In a cluster randomised controlled trial, community teams caring for patients with psychotic disorders in the community, currently AOTs, will be randomly allocated to the intervention group or control condition. The allocation of teams, and not individual patients, will prevent contamination of practice within teams and facilitate the assessment of overall experiences in teams with the practice. It will also make it possible for teams in the experimental group to offer financial incentives to further patients outside the study, without compromising the study design. Teams might consider this to avoid a sense of unfairness among patients cared for by the same team or simply to have more patients benefiting from the intervention. Outcomes will be analysed on the level of individual patients. The effect of clustering of patients within teams will be controlled for in a mixed effects model. The trial will not be 'blind', as masking of patients and clinicians is impossible. Yet, the primary outcome criterion (percentage of injections taken) and secondary outcomes, with the exceptions of global clinical improvement which is rated by clinicians and subjective quality of life and treatment satisfaction which is rated by patients, can be obtained objectively and are taken from the medical records, and should therefore not be influenced by lack of masking.
AOTs will first be approached and informed about the study through the National Forum for Assertive Outreach and local collaborators at study centres in London, Oxford and Liverpool. We will approach around 100 AOTs that are based within reasonable distance of the study sites so that regular travelling to the teams is realistic. AOTs will receive information about the study on regional and national meetings of AOTs and material circulated through email. This will be followed up by direct telephone calls of the director of the National Forum for Assertive Outreach and other members of the research team including the research assistants. Although we expect that a number of teams will object to either the practice of offering financial incentives or being randomly allocated within a research design or both, informal consultations showed that we can expect more than 36 teams to volunteer for participation in the study. To include an AOT in the pool of eligible teams we will ask for preliminary informed consent by the team manager. AOTs already practising a financial incentive scheme will be excluded. Yet, a survey conducted in 2006 [18
] identified only one AOT in England using financial incentives at the time, and this number is unlikely to have increased substantially since. We will then randomly select 36 teams out of the pool of volunteering teams, allowing for two teams to drop out in the further procedure before the trial begins.
All recruited teams will be visited by a member of the research team to explain the nature of the intervention and the study. Clinicians and managers in all teams will receive a structured presentation addressing the research background, the design of the trial, and ethical as well as practical issues of implementation. Written informed consent to participate will then be obtained from team managers and psychiatrist consultants.
The next step will be to identify patients in each team fulfilling the inclusion criteria. We expect the number of patients in most teams to vary between 5 and 8. We expect to recruit 4 patients per team and will randomly select patients if required. These patients will be informed about the study by a clinician and then approached by a researcher for written informed consent for their data to be used in research and for participating in a trial, in which patients in some but not all teams are offered financial incentives to improve medication adherence. If patients cannot be contacted initially or do not provide written informed consent, further patients fulfilling the inclusion criteria will be recruited from the participating teams until the total sample size is reached. Selecting and recruiting patients before randomisation is essential to avoid a possible bias in the selection and recruitment procedures based on awareness of whether patients will be in the experimental or control group. After this one-off contact between the patient and a research assistant, there will be no requirement for further contacts between research assistants and patients in either group. Following the initial interview, patients are not required to participate in any research interviews or assessments at any point of time. This simple and non-intrusive procedure is meant to minimise the number of non-consenting patients (which always is a problem with research in challenging patients in AOTs) and avoid a selection bias as far as possible. Only if patients volunteer to be contacted at the end of the trial again, a researcher will attempt such contact (possibly via telephone) to ask 11 questions on patient reported outcomes.
After the recruitment of patients, 34 AOTs will be randomly allocated to the intervention or control condition stratifying for the type of catchment area (i.e. inner city, suburban or rural).
Patients in the AOTs that have been allocated to the intervention will be offered a financial incentive for each depot injection of anti-psychotic medication for a 12 month period. Patients will receive £15 for one injection with the total sum not exceeding £60 for a four-week period (the maximum number of injections is 4 per month). The administering clinician will give the money in cash directly after the injection. Patients will sign a receipt. There are several reasons to set a standard sum of £15 for each depot injection:
• A fixed sum per injection simplifies the practice and makes it transparent for all clinicians and patients involved.
• The sum of £15 is in line with the successful open pilot study in East London.
• The sum is below the limit of £20 per week which would interfere with patients' disability benefits. Most patients eligible for the study receive Disability Living Allowance, Income Support with Disability Premium, or Incapacity Benefit. In all of these cases, patients are not entitled to have a separate income of more than £20 (including therapeutic earnings and income through research participation) without having their benefits reduced.
• £15 per injection is intended to be an incentive helping persuade otherwise ambivalent patients. Yet, it is important to limit the total sum to a maximum of £60 per four weeks so that patients do not become financially dependent on the additional income. The money is intended to provide an incentive, but not lead to financial dependence on the scheme.
Otherwise all patients will receive treatment as usual. The type, frequency and dosage medication and all other interventions will not be affected by participation in the study.
Members of the research team will attend meetings of each AOT in the intervention group and discuss again the practice of offering financial incentives and the nature of the study. Following that there will be a brief training programme on the exact procedure. The procedure of the intervention will also be outlined in a written manual. All teams will then be regularly visited by the research assistants and, if required, also by members of the team of applicants. A discussion of the practice at a team meeting will be repeated after 6 months of the intervention period.
The only inclusion criterion for teams is that they care for patients with psychotic disorders who have problems adhering to antipsychotic maintenance medication. These are currently dedicated AOTs with a corresponding policy. The only exclusion criteria are lack of willingness to participate and an already existing practice of offering financial incentives to patients with problematic medication adherence.
For patients in the AOTs there are the following inclusion criteria:
• being cared for in the AOT for at least 4 months,
• between 18 and 65 years of age,
• capacity to give informed consent to participate in the study and actual written informed consent,
• an established diagnosis of schizophrenia, schizo-affective psychosis, or bipolar illness according to ICD-10,
• being prescribed depot injections of anti-psychotic medication,
• poor adherence to anti-psychotic medication, i.e. missed 50% or more of prescribed depot injections over the last 4 months (so that the percentage of taken depots is based on a minimum of 4 prescribed depots), and
• failure of all other methods available to the team to ensure adherence to medication.
Exclusion criteria are:
• learning difficulty
• poor command of English so that clinical communication and discussion of agreements is impaired
The primary outcome is adherence to anti-psychotic maintenance medication during the 12 month trial period. Adherence will be measured, objectively, as the percentage of prescribed depot injections actually taken. As the primary outcome, the percentage will be used as a continuous variable. However, we will also analyse the percentage in a dichotomised way, comparing the ratio of patients with 'good' adherence (i.e. =80% of prescribed depots taken [20
]) in the two conditions.
Further secondary outcomes are:
a) The time 'slippage' of taking depots, defined as the percentage of the prescribed time interval that has expired before the depot is taken;
b) Clinical improvement as assessed on the Clinical Global Impression Scale (CGI) [21
] by the treating consultant psychiatrist at the end of the 12 month period;
b) Number of involuntary and voluntary hospital admissions during the trial period;
c) Costs of care: data on the use and frequency of use of inpatient care, outpatient care (including home visits, home treatment), and other health services during the 12 month treatment period will be obtained from case notes and electronic administrative data bases. Costs for the intervention will be estimated for each participating team from information provided by staff. Established national unit costs will be used to estimate direct health care.
d) The number of attempted and completed suicides, incidences of physical violence, police arrests and days spent at work/training/education will also be recorded over the 12 month trial period.
e) Subjective quality of life and satisfaction with medication which will be assessed at the beginning and end of the intervention period using the 11 item scale established in the DIALOG trial [22
]. The scale contains 11 items asking patients to rate their satisfaction with 8 life domains and 3 treatment aspects, one of which is medication, on a scale ranging from 1 (lowest satisfaction) to 7 (highest satisfaction).
f) Continuation with financial incentives (in intervention group only) and adherence during a 6 month follow up period will be taken from the medical records.
g) Teams in the intervention group will be asked after 6 months, 12 months and 18 months about all aspects of experiences with the scheme including whether patients on the scheme asked for an increase of the incentive, and whether other patients with hitherto good adherence also asked for financial incentives and/or became poorly adherent in order to be eligible for the incentives. This will be done using open questions with a written documentation of the answers.
Simple measures of subjective quality of life and satisfaction with medication are the only patient reported outcome criteria used in the study. They have been included to obtain a subjective outcome that reflects the user perspective. However, this will be an element that patients can participate in or not. If they do not consent to be contacted for completing the scale at the end of the intervention period, they will still participate in the trial, and there are no mandatory patient rated or interview based criteria. The patients to be recruited for the trial have been 'difficult to engage' in care, and many may refuse participating because they do not want to be interviewed by a researcher or complete questionnaires. This would result in difficulties to recruit and - more importantly - a significant selection bias.
Risks and anticipated benefits for trial participants and society including how benefits justify risks
There are potential risks linked to offering financial incentives for patients in the intervention group. These include that patients a) become financially dependent on the incentive, b) demand more money over time, c) will not want to terminate the scheme although they might be prepared to adhere to medication even without the incentive, and d) spend the additional income on illegal drugs. Also, other patients who have been adherent so far might ask to be offered financial incentives as well and/or decrease their adherence to become eligible. Based on 5 years experience with the intervention in the AOT in the East London Borough of Newham, one can expect most of these risks to be limited. No patient with good adherence has ever asked to receive financial incentives as well (and to our knowledge none has ever become poorly adherent in order to be eligible for the scheme). One patient receiving the intervention has once asked for the money to be increased which was declined without any negative consequences. The financial dependence is difficult to judge, but the maximum overall amount of £60 per 4 week period is rather small to induce dependence. We cannot guarantee whether patients spend the additional income on illegal drugs, but all patients have civil rights and the capacity to decide on what they want to spend their money and, on a practical level, the amount of incentives is not sufficient to fund a significant use of illegal drugs.
The anticipated benefits for the patient include a much better quality of life with reduced distress, lower suicide risk, fewer problems with the justice system, lower rate of compulsory treatment and less time spent in psychiatric in-patient units. Some patients may see the benefit of the medication, change their attitude towards it and later take it without financial incentives [23
The potential benefits to society include a reduced risk of patients to harm others and much lower costs in terms of input of health services and other services in the society including the police and the justice system.
For patients in the control group there are no discernible risks or benefits. They will only be asked whether they consent to their data being used for research and would consider in principle an offer of financial incentives to take their medication. Their care will not be altered at all. For patients in either group there are a maximum of eleven satisfaction ratings on one scale, but no potentially distressing interviews or assessments. In the intervention group, patients get offered financial incentives, but can refuse further financial incentives and medication itself at any point of time.
Trial Steering Committee and Data Monitoring Committee
A trial steering committee (TSC) will be established with an independent chair, a user representative, and at least two further independent experts.
Although the amount of data collection is limited in the trial, we will also establish a Data Monitoring and Ethics Committee (DMEC) because of the ethically sensitive nature of the intervention. The DMEC will be independent of the applicants and report to the TSC. It is suggested to have a joint TSC/DMEC meeting at the beginning of the study, and subsequently arrange DMEC meetings before the TSC meetings. The meetings of both groups will be scheduled for times immediately following the expected delivery of major milestones, i.e. in month 6, 12 and 25 of the study
The study has been approved by Ealing and West London Research Ethics Committee (REC reference number: 09/H0710/35). All data will be anonymised and stored securely in line with the Data Protection Act. No published data will contain patient identifiable information.
For analysis, we will use generalized linear models as appropriate to the outcome, with random effects for groups, and sensitivity analyses to explore the impact for missing data. A detailed Statistical Analysis Plan will be agreed by the TSC prior to analysis of un-blinded data.
A cost-effectiveness analysis will be conducted from an NHS perspective, using data on health service use, national unit cost figures and the main outcomes in turn (adherence, time 'slippage' of taking depots, CGI). Incremental cost-effectiveness ratios and cost-effectiveness acceptability curves will be estimated and employed as necessary, generated from the net benefit approach and using bootstrap regression for a range of values of willingness to pay for incremental outcome changes. Sensitivity analyses will examine the impact of altering key assumptions and parameter values. It is usual in any trial to find differences in service access, treatment adherence, baseline characteristics, changes in outcome dimensions over time, cost and cost-effectiveness. In the present study, these variations would be of particular interest, and we therefore plan to analyse patterns within the samples in order to examine whether there are identifiable patterns of inequity with respect to need, socioeconomic group, and key demographic characteristics. The concentration index approach, now quite widely used in health economics for example, offers a robust and informative methodology [24
Proposed sample size
We will recruit 34 AOTs in England (initially 36 to allow for two teams to drop out between recruitment and beginning of trial), and 4 patients within each team. Seventeen teams each will be randomly allocated to the experimental group and the control intervention, i.e. 'treatment as usual'. We aim to have 68 patients in each arm of the trial, allowing for one patient per team to be lost between recruitment and one year follow up. This estimate of a loss of one patient per team may be rather pessimistic, but enables us to have a minimum of 52 patients per arm (assuming that in at least one team per arm there will be no loss to follow up) included in the intention-to-treat analysis. Dropping out of the study and the intention-to-treat analysis will occur only because of a) death, b) long-term imprisonment, c) long-term hospitalisation, d) unknown whereabouts with no chance to obtain outcome data, or e) withdrawal of consent for the data to be used for research. Patients in the intervention group may discontinue with the intervention within the one year study period, because their clinicians think that maintenance medication is not appropriate anymore or patients themselves decide to come off the scheme. Such patients will still be included in the intention to treat analysis, and discontinuing with the scheme will not compromise the availability of outcome data. Refer to Figure . CONSORT Flow Diagram for details.
This figure displays the CONSORT Flow Diagram.
According to the definition of good adherence as taking at least 80% of prescribed medication, the study is powered to detect a difference in adherence from 25% in the TAU arm to 65% in the experimental arm with 90% power for 5% significance. To convert this to a continuous measure requires an estimate of the standard deviation of the percentage of medication taken: Remington et al 2007 [20
] estimate this as 31%. This estimate may appear high, which makes our power calculation rather conservative. Assuming the 31% standard deviation pertains to both arms, the original assumptions are then equivalent to assumed means of 60% of prescribed medicine on TAU, and 92% on treatment. In fact, the mean in Remington et al [20
] was 66%, so the revised sample size calculations on the continuous measure are powered for a more modest increase from 65% to 85% (an absolute difference of 20%). This would require 47 per group in an individually randomised study. This then has to be inflated to allow for clustering. Assuming an ICC of 0.05 and an average of 3 patients per team gives an inflation factor of 1.1 [25
] or 52 per group. We will therefore aim to have one year follow up data from at least 3 patients each from 17 teams per arm (and for 4 patients in one team in each arm). To allow for potential dropout we will actually recruit 4 per team. We do not propose a correction for variable group size. The numbers recruited per team are under our control and the loss to follow-up rate is likely to be low with small differences between teams. Thus, the coefficient of variation of the group sizes is unlikely to exceed 23% [26