Chapter 6OCTET Trial

Clinicians and approved social workers

The long and heated debate preceding the introduction of CTOs nationally was reflected in the views raised during our consultations. Some were concerned that CTOs constituted ‘incarceration in the community’ and that CTOs might lead to a lowering of the threshold for compulsion so that a larger proportion of people at relatively low risk might be subject to compulsion. Others commented that, in practice, the new provision did not represent much change and that ‘we are doing it anyway with Section 17 as a long leash: at least a CTO is honest’.

Clinicians noted that attitudes to CTO use had changed since the new legislation had been passed by Parliament. Much of the opposition had dissolved and CTOs were rapidly becoming accepted as part of the provision that was soon to be implemented. This was expressed in concerns or even fear associated with not complying with the new legislation: ‘As a clinician, I will opt for a CTO because of the new law [when it takes effect]. We would be frightened what will happen if we do not.’ They were also concerned about public harm and liabilities in the event that tribunals would discharge patients in the non-CTO group who subsequently committed crimes. As such, CTOs were considered the more restrictive option, with more control remaining with the clinician.

At the time of the consultation, few mental health professionals had received training on the amended MHA and how it would impact on services. There was considerable uncertainty about how CTOs would work in practice and whether or how Section 17 Leave would change. Prolonging Section 17 Leave was seen as being likely to be challenged by tribunals, as the new Code of Practice indicated that this would not constitute ‘good practice’. Psychiatrists were concerned that mental health lawyers might be eager to test the amended MHA, and worries were expressed about how to explain to tribunals why an RC had repeated Section 17 Leave instead of placing the person on a CTO. Several psychiatrists stated ‘I don’t want to be involved in the first judicial review’.

Mental health lawyers

Lawyers representing patients in tribunals emphasised the right of patients to make decisions about their treatment, including taking part in trials; they suggested that legal representatives would be obliged to take a client’s participation in any study into account when representing them at a hearing. They strongly emphasised that patients must be fully informed about the trial and that the different mechanisms for treatment following from randomisation should be made clear to patients prior to enrolment in the study.

The legal representatives we spoke to expressed surprise at the degree of fear of the MHRT among clinicians. Their view was that a tribunal only recommends treatment and that it is a matter for clinicians to ‘stick to their guns’ regarding what they consider an appropriate course of action.

Academic legal experts

Unlike many clinicians, who believed that the clinical effectiveness of CTOs needed to be tested, some of the academic legal experts we consulted saw the overall research question as being of limited interest to the law, since the amended MHA had already been passed by Parliament and CTOs were soon to become part of the MHA.

Some legal experts expressed uncertainty whether or not the new Act presented the option of using Section 17 Leave as the control arm of the trial, given the directions in the Code of Practice; they anticipated that MHRT hearings would insist on the Code being followed. They thus saw the MHRT, despite its inability to go beyond recommending treatment, as potentially exerting influence over the trial.

The legal experts also identified a worry about a shifting threshold for detention, and they described the new Act as ‘vague’ on this point. Some believed that the introduction of CTOs might lead to an increasingly ‘defensive practice’ in which clinicians might leave it to tribunals to discharge patients from CTOs. Moreover, would some such ‘defensive’ clinicians, if a patient were randomly allocated to be managed without CTO, end up using Section 17 Leave for longer periods than they would have done otherwise and thus threatening the naturalistic design of the trial? Some experts were also sceptical as to whether or not clinicians would be willing to put patients forward given that the randomisation would lead to some patients who were deemed appropriate for CTOs not being put on to CTOs.

The legal experts raised two related issues regarding clinical equipoise. First, would there be genuine equipoise when the clinical judgement was that CTO would be appropriate? Second, would clinicians who were willing to put patients on to a CTO be in equipoise? In other words, once a clinician deemed that a CTO was appropriate for a particular patient, would they then deem CTOs to have benefits and by definition not be in equipoise?

Mental Health Review Tribunal

Those representing the MHRT expressed a different view of the Code of Practice compared with some academic lawyers. Although, from the point of view of Government, there would be little point in issuing a new Act with provision for CTOs without a steer as to their use, they argued that case law illustrates that the Code does not constitute a statute; making decisions contrary to the Code might therefore be in adherence with the MHA.⁹⁵ Moreover, they pointed out that the Code is directed towards clinicians rather than tribunals.

Regarding the appropriate treatment arms for a study, the MHRT stakeholders raised the issue of what would constitute the least restrictive option. Established case law and European law require a tribunal to apply the least restrictive option. As a matter of law, the MHRT stakeholders considered that if a patient were subject to Section 17 Leave, this would mean that Section 3 was still active and the patient was still detained, whereas a CTO would represent a form of discharge from Section 3 (unless the patient was recalled), so that the patient would not be considered to be detained despite the restrictions to their personal freedom. Contrary to the impression of most clinicians, they thus saw CTOs as the least restrictive option, based on Article 5 of the European Convention on Human Rights (albeit not yet tested in a court of law).⁹⁶ Following from this, they considered that it would be difficult for a tribunal to believe simultaneously both that Section 17 Leave was appropriate and that a CTO was appropriate for a particular patient. They expressed a degree of surprise at what they saw as ‘liberal’ views and uses of Section 17 Leave among clinicians, as, from their point of view, it constituted part of Section 3.

This position on the relative restrictedness of CTO and Section 17 Leave was seen as relevant to the equipoise of any clinician participating in a trial. The MHRT stakeholders explained that when a tribunal looked at a patient’s circumstances (e.g. treatment, potential risks) and decided whether or not, on balance, continued detention or a CTO was justified, only clinical arguments would be considered. If the clinician were in genuine equipoise, it was suggested, they might be unable to provide such an argument. They advised us to seek separate legal advice to address this issue.

From the viewpoint of the MHRT stakeholders, it was mental health lawyers who might provide a challenge to the trial: they might be eager to test the new provision via judicial review should they believe that a tribunal had been influenced in any way by any factors other than Statute. They also noted, however, that a robust clinical argument would in most cases suffice for the use of either option; while the tribunal might recommend a CTO, clinicians did not have to follow their recommendation. They commented that perhaps some clinicians ‘fear the tribunals too much’.

Service user representatives

There was less enthusiasm for taking part in the consultations among service user representatives and mental health organisations than among the other groups of stakeholders and thus fewer people took part or expressed their views. Overall, there was strong support for testing the effectiveness of CTOs, but they expressed concern over the lawfulness of randomisation and whether or not clinicians would find it ethical to randomise. They also reiterated the importance of fully informing patients about what taking part would entail and of their right to withdraw, and recommended using a wide range of outcome measures.

Key issues raised in the consultation exercise

The consultation exercise thus elicited a range of views as to the various legal, clinical and practical aspects of the proposed protocol. In particular, although many clinicians considered CTO to represent a more coercive alternative than Section 17 Leave, many lawyers considered it to be less restrictive. Three issues emerged overall:

1. what would constitute appropriate treatment arms of the study
2. issues of equipoise
3. whether randomisation between different legal statuses would be lawful.

These are summarised here along with the conclusions we reached about our trial design based on the consultation.

Treatment arms It was clear that to randomise between a CTO and voluntary status, as the two US RCTs had been able to do, would be unacceptable to clinicians and ethicists, and might be considered unlawful should it be tried in the courts. The consultation exercise confirmed that patients in the control arm of our trial should instead leave hospital via Section 17 Leave. Many consultees were concerned about the lawfulness of prolonged use of Section 17 Leave or how it would impact on the research. To preclude the threat of prolonged use of Section 17 Leave during the trial, which would serve as a de facto CTO and contaminate the results, we decided that we would recruit only from clinicians who used Section 17 Leave as intended by the Code of Practice that accompanies the amended MHA. This states that if extending Section 17 Leave beyond seven consecutive days, the RC should consider a CTO instead. This gives a clear indication that Section 17 Leave should be treated as a short-term measure, whereas prolonged community compulsion should be managed by a CTO.

As the legal analysis indicated, for the trial to be lawful, patients needed to be randomised to two equally restrictive positions. Therefore, this confirmed that it would not be lawful to randomise a patient either to be put on a CTO or to be discharged to voluntary status. The solution, which our experts deemed to be both ethically and legally valid, was to randomise the patients to leave hospital either on a CTO or via Section 17 Leave. Those in the control arm (henceforth referred to as the ‘non-CTO arm’) would then proceed to voluntary care.

Equipoise The issue of equipoise is central to all RCTs, as no clinician should put forward a patient to a trial if he or she considers treatment in one arm to be inferior to that in the other. Given the fact that there was no robust evidence for CTO effectiveness at the time and that English psychiatrists did not have any clinical experience on which to base an opinion, by definition a state of equipoise existed. We decided to recruit only from psychiatrists who accepted this. This was in order to address the concern that a psychiatrist might be willing to refer one patient but not another (cherry-picking the RCT patients). Given the state of the evidence, if one were in equipoise in one case, one must be in equipoise on all cases. We did accept, however, that there might be exceptional circumstances (such as insistence from family) when an eligible patient was not put forward. Moreover, there were psychiatrists who, perhaps as a result of the heated debate preceding the introduction of CTOs, were convinced that CTOs would be the best option, and there were also psychiatrists who were equally convinced that CTOs were not effective or should not be used for ethical reasons. It was considered unlikely that either of these groups of psychiatrists would participate in the trial. This is a limitation that applies to most RCTs. (For further details about the state of equipoise, see Chapter 11.)

Lawfulness of randomisation The legal experts involved in our consultation exercise identified a need for further work on understanding the legal issues arising from our proposed trial design and this issue was also raised by the Research Ethics Committee. Following their initial review, and the stakeholder consultation, we therefore commissioned a formal legal opinion from John Dawson, Professor of Law, and Marion Rickman, mental health lawyer (and subsequently tribunal judge). This concluded that our protocol was lawful. To arrive at this decision, they examined three features of the law in detail to answer the questions:

• Do the legal criteria for the two chosen treatment arms overlap? Is there therefore a group of patients who may simultaneously meet the legal criteria for both treatments?
• Are both the treatment arms equally restrictive in terms of the patient’s liberty?
• Is there a situation in which RCs can be genuinely uncertain as to which of the two treatments is the more appropriate option for an individual patient?

They concluded, respectively, that:

• there would be a number of patients fulfilling the criteria for detention under both regimes
• it was difficult to ascertain which of the two treatment arms would constitute a more or less restrictive option
• genuine equipoise existed, making it both ethical and lawful for RCs to allow a specific group of patients to be randomised between a CTO and Section 17 Leave.

Thus, within particular constraints, their legal opinion was that an RCT of CTOs would be feasible. (The legal analysis is reported in detail in Chapter 11.)

The legal analysis also concluded that the period soon after the change in the MHA would provide a window of opportunity to study the outcomes for patients of the varying clinical practices that were likely to emerge from the position of equipoise in which clinicians would find themselves at that point, because of genuine uncertainty about both treatment efficacy and the proper application of the law. The legal analysis also advised on the correct procedure for randomisation and advised that clinical decision-making should not be influenced by the patient’s participation in the trial or randomised allocation (see Overview of the OCTET Trial).

The exploration of legal, ethical and feasibility issues during the consultation period shaped the final research protocol. In addition to clarifying the randomisation process, it also helped shape the design of eligibility criteria, outcome measurement and strategies for recruitment and data collection, all of which we describe next.

Design

The OCTET Trial was a single-outcome, parallel-arm, non-blinded randomised trial. Its primary objective was ‘to conduct the most rigorous trial possible of CTOs with prolonged, high-quality care incorporating a broad range of outcomes, and identify patient and service predictors of response’. Our aim was to compare CTO use to voluntary outpatient treatment. This aim was modified on the basis of the legal analysis, as explained above. Our trial therefore randomised the patients to leave hospital either on a CTO or via Section 17 Leave. There was an understanding that for patients in the non-CTO arm, Section 17 Leave was to be used according to the MHA Code of Practice⁴⁸ and be restricted to a short period of days, or at most weeks, before discharge to voluntary care took place. The trial thus randomly assigned patients to receive one of two forms of mandatory outpatient care at the point of discharge from inpatient care: the two main options for clinicians regarding patients who needed ongoing supervision in the community from November 2008 onwards. Patients were assessed at baseline, and at 6 and 12 months, and the primary outcome measure was readmission to hospital, as detailed below and represented in the OCTET flow chart (Figure 3).

FIGURE 3

The OCTET Trial procedure.

Taking into account the findings from the North Carolina RCT,⁶⁴ which suggested a correlation between outcomes and the combination of duration of CTO and intensity of services, we designed the OCTET Trial so that patients in both arms would receive approximately weekly contact with services. This was in line with common practice for this group of patients in English services.

Eligibility

Patients were deemed eligible for the study when the RC and AMHP agreed that they needed ongoing supervision in the community, but (after having considered the relevant legal standards, including the guidelines in the Code of Practice) when clinicians recognised the uncertainty as to which option was more appropriate (CTO or non-CTO). Potential referring clinicians who held strong views in favour of certain patients being likely to be more or less well on a CTO were actively discouraged from referring any patients to the trial, so that the referring clinicians would be those, reflecting the evidence base, with a neutral view (accepting the state of equipoise), thus avoiding cherry-picking. Clinicians who used Section 17 Leave as a de facto CTO were excluded from recruiting to the trial.

During the preparations for the amendment of the MHA, discussions around CTOs focused almost exclusively on the situation of so-called revolving-door patients, that is, patients experiencing frequent relapse and readmission to hospital. Both expert views and the evidence from other jurisdictions suggested that CTOs were likely to be mostly used for patients with psychosis in adult mental health services. The amendment did, however, pave the way for the use of CTOs for a much wider group of patients, including those with learning disabilities and in forensic detention. Given the potentially different clinical uses of CTOs for different patient groups, we decided to design the trial to focus on a relatively homogeneous group of patients. We therefore decided to exclude those on restricted forensic sections and those without a primary diagnosis of psychosis, as well as those in adolescent or older adult services. Including the other groups would require a significantly larger sample size, which could prove unachievable given restrictions of time and resources. The group included matched those studied in the majority of CTO studies internationally, to facilitate comparisons.

Clinical decision-making

Once we had recruited a patient and conducted the baseline interview, legal advice indicated that the correct procedure was for the independent statistician to advise the RC on the random allocation, who would then allocate the patient to either the CTO group or the non-CTO group. In practice, RCs referring to the trial undertook to allocate the patient according to the statistician’s advice; this is no different from the usual randomisation procedure in any RCT. This procedure is referred to simply as ‘randomisation’ below.

After patients had been recruited and randomised, it was stipulated that when RCs made decisions about them, they should strictly adhere to the statutory process and criteria in every case, as the law required, and should fully consider every option open to them within the legal regime and the relevant factors listed in the statute or the Code of Practice. We thus did not exercise control over clinical decisions subsequent to randomisation. Patients in both groups retained all their usual legal rights, including their right of access to the MHRT, which might discharge them from compulsory treatment. No attempt was made to influence how clinicians subsequently managed these patients, including whether the clinicians decided to renew a patient’s treatment on Section 17 Leave or renew their CTO. When a patient was discharged from involuntary treatment by the RC or the MHRT, or transferred from Section 17 Leave to a CTO, for instance, as required by law, this would simply be counted as one outcome of the process for that patient.

Patients in the CTO arm of the study might therefore, during the follow-up period, be recalled to hospital, have their CTO revoked or be discharged from it. Patients in the non-CTO arm might be discharged to voluntary status after the initial period of leave, or might have it renewed, or they might, in exceptional cases, be put on a CTO. (Details of which of these scenarios we handled as protocol violations are given below, under Methods, while actual protocol violations appear under Results.) Given the Code of Practice, there was a reasonable expectation that the majority of patients randomised to the non-CTO arm of the trial might proceed fairly swiftly to voluntary status. We collected data on changes to legal status systematically as part of the trial.

Procedures and NHS permissions

Permission to carry out the trial needed to be obtained for each trust, and each research assistant needed to obtain personal permission to access professionals, patients and medical records in each trust. We sought permission to conduct the trial locally from R&D offices in around 60 NHS trusts that provided mental health services in England. All trusts within a reasonable travelling distance were approached.

Having secured R&D approval, we carefully mapped that trust’s services before embarking on extensive information sharing and a series of meetings with the mental health teams to secure their collaboration.

We then identified potentially eligible patients from a network of CMHTs and AOTs agreeing to participate in the trial. In NHS trusts with a split between inpatient and outpatient services, patients were recruited via the inpatient service. This happened, however, in agreement with the outpatient service to which the patient would be transferred when leaving hospital. All community teams had to be able to provide approximately weekly contact with the patient.

Following identification of an eligible patient, a member of the care team then approached the patient and secured agreement that one of the study researchers could make contact. This researcher then assessed and recorded the patient’s capacity to participate in research and sought written informed consent for inclusion in the study. Once the consent form had been signed and the baseline assessment completed, the patient was randomised.

For the follow-up interviews, the researcher contacted the mental health team to confirm that the patient was still in their care or, if not, where they were then based; and whether they believed that it was appropriate to contact the patient.

Obtaining access in NHS trusts

Obtaining permission in the NHS trusts we approached proved much more time-consuming than expected. The time required to deal with our applications varied considerably. Although one trust was able to process the necessary paperwork in a matter of days, for others it took nearly a year; in some cases, it took longer and we were eventually obliged to give up. The extent of documentation required varied and, although this was usually easily dealt with, it caused some delays, in particular in the cases when a trust required specific contracts drawn up between themselves and the host trust. One trust refused collaboration because the service leads did not give their support to the study. Two others refused as a result of their own internal peer reviews of the protocol: one review concluded that there was no immediate benefit to that trust; the other deemed the protocol unethical. It took us between 2 days and, in one case, 11 months, to gain R&D permissions in the 44 trusts. Of these 44 trusts, 32 trusts – predominantly in the Midlands and southern England – recruited to the trial (see Results, below).

Each of our research assistants (14 in all, over the 28 months of recruitment and data collection) was required to obtain individual permissions in each of the collaborating trusts in which they were working. We found that a wide range of procedures for obtaining such letters of access was required. Despite NHS guidelines, a number of trusts required new Criminal Records Bureau (CRB) checks to be conducted, which caused several months delay in each case. In total, 607 letters of access, 56 honorary contracts, 17 research passports and seven clinical licences were processed and issued to our researchers: on average, 21 contracts per trust. In addition to this, 51 CRB checks were also performed (on top of those issued on behalf of the employing trust).

Getting mental health teams and their patients on board

We initially focused on recruiting from AOTs, so began by mapping the location and contact details of these teams in the participating trusts. It proved more difficult than we had expected to find out the number and location of AOTs, as it was hard to find people within each trust with an overview of how services were organised and divided both geographically and functionally. This process resulted in a largely complete map of AOTs across our chosen trust areas, of which there were approximately 200.

We had initially envisaged that the vast majority of the patients placed on CTOs in England would come from AOTs, as both AOTs and CTOs were designed to be employed for revolving-door patients. We quickly realised, however, that a considerable proportion of patients from regular adult CMHTs were also being placed on CTOs, and we therefore made the decision to recruit from these CMHTs as well as AOTs. This was a huge undertaking. Not only were there on average three times the number of CMHTs as AOTs in each trust, but also the functional split between inpatient and outpatient care meant that all of the inpatient services into which CMHTs fed had to be contacted separately. This meant that we had a total of approximately 1000 potential teams, including CMHTs and AOTs, as well as approximately 500 inpatient units.

We contacted by e-mail and telephone the team leaders, the consultant psychiatrist for each team and the medical directors for each trust. It was at this point that it became apparent to us that services were increasingly being split into those providing care for inpatients and those providing care for outpatients. This situation differed from the pre-existing arrangement in which a consultant psychiatrist was attached to a particular geographical area, within which he or she had responsibility for the care of patients regardless of whether they were currently in hospital or in the community. This increased the number of consultant psychiatrists we were to contact considerably.

Once contact had been established with the mental health teams and information about the OCTET Trial had been sent to them, we used a combination of large meetings, forums, academic meetings and individual team meetings to inform teams and doctors about the study and agree a plan for the recruitment of their patients to the study by their own referral. We also met with ward managers and inpatient consultants to secure their agreement for us to monitor the inpatient wards for eligible patients. We then e-mailed or telephoned the clinicians responsible for potentially eligible candidates to seek their permission to meet the patient and proceed with the consent process. Our recruitment therefore proceeded along two routes: both through the direct referral of patients by the consultant and team and through the identification of eligible patients from wards.

In getting teams on board, holding large meetings in which many teams and doctors were present was most successful. In these meetings, the issues surrounding the OCTET Trial were thoroughly discussed. In each case, we then held a series of smaller meetings with individual interested teams and consultants, to discuss practicalities and specifically to ascertain whether or not there were any suitable patients known to the team. This ensured that research assistants mainly travelled to individual teams that had already expressed interest in the study and that were aware of the broad outline of the study. Even so, many teams were unable to find eligible candidates once they heard the inclusion criteria in more detail, even when they were happy to do so in principle. In these cases, the inpatient ward monitoring process was designed to ensure that should any eligible candidates exist in the future, they would be identified.

Ultimately, recruitment was most successful with teams and doctors who had met a member of the team in person and had the opportunity to discuss any issues and concerns in detail. Owing to the number of teams involved, this initial setting-up phase of the study was both resource heavy and time-consuming. Once the teams and doctors were familiar with the study team, however, they were often happy to receive further e-mails prompting them to refer or asking permission to see a patient identified on the ward. This system made recruitment reliable and efficient, despite requiring a considerable input of effort initially. Because of the complexity of the study, the recruitment of patients by a less personal process, for example solely by e-mail, often resulted either in a refusal to recruit, or in confusions and misunderstandings about the study process, and ultimately wasted time. As teams were located in such disparate locations as Cornwall, Lincolnshire, Greater Manchester and Birmingham, one of the biggest challenges in the OCTET Trial was to use this personal and hands-on approach with a team of six researchers, all of whom were based in Oxford.

Potential obstacles to referral were many and varied. First, some team members or consultants were not in equipoise, so were opposed in principle to allocating their patients randomly to one of two different legal statuses. Both the multidisciplinary nature of mental health teams and the functional inpatient–outpatient split increased the number of professionals involved in, and responsible for, the care of individual patients, which ultimately increased the likelihood of at least one of those professionals objecting to the patient’s involvement in the trial. This proved to be the biggest problem: when the parties in disagreement were the inpatient and outpatient clinicians for one patient, one of whom would implement the CTO, or not, depending on the outcome of randomisation, and the other of whom would manage the patient in the community. If these parties could not agree on the acceptability of randomisation then recruitment could not go ahead. Once a patient had been identified as eligible and permission had been gained from both the inpatient and outpatient consultants, and the relevant team members where necessary, we would be able to approach the patient.

This final stage of recruitment also posed several problems. First, access to wards in trusts that were remote from our own proved at times unreasonably difficult, given that we had official researcher access to all relevant sites. Second, a lack of communication between doctors and ward staff, or within the ward, resulted in a few cases in patients being sent on leave or discharged before we could see them. This was despite the rapidity with which the research team responded to patient referrals, often seeing patients no more than a few hours after they had been referred. A third issue was premature referral, which resulted in some patients being too unwell at the time we saw them to give informed consent. This did not often prevent the patient being recruited, as the researchers would return to the ward when they were well enough, but it did waste time and resources. Last, patients themselves sometimes refused to take part or were judged by the research team to lack the capacity to consent. We stayed in contact with the recruiting teams regularly via e-mails or phone calls, and we produced a monthly newsletter in which we provided information about the progress of the trial and relevant information about CTOs. We stayed in touch with patients via postcards at Christmas and made contact again some weeks before each 6-monthly interview.

In these extensive recruitment activities, we got considerable support from the Mental Health Research Network, the clinical support officers of which often provided considerable practical assistance in local NHS trusts, such as making contact with hospital wards and local clinicians, and in some cases helping with obtaining R&D access and accessing medical records.

In total, the study team assessed 442 patients, of whom 336 were randomised, as we detail in Results.

Demographics and psychiatric history

• The OCTET Socio-demographic Schedule This collected data on three areas:
  • ‘Self and Home’ (basic information including age, sex, ethnicity and educational achievement, employment, family, including marital status; living situation).
  • ‘Clinical History’ (diagnosis, psychiatric history, current psychiatric medication), which was corroborated from medical records.
  • ‘Legal History’ (criminal convictions, imprisonments).

Clinical and social outcomes

• Symptom severity The Brief Psychiatric Rating Scale (BPRS),⁹⁷ a researcher-rated instrument, was used to assess symptom severity on a scale from ‘1’ (not present) to ‘7’ (extremely severe) over the last 2 weeks. The total score (range 24–168) is a sum of ratings across 24 symptom domains.
• Insight The Insight and Treatment Attitudes Questionnaire (ITAQ)⁹⁸ is an 11-item questionnaire assessing patient awareness of illness and need for treatment on a three-point Likert scale (no = 0, possibly = 1, yes = 2). Total scores range from ‘0’ (no insight) to ‘22’ (full insight). Its two subscales are the Awareness of Illness Scale and the Attitude to Treatment Scales (range 0–10 and 0–12, respectively).
• Substance misuse The CAGE^99,100 is a screening questionnaire for alcohol and drug misuse. Four items are rated for each area, covering the last 30 days, with yes/no responses. Two or more positive responses indicate a drink/drug problem. Its four questions focus on ‘Cutting down’, ‘Annoyance by criticism’, ‘Guilty feeling’ and ‘Eye-openers’, providing the acronym for the scale.
• Social functioning The Global Assessment of Functioning (GAF),¹⁰¹ a researcher-rated, single-item scale, was used to assess impairment in functioning over the previous 2 weeks. Scores range from ‘1’ (severe, life-threatening impairment) to ‘100’ (superior functioning). The rating is based on the information collected during the course of the interview.
• Overall social outcomes The Objective Social Outcomes Index (SIX)¹⁰² summarises objective indicators of social outcomes (employment, housing, living status and social contacts) in one overall score (range 0–6), with higher scores indicating better social outcome.
• Health-related quality of life The EuroQol-5 Dimensions (EQ-5D),¹⁰³ a non-disease-specific five-item scale, is commonly applied in health economics research and was used to ascertain the patient’s self-description and valuation of their health status. Data from this measure are reported under OCTET Economic Evaluation (see Chapter 7).

Medication

• Type of medication We collected data on prescribed psychotropic medication from medical records.
• Attitudes and adherence to medication The Drug Attitude Inventory, 10-item version (DAI-10)^104,105 assesses experiences of, and attitudes towards, medication. Patients rate 10 statements as true or false (variously scored as ‘–1’ or ‘1’). The total score ranges from ‘–10’ (negative subjective response/adherence) to ‘10’ (positive subjective response/adherence).

Experiences of services

• Therapeutic relationships The Scale To Assess Therapeutic Relationship–Patient Version (STAR-P)¹⁰⁶ assesses community patients’ relationships with their care co-ordinators. Patients rate, on a five-point Likert scale (0–4), the frequency with which communication, consultation and trust is present in interactions across 12 items, which constitute three subscales: Positive Collaboration (range 0–24), Positive Clinician Input (range 0–12) and Non-Supportive Clinician Input (range 0–12). The total STAR-P score ranges from ‘0’ to ‘48’. Higher scores indicate better relationships on all scales (with Non-Supportive Clinician Input being reverse scored).
• Satisfaction with services The Client Satisfaction Questionnaire (CSQ-8)¹⁰⁷ assesses satisfaction with health services on eight items using a four-point Likert scale (1–4). It yields a total score ranging from ‘8’ (low satisfaction) to ‘32’ (high satisfaction).
• Preference for joint decision-making and information-seeking The Autonomy Preference Index (API),¹⁰⁸ adjusted to the mental health setting,¹⁰⁹ is a 23-item questionnaire on preferred autonomy (14 items were utilised for the purposes of our analysis). Its two subscales are the Decision Making Preference Scale, which measures patients’ desire for their own rather than their psychiatrists’ involvement in clinical decision-making, and the Information Seeking Preference Scale, which measures patients’ desire to be informed about their illness and treatment. Both use five-point Likert scales (strongly agree = 1, strongly disagree = 5). Adjusted scores for each subscale range from ‘0’ to ‘100’, where ‘0’ is a lack of desire to be involved/informed, ‘100’ the strongest possible desire and ‘50’ indicates a neutral stance.

Experience of leverage and informal coercion

• Experience of coercion The MacArthur Leverage Interview⁷ ascertains patients’ experience of leverage. Leverage was defined as making support to obtain housing, money and child custody conditional on treatment adherence or reducing or dropping criminal charges if patients adhered. It is a semistructured interview designed to ascertain experience of leverage during the 6 months prior to interview and across the patient’s lifetime. Here we report experience over only the previous 6 months. We adapted it for use in the English setting for the ULTIMA Study, including adding a section asking about child access. Questions test for both access to, and potential withdrawal of, benefits. Any positive response within a specific coercion area counts as ‘reported’ (Box 2).
• Perceived coercion The MacArthur Admission Experience Survey (AES),¹¹¹ adapted for outpatient use,¹¹² contains 14 statements rated on a five-point Likert scale (1 = strongly agree, 5 = strongly disagree). There are three subscales:¹¹³ the Perceived Coercion Scale assesses perception of influence and control in treatment decisions (range 5–25, with a high score indicating a high level of perceived coercion); the Negative Pressures Scale assesses experienced threats and force (range 6–30, with a high score indicating higher negative pressure after reverse scoring); and the Procedural Justice Scale assesses experience of having a say in one’s care (range 3–15, with a high score indicating feeling less involved in one’s care). No total score for the AES was utilised.
• Fairness and effectiveness of pressure The Index of Fairness assesses patient agreement with statements on the fairness of any treatment pressure they have experienced in the last 6 months, and the Index of Effectiveness assesses their views of how effective this pressure has been in making them stay in treatment and gain control.^114,115 It is rated on a five-point Likert scale (1 = strongly agree, 5 = strongly disagree) with a total score for each scale (range 4–20), with higher scores indicating that the pressure is viewed as more fair/effective.

BOX 2

Questions used to identify experienced leverage

Service usage

• A modified version of the Client Service Receipt Inventory (CSRI),¹¹⁶ which measures the number and duration of contacts the patient has had with a range of health and social care professionals during the 6 months prior to interview and the location of these meetings (24 items). (Some data from this measure are reported below but it was most extensively used for the OCTET Economic Evaluation reported in Chapter 7.)

(The full Schedule of Procedures and the instruments listed here are available from the authors on request.)

Intention-to-treat population

Analysis was carried out on an intention-to-treat basis (n = 336), except for the tertiary analyses for which the sample was 333. The intention-to-treat population included all randomised patients; we thus analysed data from dropouts or protocol violators according to their randomised group.

Baseline characteristics, interview refusers and loss to follow-up

We assessed the baseline comparability of the two randomised groups by tabulating patient characteristics and treatment experiences; we did not perform any statistical tests on baseline data. We compared the baseline characteristics of any patients who refused to participate in follow-up interviews (or who had inadequate English language for this) to those of patients completing the follow-up interviews.

Primary analysis

The primary analysis was a test of the difference in the proportion of patients readmitted to a psychiatric hospital during the 12-month follow-up period between the CTO arm and the non-CTO arm. We analysed the primary outcome using a log-binomial regression model adjusted for the stratification factors (sex, schizophrenic status and duration of illness). Results are presented as the relative risk of readmission for the CTO group compared with the non-CTO group, with appropriate 95% confidence intervals (CIs) and two-sided p-values.

Secondary analyses

We conducted secondary analyses using the intention-to-treat population. There were no missing data for secondary outcomes. We analysed secondary outcomes in the same way as primary outcomes, using multiple regression models with adjustment for stratification factors. The type of regression model depended on the data distribution. All model assumptions were assessed.

Number of readmissions and number of nights in psychiatric hospitalisation are count outcomes, and we analysed these using adjusted zero-inflated Poisson and negative-binomial regression models, respectively. Results are presented as adjusted incidence–density ratios (IDRs) with 95% CIs, and interpreted in the same way as relative risks.

The number of nights from first discharge to first readmission and the time spent under compulsion are time-to-event outcomes. We therefore performed these analyses using adjusted proportional hazards regression, and present the results as hazard ratios (HRs) with 95% CIs. Kaplan–Meier plots are also presented and the median time to readmission is calculated with 95% CIs. We used the log-rank test to compare the median time under compulsion between the two arms, whereas we used the Wilcoxon rank-sum test for comparison of time to first readmission, as this variable violated the log-rank test assumptions.

Tertiary analyses

We conducted tertiary analyses using the population of 333 patients (omitting the three exclusions). Thus we included all available data from all of the patients, and imputed missing values intrinsically within the model rather than requiring multiple imputations.

Sensitivity analyses

We conducted a repeated measures sensitivity analysis for end points measured at multiple time points using multivariable mixed-effects regression models. We used the Stata command xtmixed for continuous end points, xtlogit for binary end points and xtmelogit for multiple category end points. We included all available data from all patients, with missing values intrinsically imputed within the model rather than requiring multiple imputations. We entered treatment, stratification factors and time point (time since randomisation) into the model as fixed effects, and the model contained a patient-specific random intercept. We treated an interaction between time point and treatment group as a fixed effect to allow estimation of treatment effect at each time point. We also used likelihood ratio tests to assess whether or not time should be included in the model as a random effect. We also explored different covariance structures.

As a sensitivity analysis, we imputed missing data in self-reported outcomes using multiple imputations¹¹⁷ in the mi routine in Stata. Subsequently, we used regression modelling to estimate the association between change from baseline to 12 months and study arm, adjusting for baseline end point and randomisation factors. We chose the regression models according to the distribution of the end points (logistic models for CAGE, leverage and employment; linear models for CSQ-8, STAR-P, BPRS, GAF, ITAQ, DAI-10, API, SIX, AES subscales, Index of Fairness and Index of Effectiveness).

We performed safety analyses according to the statistical analysis plan (see Appendix 2).

Subgroup analysis

Subgroup analyses are designed to explore whether or not any treatment effect tested in an RCT varies across subgroups defined by baseline patient characteristics.¹¹⁸

We defined binary subgroups a priori, as identified in the literature to be related to outcomes. We performed subgroup analysis for the primary outcome, all of the secondary outcomes apart from time under compulsion and for the main clinical outcomes, BPRS and GAF. Potential errors (i.e. increased type 1 error and decreased type 2 error) introduced by a subgroup analysis were controlled by selecting the subgroups and stabilising the hypothesis to test (i.e. equal direction of effect as for the overall sample) prior to accessing the data and thus performing the analysis. With 13 subgroups there is a 12% chance of at least one subgroup resulting significant at the 5% level. All conclusions were written paying due consideration to this fact.

We evaluated 13 groups in this analysis, the first three of which were used as stratification factors during randomisation:

• diagnosis: schizophrenia spectrum versus other psychoses
• duration of illness: < 2 years versus ≥ 2 years
• gender: male versus female
• age: ≤ 40 years versus > 40 years
• ethnicity: white, black, Asian, mixed race and ‘other’
• immigration history: UK versus other
• marital status: married/co-habiting versus single/separated/divorced
• living status: living alone (including homelessness) versus living with others
• accommodation status: independent versus supported/homeless
• years of education: ≤ 12 years versus > 12 years (as 12 years of education is compulsory in England)
• tertiary education: ‘yes’ versus ‘no’
• BPRS: ≤ 33 versus > 33
• GAF: ≤ 49 versus > 49.

We performed the subgroup analysis by fitting the same models as for primary and secondary outcome measures plus an additional interaction effect for interactions between study arm and the relevant subgroup variable. The p-value of interest is that for the interaction test.

Family carers questionnaire

According to the original protocol, we were to ask the patients to identify family carers and ask their consent to contact the carer. These carers were then to be sent a carer questionnaire by post, enclosing a letter with full information about the study. We initially undertook this part of the trial but it had a poor response rate. In July 2010, of the 135 patients who had reached the 6-month follow-up point, 38 had agreed to a carer being contacted and only two of those carers had returned a questionnaire. It was anticipated that many of the patients who were eligible for the trial had limited contact with their family, so this result was not surprising. With a response rate of 3%, we decided not to pursue this part of the study, as the results would not be representative or robust. Some of the items covered in the questionnaire, such as carer strain and perceptions of patient well-being, were included in the topic guide for the qualitative interviews with family carers.

Clinical staff questionnaire

According to the original protocol, care co-ordinators were to be asked to complete two validated instruments covering demographics, the therapeutic relationship with the patient, and their assessment of the patients’ health-related quality of life, at both baseline and 12 months. The questionnaires were handed to the relevant staff, or were posted to them, or both. A poor response rate at baseline, however, was compounded by multiple changes of staff, which made identifying the appropriate staff group at follow-up almost impossible. The information sought was designed to be analysed in conjunction with patients’ views of their own health, quality of life and therapeutic relationships, and was thus time sensitive and dependent on a response from a particular individual. Given the difficulties in obtaining this, we therefore decided not to pursue this part of the study.

Statistical analysis

We conducted an additional sensitivity analysis for the primary outcome. This analysis excluded all protocol violators. It was not a prespecified analysis but was conducted to address concerns raised about the number of protocol violators, particularly those who had been discharged to the wrong arm of the trial.

The analyses undertaken were all in line with the spirit of the analyses detailed in the original proposal, but we made some changes to the original proposal when writing the statistical analysis plan: these were changes to the types of statistical models or tests performed to allow for more sophisticated adjusted regression models to be used as the primary comparisons, with the simpler unadjusted tests used for secondary sensitivity analyses. Adjusted regression models have more statistical power (i.e. are more precise) than unadjusted tests such as t-tests, and thus make better use of the data collected.

The original protocol makes reference to minimisation factors. No minimisation procedure was carried out during treatment assignment. Instead, a stratified block design was used. Tertiary outcomes were specified. In particular, we decided to treat satisfaction with service as a tertiary outcome instead of a secondary one. We did not analyse the effect of the discipline of the clinical supervisor in readmissions, as originally planned, as information on this outcome was not available.

We added type of medication as a tertiary outcome, as we considered this to be clinically relevant, given a reported association between CTOs and the use of depot injections.⁴ We included this before any of the data analyses were conducted. Patient-rated adherence to medication could not be adequately measured at baseline (patients were recruited while detained in hospital for treatment) and data collected on this during follow-up were of insufficient quality to be included as an outcome. We therefore used the DAI-10, which correlates with clinician-rated adherence,¹⁰⁴ as the reported measure for adherence. Missing data made it impossible to analyse different types of leverage separately or divide leverage experience into types of pressure. We originally planned to analyse tertiary outcomes by type of service (e.g. ACT, early intervention, crisis teams), but ongoing service reconfigurations rendered this analysis redundant.

Primary and secondary outcome data

We based all of our analyses (apart from the tertiary analyses) on all 336 patients, as we conducted them on an intention-to-treat basis. We had data on the primary and secondary measures for all 333 patients at baseline and 12 months. We included the three patients who either withdrew or were ineligible, with their data missing for all variables, apart from data on the inclusion criteria and the randomisation factors.

Interview data

Of the 336 patients in the trial, 14 were not interviewed at baseline (10 because they wanted to take part but did not want to be interviewed and four because of inadequate English). The remaining 322 patients completed baseline interviews. We compared the baseline characteristics of the 14 patients not interviewed to those of the interviewed patients; there were no obvious differences that might have skewed the subsequent analyses.

At 12 months, we interviewed 241 patients [72%: 125 (75%) of the CTO group and 116 (69%) of the non-CTO group], of whom 189 (56%) completed all three interviews [98 (59%) of the CTO group and 91 (54%) of the non-CTO group]. Those not interviewed at 12 months either refused or did not attend [61 (18%) overall: 27 (16%) of the CTO and 34 (20%) of the non-CTO group], were non-contactable (20 (6%) overall: 8 (5%) of the CTO and 12 (7%) of the non-CTO group), had inadequate English language [4 (1%) overall: 2 (1%) of each group], were deceased [5 (1%) overall: 3 (2%) of the CTO and 2 (1%) of the non-CTO group] or were not interviewed because the clinical team advised against it (one patient in the CTO group) or for other reasons (one patient in the non-CTO group).