Clinical effectiveness, cost-effectiveness and acceptability of low-intensity interventions in the management of obsessive compulsive disorder: the Obsessive Compulsive Treatment Efficacy randomised controlled Trial (OCTET)

Cognitive–behavioural therapy and low-intensity interventions

Therapist-delivered CBT, including exposure and response prevention (ERP), is a mainstay of psychological treatment for OCD. 10 CBT is a ‘talking therapy’ that aims to change unhelpful thoughts and behaviour. ERP are specific CBT techniques used in the treatment of OCD. Exposure therapy involves confronting the feared stimuli until fear subsides, through a process known as extinction. Response prevention means resisting the need to ritualise (i.e. perform a compulsive reaction).

One of the major problems related to the management of OCD is access to CBT, with many patients traditionally facing significant delays and spending large amounts of time on waiting lists for this treatment. This reflects a lack of trained therapists and the significant amount of therapist time traditionally seen as necessary for each individual treatment.

There has been significant interest in ways of overcoming this issue that are sustainable for health-care systems. A key development has involved ‘low-intensity’ interventions. Although there is a lack of consensus over precise definitions, common features of low-intensity interventions include:11

• typically CBT based

• communicate CBT principles in accessible ways and delivered in flexible forms (e.g. telephone, e-mail)

• involve a health technology (e.g. a computer package or book) to deliver key ‘active ingredients’ of therapy

• usually delivered by a low-intensity worker [psychological wellbeing practitioner (PWP)]

• use fewer resources (in terms of therapist time) than conventional CBT.

Low-intensity interventions can be used by patients unsupported (reducing therapist time to zero). However, they are often used in a ‘guided’ or ‘supported’ form, with some therapist time, often with a therapist who is fully trained but has less experience than a conventional CBT therapist. Two key forms of low-intensity interventions are ‘guided self-help’ (use of a book providing instruction in CBT principles), often supported via telephone by a professional; and computerised CBT (cCBT), using computer- or web-based systems to provide instruction in CBT principles, again supported by a professional.

An advantage of low-intensity interventions that use health technologies (cCBT) or are delivered by technology (telephone) is that they are more readily accessed by those who are unable to attend scheduled clinic appointments because of geographical, social, physical or psychological difficulties.

Low-intensity interventions complement, rather than replace, existing CBT provision. The way in which conventional therapist-delivered CBT and low-intensity treatments are delivered as part of a system of care is the subject of the next section.

National Institute for Health and Care Excellence guidelines

The NICE guidelines for OCD12 recommend management using a ‘stepped-care’ approach. 13 Stepped care is viewed as a potential solution to poor access to traditional treatments by increasing the efficiency of service provision and better-matching treatment to patient need.

In stepped-care models, different ‘steps’ or levels of treatment intensity exist. However, there are two ways in which patients are assigned to different steps.

1. Stepped care: a sequential approach in which all patients move through steps in a systematic way irrespective of illness severity, patient need or choice. Under this model all patients will initially receive low-intensity treatments and only ‘step up’ if and when these first-line interventions fail.

2. Stratified care: a targeted approach in which patients are assessed and referred to an appropriate step of treatment based on severity, complexity or other agreed criteria. Patients can thus access an appropriate level of treatment and, if needed, access a more intensive intervention without first having had to receive a lower-intensity treatment. 14

Although these can be considered as alternatives, models of care can include aspects of both models (i.e. the majority of patients use the stepped model, but patients with certain characteristics are stratified).

The NICE model uses stratification based on (1) severity of OCD symptoms and (2) functional impairment. The model consists of six steps divided into three stages: awareness (step 1), recognition and assessment (step 2) and treatment (steps 3–6). Our focus is on steps 3–6.

• At step 3, adults with OCD with mild functional impairment, low-intensity psychological (CBT) treatments are recommended. Low-intensity treatments are delivered by a PWP and defined as < 10 hours of therapist time per patient. Treatments include (1) brief individual CBT using structured self-help materials, (2) individual CBT delivered by telephone or (3) group CBT.

• At step 4, adults with OCD who have moderate functional impairment (or who have mild functional impairment, but have failed to engage or improve with lower-intensity treatments) will usually be offered a selective serotonin reuptake inhibitor and/or higher-intensity CBT delivered by a high-intensity therapist (HIT).

• Step 5 is recommended for those with severe functional impairment and delivered within secondary care by mental health professionals with expertise in OCD. Treatment recommendations include a combination of a selective serotonin reuptake inhibitor and CBT.

• Step 6 is reserved for people presenting with severe and chronic functional impairment, treatment refractory, and/or a risk to self or others. Specialist inpatient treatment and care may be offered at this step.

The NICE guidelines for OCD have been in existence since 2005. 12 An evidence update was set for February 2014, but no significant evidence was identified to change the recommendations. During the completion of the evidence update, the effectiveness of technology-enhanced treatment for OCD was identified as an evidence gap and, as a result, was registered on the UK Database of Uncertainties about the Effects of Treatments. 15

Service delivery models for psychological therapies in the UK

Psychological interventions for OCD are usually delivered in Improving Access to Psychological Therapies (IAPT) services. 16 IAPT commenced in 2007 with the aim of achieving a better balance between supply of psychological therapies and demand.

Since its inception, > 5000 new psychological therapists have been trained. IAPT training, which follows a national curriculum, produces two types of therapists:

1. low-intensity therapists, also known as PWPs

2. HITs.

Psychological wellbeing practitioners are tasked with facilitating a range of low-intensity interventions including guided self-help, cCBT and psychoeducational groups for mild to moderate depression and anxiety disorders. HITs provide predominantly CBT interventions for moderate to severe cases. The amount of time allocated to individuals varies but, on average, PWPs provide 3–4 hours of therapeutic support per person over 6–8 weeks, whereas HITs provide weekly 60-minute sessions over 8–16 weeks.

Given the NICE guidelines, it would be sensible to assume that low-intensity interventions for OCD (guided self-help or cCBT) would be supported by a PWP and provided in IAPT services for people with mild functional impairment.

However, the NICE model is not being implemented precisely, as designed in practice,17 for a number of reasons:

• In most IAPT services, people with OCD are referred directly to, and treated by, HITs regardless of their illness severity. OCD is perceived as a complex mental health disorder and hence the PWP IAPT National Curriculum18 does not address OCD knowledge or treatment.

• Although the NICE guidelines recommend low-intensity interventions for mild OCD, there is evidence that people with OCD with mild functional impairment do not present to services. Consequently, a substantial proportion of patients seen by IAPT practitioners will be experiencing OCD with at least a moderate level of functional impairment.

• Discrepancies exist between the definitions of low-intensity treatments adopted by IAPT services and NICE. Although there is no formal consensus regarding the precise number of therapist hours required by low-intensity interventions, NHS and IAPT services typically presume no more than 3–6 hours of input. The NICE guidelines advocate up to 10 hours of support for low-intensity interventions for OCD, thereby positioning at least some of these treatments within the remit of higher-intensity services.

Summary of the evidence for low-intensity interventions for obsessive–compulsive disorder

Despite NICE’s recommendations for low-intensity interventions for OCD, evidence of the clinical effectiveness, cost-effectiveness and acceptability of these interventions is limited. Effective synthesis is hampered by confusion over the scope of low-intensity interventions.

Lovell and Bee19 completed a systematic review of 13 studies (n = 492 participants) of CBT-based treatments that used health or communication technology in adults with OCD, including self-help manuals and cCBT alongside telephone and videoconferencing. Heterogeneity of populations, interventions and outcomes across the studies prevented meta-analysis. Self-help manuals were assessed in five small, uncontrolled, quasi-experimental studies and found moderate to large improvements in OCD symptoms; however, the absence of controlled trials means that the generalisability of the findings is unclear. cCBT was assessed in five studies, four of which were of BT Steps [now called OCFighter version 1.0 (CCBT Ltd, Birmingham, UK)]. The results revealed significant moderate to large effects on OCD symptoms score in favour of BT Steps. However, all studies were undertaken by the developers of the software and no independent analysis was available. The authors concluded that preliminary data support the idea that technology holds promise in treatment for OCD. Nevertheless, definitive conclusions about the relative efficacy of using health technologies as a replacement for therapist contact needed stronger evidence from rigorous randomised controlled trials (RCTs).

Herbst et al. 20 similarly examined telemental health applications for adults and children with OCD and included studies delivered by computer, the internet, telephone or self-help literature either delivered alone or supported by a health professional. No meta-analysis was completed and individual effect sizes ranged from 0.46 to 2.5. Of the 24 studies (n = 839 participants) included in this review, seven used written self-help materials (guided self-help), 11 were delivered by telephone, three were computer assisted, one comprised an online self-help group and two used videoconferencing. The review suggested that telemental applications may have promise but, once again, clinical and methodological heterogeneity, and small sample sizes precluded definite conclusion.

A more recent review21 looked at the efficacy of all types of technology-delivered CBT for OCD versus control conditions and in comparison with high-intensity CBT. Eight RCTs (n = 420 participants) were included and results found that CBT delivered by technology was superior to the control intervention in reducing OCD symptoms but not on comorbid depression. There were no differences in reductions in OCD symptoms between CBT delivered by technology and therapist-delivered CBT. Similar to previous reviews, this review concluded that further RCTs are warranted to examine the efficacy of technology-delivered CBT.

Results of the internal pilot

In terms of question 1, the internal pilot proved that recruitment was feasible (with the addition of more sites) and that the planned sample size could be achieved in principle. The pilot recruitment graph is presented in Appendix 1.

However, evaluation of questions 2 (time patients spent on the high-intensity CBT waiting list) and 3 (timing of a 3- or 6-month outcome for the primary assessment point) demonstrated significant challenges. As highlighted in the introduction, one of the key reasons for conducting OCTET was that the need for psychological therapy services for OCD exceeds demand, meaning that patients often face long waiting lists for high-intensity CBT treatment. Low-intensity interventions are designed to allow more rapid access and better management of demand.

The OCTET recruitment method was to screen existing high-intensity CBT IAPT waiting lists for potentially eligible patients where OCD was known or noted in the referral letter. Similar methods (large-scale screening of existing lists of potentially eligible patients) have underpinned a number of our successful large multicentre mental health trials [e.g. the Randomised Evaluation of the Effectiveness and Acceptability of Computerised Therapy (REEACT trial),36 cost-effectiveness of collaborative care for depression in UK primary care trial (CADET)37].

Extensive mapping of IAPT high-intensity CBT waiting lists at recruitment sites was conducted prior to the trial commencing, which identified waiting lists of between 60 and 800 patients and between 40 and 100 OCD referrals per year. Therefore, it was predicted that we would be able to achieve and maintain the required recruitment rate.

During the internal pilot, the Department of Health informed all IAPT sites in England that patients should be offered more rapid access to treatment as part of the IAPT initiative. This had two major effects:

1. A proportion of participants randomised to a low-intensity intervention were offered their high-intensity CBT appointment prior to commencing the intervention or before completing the low-intensity intervention. One waiting list reduced from in excess of 700 patients to < 50 patients in 4 weeks.

2. A proportion of participants were ‘removed’ from the high-intensity CBT waiting list and offered access to a form of low-intensity intervention. In many cases, these interventions were in line with current good practice (including written material focusing on understanding and managing anxiety, supported by a PWP), but were not OCD specific or recommended by the NICE OCD guidelines. However, having been removed from the waiting list, such patients were lost to the OCTET recruitment procedure, significantly reducing the numbers of potentially eligible patients who could be contacted.

These changes to the high-intensity CBT waiting list procedure had significant implications for OCTET. In response, we made the following changes:

• As a proportion of participants were reaching the top of the CBT waiting list prior to 12 weeks, we proposed to services that regardless of the length of waiting list, patients would be asked to wait at least 12 weeks before high-intensity CBT was offered. This proposal was supported by the Health Technology Assessment programme, the Data Monitoring and Ethics Committee (DMEC) and Trial Steering Committee (TSC), and was approved by the National Research Ethics Service Committee North West Lancaster. Despite this, only one site agreed to implement this change because of concerns that such a strategy contradicted the requirements of the Department of Health.

• We initiated a new strategy in some sites, whereby all patients on waiting lists for high-intensity CBT were mailed with an invite to OCTET rather than the original strategy that restricted mailing to patients with an indication of OCD on NHS records. The rationale was that OCD is often underdiagnosed and in consultation with clinical colleagues it transpired that many referrals from general practitioners (GPs) did not indicate OCD on the referral letter. The trial consent procedures and inclusion criteria remained the same. This strategy proved successful in increasing numbers, but was more time-consuming as a number of people who self-reported OCD did not meet our trial inclusion criteria.

• We extended self-referral opportunities in two sites (as recruitment was low at these particular sites) by placing adverts in local newspapers, on social media [e.g. local OCD group Facebook page (Facebook, Inc., Menlo Park, CA, USA)], in GP surgeries, community centres, trust magazines and in other local publications in one site. The trial consent procedures and inclusion criteria remained the same.

• We engaged additional IAPT services within 10 NHS trusts in addition to our existing sites. Despite increasing the number of potential participants, this raised logistical and resource challenges, as additional recruitment and training of researchers and PWPs was required.

• Given the directive from the Department of Health regarding high-intensity CBT waiting lists, it was not feasible to use the proposed 6-month follow-up as the primary outcome assessment point, and hence it was agreed to retain the 3-month assessment as the primary outcome assessment point. Given that removing the 6-month assessment would result in a variety of ethical issues and could potentially reduce participant engagement at 12 months, it was decided that the 6-month assessment should be retained.

Ethics and governance

Ethics approval for the study was granted by National Research Ethics Service Committee North West – Lancaster (reference number 11/NW/0276). Site-specific approvals were obtained from the relevant local research governance offices covering the trusts involved in the trial. The trial was registered with the International Standard Randomised Controlled Trial Number Register (ISRCTN73535163).

Low-intensity interventions

Psychological wellbeing practitioner training and supervision

Psychological wellbeing practitioners were trained in both the guided self-help and supported cCBT interventions. Training was delivered in all IAPT services in the 15 NHS sites. Top-up training was provided to PWPs at sites where a large delay between training and participant recruitment occurred, and because of the high turnover of PWPs the training was repeated at a number of sites.

The standardised training consisted of 3 days, and included 1 day explaining the nature, features, treatment and NICE recommendations for the management of OCD. In addition, day 1 also explained OCTET in terms of study design, rationale and trial procedures.

Day 2 focused on guided self-help and involved small- and large-group work and skills practice of ERP with specific feedback using exemplar cases. Training was delivered by the chief investigator and two coapplicants. cCBT training for OCFighter was delivered at the participating sites for 1 day by CCBT Ltd (the commercial producers of OCFighter). The training was standardised, using the same trainers (OCTET chief investigator, coapplicants and CCBT Ltd), materials and intervention manuals.

Training manuals for PWPs were developed by the trial team for guided self-help and by CCBT Ltd for the supported cCBT arm. A reference manual for PWPs was also generated that included general information about OCTET, recruitment, randomisation and allocation procedures, recording and storage of session recordings and monitoring of allocated participants.

Psychological wellbeing practitioners delivering the interventions were provided with telephone supervision on a 2-weekly basis for between 10 and 30 minutes (dependent on the number of patients to be discussed). Supervision was delivered by OCTET applicants or CBT therapists within IAPT services. All supervisors (two senior clinicians within a service and coapplicants) were required to attend the 3-day training).

Adherence and fidelity

To ensure the adherence to the guided self-help and supported cCBT interventions, and potentially enhance the reliability and internal validity of the trial, a number of strategies were implemented during trial development and completion.

Treatment adherence was examined by requesting PWPs complete contact sheets detailing dates of all sessions attended, length of sessions and mode of contact (face to face, telephone or e-mail). CCBT Ltd provided automated recordings of frequency and duration of supported cCBT use.

Fidelity was examined by asking PWPs to record all face-to-face and telephone sessions (with participant consent), using a digital recorder and (when required) a telephone-recording device.

These recordings were used to examine fidelity to the low-intensity interventions. A rating scale was developed based on the low-intensity intervention PWP manuals, which defined specific tasks to be carried out in session 1 and in subsequent sessions for both supported cCBT and guided self-help. Criteria for the fidelity scale were established via discussions within the OCTET team. The components extracted for fidelity were rated as ‘implicit’, ‘explicit’ or ‘absent’, and an overall rating generated using a 5-point Likert scale from ‘unacceptable’ to ‘excellent’. Fidelity was evaluated by a rater (a PWP independent of OCTET), who was blind to the treatment outcome.

Site recruitment

Mental health trusts across four UK centres in England (Manchester, York, East Anglia and Sheffield) were involved in the trial. A dedicated site lead was allocated at each centre. Trusts were recruited throughout the trial and further support was sought from the Mental Health Research Network (MHRN) to engage with additional trusts.

Patient recruitment

Randomisation, concealment of allocation and blinding

Patients were randomised (in a ratio 1 : 1 : 1) into one of the three arms using a central randomisation service, via a secure web-based system administered by the York Trials Unit (YTU).

Allocation involved minimisation on the following factors:

• OCD severity on the Y-BOCS-SR (16–23, moderate; 24+, severe/very severe)

• current antidepressant medication use (yes/no)

• depression on the Patient Health Questionnaire-9 (PHQ-9) (< 10, mild depression; 10–14, moderate depression; 15–19, moderate to severe depression; > 20, severe depression)

• duration of OCD (0–5 years; 6–10 years; > 10 years).

Follow-up assessments

Follow-up assessments were conducted 3, 6 and 12 months following randomisation. Participants were contacted approximately 1 month to 1 fortnight prior to the follow-up time point to arrange a suitable time and location to meet (either at the clinical site or in their own home). To reduce the chance of the researcher becoming unblinded, the Adult Service Use Schedule (AD-SUS) self-complete, Client Satisfaction Questionnaire-8 (CSQ-8) and Pathway questionnaire were posted prior to the visit for return in a Freepost envelope to the YTU. Participants were given a £5 shopping voucher to thank them for their time for each follow-up they completed or partially completed.

Outcome assessments

As the primary outcome was the Y-BOCS-OR, the collection of data was face to face. However, we acknowledged that achieving this with all follow-ups could prove difficult. Therefore, we developed a highly structured standardised operating procedure (SOP) to assist researchers with participants who were difficult to contact, to ensure retention rates were maximised. These procedures included contacting participants using a scaling-down approach (i.e. if not willing/unable to attend face to face we offered telephone assessment using Y-BOCS-SR and if this failed, to post the primary outcome with a Freepost envelope for return to the YTU).

Measures

Researcher training

Researchers and clinical study officers (CSOs) from the MHRN involved in the recruitment of participants were provided with a 1-day training session covering trial procedures and completion of eligibility, baseline and follow-up interviews. The trial manager provided training. A proportion of the day was spent equipping individuals with the necessary skills to complete the Y-BOCS-OR. Inter-rater reliability among the researchers/CSOs was employed to measure agreement between ratings for the Y-BOCS-OR. Top-up training was provided throughout the trial.

Sample size calculation

At each time point three pairwise comparisons were carried out between the three intervention options: supported cCBT, guided self-help and a waiting list for high-intensity CBT. The study was therefore powered using a 1.67% significance level.

The comparison of either supported cCBT or guided self-help is a partially nested design for which the sample size calculation needs to consider the intracluster correlation coefficient (ICC) for therapist. The comparison of supported cCBT with guided self-help is a crossed therapist design, as support for both treatments was delivered by the same therapists. Sample size for crossed therapist design depends on the ICC for therapist for treatment within therapist, which is smaller than the ICC for therapists. Formulae for this calculation are given in Walwyn and Roberts. 67 In the absence of estimates of the two ICCs required for the two calculations, sensitivity of study power to larger values was considered in the calculation below.

Assuming a standard deviation (SD) for the primary outcome (Y-BOCS-OR) at 6 months of 7.3 units, a correlation between baseline Y-BOCS-OR and 6-month Y-BOCS-OR of 0.43,41 a study with 366 service users followed up to the primary end point has a power > 80% to detect a difference of 3 Y-BOCS points for each comparison. We were unable to find evidence for a ‘clinically important difference’. A reduction of 3 points was agreed based on clinical consensus with the study team. This calculation assumed that supported cCBT and guided self-help were delivered by 24 therapists. It also assumes that the ICC for therapists was 0.06 and an ICC for treatment within therapist was 0.015, which implies that the correlation between the random effect for supported cCBT and guided self-help is 0.75. The design effects, sometimes called the sample size inflation factor, were 1.1225 and 1.06125 for the partially nested and crossed designs, respectively. We considered these values of the ICC to be plausible, but in the event that the ICC for therapist was as large as 0.1 and the ICC for treatment within therapist was 0.05, the power of the trial is still > 75% for all three comparisons.

Based on an 85% follow-up rate, the target sample size for the trial was set at 432 in the initial application. Monitoring during the trial suggested that the follow-up rate to the primary end point (3 months) was likely to be lower rather than the assumed 85%. Assuming a follow-up rate of 78%, total sample size was increased to 472.

Low-intensity intervention uptake

Patient engagement with the treatment process was summarised and reported descriptively. There is no consensus for the level of uptake of low-intensity treatments that might define an appropriate ‘dose’.

Analysis of the primary outcome measure and secondary quantitative outcome measures

Cleaning of outcome and baseline data was conducted without the treatment group allocations in view. Summary statistics from these preliminary analyses were reviewed by the trial research team to identify data errors.

Preliminary analyses compared the characteristics of subjects with and without complete data at follow-up time points, by treatment group, using a logistic regression model. This was carried out for the primary outcome at the three follow-up assessment points. This analysis was used to develop an understanding of the missing data mechanism and to determine the appropriate methods for dealing with missing outcome data.

Statistical analyses of the primary outcome measure, Y-BOCS-OR, are based on a linear mixed model with random effects for supported cCBT and guided self-help therapist using restricted maximum likelihood. As therapist is crossed with treatment, separate random effects were included for each treatment enabling the estimation of the intracluster correlation for supported cCBT and guided self-help. In addition to treatment, the following fixed baseline or demographic covariates were included to improve statistical efficiency:

1. OCD duration (by categories 0–5, 5–10, > 10 years)

2. OCD severity (as measured by Y-BOCS-OR at baseline)

3. anxiety (as measured by GAD-7)

4. depression score (as measured by PHQ-9)

5. antidepressant drug use (yes/no)

6. sex.

A small number of baseline covariates were missing for those covariates not used in the minimisation. To maximise the number of subjects included in the model, these values were imputed by single imputation using other covariates in keeping with the method suggested by White and Thompson. 68 Using this procedure, statistical modelling can be carried out on all participants with outcome data. A logistic regression model was used, fitted to CBT uptake, with treatment allocation, sex, duration of OCD, baseline Y-BOCS-OR, GAD-7, PHQ-9 and antidepressant medication on entry into the trial as covariates.

The same analyses were carried out for quantitative secondary outcomes. These analyses assume that subjects are missing at random.

Uptake of high-intensity cognitive–behavioural therapy

Uptake of high-intensity CBT was recorded at the 3-, 6- and 12-month follow-ups. A logistic regression model was used to estimate the adjusted odds ratio (OR) for uptake (as a binary outcome of a patient attending at least one CBT appointment), comparing the two low-intensity interventions separately with a waiting list for high-intensity CBT. The models included treatment allocation, sex, duration of OCD, baseline Y-BOCS-OR, GAD-7, PHQ-9 and antidepressant medication on entry into the trial as covariates. Where presented, tables give the adjusted ORs with confidence intervals (CIs) and p-values and the exponent of the model constant.

Recovery and remission

Recently, expert consensus guidelines for defining treatment response and remission in OCD have been published based on YBOC-OR. 69 These are defined as ≥ 35% reduction on the Y-BOCS-OR for response and Y-BOCS-OR score of ≤ 12 for remission. A logistic model was fitted to responses at 12 months adjusting for sex, baseline GAD-7 and PHQ-9 scores, antidepressant medication use at randomisation and duration of OCD. The analysis is additional to the statistical analysis plan as this was signed-off prior to publication of the guidelines.

Moderator analyses

A subgroup analysis by severity specified in the trial protocol was conducted by adding a treatment severity interaction term to the analysis of the primary outcome at the primary end point (3 months). In addition, analysis of treatment effect moderation by both age and chronicity of OCD was carried out.

Two types of interactions between moderators and treatment can occur. If the moderator is just affecting the magnitude of the treatment effect it is called a quantitative interaction. If the moderation causes a reversal of the treatment effect it is called a qualitative interaction. Interactions of treatment with both chronicity and severity were hypothesised to be quantitative, whereas interactions of treatment with age were hypothesised to be qualitative (representing potential difficulties of older patients in engaging with supported cCBT relative to guided self-help). Thus, a significant treatment effect would be required before assessing treatment moderation via severity and chronicity of OCD.

The analyses were carried out by adding a treatment with moderator interaction terms to the primary analysis model. For severity and chronicity, an overall test of the interaction was carried out. The hypothesis related to age concerned only the low-intensity interventions, so the contrast between guided self-help and supported cCBT was estimated.

Data were analysed using Stata version 13 (StataCorp LP, College Station, TX, USA).

Perspective

The primary perspective of the economic evaluation was the NHS/Personal Social Services perspective preferred by NICE. Secondary analyses included all additional resources likely to be relevant to a societal perspective in this population: productivity losses (as a result of time off work resulting from illness), and out-of-pocket expenses and savings.

Data collection

An adapted version of the AD-SUS was used to measure individual-level resource use. The AD-SUS is used to collect service use and related data, and has been successfully used in a range of adult mental health populations. 51–53 The AD-SUS was adapted for OCD, as described above (see Secondary outcomes), and used to record all-cause hospital and community-based health- and social-care services, medication and out-of-pocket expenses and savings. Information on medications used including drug, dose and duration were collected for economic purposes only. Productivity losses were recorded using the World Health Organization’s HPQ. 54,55 The AD-SUS and HPQ were administered by interview at baseline, and the 3-, 6- and 12-month follow-ups, and covered the previous 6-month period at baseline interview and the time since last interview at each follow-up point. Data on the number and duration of supported cCBT and guided self-help contacts were recorded on a session-by-session basis by PWPs using an intervention proforma. Use of all other psychological therapies, including high-intensity CBT, was collected and self-reported by participants in a separate self-complete proforma kept separate from the AD-SUS in order to ensure that interviewers remained blinded to randomisation allocation status. Participants completed the form alone and placed it in a sealed envelope before handing it to the interviewers. Data on the number and duration of intervention contacts were recorded on a session-by-session basis by PWPs using an intervention proforma.

Costs

All costs are reported in pounds sterling at 2013/14 prices. Discounting was not relevant, as the follow-up did not exceed 12 months. Unit costs were applied to individual-level resource use data to calculate total costs per participant and are detailed in Appendix 2. In summary, unit costs for most hospital and primary care services were obtained from NHS Reference Costs,70 Unit Costs of Health and Social Care71 and the British National Formulary for medications. 72

Productivity losses because of OCD were calculated using the human capital approach by multiplying days off work attributable to illness by the individual’s salary and not accounting for early retirement resulting from illness. 73 Lost productivity costs were capped at 5 days per week (maximum of 130 days for the 6-month period and 65 days for the 3-month period).

Intervention costs

Psychological wellbeing practitioner sessions were costed using published data on the cost of low-intensity IAPT interventions,74 inflated to 2013/14 prices using the Hospital and Community Health Services Pay and Prices Index. 71 The approach is outlined in Appendix 3.

Training for PWPs to provide supported cCBT and guided self-help support was provided to all PWPs, and associated costs were calculated as outlined in Appendix 4.

For the supported cCBT arm, the cost to the trial of the OCFighter program was £10,000. This was divided by the 157 participants who were randomised to supported cCBT to give a cost per participant in this group of £63.69. In the guided self-help arm, participants were provided with a self-help manual (£1.87 per manual for printing costs only, excluding development and design costs as these are sunk costs), photocopied worksheets (£0.81 per set of photocopied sheets per participant) and a CD (£1.18 per CD, one per participant), plus £1.68 postage charge. Thus, the total cost of guided self-help materials per participant was £5.54.

Access to high-intensity CBT, for all groups, was recorded using the intervention proforma. Unit costs for CBT are reported in Appendix 2.

Analysis

Data were analysed using Stata. Participants were analysed on an intention-to-treat basis (i.e. according to the group to which they were randomised regardless of intervention compliance).

Costs and outcomes were compared at baseline, 3, 6 and 12 months and are presented as mean values by arm with SDs. Mean differences and 95% CIs were obtained by non-parametric bootstrap regressions (1000 repetitions) to account for the non-normal distribution commonly found in economic data, with adjustment for clustering at the therapist level, using the ‘cluster’ option in Stata. To provide more relevant treatment-effect estimates75 regressions to calculate mean differences in costs were repeated with the further inclusion of covariates for the baseline value of the relevant variable (costs or EQ-5D-3L utility) plus variables thought to influence costs and outcomes: antidepressant drug use; anxiety (GAD-7); depression score (PHQ-9); sex; OCD duration (0–5, 5–10 and > 10 years); and Y-BOCS-OR score.

The primary analysis was a complete-case analysis (i.e. excluding those lost to follow-up or with missing ADSUS and/or EQ-5D-3L data at a particular time point). To explore the potential impact of excluding non-responders, we examined the sociodemographic and clinical characteristics of those included in the analyses and those in the full sample. A secondary analysis was carried out with missing baseline, 3-month and 12-month total costs and outcomes imputed using the input imputation command in Stata (version 11) and including the baseline variables described above.

Cost-effectiveness was explored in terms of QALYs calculated using the EQ-5D-3L measure of health-related quality of life, assessed at baseline, 3, 6 and 12 months. Appropriate utility weights were attached to health states76 and QALYs were calculated using the total area-under-the-curve approach with linear interpolation between assessments. 77

Two sets of cost–utility analyses were conducted: one compared groups at 3 months and one compared groups at 12 months (primary end point). Incremental cost-effectiveness ratios (ICERs) were calculated, defined as the mean difference in cost between two groups divided by the mean difference in effect.

Uncertainty was explored using cost-effectiveness planes and cost-effectiveness acceptability curves (CEACs) based on the net-benefit approach. 78 Cost-effectiveness planes illustrate the uncertainty around the estimates of costs and effects by plotting the bootstrapped cost and effects, with points in each quadrant indicating a different implication for economic evaluation. CEACs are an alternative to CIs around ICERs and show the probability that one intervention is cost-effective compared with the other, for a range of values that a decision-maker would be willing to pay for an additional unit of an outcome. A series of net benefits were calculated for each individual for a range of values for willingness to pay for a QALY. After calculating net benefits for each participant for each value of willingness to pay, coefficients of differences in net benefits between the trial arms were obtained through a series of bootstrapped linear regressions (1000 repetitions) of group upon net benefit, which included the same covariates used for comparisons of outcomes in the primary economic analyses. The resulting coefficients were then examined to calculate the proportion of times that the intervention group had a greater net benefit than the control group for each value of willingness to pay. These proportions were then plotted to generate CEACs for all cost–outcome combinations.

This analysis takes a decision-making approach, ignoring statistical significance and focusing instead on the probability of one intervention being cost-effective compared with another intervention, given the data available. This is the recommended approach to economic evaluation, preferred over traditional reliance on arbitrary decision rules regarding statistical significance, which are being increasingly criticised as irrelevant in a decision-making context. 79,80 Instead, it is argued that the decision to adopt one intervention over another should be based on the expected cost-effectiveness of the intervention, or the probability of making the correct decision.

Trial monitoring

Trial-specific procedures

To standardise processes across all sites and to maximise data quality, trial-specific SOPs were generated for all individuals involved in the trial covering all procedures and frequently asked questions.

Researcher trial-specific SOPs covered booking eligibility; baseline and follow-up interviews; recruitment procedures; dealing with difficult to contact participants; retention procedures; conducting and reporting risk assessments; AEs recording and reporting; managing participants; and own distress and blinding.

Checklists and letter templates were included to assist with all aspects of the trial and the trial procedure document was updated and disseminated to all researchers as and when necessary throughout the trial.

Recruitment

Psychological wellbeing practitioners were recruited from IAPT services within NHS trusts to deliver the trial interventions. In total, 204 PWPs from 15 NHS trusts attended the 3-day OCTET training. The number of PWPs trained from each trust ranged from 1 to 27, with an average of 13 per trust (see Table 3).

Psychological wellbeing practitioner characteristics

Of the 93 PWPs allocated OCTET patients, 68 returned a questionnaire detailing demographics, qualifications, OCD training and previous experience (Table 4).

Most PWPs (87%) were female, aged between 24 and 61 years, with a mean age of 34 years. Their highest educational qualification was a postgraduate certificate/diploma (81%) and the majority (66%) were experienced in delivering psychological interventions with > 2 years in post.

As expected, most PWPs had experience of delivering guided self-help (94%), with a lower proportion reporting experience of delivering supported cCBT (69%). The types of low-intensity interventions most frequently delivered were behavioural activation (91%) and structured problem-solving (79%).

More than half of the PWPs had received some training in OCD as part of their IAPT training, but this had largely focused on identification rather than treatment.

Revised sample size

Our original sample was 432; however, because of concerns about meeting the proposed retention rate (85%), we increased the target sample size to 472 (see Chapter 2, Sample size calculation). Subsequently, there was a small overshoot of recruitment and randomisation to 475, but there were also two post-randomisation exclusions because of a participant being aged < 18 years (n = 1, supported cCBT) and issues relating to participant suicide risk (n = 1, waiting list). Both exclusions were ratified with the DMEC, giving a trial sample of 473 participants. The Consolidated Standards of Reporting Trials (CONSORT) diagram for OCTET is shown in Figure 1. Of the 473 patients recruited, 158 were randomised to guided self-help, 157 to supported cCBT and 158 to a waiting list for high-intensity CBT. The number of patients recruited by site is shown in Table 5.

FIGURE 1.

The CONSORT flow chart illustrating recruitment participants into OCTET. a, Post-randomisation exclusion required because of a participant aged < 18 years (n = 1, cCBT) and risk issues relating to increased risk and not a change in supervisor decision (n = 1, waiting list). Source: Lovell et al. 84 This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Retention

Retention rates were 81% at 3 months, 75% at 6 months and 71% at 12 months (Figure 2). Retention rates by group at 3 months were broadly similar: guided self-help was 82% (n = 130), supported cCBT was 77% (n = 121); and a waiting list for high-intensity CBT was 84% (n = 133). At 12 months, retention rates were: guided self-help at 72% (n = 114), supported cCBT at 67% (n = 105); and a waiting list for high-intensity CBT at 73% (n = 115). Participants who withdrew from the study are reported in the CONSORT flow chart (see Figure 2); more specific reasons for withdrawal are detailed in Appendix 5.

FIGURE 2.

The CONSORT flow chart illustrating retention of participants in OCTET.

Demographic characteristics

Baseline demographic characteristics are presented in Table 6 (variable counts are because of missing data items).

Of the 473 patients recruited, 95% were white, 60% were female and the median age was 33 years. Around three-quarters were employed, half were married or living with a partner, and 28% had been educated to degree level or above.

Just over 50% of participants reported previous professional help with OCD and around half were currently using antidepressant medication. Patients reported that the duration of their OCD was > 10 years (55%), 5–10 years (12%) or < 5 years (33%).

The groups were largely comparable in terms of employment status and receipt of statutory sick pay. Marginally fewer instances of long-term sickness/disability and being in receipt of associated benefits were found in the waiting list for high-intensity CBT group than in the intervention groups (9.5%, waiting list for high-intensity CBT; 14.0%, supported cCBT; 14.6%, guided self-help).

Patients detailed their treatment preferences prior to randomisation. They were asked whether they preferred to stay on the waiting list for high-intensity CBT, had no preference or preferred to receive a low-intensity intervention. Of the total sample, 42% preferred allocation to the intervention groups, 47% had no preference and 11% preferred to stay on the waiting list for high-intensity CBT.

Patients who were allocated to a low-intensity intervention were also asked if they would prefer supported cCBT or guided self-help. Preferences between the two low-intensity treatments were evenly balanced, with 50% not having a preference (Table 7).

There were no striking differences in demographic characteristics at baseline.

Clinical characteristics

Baseline clinical characteristics are presented in Table 8 (variable counts are because of missing data). The mean baseline Y-BOCS-OR score was approximately 25 at baseline (indicating severe OCD). The mean PHQ-9 score was approximately 12 at baseline (indicative of mild to moderate depression), whereas GAD-7 scores were approximately 12.5 (indicating moderate anxiety).

Just over half of the patients were currently taking antidepressant medication and 69% of patients had a primary comorbid diagnosis, as derived from the CIS-R.

There were no striking differences in clinical characteristics at baseline.

Allocation of patients to psychological wellbeing practitioners

Of the PWPs trained (n = 204), 93 (46%) were allocated an OCTET patient (Table 9). The number of PWPs allocated patients per trust ranged from 0 to 14, a range of 0 to 100% of those trained within each trust. PWPs were not allocated patients for various reasons: leaving the service, high clinical workload or insufficient patients compared with the number of therapists in each site. In total, 315 patients were allocated to the 93 PWPs. The number of patients allocated to PWPs ranged from 1 to 18. Nearly half the PWPs (46%, n = 42) were allocated patients in both supported cCBT and guided self-help; 33% (n = 30) were allocated only guided self-help patients, and 20.9% (n = 19) were allocated only supported cCBT patients.

Uptake of interventions

As detailed in Chapter 2, PWPs completed a contact sheet for each allocated participant and recorded session date, number and the length of the session. Of the 158 patients allocated to guided self-help, 155 (98%) contact sheets were received and of 157 patients allocated to supported cCBT, 154 (98%) contact sheets were available.

Fidelity of delivery of guided self-help and supported computerised cognitive–behavioural therapy

Psychological wellbeing practitioners were asked to record all face-to-face and telephone sessions using a digital recorder and, when required, a telephone-recording device.

Owing to an ethics breach (a digital recorder was lost at one site), all recordings were stopped for 5 weeks. Even with this lapse, adherence to recording was low, with only 169 (26%) of a possible 648 guided self-help sessions recorded, and only 61 (17%) of a possible 350 supported cCBT sessions recorded.

All available session recordings were given a random number using an online random number generator and the independent rater randomly selected 25% of the guided self-help and supported cCBT recordings. The independent rater randomly selected 43 recordings from guided self-help sessions 1–12, and 17 recordings from supported cCBT sessions 1–6 to ensure an equal representation of each of the different stages of the interventions.

Of the guided self-help sessions, nine (21%) were rated as average, 24 (56%) were rated as good and 10 (23%) were rated as excellent. Of the supported cCBT session recordings, 11 (65%) were rated as good and six (35%) were rated as excellent.

The length of sessions on the digital audio files was examined to see if they were consistent with the self-report data available, as recorded on the contact sheets. For guided self-help the mean duration of session 1 (assessment) was 53 minutes (60 minutes was detailed in the PWP manual), the mean duration of session 2 (treatment rationale) was 35 minutes (30 minutes recommended), the mean duration of sessions 3–9 (setting weekly goals and support) was 26 minutes (30 minutes recommended) and the mean duration for sessions 10–12 (relapse prevention/staying well) was 19 minutes (recommended 30 minutes). The mean duration of sessions 1–6 for supported cCBT was 15 minutes (recommended 10 minutes).

In summary, PWP adherence was satisfactory, although the mean number of sessions for both guided self-help and supported cCBT was substantially less than recommended. A small number of patients receiving guided self-help had more sessions than in the protocol, although duration, as measured by self-report and digital audio files, was within the recommended limits. No extra sessions were delivered in the supported cCBT but both self-report and digital audio files revealed that the mean session length was 13 minutes, just over the recommended 10 minutes. Fidelity to the interventions was good, although only a limited number of digital recordings were available.

Uptake of cognitive–behavioural therapy prior to primary outcome assessment (3 months)

As designed, patients in OCTET were not expected to receive high-intensity CBT prior to the 3-month outcome assessment. However, because of changes in service delivery outside the control of OCTET, a number of patients started to receive high-intensity CBT prior to the 3-month assessment.

The Pathway questionnaire was used at each follow-up to collect data on high-intensity CBT uptake. Owing to the nature of the delivery and completion of this questionnaire (left with the participant to complete independently prior to postal return to the YTU), there were high levels of missing CBT uptake data when compared with the number of visits completed at each time point (28% missing at 3 months, 29% at 6 months and 30% at 12 months).

In addition, the CBT uptake data, as reported by the participant, did not always correspond to other patient-reported measures collecting treatment access data, nor did the data necessarily correspond to information provided by the IAPT services. As a result, to ensure validity and accuracy of the CBT uptake data, all sites were contacted to determine which participants had received a minimum of one session of high-intensity CBT prior to each follow-up time point.

In total, 67 (42%) of the patients allocated to the waiting list for high-intensity CBT group started full CBT prior to the primary outcome assessment, compared with 21% in supported cCBT and 23% in guided self-help (Table 12).

Patients in waiting list for high-intensity CBT group had a significantly higher uptake of high-intensity CBT than supported cCBT (adjusted OR 0.36, 95% CI 0.19 to 0.68; p = 0.001) and guided self-help (adjusted OR 0.43, 95% CI 0.22 to 0.84; p = 0.014). There were no differences between supported cCBT and guided self-help (adjusted OR 0.84, 95% CI 0.47 to 1.51; p = 0.562). No other factors predicted uptake of full CBT (Table 13).

Aim 1: the clinical effectiveness and cost-effectiveness of low-intensity interventions (guided self-help and supported cCBT) versus waiting list for high-intensity CBT in the management of OCD at 3 months

As noted earlier, the 3-month follow-up point was taken as the primary assessment point following the internal pilot. As described in Chapter 1, Results of the internal pilot: (1) we retained the 6-month follow-up point and report these data for completeness and (2) valid values of Y-BOCS-OR where obtained for 346, 308 and 283 services users at 3, 6 and 12 months, respectively. Substitution of Y-BOCS-SR for missing values increased the sample available for analysis to 383 at 3 months, 356 at 6 months and 332 at 12 months.

Analysis of the primary outcome (Y-BOCS-OR) (Table 14) showed that the benefit of supported cCBT over waiting list for high-intensity CBT was less than 1 point on the Y-BOCS-OR (adjusted mean difference –0.71, 95% CI –2.12 to 0.70; p = 0.325) at 3 months.

There was evidence of benefit of guided self-help over a waiting list for high-intensity CBT (adjusted mean difference –1.91, 95% CI –3.27 to –0.55; p = 0.006) at 3 months. This was statistically significant at the Bonferroni-corrected significance level (1.67%). This was < 3 points on the Y-BOCS-OR defined as a ‘clinically important difference’ in the sample size calculation, although the 95% CIs did include an effect of this magnitude.

Comparison of guided self-help with supported cCBT showed a reduction of around 1 point on the Y-BOCS-OR for guided self-help as compared with supported cCBT (adjusted mean difference 1.2, 95% CI –0.22 to 2.61; p = 0.097) at 3 months.

Similar trends were observed for Y-BOCS-SR, although no significant differences between treatments were observed at the 1.67% significance level used.

Patient satisfaction

Analysis of the satisfaction outcome (CSQ-8) showed no differences in the satisfaction of patients receiving supported cCBT compared with those allocated to the waiting list for high-intensity CBT (adjusted mean difference –0.31, 95% CI –2.07 to 1.45; p = 0.732) at 3 months (Table 16).

Analysis of the satisfaction outcome (CSQ-8) showed that patients receiving guided self-help were more satisfied than those allocated to a waiting list for high-intensity CBT (adjusted mean difference 1.69, 95% CI –0.04 to 3.42; p = 0.055) at 3 months. The effect estimate did not reach statistical significance according to the corrected significance level (1.67%).

Analysis of the satisfaction outcome (CSQ-8) showed that patients receiving supported cCBT were less satisfied than those receiving guided self-help (adjusted mean difference –2.00, 95% CI –3.63 to –0.37; p = 0.016) at 3 months. The effect estimate did reach statistical significance according to the corrected significance level (1.67%).

Aim 2: the clinical effectiveness and cost-effectiveness of low-intensity interventions (guided self-help and supported cCBT) plus high-intensity CBT versus waiting list for high-intensity CBT plus high-intensity CBT in the management of OCD at 12 months

Analysis of the primary outcome (Y-BOCS-OR) showed that the benefit of supported cCBT over a waiting list for high-intensity CBT was around 1.5 points (adjusted mean difference –1.37, 95% CI –3.59 to 0.84; p = 0.224) at 12 months (Table 17).

Analysis of the primary outcome (Y-BOCS-OR) showed that the benefit of guided self-help over waiting list for high-intensity CBT was around 2.5 points (adjusted mean difference –2.37, 95% CI –4.37 to –0.38; p = 0.02) at 12 months, although the effect estimate did not reach statistical significance according to the Bonferroni-corrected significance level (1.67%).

The benefit of guided self-help over supported cCBT was 1 point on the Y-BOCS-OR, but this was not statistically significant (adjusted mean difference –1.00, 95% CI –1.19 to 3.19; p = 0.37).

Intracluster correlation coefficients caused by therapist variation and design effects

Sample size calculation and statistical analysis allowed for ICC for therapist variation. Estimates of the ICC are given in Table 18 for the primary outcome and quantitative secondary outcome measures. These were calculated from the variance components of the linear mixed-model analyses. The estimation procedure (restricted maximum likelihood) constrains estimates of the random-effects variance to be positive. Twenty of the 52 estimates had a magnitude < 0.001. Estimates of the ICC were highly variable with a maximum of 0.225 for SF-36 mental functioning.

When designing the trial we assumed that the interventions would be delivered by 24 therapists, each with an average caseload of six supported cCBT and six guided self-help patients. The number of therapists was somewhat larger (see Table 9) making the study power less sensitive to larger values of the ICC.

The observed design effect can be calculated as the ratio of the variance of the treatment effect estimate when cluster is accounted for in a random-effects model to the same variance in an ordinary least squares model that takes no account of clustering. As there are three treatment comparisons there are three design effects for each end point. For the primary outcome measure, the largest value of the three design effects was 1.035 at 3 months when supported cCBT was compared with a waiting list. At 6 months, design effects were 1 as the ICC was negligible (< 0.001). At 12 months, the largest design effect was 1.129, again when supported cCBT was compared with a waiting list. The design effects of the sample size calculation (see Chapter 2, Sample size calculation) were 1.123 for the comparison of supported cCBT or guided self-help with a waiting list and 1.061 when supported cCBT was compared with guided self-help. The design effect of the study was similar to that hypothesised in the trial protocol.

A 98.33% CI corresponds to a 1.67% significance level that we have used for hypothesis testing. For the comparison of cCBT against a waiting list, the 98.33% CI is –4.07 to 1.33, and for guided self-help against a waiting list it is –4.81 to 0.06. Given that the upper limits are small, we can conclude that both active interventions are non-inferior to waiting list.

Supported computerised cognitive–behavioural therapy and guided self-help

Table 33 provides details of resources and costs in the two low-intensity interventions. The mean cost of guided self-help (£383) was over twice that of supported cCBT (£155) as a result of the greater number and length of the PWP contacts.

Health- and social-care services

Cost components over the three follow-up periods (from baseline to 3 months, between 3 and 6 months and between 6 and 12 months) are summarised in Table 34 and cost comparisons between groups are presented in Table 35.

Health- and social-care services (the largest cost component in each arm) were lowest in the guided self-help group and highest in the waiting list for high-intensity CBT group over all three time periods. At the adjusted p-value threshold of 0.0167, only the intervention costs were significantly different between the groups, with the intervention costs in the supported cCBT group (unadjusted mean difference £155, 95% CI £134 to £175; p < 0.001; adjusted mean difference £158, 95% CI £137 to £180; p < 0.001) and the guided self-help group (unadjusted mean difference £383, 95% CI £316 to £449; p < 0.001; adjusted mean difference £383, 95% CI £313 to £450; p < 0.001) being higher in the adjusted and unadjusted analyses, and with the cCBT group having lower intervention costs than the guided self-help group (unadjusted mean difference –£228, 95% CI –£299 to –£157; p < 0.001; adjusted mean difference –£223, 95% CI –£298 to –£149; p < 0.001).

Productivity losses and out-of-pocket expenses and savings

Productivity losses were highest in the guided self-help group between baseline and 3 months, in the supported cCBT group between 3 and 6 months and in the waiting list for high-intensity CBT group between 6 and 12 months, but there were no significant differences between either supported cCBT or guided self-help compared with the waiting list for high-intensity CBT over any of the time periods. Out-of-pocket expenses also varied greatly between groups and over the three time periods, but there were no significant differences. Out-of-pockets savings were minimal over all time periods in all arms (see Tables 34 and 35).

Table 36 reports total health- and social-care costs and total societal costs (including productivity losses and out-of-pocket expenses and savings) for the two economic analysis periods: baseline to 3 months and baseline to 12 months.

Table 37 details the results of the statistical comparisons between the two self-help interventions and between the waiting list for high-intensity CBT group and the supported cCBT and guided self-help groups.

Over the baseline to 3-month period, guided self-help was the most expensive group and the waiting list for high-intensity CBT group was the least expensive at the 0.0167 p-value threshold. The guided self-help group was significantly more expensive in unadjusted and adjusted analyses at 3 months than the waiting list for high-intensity CBT group in terms of both health- and social-care costs and societal costs, but no other significant differences were detected.

Over the full baseline to 12-month period (primary end point), differences between the groups had disappeared for health- and social-care costs, which were almost identical between the three groups. In terms of societal costs, guided self-help was the cheapest group and the waiting list for high-intensity CBT group the most expensive, although these differences were not significant. Tables 38 and 39 report the same results but with missing data imputed, and show very similar results with no changes in terms of statistical significance.

Incremental cost-effectiveness ratios

Table 42 presents the ICERs for all three pairwise comparisons at 3 and 12 months. Results are reported for both the health- and social-care perspective, and the societal perspective.

At 3 months, cost-effectiveness ratios for both supported cCBT and guided self-help compared with waiting list for high-intensity CBT are greater than £30,000 per QALY for both the health- and social-care perspective and the societal perspective. At 12 months (primary end point), however, ratios for supported cCBT compared with the waiting list for high-intensity CBT are negative for both perspectives, indicating, on average, that the supported cCBT group is both cheaper and more effective than waiting list for high-intensity CBT.

Ratios for guided self-help compared with the waiting list for high-intensity CBT are negative from the societal perspective (on average guided self-help is cheaper and more effective than the waiting list for high-intensity CBT) and approximately £4000 per QALY from the health- and social-care perspective (on average guided self-help is more expensive but also more effective), which is below the NICE willingness-to-pay threshold level of £20,000–30,000 or less. However, as ICERs are based on point estimates, we must examine the cost-effectiveness planes and CEACs to account for the variability and uncertainty around these estimates.

Supported computerised cognitive–behavioural therapy versus the waiting list for high-intensity cognitive–behavioural therapy

Figures 3–6 show the bootstrapped replications for cost and effect pairs on the cost-effectiveness plane for supported cCBT versus a waiting list for high-intensity CBT at 3 and 12 months (health- and social-care perspective and societal perspective). At 3 months, a greater proportion of scatter points lie to the right of the vertical axis where the bootstrap replications represent points where cCBT is more effective than the waiting list. In addition, a greater proportion of scatter points lie above the horizontal axis where the bootstrap replications represent supported cCBT having higher costs than the waiting list. This is the same from both perspectives and generates a trade-off scenario – there are better outcomes but for a greater cost.

FIGURE 3.

Cost-effectiveness plane for supported cCBT vs. the waiting list for high-intensity CBT at 3 months from the health- and social-care perspective.

FIGURE 4.

Cost-effectiveness plane for supported cCBT vs. the waiting list for high-intensity CBT at 3 months from the societal perspective.

FIGURE 5.

Cost-effectiveness plane for supported cCBT vs. the waiting list for high-intensity CBT at 12 months from the health- and social-care perspective.

FIGURE 6.

Cost-effectiveness plane for supported cCBT vs. the waiting list for high-intensity CBT at 12 months from the societal perspective.

Figure 7 shows the CEACs for supported cCBT versus the waiting list for high-intensity CBT at 3 months from both perspectives. The probability of supported cCBT being cost-effective compared with the waiting list at the NICE willingness-to-pay level of £20,000–30,000 is around 40–50% from the NHS/Personal Social Services perspective and 25–35% from the societal perspective.

FIGURE 7.

Cost-effectiveness acceptability curves for supported cCBT vs. the waiting list for high-intensity CBT at 3 months from the health- and social-care perspective and societal perspective.

At 12 months, from both perspectives, a greater proportion of scatter points lie to the right of the vertical axis, where the bootstrap replications represent supported cCBT being more effective than the waiting list, while a slightly greater proportion of scatter points fall below the horizontal axis where the bootstrap replications represent cCBT being less costly than the waiting list.

Figure 8 shows the CEACs for supported cCBT versus the waiting list at 12 months (primary end point) from both perspectives. The probability of supported cCBT being cost-effective compared with the waiting list at the NICE willingness-to-pay level of £20,000–30,000 is around 70% from both perspectives.

FIGURE 8.

Cost-effectiveness acceptability curves for supported cCBT vs. the waiting list for high-intensity CBT at 12 months from the health- and social-care perspective and societal perspective.

Guided self-help versus the waiting list for high-intensity cognitive–behavioural therapy

Figures 9–12 show the bootstrapped replications for cost and effect pairs on the cost-effectiveness plane for guided self-help versus the waiting list for high-intensity CBT at 3 and 12 months (health- and social-care perspective and societal perspective). At 3 months, the majority of the scatter points lie to the right of the vertical axis, where the bootstrap replications represent guided self-help being more effective than the waiting list, and the majority of the scatter points lie above the horizontal axis where the bootstrap replications represent guided self-help having higher costs from both perspectives, indicating a trade-off.

FIGURE 9.

Cost-effectiveness plane for guided self-help vs. the waiting list for high-intensity CBT at 3 months for the health- and social-care perspective.

FIGURE 10.

Cost-effectiveness plane for guided self-help vs. the waiting list for high-intensity CBT at 3 months for the societal perspective.

FIGURE 11.

Cost-effectiveness plane for guided self-help vs. the waiting list for high-intensity CBT at 12 months from the health- and social-care perspective.

FIGURE 12.

Cost-effectiveness plane for guided self-help vs. the waiting list for high–intensity CBT at 12 months from the societal perspective.

The guided self-help versus the waiting list for high-intensity CBT CEACs in Figure 13 indicate that the probability of guided self-help being cost-effective compared with the waiting list at the NICE willingness-to-pay level of £20,000–30,000 is 0–5% from the health- and social-care perspective and 5–10% from the societal perspective.

FIGURE 13.

Cost-effectiveness acceptability curves for guided self-help vs. the waiting list for high-intensity CBT at 3 months from the health- and social-care perspective and societal perspective.

At 12 months (primary end point), from a health- and social-care perspective, slightly more of the scatter points lie to the right of the vertical axis where the bootstrap replications represent guided self-help being more effective than the waiting list. In addition, slightly more of the scatter points lie below the horizontal axis, where the bootstrap replications represent guided self-help having lower costs than the waiting list. From a societal perspective, more of the scatter points lie to the right of the vertical axis, where the bootstrap replications represent guided self-help being more effective than the waiting list, while the majority of the scatter points lie below the horizontal axis, where the bootstrap replications represent guided self-help having lower costs.

This is reflected in the guided self-help versus the waiting list CEACs in Figure 14, which show that the probability of guided self-help being cost-effective compared with the waiting list group at £20,000–30,000 per QALY is around 60% from the health- and social-care perspective and below 80% from the societal perspective.

FIGURE 14.

Cost-effectiveness acceptability curves for guided self-help vs. the waiting list for high-intensity CBT at 12 months from the health- and social-care perspective and societal perspective.

Supported computerised cognitive–behavioural therapy versus guided self-help

Figures 15–18 show the bootstrapped replications for cost and effect pairs on the cost-effectiveness plane for supported cCBT versus guided self-help at 3 and 12 months, and from the health- and social-care perspective and societal perspective.

FIGURE 15.

Cost-effectiveness plane for supported cCBT vs. guided self-help at 3 months from the health- and social-care perspective.

FIGURE 16.

Cost-effectiveness plane for supported cCBT vs. guided self-help at 3 months from the societal perspective.

FIGURE 17.

Cost-effectiveness plane for supported cCBT vs. guided self-help at 12 months from the health- and social-care perspective.

FIGURE 18.

Cost-effectiveness plane for supported cCBT vs. guided self-help at 12 months from the societal perspective.

At 3 months, for both the health- and social-care perspective and societal perspective, more bootstrap replications lie below the horizontal axis, where the replications represent supported cCBT being less costly than guided self-help. In addition, more replications lie to the left of the vertical axis indicating lower effectiveness of supported cCBT indicating a trade-off scenario.

The supported cCBT versus guided self-help CEACs in Figure 19 indicate that supported cCBT has an 80% probability of being cost-effective compared with guided self-help at the NICE willingness-to-pay levels of £20,000–£30,000 from the health- and social-care perspective, and over a 60% probability from the societal perspective.

FIGURE 19.

Cost-effectiveness acceptability curves for supported cCBT vs. guided self-help at 3 months from the health- and social-care perspective and societal perspective.

At 12 months (primary end point), from both the health- and social-care perspective and the societal perspective, slightly more replications lie to the right of the vertical axis, indicating that the effectiveness of supported cCBT is higher than that of guided self-help. From a health- and social-care perspective, slightly more bootstrap replications lie below the horizontal axis, where the replications represent supported cCBT being less costly than guided self-help. In addition, from a societal perspective, slightly more replications lie above the horizontal axis, where the bootstraps indicate supported cCBT being more costly.

The supported cCBT versus guided self-help CEACs in Figure 20 suggest that the probability of supported cCBT being cost-effective compared with guided self-help is around 65% from the health- and social-care perspective and around 50% from the societal perspective.

FIGURE 20.

Cost-effectiveness acceptability curves for supported cCBT vs. guided self-help at 12 months from the health- and social-care perspective and societal perspective.

Supported computerised cognitive–behavioural therapy versus guided self-help versus the waiting list for high-intensity cognitive–behavioural therapy

The three interventions were compared head to head in a three-way comparison at 12 months (the primary end point). The CEAC in Figure 21 shows that the probability of supported cCBT being cost-effective at the NICE willingness-to-pay level of £20,000–30,000 is between 40% and 50% from the NHS/Personal Social Services perspective, compared with between 30% and 40% for guided self-help and around 20% for the waiting list for high-intensity CBT. The CEAC in Figure 22 shows that the probability of supported cCBT being cost-effective at the NICE willingness-to-pay threshold of £20,000–30,000 is between 35% and 40%, compared with 50–55% for guided self-help and around 10% for the waiting list for high-intensity CBT.

FIGURE 21.

Cost-effectiveness acceptability curves for supported cCBT vs. guided self-help vs. the waiting list for high-intensity CBT at 12 months from the health- and social-care perspective.

FIGURE 22.

Cost-effectiveness acceptability curves for supported cCBT vs. guided self-help vs. the waiting list for high-intensity CBT at 12 months from the societal perspective.

Methods

Results

Summary of findings

Qualitative research was nested in the intervention arms of OCTET to explore participants’ views of two low-intensity interventions for OCD. Group consensus suggested that the provision of low-intensity psychological interventions may confer substantial benefits in terms of increasing the perceived accessibility of psychological treatments for this population. Both guided self-help and supported cCBT were professed to increase perceived service flexibility, overcome intervention access barriers and sustain, where desired, a sense of anonymity or privacy in care.

Both guided self-help and supported cCBT are based on CBT principles. Integral to both interventions is the concept of problem-solving, through which participants were encouraged to work in a time-limited process towards specific, agreed goals. Much of the rise in the popularity of low-intensity interventions has been attributed to their ability to convey these empirically grounded techniques (CBT) with limited therapist resource. Paradoxically, however, the findings of our current study suggest that a minimum amount of therapeutic contact may be necessary to optimise the acceptability of these interventions.

Intervention acceptability emerged as a complex construct that extended beyond a basic concept of accessible care. Key differences in the two interventions were observed in terms of their capacity to provide the type or components of therapy most valued by service users, the ability to tailor therapy to individual symptoms and experiences, their capability to respond dynamically to participant need and to sustain participant motivation for therapy engagement. Although not guaranteed, greater amounts of practitioner contact appeared to increase the likelihood that these requirements would be met, and thus that the intervention would be positively appraised by study participants. With the two low-intensity interventions differing in the amount of professional support available, guided self-help benefited from greater PWP support and personalisation and thus was more likely to be the subject of positive feedback.

As with any study determining the acceptability of a new health intervention, variation in patient preferences and values will be observed. For a minority of individuals in the current study, access to evidence-based techniques was a priority and was in itself sufficient to confer treatment satisfaction. cCBT, with proportionally less practitioner support, thus still attracted a level of positive participant feedback, albeit less frequently and among the most self-efficacious participants.

Increasingly, innovations in treatment delivery are being proposed by health-care providers to maximise the availability, accessibility and cost-effectiveness of mental health care. Differences in the underlying treatment experiences of different individuals raises the question of whether the acceptability of low-intensity interventions is determined more by the features of the delivery model itself or the characteristics and underlying treatment preferences of the patients being referred. The current study identified six different characteristics of ‘good’ therapy, against which the acceptability of the trial interventions was judged. Future generations of low-intensity treatments for OCD, whether computerised or manual based, should seek to encompass and optimise features to maximise therapeutic uptake and population reach. Greater recognition of each of these characteristics and of their potential loss in current incarnations of low-intensity interventions would ultimately enable service users to achieve their desire for personalised treatment with a more rapid, anonymised and less stigmatising pathway to care.

Strengths and limitations of qualitative study 1

Qualitative exploration of the views of trial participants provides critical insight into the experiences of low-intensity intervention users. However, it is nonetheless subject to many limitations inherent in qualitative research, particularly with respect to generalisability. Nesting a qualitative study within a trial inevitably raises the possibility of selection bias. All interviewees had consented to participate in a randomised trial of low-intensity interventions for OCD and thus may be argued to display a level of openness towards low-intensity interventions atypical of a broader service population. OCTET recruited participants from waiting lists for high-intensity CBT and it is, therefore, possible that satisfaction with the trial interventions was elevated because of the difficulties participants were experiencing in accessing usual care. Conversely, it may also be argued that participants’ existing treatment expectations could negatively bias attitudes towards alternative and less-intensive forms of care. OCTET offered low-intensity interventions as a precursor, rather than a replacement, to high-intensity CBT. This design reflected the ethos of the NICE-recommended stepped-care models for adults with OCD. Within stepped-care models, low-intensity interventions are advocated as front-line treatments for mild impairments, with the option of ‘stepping up’ to higher-intensity care if initial intervention is ineffective. Hence, our nested trial design offered an opportunity to explore important tensions between users’ treatment priorities and preferences, and the economic and political drivers behind the routine integration of low-intensity interventions into mental health care.

By employing a purposive sampling approach, we ensured maximum variation in the ages and geographical locations of study participants. Heterogeneity in ethnicity was more limited. Our sample was predominantly white British, reflecting an underlying bias in our trial population. Had it been possible to include the views of individuals from different ethnic backgrounds, this may have contributed substantially to informing culturally relevant and sensitive care.

The majority of the interviews were conducted by a single researcher, who had also conducted trial follow-up visits for the purposes of quantifying intervention outcomes. Although this may have hindered as much as facilitated participants’ willingness to discuss their treatment experiences, the broad array of views expressed in relation to supported cCBT and guided self-help go some way to negating this concern. Independent data coding, undertaken by three academic researchers in collaboration with a service user representative, raises confidence in the rigour of our analysis. The views of those allocated to a waiting list for high-intensity CBT were not explored qualitatively; as such, data were unable to illuminate treatment experiences. They may, however, have offered an objective insight into the treatment expectations and preferences of a clinical sample, not affected by recall bias or the post hoc reconstruction of events. Baseline treatment preferences were measured quantitatively in the trial to redress this knowledge gap.

Methods

Results

Summary of findings

This qualitative study aimed to explore PWP perceptions of two low-intensity interventions, guided self-help and supported cCBT, for OCD. Intervention acceptability was judged both from the perspective of the professionals concerned, and their personal interpretations of patient need. Practitioners were consistent in acknowledging the advantages of low-intensity interventions at a population level. Both guided self-help and supported cCBT were advocated to overcome long-standing barriers to the delivery and receipt of mental health care, improving accessibility via enhanced service flexibility and patient choice.

At an individual level, differences in practitioners’ perceptions of patient need emerged as a key factor influencing their intervention acceptability judgements. Group consensus identified OCD as a complex and chronic mental health disorder with variable symptoms profiles and presentations. Concerns were thus initially raised regarding the ability of low-intensity interventions to meet the needs of this client group, with different levels of professional confidence generating different levels of readiness to deliver these treatments. Dedicated training and the provision of essential therapist resources were regarded as vital precursors of a service roll-out, and a particularly advantageous element of OCTET. Practitioner training was perceived to be instrumental in changing practitioner attitudes, reducing fears and ensuring the fidelity and delivery of the trialled interventions.

Potential challenges to national implementation nonetheless remain. Practitioners in the current study often referred to their own professional role and responsibilities, contrasting these with the remit of higher-intensity, accredited cognitive–behavioural therapists. Differentiation between these two service levels gave rise to a distinction in the types of patients and disorders traditionally seen by each group and thus to a prospect of wider professional resistance. A shared fear among several PWPs was that OCD was a disorder that conventionally fell within the remit of higher-intensity psychological services and thus was unlikely to be sanctioned for treatment by paraprofessionals. Even within the context of entrenched models of PWP, existing service philosophies and protocols appeared to affirm the importance and centrality of the therapeutic relationship, and thus sometimes brought into question the legitimacy of reduced client contact achieved via the instigation of self-managed therapy packages. Trial protocols stipulated the amount and nature of contact permitted for the guided self-help and supported cCBT intervention. Longer-term integration into services demanding remote working practices or seeking additional resource efficiency savings may only exacerbate this problem.

Studies of remote psychiatric consultations suggest that the normalisation of any new health technology into statutory health care is likely to be mediated both by the properties of the technology itself and the sociological orientation of its users. 89 With the two low-intensity interventions differing in the amount of professional support available, guided self-help aligned more readily with practitioners’ own expectations of their role. Differences in the presentation and functioning of guided self-help and supported cCBT resources exacerbated this effect, influencing the perceived suitability of the two different intervention models for the tasks in hand. Although supported cCBT was initially perceived to be an innovative solution to service access, technical difficulties were found to limit its reach. The importance of providing high-quality intervention resources was a key theme emerging from practitioner discourse, and a key feature influenced the perceived credibility and acceptability of low-intensity interventions. Future generations of supported cCBT and guided self-help must, therefore, ensure that they are optimised in this regard.