5.1. INTRODUCTION
5.1.1. Recognition of depression
As was described in Chapter 2, most people with depression who consult their GP are not recognised as depressed, in large part because most such service users are consulting for a somatic symptom and do not consider themselves as depressed despite the presence of symptoms of depression (Kisely et al., 1995). Symptoms such as fatigue, insomnia and chronic pain, which are associated with depression, may be missed because they are attributed to a physical disorder (a disorder that is actually comorbid with depression).
It has been shown that only around 30% of people presenting with depressive disorder are diagnosed and offered treatment (Marks et al., 1979). This is a source of concern, although it is probably more likely for mild rather than more severe disorders (Kessler et al., 2002a). The consequences of this poor recognition are uncertain. In a large multicentre WHO study of mental disorders, episodes of depression that were either untreated by the GP or missed entirely had the same outlook as treated episodes of depression; however, they were milder at initial consultation (Goldberg et al., 1998). A smaller UK study (Kessler et al., 2002b) found that the majority of people with undetected depression either recovered or were diagnosed during the follow-up period; however, almost 20% of individuals included in the study remained undetected by a healthcare professional and unwell after 3 years.
People with depression with more severe disorders or who present with psychological symptoms are more likely to be recognised as depressed. This is an undesirable state of affairs as large numbers of people experience depression each year, which has major implications for primary care.
5.1.2. Recognition of anxiety disorders
As with depression, anxiety symptoms are also often not recognised by primary healthcare professionals because service users may not complain of them directly (Tylee & Walters, 2007). Modes of presentations for anxiety disorders, which initially may not be recognised as being due to anxiety, include frequent attendance with multiple symptoms that may initially be considered as possible symptoms of a physical disorder (Blashki et al., 2007).
Given that anxiety disorders are often associated with physical symptoms, it is not surprising to find that they are also commonly found among people receiving care in non-psychiatric secondary care settings. As in primary care, these frequently go unrecognised unless clinicians specifically look out for these disorders during routine consultations (Kroenke et al., 1997).
Based upon surveys in hospital settings, OCD is also common among people with chronic physical health problems. For example 20% of UK dermatology outpatients (Fineberg et al., 2003) and 32% of people presenting to rheumatologists and dermatologists with systemic lupus erythematosis (Slattery et al., 2004) met criteria for OCD.
The nature of some symptoms of panic attacks, such as palpitations, tachycardia, shortness of breath and chest pain, may lead some individuals to think that they are experiencing a potentially life threatening illness such as a heart attack. This often results in presentation to accident and emergency departments. It has been estimated that between 18 and 25% of patients who present to emergency or outpatient cardiology settings meet the criteria for panic disorder (Huffman & Pollack, 2003), which is often not recognised.
The problem of under-recognition for anxiety disorders has recently been highlighted by evidence that the prevalence of PTSD is significantly under-recognised in primary care (Ehlers et al., 2009). Many individuals will consult their GP shortly after experiencing a traumatic event, but will not present a complaint or request for help specifically related to the psychological aspects of the trauma; for example, an individual who has been physically assaulted, or involved in a road traffic accident or an accident at work might present requiring attention to the physical injuries sustained. Similarly, individuals who have been involved in traumatic life events often present at local emergency departments, notification of which is sent to GPs.
In both anxiety and depressive disorders, the initial presentation and complaint may take the form of somatic symptoms alone, such as lethargy or poor sleep in the case of depression and palpitations or muscular tension in the case of anxiety disorders. In light of this fact, consideration should be given to these symptoms as possible indicators of a common mental health disorder, in particular where no physical cause of these symptoms is apparent. Finally, as was also noted in Chapter 2, one major reason for poor recognition of common mental health disorders has been found to be a lack of effective consultation skills on the part of some GPs.
5.2. CASE IDENTIFICATION
5.2.1. Introduction
The first NICE guideline on depression, Depression: Management of Depression in Primary and Secondary Care (NICE, 2004b; NCCMH, 2004a), in addition to other NICE mental health guidelines, considered the case for general population screening for a number of mental health disorders, and concluded that it should only be undertaken for specific high-risk populations where benefits of early identification outweigh the downsides, such as people with a history of depression, significant physical illnesses causing disability or other mental health problems such as dementia. The criteria by which NICE judged the value of this approach was set out by the UK NHS National Screening Committee. Additional experience on the use of case identification strategies for depression has emerged since this approach was endorsed and incentivised under the NHS primary care QOF (British Medical Association & NHS Employers, 2006).
Recently, the updated edition of the Depression: the Treatment and Management of Depression in Adults guideline (NCCMH, 2010b) and the guideline on Depression in Adults with a Chronic Physical Health Problem (NCCMH, 2010a) reviewed available case identification instruments for depression. These guidelines recommended that healthcare professionals should be alert to possible depression (particularly in people with a past history of depression or a chronic physical health problem with associated functional impairment) and consider asking people who may have depression two questions, known as the ‘Whooley questions’:
During the last month, have you often been bothered by feeling down, depressed or hopeless?
During the last month, have you often been bothered by having little interest or pleasure in doing things?
If a person answers ‘yes’ to either of these questions, then the guidelines recommend that a practitioner who is competent to perform a mental health assessment should review the person's mental state and associated functional, interpersonal and social difficulties. Furthermore, when assessing a person with suspected depression, the guidelines recommend that practitioners should consider using a validated measure (for example, for symptoms, functions and/or disability) to inform and evaluate treatment.
Compared with depression, the case for the routine identification of anxiety has received less attention. Nevertheless, the epidemiology and unmet needs attributable to anxiety are similar to that of depression. There has been less experience to draw upon relating to the use of routine case identification measures in primary care in the UK, since their use has not previously been endorsed by the UK QOF. A decision was taken when formulating the scope of this guideline that the use of brief anxiety questionnaires should be considered alongside depression questionnaires. The classification of anxiety disorders is more complex compared with depressive disorders; however, the view of the GDG was that the development of separate case identification questions for each type of anxiety disorder would very likely be impractical and have no utility for routine use in primary care. The GDG's preference was to explore the possibility that a small number of case identification questions with general applicability to a range of anxiety disorders should be the starting point for a review in this area. As with depressive disorders, a potentially positive response would then prompt a further assessment.
Given that case identification instruments were recently reviewed for the two guidelines on depression, the present review will focus on case identification instruments for anxiety or GAD.
Definition
Case identification instruments were defined, for the purposes of this review, as validated psychometric measures that were used to identify people with anxiety. The review was limited to instruments likely to be used in UK clinical practice, that is, ‘ultra-brief instruments’ (defined as those with one to three items) or ‘longer instruments’ (defined as those with four to 12 items). The identification instruments were assessed in consultation samples (which included primary care and general medical services) and community populations. ‘Gold standard’ diagnoses were defined as DSM or International Classification of Diseases (ICD) diagnosis of anxiety; studies were sought that compared case identification with an ultra-brief or longer instrument with a gold standard. Studies that did not clearly state the comparator to be diagnosis by DSM or ICD, used a scale with greater than 12 items, or did not provide sufficient data to be included in the review were excluded.
5.2.2. Clinical review protocol
The review protocol, including the review questions, information about the databases searched, and the eligibility criteria used for this section of the guideline, can be found in (further information about the search strategy can be found in Appendix 6).
Clinical review protocol for the review of case identification tools.
5.2.3. Studies considered
The literature search for observational studies yielded 3,849 papers. Further inspection of these identified a total of 20 studies (N = 15,344) that met the eligibility criteria for this review: BYRNE2011 (Byrne & Pachana, 2011), CAMPBELL2009 (Campbell-Sills et al., 2009), DENNIS2007 (Dennis et al., 2007), EACK2006 (Eack et al., 2006), GILL2007 (Gill et al., 2007), HALL1999 (Hall et al., 1999), HAWORTH2007 (Haworth et al., 2007), KRASUCKI1999 (Krasucki et al., 1999), KREFETZ2004 (Krefetz et al., 2004), KROENKE2007 (Kroenke et al., 2007), LANG2009 (Lang et al., 2009), LOVE2002 (Love et al., 2002), MEANS-C2006 (Means-Christensen et al., 2006), NEWMAN2002 (Newman et al., 2002), POOLE2006 (Poole & Morgan, 2006), SMITH2006 (Smith et al., 2006), STARK2002 (Stark et al., 2002), WEBB2008 (Webb et al., 2008), WHELAN2009 (Whelan-Goodinson et al., 2009) and WILLIAMSON2005 (Williamson et al., 2005). Of the 20 studies, five were conducted using a sample of older people, 15 were conducted using consultation samples (six of these were in primary care), five were conducted using a community sample and eight were conducted using people with chronic physical health problems. All studies were published in journals between 1999 and 2010. Of the excluded studies, a number did not meet one or more eligibility criteria (112 were about depression only, 30 did not use an appropriate gold standard, 74 were about an instrument with more than 12 items and 294 reported data for a non-English language instrument) or could not be evaluated (438 reported insufficient data and 53 were not available as full text). Further information about both included and excluded studies can be found in Appendix 14.
5.2.4. Evaluating identification instruments for anxiety
Review Manager Version 5.0 (Cochrane Collaboration, 2008) was used to summarise diagnostic accuracy data from each study using forest plots and summary ROC plots.13 Where more than two studies reported appropriate data, a bivariate diagnostic accuracy meta-analysis was conducted using Stata 10 (StataCorp, 2007) with the MIDAS (Module for Meta-analytical Integration of Diagnostic Test Accuracy Studies; Dwamena, 2007) command to obtain pooled estimates of sensitivity, specificity, likelihood ratios and diagnostic OR (for further details, see Chapter 3). To maximise the available data, the most consistently reported and recommended cut-off points for each of the scales were extracted (see ).
Cut-off points used for each of the case identification instruments.
Heterogeneity is usually much greater in meta-analyses of diagnostic accuracy studies compared with RCTs (Cochrane Collaboration, 2008; Gilbody et al., 2007). Therefore a higher threshold for acceptable heterogeneity in such meta-analyses is required. However when pooling studies resulted in I2 >90%, meta-analyses were not conducted.
Only three instruments were evaluated for diagnostic accuracy by more than one study (HADS-A, eight studies; GHQ-12, two studies; GAD-Q-IV, two studies), but only the HADS-A studies could be meta-analysed. and summarise the diagnostic accuracy data for each of the ultra-brief instruments. shows a forest plot of the sensitivity and specificity for the ultra-brief instruments. , , and summarise the diagnostic accuracy data for each of the longer instruments. and summarise the results of the meta-analysis of the HADS-A.
Evidence summary table for ultra-brief instruments (one to three items).
Summary ROC plot of tests for ultra-brief instruments.
Forest plot of sensitivity and specificity for ADD (GAD item), GAD-Q-IV item 2, VAS, GAD-2 and ADS-GA (three items).
Evidence summary table for longer instruments (four to 12 items).
Forest plots of sensitivity and specificity for ADS-GA (four items), OASIS, GAI-SF, BSI-Anxiety – six item, BAI-FS and GAD-7.
Forest plots of sensitivity and specificity for PHQ-A, RSCL, PSWQ-A, GAD-Q-IV, ADS-GA, EPQ-N, GAD-12 and MCS-12.
Summary ROC plot of tests for longer instruments.
Forest plot of sensitivity and specificity for HADS-A.
Summary ROC plot of tests for HADS-A, by age group.
5.2.5. Clinical evidence summary
Only the HADS-A had enough evidence to synthesise the results using metaanalysis, although it should be noted that no studies were conducted in primary care. Therefore, the clinical utility of all instruments should be interpreted with some caution.
With regard to ultra-brief instruments (defined as those with one to three items), for the identification of any anxiety disorder or GAD in adults the GAD-2 had the best diagnostic accuracy for use in primary care. For the identification of GAD in older adults, the three-item version of the ADS-GA had the best accuracy. In secondary care only the VAS has been evaluated, but diagnostic accuracy was poor.
With regard to longer instruments (defined as those with four to 12 items), for the identification of GAD in adults, the GAD-7 had the best diagnostic accuracy for use in primary care. For the identification of GAD in older adults, the four-item version of the ADS-GA had the best accuracy for use in primary care. In secondary care, the GAI-SF has adequate accuracy for the identification of GAD in older people. In a community sample, the GAD-Q-IV had good accuracy for the identification of GAD in adults. No other instrument had adequate accuracy, including the HADS-A, although, as can be seen in , the HADS-A may have better diagnostic accuracy when used with older adults.
5.2.6. Health economic evidence
The systematic search of the economic literature undertaken for the guideline identified only one eligible study on identification methods of postnatal depression (Hewitt et al., 2009). This was an HTA and the validity (diagnostic accuracy), acceptability, clinical effectiveness and cost effectiveness of methods to identify postnatal depression were assessed. Although Hewitt and colleagues (2009) conducted an extensive systematic literature review, none of the studies retrieved comprised a full economic evaluation of a postnatal depression identification method. A model was therefore constructed to assess the costs and outcomes of different identification strategies. The analysis was conducted from the NHS/personal social services (PSS) perspective. Although Hewitt and colleagues (2009) found 14 different strategies to have been validated in the literature, a review analysis of the validity of the diagnostic accuracy of the different identification methods was conducted in the HTA report and those results were used to determine the identification strategies considered in the economic analysis. Specifically the analysis considered the Edinburgh Postnatal Depression Scale (EPDS) (cut-off points seven to 16) and the BDI (cut-off point ten), which were compared with current practice (that is, routine case identification without the formal use of a diagnostic instrument). The model consisted of two parts: (1) an identification model reflecting the diagnostic performance and administration costs of the alternative identification strategies; and (2) a treatment model that evaluated the subsequent costs and outcomes, expressed in quality-adjusted life years (QALYs).
Women were assumed to enter the model at 6 weeks postnatally. The source of clinical effectiveness data was a systematic review and meta-analysis. Resource use estimates were based on assumptions and the NICE Antenatal and Postnatal Mental Health guideline (NCCMH, 2007). National unit costs were used and were expressed in 2006–07 prices.
The analysis estimated that routine care was the least costly and least effective strategy. Strategies were ranked in terms of costs (from the least expensive to the most expensive); the EPDS cut-off points 7 and 13 and the BDI cut-off point 10 were dominated because they were more expensive and less effective than the previous strategies and were excluded. The incremental cost-effectiveness ratios (ICERs) of the non-dominant strategies were calculated and the ICER for the EPDS cut-off point 15 was found to be higher than that of the next more effective strategy on the ranked list, and thus was excluded due to extended dominance. Of the remaining non-dominated identification strategies, the EPDS at a cut-off point of 16 when compared with routine care resulted in an ICER of £41,103 per QALY. The ICER of the EPDS at a cut-off point of 14 was £49,928 per QALY compared with the EPDS cut-off point of 16. The ICER of the EPDS at lower cut-off points (for example, eight, nine to 11) exceeded £100,000 per QALY. At each of the three willingness-to-pay thresholds examined, namely £20,000, £30,000 and £40,000, the strategy with the highest individual probability of being cost effective was routine current practice, with probability reaching 88%, 59% and 39%, respectively. Formal identification strategies had a combined probability of being cost effective that increased with higher willingness to pay; this combined probability exceeded the respective probability of routine current practice at a willingness-to-pay between £30,000 and £40,000 per QALY. However, individual probabilities of each strategy were low, indicating a high uncertainty among the different formal identification strategies as to which is the optimal strategy in terms of cost effectiveness. The extensive sensitivity analyses performed demonstrated that the results were sensitive to the cost of treating false positive cases. This economic study has also been reported elsewhere (Paulden et al., 2010) and limitations of the model, and consequently the validity of the conclusions drawn, have been pointed out (Pilling & Mavranezouli, 2010) including the assumptions that a positive response to the case identification questions leads directly to the provision of a psychological intervention; the use in the model of only one type of intervention (the most costly); and the fact that the model assumes zero false positives in routine practice when the actual rate is likely to be nearer 15% (Mitchell et al., 2009).
Details on the methods used for the systematic review of the economic literature are described in Chapter 3; the evidence table with details of the study is presented in Appendix 10. The completed methodology checklist of the study is provided in Appendix 12.
5.2.7. Economic modelling
Introduction: the objective of economic modelling
The cost effectiveness of different identification methods for anxiety disorders was considered by the GDG as an area with likely significant resource implications. In addition, it was an area where available clinical data were adequate to allow the development of an economic model. Therefore, an economic model was constructed to assess the relative cost effectiveness of identification methods for people with anxiety disorders in the UK. In constructing this model, the GDG was concerned to model an element of the case identification and assessment pathway. Specifically, the model was designed to test whether the use of a brief case identification tool (the GAD-2), followed by the use of a more formal assessment method (the GAD-7), was more cost effective than standard care in the identification and initial assessment of anxiety disorders. In this case ‘formal assessment’ refers to the use of an additional psychometric measure (the GAD-7). ‘Further assessment’ refers to the routine clinical assessments that healthcare professionals would undertake to arrive at an informed and consensual decision with the person about the choice of treatment.
Economic modelling methods
Interventions assessed
The choice of interventions assessed in the economic analysis was determined after reviewing available relevant clinical data included in the guideline systematic literature review and the expert opinion of the GDG. Based on these, the following identification methods were assessed in the economic analysis: the use of the GAD-2 (cut-off point of three or more) followed by GAD-7 (cut-off point of eight or more), compared with GP assessment (that is, routine case identification without the formal use of a diagnostic instrument)14. It should be noted that the economic model focused on GAD because there were better data available relative to other anxiety disorders and it is one of the more commonly presenting anxiety disorders.
Model structure
A decision-analytic model in the form of a decision-tree was constructed using Microsoft Office Excel 2007 (Microsoft, 2007). According to the model structure, two hypothetical cohorts of people with GAD were initiated on one of the two identification strategies. People found positive for GAD with the GAD-2 were further assessed using the GAD-7. Depending on whether people undertaking the test did or did not have GAD and the outcome of the identification test, four groups of people were formed: true positive, true negative, false positive and false negative. Each of the four groups was assigned to a care pathway and followed up for 34 weeks. The care pathways for people identified as having GAD reflect the pathways described in the NICE Generalised Anxiety Disorder in Adults guideline (NCCMH, 2011a). People who were found to be true positive for GAD were assumed to receive one of the following treatment options, in proportions determined by the expert opinion of the GDG: (1) active monitoring (10%); (2) low-intensity psychological interventions (55%); (3) high-intensity psychological interventions (24.5%); and (4) pharmacological treatment (10.5%). Based on the Generalised Anxiety Disorder in Adults guideline (NCCMH, 2011a), low-intensity psychological interventions consisted of non-facilitated self-help, guided self-help and psychoeducational groups in equal proportions; high-intensity interventions consisted of CBT and applied relaxation; drug treatment consisted of sertraline for 8 weeks followed by 26 weeks of maintenance therapy with sertraline. Based on the duration required for pharmacological treatment, the time horizon of the analysis was 34 weeks. People who were found to be false positive for GAD received the same treatments in the same proportions as described for those who were found to be true positive, but were assumed to stop treatment earlier, and hence consumed only 20% of the healthcare resources (and consequently incurred 20% of the respective costs). People who were found to be false negative were assumed to receive no formal treatment, but incurred health and social care costs. People who were found to be true negative were assumed to receive no treatment and incur no health or social care costs. A schematic diagram of the model is presented in .
Schematic diagram of decision-tree constructed for the assessment of the relative cost effectiveness of different identification strategies for people with GAD – costs and outcomes considered in the analysis.
The economic analysis adopted the perspective of the NHS and PSS, as recommended by NICE (2009d). Costs consisted of identification costs (GP time), treatment costs for those identified as having GAD (low- and high-intensity psychological interventions as well as pharmacological interventions), and health and social care costs incurred by people with GAD that were not identified by one of the alternative strategies, or that were identified but did not respond to treatment. The measure of outcome was the QALY.
Clinical input parameters of the economic model
Clinical input parameters of the identification model included the sensitivity and specificity of the identification methods (GP assessment, GAD-2 and GAD-7). Sensitivity and specificity of the two formal identification methods were obtained from KROENKE2007, while the sensitivity of the GP assessment was obtained from Wittchen and colleagues (2002), and Davidson and colleagues (2010). The respective specificity of the GP assessment was a conservative estimate based on the expert opinion of the GDG. It must be noted that the model assumed that diagnostic characteristics of GAD-2 and GAD-7 were independent from each other, although administration of GAD-7 followed that of GAD-2. Regarding treatment, response rates were obtained from studies included in the systematic review and meta-analysis conducted for the Generalised Anxiety Disorders in Adults guideline (NCCMH, 2011a). Response rates for psychological interventions were estimated using an intention-to-treat approach. This means that response estimates accounted for the total number of people receiving psychological therapy, including dropouts. In contrast, response rates for people receiving pharmacological treatment did not account for people who discontinued treatment due to intolerable side effects, in accordance with the respective network meta-analysis undertaken for the Generalised Anxiety Disorders in Adults guideline. Given that the proportion of people receiving pharmacological therapy in the model is relatively low (10.5%), non-consideration of discontinuation due to intolerable side effects following pharmacological treatment was not expected to have significantly affected the results of the economic analysis. People under active monitoring (10% of people identified as GAD positive) were all assumed to improve because the GDG considered that initially a higher proportion of people who were identified as GAD positive would be assigned to active monitoring in UK routine clinical practice; however, some people would not improve and would, in reality, be offered one of the other treatment options described in the model. For simplicity, the model assumed that a lower percentage of people were offered active monitoring compared with routine practice, but all of them improved. The remaining people, who in routine practice would be transferred from active monitoring to other treatments following non-improvement, were assumed in the model to be initiated on other treatment options immediately following the identification of GAD.
Utility data and estimation of QALYs
To express outcomes in the form of QALYs, the health states of the economic model needed to be linked to appropriate utility scores. Utility scores represent the HRQoL associated with specific health states on a scale from 0 (death) to 1 (perfect health); they are estimated using preference-based measures that capture people's preferences on the HRQoL experienced in the health states under consideration. The utility scores for specific health states associated with GAD were obtained from Allgulander and colleagues (2007), consistent with the Generalised Anxiety Disorder in Adults guideline (NCCMH, 2011a).
Cost data
Costs associated with the identification methods were calculated by combining resource use estimates (GP time) with respective national unit costs (Curtis, 2009). It was assumed that administration of GAD-2 and GAD-7 required 10 and 15 minutes, respectively, whereas routine GP assessment required on average one GP visit.
Costs of psychological treatments (low- and high-intensity psychological interventions) were estimated using average estimates of the respective costs from the Generalised Anxiety Disorder in Adults guideline (NCCMH, 2011a). Costs of pharmacological treatment were also based on the NICE guideline (NCCMH, 2011a), using treatment with sertraline as reference. People who were falsely detected as having GAD were assumed to incur 20% of the treatment cost of a true positive person, according to the GDG's estimate. Health and social care costs of people falsely negative (that is, people having GAD but not identified by the methods assessed in the model), as well as the respective costs of people not responding to treatment, were taken from the NICE guideline (NCCMH, 2011a), based on resource use data reported in the most recent adult psychiatric morbidity survey in England (McManus et al., 2009), further expert opinion (GDG for the Generalised Anxiety Disorder in Adults guideline) and national unit costs (Curtis, 2009).
All costs were expressed in 2009 prices. Discounting of costs and outcomes was not necessary since the time horizon of the analysis was shorter than 1 year.
reports the values of all input parameters utilised in the economic model, and provides details on the sources of data and methods that were used at the estimation of input parameters.
Input parameters utilised in the economic model of pharmacological treatments for people with GAD.
Data analysis and presentation of the results
A sensitivity analysis was undertaken where data were analysed as point estimates; results were presented as mean total costs and QALYs associated with each identification method. Subsequently, the ICER is calculated, which expresses the additional cost per additional unit of benefit associated with one identification method relative to its comparator. Estimation of such a ratio allows consideration of whether the additional benefit is worth the additional cost when choosing one treatment option over another.
One-way sensitivity analyses explored:
The impact of the uncertainty characterising the sensitivity and specificity of the identification methods assessed. A scenario of 10%, 15% and 25% change in these estimates was tested to investigate whether the conclusions of the analysis would change. Furthermore, two-way sensitivity analyses on sensitivity and specificity was also performed to further investigate uncertainty around those parameters. A scenario of 10% and 20% simultaneous change in those parameters was tested.
The impact of changes in the consultation time necessary for the performance of the GAD-2 and GAD-7, respectively. A scenario of 25% change in these estimates was tested to investigate whether the conclusions of the analysis would change.
The impact of the cost incurred by those falsely detected as having anxiety. A scenario of 25% change of the estimate used was tested to investigate whether the conclusions of the analysis would change.
The impact of uncertainty characterising treatment costs was assessed. A scenario of 10%, 15% and 25% change of the estimates was tested to investigate whether the conclusions of the analysis would remain robust.
Last extreme scenarios were tested as to the percentage of those who were true positive for GAD who received the different treatment options. Holding the percentage of those under
active monitoring stable, the percentage that followed low-intensity psychological interventions, high-intensity psychological interventions and pharmacological treatment varied between 15 and 60%.
5.2.8. Economic modelling results
According to deterministic analysis, accounting for both identification and treatment costs, identification of anxiety using formal identification methods (namely GAD-2 and GAD-7), was estimated to be a cost-effective option because it resulted in a higher number of QALYs gained and lower total costs when compared with GP assessment without using a formal identification tool. Therefore, there was no need to estimate an ICER.
provides mean costs and QALYs for every identification option assessed in the economic analysis.
Mean costs and QALYs for each identification option for people with GAD assessed in the economic analysis – results per 1000 people.
Results were robust under all scenarios examined in one-way and two-way sensitivity analyses: formal identification of anxiety using GAD-2 and GAD-7 either remained the dominant strategy or resulted in a ICER when compared against GP assessment without an identification, which was well below £20,000 per QALY, the lower cost effectiveness threshold set by NICE (NICE, 2009d).
Discussion – limitations of the analysis
The results of the economic analysis suggest that the use of GAD-2 followed by GAD-7 as a tool for the identification of people with anxiety is a cost-effective option when compared with GP assessment alone (without using any formal identification tools). The cost effectiveness of the identification methods is mainly attributed to their diagnostic accuracy combined with the fact that they can be easily and quickly performed by GPs, resulting in relatively low intervention costs.
One of the limitations of the economic analysis is that, due to lack of available evidence, a number of the estimates used were based on GDG assumptions. Despite the fact that impact of their variability was assessed in sensitivity analysis, further research is required to compare the diagnostic performance of different identification tools for anxiety, case finding questions and generic anxiety measures, and to evaluate their impact on costs, resource use and health outcomes. As the treatment model was adopted from the Generalised Anxiety Disorder in Adults guideline, it is subject to the same limitations reported there (NCCMH, 2011a), along with the limitation that although GAD is one of the most commonly presenting anxiety disorders in primary care, the treatment outcomes that informed the model will be different for other disorders.
5.2.9. Overall conclusions from economic evidence
Existing economic evidence is particularly limited in the area of identification methods for people with common mental health disorders. The economic analysis undertaken for this guideline suggests that the use of formal identification tools (GAD-2 followed by GAD-7) comprises a cost-effective option when compared with GP assessment alone (without using formal identification tools) for people with GAD (as a proxy for the anxiety disorders), because it appears to result in better outcomes (more people identified and higher number of QALYs) and lower total costs.
5.2.10. From evidence to recommendations
The GDG agreed that diagnostic accuracy would be assessed using sensitivity, specificity, ROC curves, negative and positive likelihood ratios, and diagnostic ORs. Based on the approach taken for the updated editions of the Depression guidelines (NCCMH, 2010a and 2010b), when describing the sensitivity and specificity of the different instruments, the GDG defined values above 0.9 as ‘excellent’, 0.8 to 0.9 as ‘good’, 0.5 to 0.7 as ‘moderate’, 0.3 to 0.5 as ‘low’ and less than 0.3 as ‘poor’. For likelihood ratios, a value of LR+>5 and LR− <0.3 suggests the test is relatively accurate. For diagnostic ORs, a value of 20 or greater suggests a good level of accuracy.
The GDG aimed to develop recommendations that promoted the cost-effective identification of individuals with anxiety and depressive disorders. The recently completed NICE Depression guideline performed a systematic review of case identification methods for depression, and the GDG adapted the recommendations from the guideline for case identification (see NCCMH, 2010b, for a full description and discussion of the relevant evidence). In contrast, none of the other NICE guidelines that focused on anxiety disorders had developed recommendations for the identification of anxiety. In developing the recommendations for case identification for anxiety disorders, the GDG was mindful that the main focus of this guideline was on primary care services and the requirement to develop a method that not only had good sensitivity and specificity for all anxiety disorders but one that was also feasible (that is, had good clinical utility). The GAD-2 instrument performed best, and was the one measure that met the two key criteria of good diagnostic accuracy and feasibility. The GDG concluded that, although some of the longer instruments had good diagnostic accuracy, these instruments would not be feasible in the context of primary care. The GDG therefore decided to adopt the GAD-2 as the recommended measure. However, the GDG was concerned that the GAD-2 focused on anxiety and worry and that a number of people with an established phobic disorder would not be identified. This was thought possible because those with an established phobic disorder may avoid phobic objects or situations and, as a consequence of the avoidance, would not experience significant anxiety or worry and would therefore score low on the GAD-2. The GDG took the view that it was important to ask a subsidiary question for those people where the practitioner had a significant suspicion of an anxiety disorder but had returned a GAD score of less than three. The question, ‘Do you find yourself avoiding places or activities and does this cause you problems?’, was designed to detect a number of people with phobia whose functioning was impaired but who otherwise would not be identified by the two GAD questions. The GDG also undertook health economic modelling of the use of the two GAD questions. This modelled a larger part of the care pathway than was covered by the case identification questions alone and included the use of the GAD-7 (see Section 5.3) as part of the assessment subsequent to a positive response to the case identification questions. The economic model suggested that the combination of the case identification questions with the use of a formal assessment tool (the GAD-7) was cost effective and further supports the GDG's view that a recommendation for the case identification of anxiety disorders was appropriate. The recommendations set out below were developed in conjunction with the recommendations for assessment set out in Section 5.3.8 of this chapter.
5.2.11. Recommendations
- 5.2.11.1.
Be alert to possible depression (particularly in people with a past history of depression, possible somatic symptoms of depression or a chronic physical health problem with associated functional impairment) and consider asking people who may have depression two questions, specifically:
During the last month, have you often been bothered by feeling down, depressed or hopeless?
During the last month, have you often been bothered by having little interest or pleasure in doing things?
If a person answers ‘yes’ to either of the above questions consider depression and follow the recommendations for assessment (see Section 5.3.8)15.
- 5.2.11.2.
Be alert to possible anxiety disorders (particularly in people with a past history of an anxiety disorder, possible somatic symptoms of an anxiety disorder or in those who have experienced a recent traumatic event). Consider asking the person about their feelings of anxiety and their ability to stop or control worry, using the 2-item Generalized Anxiety Disorder scale (GAD-2; see Appendix 13).
If the person scores three or more on the GAD-2 scale, consider an anxiety disorder and follow the recommendations for assessment (see Section 5.3.8).
If the person scores less than three on the GAD-2 scale, but you are still concerned they may have an anxiety disorder, ask the following: ‘Do you find yourself avoiding places or activities and does this cause you problems?’. If the person answers ‘yes’ to this question consider an anxiety disorder and follow the recommendations for assessment (see Section 5.3.8).
- 5.2.11.3.
For people with significant language or communication difficulties, for example people with sensory impairments or a learning disability, consider using the Distress Thermometer16 and/or asking a family member or carer about the person's symptoms to identify a possible common mental health disorder. If a significant level of distress is identified, offer further assessment or seek the advice of a specialist.15
5.2.12. Research recommendations
- 5.2.12.1.
In people with suspected anxiety disorders. What is the clinical utility of using the GAD-2 compared with routine case identification to accurately identify different anxiety disorders? Should an avoidance question be added to improve case identification? (See Appendix 11 for further details.)
5.3. FORMAL ASSESSMENT OF THE NATURE AND SEVERITY OF COMMON MENTAL HEALTH DISORDERS
5.3.1. Introduction
Assessment of depression
Since April 2006, the UK GP contract QOF has incentivised GPs to measure the severity of depression at the outset of treatment in all diagnosed cases using validated questionnaires (British Medical Association & NHS Employers, 2006). The aim is to improve the targeting of treatment of diagnosed cases, particularly antidepressant prescribing, to those with moderate to severe depression, in line with the NICE guidelines.
The three recommended severity measures are the PHQ-9 (Kroenke et al., 2001), the depression subscale of the HADS (Wilkinson & Barczak, 1988; Zigmond & Snaith, 1983) and the BDI-II (Beck, 1996; Arnau et al., 2001). In general, a higher score on these measures indicates greater severity requiring greater intervention. However, the QOF guidance, again in line with the NICE guidance, also recommends that clinicians consider the degree of associated disability, previous history and patient preference when assessing the need for treatment rather than relying completely on the questionnaire score (British Medical Association & NHS Employers, 2006).
Data on the completion of the measures from the NHS Information Centre showed that they were used in a mean of 91% of diagnosed cases across all UK practices in 2007–08, up from 81% in 2006–07 (NHS Information Centre, 2008). The accuracy and utility of the measures has been questioned by GPs, however, suggesting that even if they use the questionnaires they may ignore the scores when deciding about treatment or referral (Jeffries, 2006).
Analysis of anonymous record data showed that prescriptions for antidepressants and referrals for psychiatric, psychological or social care were significantly associated with higher severity measure scores. However, overall rates of treatment and referral were very similar for service users assessed with different questionnaire measures, despite the fact that the different measures categorised differing proportions of service users as having major depression. These results suggested that practitioners (as would be expected given that factors such as associated functional impairment, duration of symptoms, patient preference and previous treatment) do not decide on drug treatment or referral on the basis of severity questionnaire scores alone (Kendrick et al., 2009).
Furthermore, qualitative interviews with GPs participating in the same study (Kendrick et al., 2009) showed that they were generally cautious about the validity and utility of identification tools, and sceptical about the motives for their introduction. The practitioners considered their practical wisdom and clinical judgement to be more important than identification tools and were concerned that the latter reduced the ‘human element’ of the consultation. Some even avoided coding patients' symptoms as depression in favour of other diagnostic labels, to avoid completing the severity measures and to save time in the consultation (Dowrick et al., 2009b). This emphasises the importance of ensuring that the introduction of new diagnostic techniques is done in such a way that it fits with existing practice and systems of care, and takes into account possible developmental or training needs of the practitioners expected to use the techniques.
Moreover, the recent Depression guidelines (NICE, 2009a and 2009b) have attempted to move away from focusing on any one aspect of the disorder, such as symptom severity, which can have the unwanted effect of leading to an oversimplified categorisation of depression, and influencing treatment choice, on a single factor such as symptom count. An important consideration was to provide a strong steer away from only using symptom counting to make the diagnosis of depression and by extension to emphasise that the use of symptom severity rating scales by themselves should not be used to make a diagnosis, although they can be important as an aid in assessing severity and response to treatment.
To make a diagnosis of depression requires assessment of three linked but separate factors: (i) severity, (ii) duration and (iii) impairment. The diagnosis of ‘major depression’ is based not only on the severity of depression, but also on its persistence, the presence of other symptoms and the degree of functional and social impairment. Individual symptoms should be assessed for severity and impact on function, and be present for most of every day. Service users who fulfil criteria for major depression of recent onset may improve spontaneously, and for those with mild depression or others whose symptom trajectory is showing improvement, it may be appropriate to ask the service user to come back for a review of symptoms in 1 to 2 weeks because a proportion will respond within a few weeks following some reassurance, psychoeducation and support from primary care staff without recourse to a formal intervention.
It is also important to emphasise that there appears to be no hard-and-fast ‘cut off’ between ‘clinically significant’ and ‘normal’ degrees of depression; the greater the severity of depression, the greater the morbidity and adverse consequences (Kessing, 2007; Lewinsohn et al., 2000). When taken together with other aspects that need to be considered, such as duration, stage of illness and treatment history, there are considerable challenges when attempting to classify depression into simple categories.
In recent years there has been a greater recognition of the need to consider depression that is ‘subthreshold’, that is, depression that does not meet the full criteria for a depressive/major depressive episode. Persistent subthreshold depressive symptoms – the preferred term in the Depression guidelines (NICE, 2009a and 2009b) – can cause considerable morbidity, and human and economic costs, and are more common in those with a history of major depression as well as being a risk factor for future major depression (Rowe & Rapaport, 2006).
Diagnosis using the three aspects listed above (severity, duration and impairment) provides a partial characterisation of the individual experience of depression. Depressed people vary in the pattern of symptoms they experience, their family history, personalities, pre-morbid difficulties (for example, sexual abuse), psychological mindedness, and current relational and social problems – all of which may significantly affect outcomes. It is also common for depressed people to have a comorbid psychiatric diagnosis, such as GAD, social anxiety disorder, panic disorder and various personality disorders (Brown et al., 2001), and physical comorbidity.
Depression is often accompanied by anxiety, and in these circumstances one of three diagnoses can be made: (i) depression (with anxiety symptoms), (ii) depression comorbid with a diagnosed anxiety disorder, or (iii) mixed depression and anxiety symptoms when the severity of symptoms for both depression and anxiety are below the threshold for either disorder.
Assessment of anxiety disorders
Compared with depression, the diagnosis, classification and epidemiology of anxiety disorders are inevitably more complex given the number of anxiety disorders. This raises particular challenges when developing simple but robust case identification and assessment strategies in this area, compared with depression.
According to the DSM-IV-TR (APA, 2000), there are six main types of anxiety disorders in adults: specific phobia, social phobia, PTSD, GAD, panic disorder (with or without agoraphobia) and OCD. Other defined anxiety disorders in adults include acute stress disorder and anxiety disorder not otherwise specified. ICD-10 (ICD, 10th revision, WHO, 1992) currently classifies disorders as phobic anxiety disorders, other anxiety disorders (including GAD and panic disorder), OCD, and reaction to severe stress and adjustment disorders. Despite their different classifications, there is consensus that the common theme throughout anxiety disorders is the overestimation of threat with anxiety that is characterised by fear and avoidance behaviour.
In the National Comorbidity Survey – Revised, the lifetime prevalence of anxiety disorders was 28.8% compared with 20.8% for mood disorders (Kessler et al., 2005a) and the median age of onset was 11 years old (compared with 30 years old for mood disorders). Specific and social phobias had the highest lifetime prevalence rate (12.5% and 12.1%, respectively), with agoraphobia without panic disorder having the lowest lifetime prevalence at 1.4% followed by OCD at 1.6%.
Unlike other disorders within DSM-IV, there are no ‘trumping’ rules within the main anxiety disorders hence it is possible to have co-occurring multiple anxiety disorders, and this is the case for a substantial minority of service users (Kessler et al., 2005b). In addition to the significant associations between the individual anxiety disorders, anxiety and mood disorders commonly co-occur (Rush et al., 2004). Approximately 50% of all people with depression meet criteria for at least one anxiety disorder and half of these individuals meet criteria for multiple anxiety disorders (Zimmerman et al., 2000). Co-occurring depression and anxiety can impact on treatment decisions (Petersen et al., 2009) and may also negatively affect treatment outcome (Brown et al., 1996).
This chapter considers the clinical utility of more formal assessments of the nature and severity of common mental health disorders (including problem specification or diagnosis). The following chapter covers the assessment of risk associated with the disorder – factors associated with response to treatment, including characteristics of the service user and their disorder, informing initial treatment choices – and also addresses the related issue of ROM.
5.3.2. Clinical review protocol
The aim of this review was to perform a narrative synthesis of existing NICE guidelines, and to supplement that synthesis using existing systematic reviews and recent RCTs that were not included in existing reviews, examining the clinical utility of more formal assessments of the nature and severity of common mental health disorders (including problem specification or diagnosis). In addition, the longer instruments reviewed for the identification of anxiety (see Section 5.2) were considered for use during formal assessment of anxiety disorders. The review protocol, including the review question, information about databases searched and the eligibility criteria used in this section of the guideline can be found in . Although the search was conducted for the period 1995 to 2010, the focus was only on systematic reviews published since 2003 (further information about the rationale for the method employed here can be found in Chapter 3 and the search strategy can be found in Appendix 6).
Clinical review protocol for the review of formal assessments.
5.3.3. Studies considered
Five existing NICE guidelines that were relevant to common mental health disorders were utilised:
Generalised Anxiety Disorder and Panic Disorder (With or Without Agoraphobia) in Adults (
NICE, 2011a) and
Generalised Anxiety Disorder in Adults (
NCCMH, 2011a)
In addition, the literature search for systematic reviews yielded 5,231 papers. Scanning titles/abstracts identified 34 potentially relevant reviews17; however, further inspection found none that met the eligibility criteria for inclusion. Evidence from supplementary searches identified one recent clinical guideline that specifically reviewed the evidence for formal assessment in people with a common mental health disorder in primary care (New Zealand Guidelines Group, 2008), and a case identification algorithm developed by the IAPT programme to support staff working in IAPT services to structure the form and content of a brief mental state review in primary care and related settings (IAPT, 2010). A list of included and excluded studies with reason for exclusion can be found in Appendix 14.
5.3.4. Summary of evidence from existing NICE guidelines
Evidence from existing NICE guidelines relevant to common mental health disorders were synthesised in two ways. The first utilised text from the body of each full guideline, while the second utilised recommendations from each guideline. For evidence from guideline text, a member of the technical team extracted any text that appeared to be relevant to assessment. As can be seen in , tabulation was then used to categorise relevant text from each guideline as relating to the clinical utility of formal assessment, as other information relevant to assessment, or as not relevant (text not shown). For recommendations, each guideline was examined for potentially relevant recommendations and these were tabulated (see ). Recommendations that were specific to risk assessment were removed from the table and can be found in the section on risk assessment in Chapter 6. Recommendations and associated text from existing guidelines were then used to develop recommendations that were either common to all common mental health disorders or disorder specific.
Preliminary summary and synthesis of relevant text (by guideline).
Preliminary summary and synthesis of relevant recommendations (by guideline).
5.3.5. Summary of clinical evidence from other sources
The New Zealand guideline on identification of common mental health disorders (New Zealand Guidelines Group, 2008) systematically reviewed the evidence for assessment instruments that were brief (less than 5 minutes) to administer. Included in their review was the first Depression guideline (NCCMH, 2004a), a review of screening for depression in adults (Pignone et al., 2002), two reviews of screening for alcohol problems (Aertgeerts et al., 2004; Fiellin et al., 2000) and 27 primary studies (some of which were included in the NICE guideline and/or the other included systematic reviews). The New Zealand guideline review concluded that the PHQ-9 appeared to have the best clinical utility for the assessment of depression, being reliable and valid for identifying depression, and sensitive to change. In addition, it was reported that other instruments with acceptable clinical utility were the GHQ-12 (Von Korff et al., 1987; Schmitz et al., 1999) and the Common Mental Disorder Questionnaire (CMDQ; Christensen et al., 2005). It was also stated that other brief tools, such as the Center for Epidemiological Studies Depression scale (CES-D; Fechner-Bates et al., 1994), the World Health Organization Wellbeing Index (WHO-5; Henkel et al., 2003) and Duke-Anxiety-Depression scale (Parkerson & Broadhead, 1997) are less accurate for routine use in primary care. The GAD-7 and the two-item version, GAD-2, (Kroenke et al., 2007; Spitzer et al., 2006) were described as valid for detecting anxiety disorders, and the GAD-7 was included as a potentially useful assessment tool. The Kessler-10 questionnaire was included as a potentially useful assessment tool, but described as only validated in secondary care (Andrews & Slade, 2001).
The IAPT screening prompts tool (IAPT, 2010) was developed on the basis of the diagnostic criteria contained in ICD-10 and also makes explicit links to the use of formal measures such as the PHQ-9. It sets out a stepwise approach to questions about the experience, duration of the symptoms and impact on functioning based on ICD-10 criteria. It is intended to be brief and can be integrated into a broader assessment of the presenting problem. It also drew on the algorithm for the differential diagnosis of anxiety and depressive disorders in the NICE Anxiety guideline (2004a). It was explicitly designed to aid IAPT staff and others working in primary care to differentiate between depression and the anxiety disorders.
5.3.6. Health economic evidence
The health economic evidence in support of the recommendations is contained in Section 5.2.6 to 5.2.9 of this chapter, using a model that was developed for the case identification questions (GAD-2) and the use of the GAD-7. The model was disorder-specific (focused on GAD). The model suggested that the combined case identification (GAD-2) and further formal assessment strategy (GAD-7) was likely to be cost effective.
5.3.7. From evidence to recommendations
The GDG aimed to provide appropriate and feasible advice for the assessment of common mental health disorders primarily focusing on primary care settings. The primary evidence source for these recommendations was drawn from existing NICE guidelines because no systematic reviews that met eligibility criteria for inclusion were found. Given the recent publication of the two Depression guidelines (NICE, 2009a and 2009b), and their recommendation for case identification and formal assessment, the GDG did not review those recommendations. However, given that there is reasonable evidence of the uptake of those recommendations (supported by the QOF), the GDG bore in mind the nature and structure of those recommendations when developing the recommendations for anxiety disorders. The recommendations drawn from individual guidelines drew on a range of different evidence sources including, where available, primary studies and systematic reviews, but also in many cases the expert advice of the GDG as high quality evidence was often lacking in this area. Therefore, the GDG took the view that the evidence developed by a range of expert groups on the best evidence available was the appropriate source from which to develop advice for assessment for this guideline. In doing so, the GDG aimed to develop recommendations that were feasible and that wherever possible had applicability across the full range of common mental health disorders, in particular the anxiety disorders. This meant that the GDG restricted their recommendations to self-completion questionnaires as the basis for any recommendations about formal rating scales because the increased time associated with the use of a clinician-rated measure would significantly detract from the use of the measure in routine care.
A number of key areas emerged where consensus and agreement was found across the five relevant guidelines. These included the manner in which the assessment should be undertaken, the content of the assessment (including the focus on the severity of symptoms), the associated functional impairment, the duration of symptoms, the use of formal rating scales (see Chapter 7), and also the previous experience of treatment and the impact on psychological factors. The GDG also considered the important area of comorbidity recognising that this is often high across the range of common mental health disorders. Taken together, this approach led to the development of recommendations about the methods by which to engage clients in the assessment process; to assess and evaluate their mental state; and the factors that may be taken into account, including previous treatment and any associated psychological and social factors. In doing so, the GDG was keen to develop recommendations that informed primary care staff on the important issue of immediate treatment and/or referral for further treatment. The GDG was also aware that they were drawing on existing evidence sources (NICE guidelines), the evidence for which had not been reviewed by the GDG. It was therefore important that the precise meaning of the recommendations drawn from other guidelines were not altered, although some rewording and restructuring of those recommendations was required to produce a coherent, clear and comprehensible set of recommendations for this guideline. The GDG was also conscious of the need to develop systems that might support the use of ROM and this was a further factor that influenced the structure and content of the assessment (again see Chapter 7 for a discussion of formal rating scales). The GDG took the view that the IAPT screening prompts tool could be of help to staff beyond those working in primary care as a way of structuring an assessment of mental state and therefore included it in the recommendations.
The GDG also considered that it was important to examine the cost effectiveness of these measures. As can be seen in Chapter 6 (Sections 6.2.6 and 6.3.6) the GDG chose to focus on the cost effectiveness of the case identification and assessment for GAD. This is one of the more commonly presenting, although under-recognised, common mental health disorders in primary care. Two key elements of the care pathway were assessed: the initial case identification for GAD and the use of the GAD-7 in addition to the standard clinical assessment. The model clearly indicated that such an approach may well be cost effective. The available data for other anxiety disorders were of poorer quality and so no other models were developed for other anxiety disorders. Given the broadly similar performance of the GAD-7 (in terms of sensitivity and specificity), and the fact that the treatment outcomes for the other anxiety disorders are also broadly comparable with, if not better than, those for GAD (NICE, 2011a and 2005b), the GDG took the view that an extrapolation from this model to other anxiety disorders was warranted. For depression, no model was developed because the approach used for case identification of depression is already well-established in the NHS and the evidence reviewed in the recent Depression guidelines (NICE, 2009a and 2009b) did not suggest any changes to current practice.
When drafting the recommendations, the GDG recognised that a number of key areas required further research. In particular, uncertainty remains about the accuracy and consequent identification of appropriate treatment by para-professionals in primary care. An assessment by a mental health professional will probably result in more accurate identification of problems and appropriate treatment, but is likely to entail greater cost and potentially significantly longer waiting times for interventions, both of which can have deleterious effects on care. In addition, a number of different ratings scales for depression and anxiety disorders are in current use, both in research studies and clinical practice. This makes obtaining comparative estimates of clinical outcomes at the individual level difficult when moving between research and clinical settings, and also between different clinical settings. A method that allows for prompt and easy ‘walking across’ between assessment instruments would have a potentially significant clinical benefit in routine care.
