 18 (IAPT, depression)

There were aspects of well-being that were identified in the review, but were much more prominent in the interviews. The first was the desire to feel calm, relaxed and peaceful, which was described across all mental health problem types and levels of severity. For one interviewee the need to feel calm and relaxed was in contrast to feelings of anger, which was an emotion affecting quality of life that was not included in the review.

I can remember when I used to be relaxed and chilled and not feeling like I do now, but I’ve felt like this for such a long time and anxious for such a long time, but I can remember what it was like not to feel like this, it’s a marvellous feeling[.]

Participant 10 (IAPT, anxiety/panic attacks)

. . . it’s hard but I have just got to lose this anger, if not I am just not going to get any better, I’m just going to carry on and I am going to end up, like I said before, I will end up either crippled, dead or locked up and I don’t want that, I just want to be able do what I want to do in the daytime and get home at night and just relax[.]

Participant 3 (IAPT, depression/anger)

Related to a desire to feel calm and relaxed was a need to feel safe. However, the strategies used to ensure these feelings of safety could inhibit other aspects of quality of life.

Where you are familiar with, where you feel alright like in [name of student area] full of students, they’re harmless, as long as you don’t bother them, that’s it, they think you are one of them because you’re young as well . . . everything matters, whether the people are safe whether the area is nice area, whether you know it properly, everything matters, yeah, every little bit contributes[.]

Participant 16 (CMHT, schizophrenia)

It is the safety issue that is most important, I have to keep myself safe . . . I can’t like go out now in case I bump into somebody in [name of district] they go home and they say ‘oh I met a young lady called so and so, in a different area‘ and they know me and my life might not be worth living . . . I am now trying to get out more, even just walk round the shops or summat, but it is hard[.]

Participant 8 (CMHT, PTSD)

Lack of concentration was a negative aspect of subjective well-being that was more prominent in the interviews. Those with severe problems, or experiencing feelings of depression at their worst, reported that they were no longer able to enjoy even the simplest of activities such as reading or watching TV because of difficulties concentrating. This also meant that they were unable to find a distraction from their problems. Those with less severe problems reported a perception that their concentration was getting worse, but not to the extent that it took away from enjoyment of hobbies or pastimes.

I used to play golf quite a lot, but I’ve not played for quite a while . . . I just don’t get the same enjoyment out of it like I used to do. It used to be a nice walk and concentrating on – I think that’s the other thing, I can’t concentrate on it anymore. It takes a lot of concentration, does golf . . . I can’t hit t’ball like I used to, just because I’ve got depression, it sounds weird, that, but I can’t play like I used to do[.]

Participant 7 (CMHT, depression)

Missing from the review was any indication of the aspects of subjective well-being affecting quality of life the most. Those who gave an indication at interview of having comorbid problems were asked this question and the overwhelming response was ‘the depression’. When asked why, they stated that it was due to its all-consuming nature over which they felt they had no control. They could obtain some relief from anxiety or the voices, but depression was an ever-present darkness that they found much more difficult to cope with.

The depression is worse than the voices, because the voices sometimes, when you are with people, like now with you, I’m not hearing them, if you’re doing serious matters, but I don’t know how long that be, with me not hearing voices, but the depression is always there . . . it’s like the English tradition, you heat milk – in the oven, in a pan, or in the microwave, and when you heat it properly you get something covering the upper layer of the milk, that’s the depression, covering all the good moods and everything, and interests . . . and it covers all your body, it affects your motivation, your energy, and you don’t want to do nothing and you end up sad . . . so like it covers up all your happiness and everything, to me, in my opinion[.]

Participant 16 (CMHT, schizophrenia)

. . . well both things stop me doing things but the depression is a lot harder to cope with, the anxiety you can take medication so that helps erm it might get severe like if I have panic attacks but panic attacks only last for a few minutes whereas depression just seems to go on and on and feels a lot harder to cope with[.]

Participant 13 (CMHT, depression)

Physical health

Physical health was not evident as a theme within many of the studies in the review. In contrast, the majority of our interviewees reported a physical health problem and this was mentioned by interviewees with all diagnoses and severities of problems covered in the sample. Physical and mental health interacted in various and complex ways.

Some reported that their mental health problems, or medication, caused or exacerbated their physical health difficulties.

It feels physical as well as mental . . . my body aches and like I think I just become really tense and that is what makes my body ache and I feel like erm I feel like my chest is being crushed and erm I can’t breathe and things like that and erm I just want to be asleep all the time to escape but I can’t sleep[.]

Participant 13 (CMHT, depression/affective personality disorder)

For others it was their physical health problems that contributed to deterioration in their mental health.

I’ve got diabetes, I’ve got high blood pressure, I’ve got arthritis, I’ve got angina, I’m on too many medications. I don’t know what they are all for and it’s just this last 6 months it has just been getting me down[.]

Participant 3 (IAPT, depression)

Participants were asked whether their physical or mental health problems affected their quality of life the most. Some found this difficult to answer, or described how the two interacted with each other. The presence of both seemed to make life particularly difficult to cope with.

. . . the two now are sort of like together on a par, I can’t cope with both of them at all . . . it is bad enough trying to cope with what’s going off in my head, but with my body in pain every day I can’t even cope, I can’t cope with that as well[.]

Participant 12 (IAPT, depression)

Self-perception

The concepts of self-efficacy, self-identity, self-stigma and self-esteem identified within the review were all revealed as having an effect on many aspects of our interviewees’ lives. This was often communicated after prompting rather than being something immediately obvious to them and explicitly stated. Consistent with the review was an acknowledgement that a lack of self-worth and confidence was preventing them from doing things that might potentially improve their quality of life, yet it was difficult to overcome.

. . . it is the area of, um, you know, not willing to take, to step out and get out of my comfort zone is what I’m the most challenged on. So I do generally feel, er, inadequate in a way because I don’t, I won’t, should I say, I ought to, sort of, you know, take control, risk, if you like, and actually get on with it rather than, you know . . . like I say, well in terms of finding a partner, it’s a confidence thing, you know, I see too much danger and, sort of, think perhaps that if you find, it would be difficult to find somebody who genuinely loved me for who I am [laughs], so, um, and I know it’s not just in – it’s in other life choices as well and er – but I’m not saying be reckless and gamble, recklessly do things but, on the other hand, um, sometimes chance favours those who will be prepared to take a chance . . .

Participant 17 (other, schizophrenia)

A negative perception of the self that affected quality of life was expressed by interviewees with all diagnoses and severities of illness. However, those who appeared most distressed at the time of the interview reported how a lack of self-esteem and confidence currently affected their lives in a detrimental way. Those with less severe problems talked about how their confidence was improving or that they were trying to address this and consequently improve their lives.

Yes it has improved a lot, if you had asked me to come here 7 week ago there was no way I would be able to come but now I have come on the bus. I feel a lot better in myself, if you had asked me 7 weeks ago I would have had to say come to my house, there was no way I could have come out. It was the thought of going out in the open I know it is horrible saying it but it knocked me for six kind of thing.

Participant 2 (IAPT, anxiety)

A further difference identified within our interviews was that for some a lack of confidence and self-esteem was something they felt had affected them all their lives and had contributed to their mental health problems, whereas for others this was a consequence of their illness.

I have got no confidence to do the things I used to do and self-esteem, no, just lack it. You see before I was a right confident person, until what happened to me happened, so I don’t know whether my self-confidence will come back, ’cos before I was very confident, I had to do hand overs at work and like talk to GPs and things like that, and it didn’t bother me, and now I don’t even like talking to people some days [laughs], I don’t know what it is, I just put it down to my illness[.]

Participant 8 (CMHT, PTSD)

The review identified how those with bipolar disorder, schizophrenia and panic disorder in particular described difficulties with having a coherent sense of self. This was also evident in our interviews; however, one interviewee diagnosed with schizoaffective disorder described how this could have a positive as well as a negative effect.

. . . it’s almost like this different personality of voices bouncing round your head, kind of, you know what you’d call your train of thought, your own, kind of, inner dialogue, that kind of, it’s like you’ve got that, but yet there’s other ones trying to, kind of, take over and become that train of thought and they’ve got their own voices and their own personalities and their own characteristics and you’re like struggling to, kind of, stop them hijacking your brain, in a way, and so you kind of, you’re having conversations in your head with them and, kind of, talk, and sometimes it’s really funny and good, it’s like having two mates round and you’re just arguing and talking and at other times, you kind of, you lose track of who you are, if you know what I mean[.]

Participant 14 (CMHT, schizoaffective disorder)

Autonomy, control and choice

Consistent with the review, our interviews revealed the importance to quality of life of the related concepts of autonomy, control and choice. The review highlighted the complex juxtaposition between support and independence: the dilemma between needing support and valuing independence while not wishing to be too dependent. This same quandary was communicated by our interviewees. They spoke of an aspiration for independence but with an acknowledgement by some that support, and sometimes dependence, were necessary during periods of illness when they were not in a position to help themselves. The need and desire for support and independence therefore changed over time.

. . . when you’re depressed, you’ve got, kind of, you’re very reluctant or unable to help yourself, and even if there’s loads on offer, you won’t find it, you kind of, you need to keep in touch with people, like, so me seeing my family or as many friends as I can and like the social worker, is vital, otherwise you could, you know, get really really down[.]

Participant 14 (CMHT, schizoaffective disorder)

In our interviews, the value of support for quality of life was expressed primarily by those with severe mental health problems, whereas those with less severe problems were more likely to be actively working towards greater independence. Of our interviewees, the person for whom independence appeared to be of the greatest importance and dependency the most troublesome had physical as well as mental health problems which affected their mobility. Being independent was important for dignity, pride and privacy, whereas being dependent resulted in feelings of guilt and being a burden which could put a strain on a good relationship.

Just to sort of do something on my own without like oh well I’ve got to get out or whatever, will you come to bus stop with me, or got to ring a taxi, just to be able to say ‘right I’m off’ not disappear but just say ‘right, I’ll see yer’ and just go without having ‘where you going, what time you coming back’ . . . I am dependent on people.

How does that make you feel?

Fed up sometimes, I mean don’t get me wrong, I would go out with my daughter or my friend but that would be by choice. You know like ‘oh I am meeting’ her rather than [daughter saying] ‘oh well I’ve got to meet her cos I’ve got to go shopping’. That sounds awful, but that’s the – do you know what I mean. You know without her thinking ‘oh I can’t do this cos I’ve got to make sure me mam’s all right’[.]

Participant 9 (IAPT, depression)

Also consistent with the review was a need for control, particularly symptom control, for those with psychosis and anxiety-related problems. For these interviewees a fear of losing control was evident. However, this prevented them from doing some of the things they would like and increased dependency upon others. This showed that although control was desired, and increased feelings of well-being, too much control could have a detrimental effect on quality of life. Our interviewees were very much aware of this but found it difficult to change and find the balance between the two. A consequence of this was that they felt bad about themselves because they were not living up to their own and others’ expectations.

I seem a bit of a control freak, I want everything to be worked out before I decide to do a certain thing, you know, I want everything to be fairly straightforward and I mean, you can’t, in a way, you can’t live life like that, and yet I still want to live life like that, do you know what I mean? . . . it’s about the stress, erm having faith or taking this, stepping out of your comfort zone, whatever you want to call it, yeah[.]

Participant 17 (other, schizophrenia)

Having choices and opportunities was also important to quality of life. As with the review, choice was particularly associated with having sufficient finances, which in turn was often linked with the availability of suitable employment. However, the opportunity of employment, or at least employment that was not demeaning to their values and expectations, was something they felt was denied, often due to perceived and experienced discrimination. Not having money meant they did not have the opportunity to pursue those things that could improve their lives in ways that were related to leisure activities, their environment or their physical and mental health care. Our interviews also showed how having choices and opportunities was cumulative, with one leading to another.

I ended up going to the halfway house [gives name] at halfway homes they had lots of opportunities . . . they were brilliant because they there was a pot of money every month that they could spend I mean I learnt to drive through that . . . it would encourage you and enable you to do these life things so I did driving, somebody else did fishing . . . I don’t think they have a training budget now which is a bit of shame because it opened my life to lots of things. I went on courses . . . I got the money for the driving, knowing that I could drive triggered off other things for me and I did get benefits at that point which allowed me a little bit more to do these things . . . So that’s how my life got built by having guidance and opportunities really . . . and I think that’s a big thing isn’t it, there’s not all the opportunities there and stuff[.]

Participant 1 (CMHT, depression/eating disorder)

Because it was difficult to find suitable work, and because of the stresses that employment entailed, some participants chose the safety of remaining on benefits. Some preferred to do voluntary work instead, which provided them with the benefits linked to employment, apart from financial, but with more flexibility and control to protect themselves from becoming ill.

Relationships and belonging

Within the review, ‘belonging’ was defined as having two dimensions: ‘valued involvement’, the experience of feeling valued, needed and accepted, and ‘fit’, the person’s perception that his or her characteristics complemented the environment they lived in. 201 The focus was on how well people perceived they fitted with society. The importance of a sense of belonging was also present in our interviews. In contrast to the review, our interviewees also felt that society did not fit with them. This was expressed as a dilemma about whether or not they wished to be part of a society that they felt had different values to their own. This sense of being ‘alien’ to society was something expressed by those with severe depression and schizotypal disorders. Those with mild to moderate anxiety/depression expressed feelings of being isolated from society but not that they did not belong to it.

I have feelings of err not belonging to the human race, like I feel very – it’s not an outcast, I just don’t feel a connection. I don’t know how else to describe that, it’s like being an alien, that is the only way I can describe it, and I know that that sounds weird but that is the only way that I can describe the feeling of it . . . I don’t cope with most people [sighs] I – [pause] my values and norms are very important to me and I know that everybody has not got the same ones[.]

Participant 4 (CMHT, depression/anxiety)

Similar to the review, there was an expression of a desire to do ‘normal’ things or to return to a normal self, particularly for those who had experienced a traumatic event. However, in contrast to the review, our interviewees communicated a sense of being different, rather than abnormal.

You know, you will not make any sense to ’em with voices or whatever you want to call them, bouncing around your head, because then that fires my, you know, it fires my imagination, but I’m talking to people that aren’t, kind of, a bit oddball or a bit, kind of, you know, a lot of people are really, kind of, straight and they’re not strange enough, you know, they don’t – I look at things and, kind of, from a certain angle, a lot of people don’t, and likewise for them, they’ll just think I’m f*****g mad or, you know, dark or depressing, which I can be, but, it’s like gallows humour, you know[.]

Participant 14 (CMHT, schizoaffective disorder)

Stigma was a strong theme in the review and though present in our interviews, was disclosed primarily by those with problems related to gender identity and schizophrenia, particularly related to employment opportunities. Stigmatisation was less of a problem for those with anxiety and depression.

Rather, a need for a better understanding of the impact of mental health problems was expressed by interviewees with all diagnoses and severities of illness. This lack of understanding resulted in them not receiving the support they needed and a feeling of not being able to talk to people about their problems.

Thinking of quality of life – what would have improved things for you at the time.

Well, more understanding from people in authority, like the council, by helping people that have got problems that need the help . . . I find a lot of people don’t understand depression and they don’t understand the way it affects people and, you know, I’ve had many people say to me, ‘oh you’re stupid, you just pull yourself together’, but they don’t understand the way it affects people, you know.

Participant 15 (IAPT, depression)

The review identified the importance of good quality relationships. These satisfied a need for support, love, care and affection, and a need to have someone they could trust and confide in, who accepts and understands them, and with whom they could experience fun and happiness. These themes were also identified in our interviews. Similarly, our interviews revealed that, particularly for those with severe and enduring mental health problems, there were difficulties forming and maintaining close, family and social relationships. The relationships that could fulfil these desired needs felt out of reach, which had a reciprocal effect on subjective well-being and self-perception. These interviewees tended to live alone with care and support provided by their mental health worker only.

I alienate myself because I don’t share. I have to take responsibility for that, but I can’t – without sharing they don’t understand why I don’t share, it’s a bit of a vicious circle. But to have that trust to share, is quite difficult for me, but, of course it is, it is not going to be easy, it takes a long time for me to be friends or something with anybody really . . . [name] from [name] she helps me a lot she is from an organisation to support people with mental health issues to do everyday things to encourage you to go out and things like that, and without [name] I think I would go insane. I only see her once a week or once a fortnight but she does understand me[.]

Participant 4 (CMHT, depression/anxiety)

In contrast, interviewees who tended to have less severe or relatively short-term problems had difficulties associated with current relationships that affected their quality of life, but these difficulties were less profound. They did not express any difficulties in their abilities to form new relationships and spoke of the support received from other family and friends.

Not evident in the review was the way in which people made negative and positive comparisons with others in order to evaluate their own quality of life. In our interviews, envious comparisons were made with people in the interviewee's immediate family or social circle. On the other hand, positive comparisons tended to be made with remote referents, people who were less fortunate than themselves with whom they might rarely come into contact. For some interviewees there was an internal struggle between the two.

I think ‘oh everybody’s at me and they’re all paired off and I haven’t got my mother and dad and everybody’s, all the world’s got a mother and dad’ even though they haven’t and then I say ‘well my sister hasn’t got a mother and dad’ she’s only same as me but she’s got a good husband, and he is a good husband. And then I say I’ve got a sister, I’ve got a brother-in-law as well . . . and I thought you know ‘pull your bloody self together, at least you’ve got legs that work to moan at, there’s some people sat in a wheelchair’ . . . and so I know I’m lucky it’s just, [sighs] oh I don’t know[.]

Participant 9 (IAPT, depression)

Activity and functioning

The findings related to activity from both the review and our interviews were very similar – that leisure and work activity had a positive effect on subjective and psychological well-being, was enjoyable, engendered a sense of belonging through social interaction, and generated feelings of self-worth, pride and a sense of achievement. Importantly, it also provided the income often needed to take part in activities that are pleasurable and aid recovery.

I went on a year’s course at engineering and I was absolutely scared about going on that, but I did it and I did it, you know, quite well . . . and at that time, I was really happy in my life and I thought well I’ve done something, I’ve achieved something here doing this . . . I think [I was happy] because I had the drive and a purpose of getting up and going out every morning and doing what, you know, normal people do, sort of thing, you know, I got into a routine which was very good, so I was happy and more stable . . . I just felt a little bit more worthwhile, you know. I don’t like to feel useless . . . I’ve missed the banter I think as well, as part of being a team, or even on your own or whatever, I miss the camaraderie[.]

Participant 10 (IAPT, anxiety/panic attacks)

Again, similar to the review, our interviews also revealed that when no enjoyment could be found from activity, it still had the benefit of providing a structure and routine to the day, relieved boredom, kept the mind active and – as expressed by many as being of particular importance – provided a distraction from problems.

[After feeling suicidal] my interest in things started coming back it has made a lot of difference because it meant that I was able to distract myself to a certain extent . . . so I was grateful that that came back a bit, rather than just being completely immersed in this world of depression, at least I got a few minutes out of, you know, it was just a few minutes at a time, but it gradually built up to the longer periods where I could go back on the computer again or I could say watch EastEnders, whereas I just completely lost interest, so that was one thing that improved[.]

Participant 13 (CMHT, depression)

Also consistent with the review was the finding that in certain circumstances activity could be detrimental to quality of life. This occurred when the activity, usually employment, felt beyond their capabilities. Reasons given for this were stress and anxiety related to pressure of work, problems concentrating, difficulties with social interaction and the exacerbation of other mental health symptoms.

. . . they systematically bullied me, unbeknown to me, I didn’t even realise because I was that stressed and sort of, out of it, my work life balance doesn’t work very well, I don’t tend to put myself first, and therefore I get iller and iller whilst enabling other people at the end of it I went off sick because I wasn’t sleeping at all, I developed symptoms of erm – it’s what I called jelly brain, and it’s when you turn your head it feels like your brain is slowly turning, even though your head has already turned, it is a very strange sensation, and it’s a sleep deprivation sign. Erm, but obviously I was still awake because of all the anxiety, it was the anxiety that usually kept me going.

Participant 4 (CMHT, depression)

Our interviews highlighted an additional issue around the dilemma people faced when choosing between short- and long-term quality of life: they could foresee the potential long-term benefits of taking part or resuming activities they previously enjoyed, but often found the prospect of doing so too daunting in the short term. The fear that their mental health and well-being might deteriorate further seemed to take priority over the potential for more long-term life benefits – ‘I’d like to but I can’t’.

. . . what the psychiatrists advocate is a relatively low stress environment, that’s how I keep my mental health on an even – you know, my mental health healthy as it were . . . part of me thinks that perhaps if I could do a job where even if it was just cleaning, and if I could, in a way, confront stress to a certain extent, be more, you know . . . I think to a certain extent, it would probably be a bit more healthy to have a bit more in my life, if you know what I mean[.]

Participant 1 (other, schizophrenia)

Our interviews also revealed some differences associated with the severity of mental health problems. Those with severe problems and who appeared most distressed at the time of interview reported finding little enjoyment in any activity, often due to an inability to concentrate. Difficulties with basic functioning and lack of self-care were reported as demonstrating a lack of interest in life itself, as was taking part in particularly risky activity, which was communicated by one participant as an illustration that she did not care whether she lived or died.

I don’t look after myself, I’m not interested in myself . . . I’m not! I’m just not interested in myself I don’t care if I am not here today gone tomorrow, I am not bothered[.]

Participant 8 (CMHT, PTSD)

In contrast, those with less severe problems spoke of leisure activities and hobbies they enjoyed when asked what improved their quality of life. These were often lone activities as, although social activities were described as potentially the most enjoyable, those activities that involved social contact appeared to be the first to diminish as problems worsened.

Hope and hopelessness

The review showed how important a positive view of the future was to quality of life. This involved having goals and aspirations, and being involved in activities that were fulfilling and had meaning and purpose. These were necessary in order to instigate change and have hope for a better future. This was also evident in our interviews, with some further insight into how difficult this was for some, particularly those with severe and enduring mental health problems.

I think whereas I was feeling completely hopeless that things would never change and I would be stuck like this for ever, and just the desperation for things to end or things to change, she just gave me that little bit of hope that things would change . . . eventually I came round to their way of thinking and it gave me hope that at least I could hang on for just a bit longer to see what happens and to see if things would change[.]

Participant 13 (CMHT, depression/affective personality disorder)

There were noticeable differences between our interviewees in how they viewed the future. This could be related to their expected level of achievement before they became ill and whether or not they felt that was something they could return to, given time.

The most negative outlook and evidently poorest quality of life was reported by those for whom previous attempts at positive action to change their situation had failed, and coping mechanisms they previously drew upon no longer had an effect. As a result, they could not conceive of their situation changing in the future.

. . . why carry on, what with my head f****d, and the pain, oh do you know, I just, I really have to ask myself, really why do I carry on . . . It is really difficult to justify yourself at night, at 3 o’clock in the morning, why are you putting up with all this, and why you just don’t commit suicide to be honest because, I don’t know, it’s just really quite exhausting, I am going to the gym to help with my back and stuff like that, but it hasn’t, but my blood pressure has come down. Hurrah! [cynically] after 3 bloody years[.]

Participant 4 (CMHT, depression/anxiety)

In the review, the emphasis was on the negative concept of hopelessness as related above. A more positive outlook tended to be expressed by our interviewees referred via the primary care services who had less severe and/or relatively short-lived mental health problems. In contrast to a perception that things were unlikely to change, they were more likely to talk about having goals and plans in place which they were actively working towards, however difficult. They were also more likely to compare themselves positively with others. Religion also provided solace for some. However, this optimism may in some respect have been due to these participants being recruited through a service specialising in cognitive behavioural therapy interventions which focus on goal setting and positive thinking techniques.

I think that I have turned the corner I’ve settled down even though I probably might be on my own now, but I’ll have to wait and see about that . . . alright I have got problems, but there are people worse off than I am, they have got more problems than me and if some people can do it I think I should be able to do it myself[.]

Participant 3 (IAPT, depression/anger)

EQ-5D

The content of the EQ-5D was presented in Table 1 . As shown in Tables 25–31 there is only a modest amount of overlap between the EQ-5D and the seven themes.

SF-36

The SF-36 covers more emotional aspects of health than the EQ-5D by including mental health, role limitations due to emotional problems, vitality and aspects of general health (see Table 2 ). This is reflected in the results in Tables 25–31 .

SF-6D

The preference-based SF-6D is derived from the SF-36 and so shares much of the same content. However, there are some important differences owing to the fact that it contains only a subset of 11 of the 36 items (see Table 3 ). The SF-6D does not cover general health perception and this means that it does not have any item for tapping the hope and hopelessness theme. It combines role limitation due to emotional problems with that due to physical problems. In addition, the SF-6D does not cover positive affect or negative energy (i.e. tiredness).

ICECAP-A

ICECAP-A is a classification like the EQ-5D with a single item per dimension. The five dimensions are feeling settled and secure; love, friendship and support; being independent; achievement and progress; and enjoyment and pleasure. The content of this instrument would seem from the summary in Tables 25–31 to be more closely aligned with the themes identified from qualitative work, and to some extent this is reassuring as they are based on a similar qualitative approach. The main difference is that the ICECAP-A was based on interviews with members of the general public rather than those with mental health problems.

Summary of findings

Our review and synthesis of qualitative research studies undertaken with people with mental health problems, reported in the previous chapter, identified six domains of quality of life: well-being and ill-being; control, autonomy and choice; self-perception; belonging; activity; and hope and hopelessness. Despite widening the types and severities of mental health problems studied, we found our interview data fitted well with the themes from the review; any differences tended to be within the themes and related to the degree of impact of the themes on different levels of severity, chronicity and diagnosis. With some exceptions, those with severe chronic difficulties were more likely to talk about losses and what took quality away from life, whereas those with moderate or relatively short-lived problems spoke of the things that added quality to life. The EQ-5D and SF-36 were found to have only a modest overlap with the seven themes identified from the research, and though the ICECAP-A had better overlap, it only considers the positive end of the spectrum.

Strengths and limitations

One of the primary aims of the interview research was to ascertain what was important to quality of life for people with a wider range of mental health disorders and levels of severity than was available within the review. To a large extent we achieved this, with interviews being undertaken with people recruited from primary and secondary services catering to the needs of people with severe and enduring problems and mild to moderate anxiety and depression. However, despite attempts, we did not receive any referrals from these services of women with psychosis-related disorders, or people with obsessive compulsive disorder. Our findings may not, therefore, be inclusive of their views. There was also an over-representation of referrals from one member of the CMHT who was a nurse. Although we are unaware of any biases that may have been introduced as a result of this, it is possible that the perspective of those being seen by a nurse in the team could be different from those whose key worker was a psychologist, occupational therapist, etc. Finally, we had hoped that practitioners would introduce the possibility of taking part in the research to all their current clients and would let them decide whether they felt able, or wished, to be interviewed. However, the number of referrals received strongly suggests that service providers acted as gatekeepers and the voices of all types of service users were not necessarily included. The implication is that further interview work could be undertaken to address these gaps.

Conclusions

The review and synthesis of qualitative research undertaken with people with mental health problems, and the interviews reported in this chapter, have together identified seven domains of quality of life: well-being and ill-being; control, autonomy and choice; self-perception; relationships and belonging; activity; hope and hopelessness; and physical health. It is important that these domains are addressed in some way in quality-of-life measures. People enter mental health services with varying degrees of distress and chronicity, and as our research shows, recovery can be a long and difficult process. However, people may perceive large improvements in their quality of life without registering on the positive end of any quality-of-life scale, although undoubtedly this should be the aim. It is therefore necessary that the full spectrum of negative through to positive aspects of each domain is included within quality-of-life measures. The EQ-5D and the SF-36 and its derivatives were found not to cover large elements of the seven themes identified by the research, and this calls into question the content validity of these measures for large parts of the population with mental health problems.

Psychometric evidence

Systematic reviews were undertaken of the psychometric literature in five mental health conditions: depression, anxiety, bipolar disorder, personality disorder and schizophrenia. Overall, the results from 91 studies identified by an exhaustive search of the literature were used to assess the construct validity and responsiveness of the EQ-5D and SF-36.

Generic measures were found to adequately reflect differences between groups or changes over time in populations diagnosed with depression. In populations with anxiety, the evidence was less clear as the differences between known groups may have been driven by comorbid depression rather than anxiety disorders themselves. For personality disorder, most studies supported the construct validity of the EQ-5D, but there was insufficient evidence on the SF-36. Within schizophrenia the evidence demonstrated known-group differences, but this was mostly limited to differences between individuals diagnosed with schizophrenia and the general population. Contradictory evidence was found in studies using clinical measures of symptom severity, where the generic measures failed to reflect differences or correlate with the clinical indicators. In bipolar disorder, generic measures reflected known differences in clinical measures of depression but not mania.

The amount of evidence found in the literature was limited in terms of size and coverage, so it was decided to expand the evidence base by undertaking further psychometric analysis of a number of existing data sets. This provided more evidence in depression, anxiety, personality disorder and schizophrenia and used more patient-based assessments in the validation. These analyses broadly supported the findings of the reviews, with the EQ-5D and the SF-6D being found to be valid in samples with mild and moderate depression and anxiety. For schizophrenia, the findings were less clear, with the EQ-5D and SF-6D not being responsive to change in comparison with the condition-specific measures.

The tests of construct validity and responsiveness tend to be rather crude, as they depend on the validity of the construct used to assess the criteria. Tests of known-group validity, for example, usually either rely on comparison with the general population or use clinical assessments of severity that may not reflect meaningful differences in quality of life from the perspective of the population with mental health problems. Such evidence cannot prove or disprove the validity of a measure; at best it can provide support for appropriateness of the measure. The findings from our review and the further analyses seem to highlight potential concerns in anxiety, bipolar disorder and schizophrenia (particularly with regard to responsiveness to treatments). It is important to judge these findings alongside qualitative evidence on the content validity of the measure.

The psychometric evidence presented an interesting and mixed picture in terms of the performance of the EQ-5D and SF-36 in populations with mental health problems. However, this quantitative evidence is not able to offer an explanation for these mixed results. For this reason, qualitative research was undertaken on the impact of mental health problems on quality of life.

Qualitative evidence

The review of previous research found 13 studies that had interviewed people with mental health problems about the way their condition impacts on their lives. It was difficult to be sure that all informative studies had been identified given differences in terminology. The findings from those studies located were synthesised using the ‘framework’ approach. 198 We identified six domains from this review: well-being and ill-being; control, autonomy and choice; self-perception; belonging; activity; and hope and hopelessness.

The complementary primary research involved semistructured interviews of people with mental health problems. These interviews expanded the scope of diagnosis and severity of illness covered by the studies in the review to include people with affective and anxiety disorders referred via NHS services, and those with severe and enduring mental health problems and mild to moderate depression/anxiety. We found that our interview data corresponded well with the themes from the review, and any differences tended to be within the themes and related to the degree of impact of the different levels of severity, chronicity and diagnosis. The only change to the themes was that physical health was found to be more important among the interviewees than suggested from the review, so this was added as a seventh theme. The review and interview data found that each theme had a spectrum of positive and negative components.

Mapping between mental health and generic scales

When generic measures are not being used, one solution is to estimate mapping functions where condition-specific measures are regressed onto generic measures to produce health-state values. We estimated mapping functions between widely used condition-specific measures (HADS, CORE-OM, PHQ-9, GAD-7, and GHQ-12) for depression and anxiety and the EQ-5D and SF-6D, using four data sets available to the investigators. Mapping functions were not estimated for other conditions as the psychometric evidence suggested that the generic measures were not appropriate.

The statistical models mapping the HADS onto the EQ-5D had poor predictive performance. The mapping functions for the SF-6D fitted the data sets better, particularly the CORE-OM onto the SF-6D, but they still suffered from some degree of over- and underprediction towards the ends of the ranges. Given the psychometric evidence that the generic measures performed satisfactorily in this group, this result is perhaps surprising. These analyses were limited by the small size of some of the data sets and the condition-specific measures available, with the EQ-5D only being mapped onto the HADS. Furthermore, most of the mental health measures only focus on mental health, and often on very narrow aspects of symptoms of mental health, which may not translate into a general health score. We conclude that these mapping functions should not be used, but that the original generic measures should be used in trials and other clinical applications in order to obtain accurate estimates.

3.1 People and track record

Applicants: The team has the expertise required to undertake the study. John Brazier is a recognised international expert on the development of preference-based measures for use in economic evaluation. His work includes the development of the SF-6D and a number of condition specific measures. He currently has an MRC grant to develop a new method for mapping between measures of HRQL. Alicia O’Cathain, is a mixed methodologist with special expertise in qualitative methods and has worked in the area of quality of life research for many years. Michael Barkham is also a clinical psychologist with an international reputation in outcome measurement in mental health including the development of the CORE-OM (Barkham et al. , 2001). Myfanwy Lloyd Jones has extensive experience in conducting systematic reviews, including in the area of qualitative evidence. Donna Rowen has expertise in estimating mapping functions. Simon Gilbody is an expert in all aspects of evidence synthesis and mental health outcome measurement, and is Editor of the Cochrane Depression and Anxiety Group. He is also an Honorary Consultant psychiatrist and psychotherapist within the local NHS trust. David Richards has a background in nursing and behaviour therapy, and has been a leading figure in the expansion of provision of psychosocial care for common mental health problems in response to the Layard agenda. He has developed and implemented low intensity-high volume care in the National Improving Access to Psychological Therapies (IAPT) demonstration site in Doncaster. Sarah Byford and Martin Knapp are senior mental health economists with considerable experience with using the EQ-5D and SF-36 across a range of MH services.

Collaborators: Glens Parry has considerable research and policy experience and is leading on the SDO-funded evaluation of the DH demonstration sites for increasing access to psychological therapies. Eleni Chambers is a user researcher with expertise in advising on issues related to public and patient involvement at all stages of the research process, from pre-protocol planning to publications/dissemination. She is a member of various national and local networks and user groups. Tony O’Hagan is an Emeritus Professor of Bayesian Statistics who has worked with JB for many years on mapping these types of data and will provide advice on key statistical aspects of the mapping.

3.2 Environment

The project will be led from the School of Health and Related Research (ScHARR) at the University of Sheffield, but is very much collaboration with the University of York and Kings College London, Institute of Psychiatry. ScHARR is a large multi-disciplinary School within the Faculty of Medicine at the University of Sheffield with over 200 staff of social scientists, epidemiologists, statisticians, public health specialists and health economists. It provides a stimulating and supportive environment in which to conduct applied and methodological health services research. The health economics groups has developed a special expertise in the development and testing of outcome measures, including preference-based measures. The Mental Health Section provides the local clinical and academic support needed to conduct the study. ScHARR is able to provide the full range of facilities, including excellent library and information resources, office space and secretarial support.

3.3 Research Plans

Part 1: Testing the properties of two generic measures in mental health

This will adopt the standard criteria for assessing measures recommended by Fitzpatrick et al. (1999) and adapted for preference-based measures by Brazier et al. (1999) that include feasibility, reliability, validity (content validity, face validity and construct validity) and responsiveness.

Feasibility is a problem general to all instruments, but some are more prone to non-response and missing data than others due to length, complexity and relevance to patients (McColl et al. , 2000). This has important implications for the power and validity of studies. Reliability over time is a desirable characteristic in any measure but it is strictly encapsulated in responsiveness. An unreliable instrument will not be very responsive. Internal reliability is an even less direct way at measuring the key property of responsiveness, though it can provide useful evidence of item functioning across different populations in different settings. Validity is the extent to which an instrument measures what it is intended to measure. In the absence of a gold standard measure for quality of life, various indirect ways of establishing validity will be used in this study including content and face validity, construct validity and concurrent or convergent validity (Streiner and Norman, 1989; Bowling, 2002). Content and face validation are best tested using qualitative methods based on interviews with the relevant population. Construct and current validation require data sets with indicators or measures of severity that can be used to assess group differences and degree of association.

Responsiveness is the ability of an instrument to measure clinically significant changes in health. Responsiveness is usually assessed statistically using measures such as the ‘effect size’, where the mean change in score is divided by either the standard deviation at baseline or the standard deviation of the change. A common assumption in the assessment of responsiveness is that for a given health change, the measure with the largest effect size is the better measure (Katz et al. , 1994). Where the objective is to minimise sample size this makes sense and maximises the ability to detect differences in the context of performance assessment. For economic evaluation this is less important, and instead it is simply a case of establishing that the descriptive content is able to detect change of significance to users. Related to this is precision which is concerned with the ability of a measure to detect changes over the range experienced by patients being assessed. This requires the items and levels of the domains to be well spread over the range of the measure.

These criteria are going to be assessed using evidence from a range of different sources: the literature, interviews with users and a number of existing data sets. They will be applied to three broad groups of people with mental health problems: 1) common mental health problems (e.g. mild to moderate depression, anxiety, obsessive compulsive disorder and panic disorders); 2) severe and complex non-psychotic disorders (e.g. personality disorder); 3) schizophrenia and other psychotic disorders. These groups are not mutually exclusive, but are designed to help achieve a sample that covers the main groups of users. It is anticipated that much of the existing evidence will be by specific mental health disorder.

Part II: Estimation of mapping functions between mental health specific measures and the generics

For those situations where the EQ-5D and/or SF-6D prove to be acceptable in terms of their psychometric properties, then the next stage in the project will be to estimate mapping functions between those mental health specific measures widely used in clinical studies onto the two generic measures. The PI recently completed a review of mapping functions (Brazier et al. , 2007) and will be drawing on the best practice identified in that review.

Regression analysis will be used to estimate the relationship between the preference-based scores of these two measures and the various mental health measures across different mental health disorders and conditions. One set of models will use the dependent variables of the EQ-5D and SF-6D indices. The explanatory variables will be the various mental health instruments entered into the models a variety of ways including dimension scores with main effects, squared terms and interaction terms. Models will also explore using item and item response variables. The inclusion of other covariates, including age and gender, will also be examined.

More complex specifications will include random effects generalised least squares (to allow for clustering in the data – usually around the individual), random effects (RE) tobit and censored least absolute deviations (CLAD) models. The RE tobit model is an appropriate alternative for estimating models using censored data (Sullivan and Ghushchyan, 2006) and as EQ-5D utility scores exhibit a ceiling effect, where a large proportion of subjects rate themselves in full health with a utility score of 1. CLAD will also used since its estimator does not depend upon assumptions of homoscedasticity and normality and hence produces consistent estimates in the presence of heteroscedasticity and non-normality (see Powell, 1984, and Sullivan and Ghushchyan, 2006). Another approach will be explored using EQ-5D and SF-6D dimensions as the dependent variables. A multinomial logit model will be used to estimate the probability that a respondent will choose a particular level for each dimension of each instrument (Grayet al. (2006)). Subsequently predicted EQ-5D level responses for each dimension are generated using Monte Carlo simulation methods and the corresponding preference-based index utility score for that health state is calculated using the relevant value sets. Bayesian approaches to modelling the data will also be developed drawing on work with modelling SF-6D using Gaussian approaches (Kharoubi et al. , 2007) that proved more flexible and better at predicting health state values.

Model performance will be mainly assessed in terms of their ability to predict EQ-5D/SF-6D index values in independent data sets, including the difference between predicted and observed values at either the aggregate health state level by calculating Mean Error (ME); or the individual level by calculating the Mean Absolute Error (MAE) or the Root Mean Squared Error (RMSE). Mean error will be examined by severity of health problem. The main purpose is to predict differences in values across sub-groups of patients, such as between arms of a trial or over time. So models will be examined in terms of the size and direction of error in their prediction of differences between sub-groups in independent data.