COSIMPO is a randomised controlled trial in which a collaborative shared care for psychosis implemented by complementary alternative providers (traditional and faith healers) and conventional primary health care providers (PHCP) is compared with care as usual in which no formal collaboration takes place between the two groups of health providers. COSIMPO is therefore a test of a complex task sharing approach for the care of patients with severe mental disorders.

Most patients with psychotic disorders receive care from traditional and faith healers (complementary alternative providers, CAPs) in the setting of COSIMPO (Nigeria, Ghana). Treatment modalities by the healers include the use of herbs and rituals but may also involve potentially harmful practices such as shackling and scarification. The objective of COSIMPO is to compare the effectiveness of a collaborative shared care program between the CAPs and PHCP in improving the outcome of patients with psychosis under the care of CAPs facilities compared with that of usual care. A cluster randomized control trial design is implemented in Nigeria and Ghana. Patients in the control arm receives usual care enhanced through the training of PHCP in the arm, but no structured or formal collaboration is implemented.

Primary hypothesis: Patients presenting to CAPs facilities engaged in collaborative shared care with PHCP will have better outcome of psychosis compared to patients in enhanced usual care at 6 months following entry into the trial as measured by a significant mean reduction in symptoms as rated with the Positive and Negative Syndrome scale (PANSS). For the purpose of this study, a 7-point difference on the PANSS total outcome scores between the two groups will be regarded as a clinically meaningful difference.

Secondary hypotheses:

1. Persons with psychosis receiving treatment from CAPs in the collaborative care program will be less likely to receive harmful treatment or be subject to human rights abuses (such as chaining, beating, starving, scarification, isolation, forced and prolonged fasting) than those receiving care as usual.
2. Persons in the intervention arm will be less likely to experience victimization and abuse from caregivers. Victimisation is defined as an act or action that exploits or treats a patient unfairly.
3. The patients in the intervention arm will have significantly less disability than those in the control arm
4. The providers (both CAPs and PHCPs) in the collaborative care program will have better attitudes to and knowledge of psychosis than those in the (enhanced) care as usual.
5. Collaborative share care will be more cost effective than usual care. Setting and location of RCT Nigeria: The study site in Nigeria is Ibadan, one of the largest cities in Africa. The inhabitants are mostly Yoruba with a mix of religious practices being prevalent. Ibadan is a cosmopolitan city with five of its eleven local governments (districts) being urban while the remaining six are rural or semi-urban. The population of Ibadan as of the last census in 2006 was 1,338,659, (Census 2006). There are on the average about 10 PHCs in each of the local governments and traditional health practice facilities are ubiquitous. All eleven local government areas in and around Ibadan are included in the study

Ghana: The Ashanti Region is located in south Ghana and third largest of 10 administrative regions. With a population of 3,612,950, the region accounts for 19.1 percent of Ghana's total population (2010 census) (Awuah-Nyamekye, Samuel, and Paul Sarfo-Mensah, 2012). Seven districts were selected for the study in Ashanti on the basis of having the size of population required for the study and being accessible to the research center.

Sampling frame The sampling frame will be all the clusters in the local government areas (LGAs) or district areas selected for the study. For the purpose of this study, a cluster will consist of a PHC clinic with its neighboring CAP facilities. That is, a cluster will have one PHC and neighboring CAP facilities, the number of the latter will be determined by distance, ease of contact, and administrative considerations. A CAP facility will be linked to only one PHC but a PHC may have more than one CAP facility linked with it. A PHC as well as at least one of its neighboring CAP facilities must agree to participate in the trial for the cluster to be included in the sampling frame before randomization.

Sample size and potential power of the trial: The primary outcome considered is the symptom severity as assessed with the PANSS score. Based on our previous naturalistic follow-up study of patients with psychosis undergoing treatment, we estimate that a mean difference of 7.5 units on the total PANSS outcome scores between the two arms will represent a clinically significant difference (given that a standard deviation of 20 was obtained for the PANSS in that study). With the target effect size of 0.38, an uninflated sample size of 112 patients per arm will be required to provide a power of 80% and at an alpha of 0.05. Drawing on the recruitment profile to the trial so far, we estimate that, over the course of approximately another 2 months (that is, until end of May 2017), we will be able to recruit an average total of 6 patients per cluster. To take account of the cluster design, we inflate the estimated sample size by 1+((k-1)*ICC)) where k is cluster size for analysis and ICC is intra-cluster correlation coefficient. We have data showing that WHO-DAS, (one of the outcome tools to be used in the study) has an ICC of 0.02 in primary care settings and we make the assumption that the ICC for the primary outcome tool, the PANSS, will not be significantly different. Using the resulting design effect of 1.10, the inflated sample size is 246.4 (224*1.10). With an estimated attrition of 20% at six months (an estimate supported by our experience with the follow-up efforts so far) we will need to recruit a total of 296 participants. With 51 clusters in the study, an average of 6 subjects per cluster will adequately permit for this target sample size to be achieved.

Informed consent for participation in the trial: Information on the nature and purpose of the study will be administered to patients, caregivers and CAPs in the local languages to aid understanding of the research study. All relevant consent forms will be translated through a process of iterative back translation to the local languages. The languages spoken in the catchment areas are Yoruba in Nigeria and Akan-Twi in Ghana. Only participants who are able to speak either of the local languages will be recruited into the study.

The investigators are particularly mindful of the need to take extra care of these issues in a multi-site study in which the subjects are otherwise vulnerable individuals. With the cooperation of the CAPs, measures will be taken to ensure privacy during consenting and screening of patients at the facilities. For the purposes of confidentiality, records of all data will be stored in locked files rooms and the keys kept by the site data manager. Only authourised persons will be allowed to have access to the data. Regular training (during monthly de-briefing) of research personnel about issues of confidentiality will be conducted to help prevent staff from violating confidentiality. Data will be stored securely at both study sites (in Kumasi and Ibadan).

THE COLLABORATIVE SHARED CARE The collaboration between the CAPs and the PHPCs taps into the skills of both types of care providers in the management of patients with psychosis. The patients that will be treated in this collaborative effort will have access to both orthodox medicine on the one hand and complementary alternative medicine (CAM) on the other hand. It is important to note that, in the design of COSIMPO, patients belong to the CAPs primarily and the PHCP acts only in advisory and supportive capacity in patient management. This is because these patients have presented primarily to the CAPs and, for the most part, are seen and managed at the CAPs facilities on either an inpatient or outpatient basis. Every aspect of the col-laboration is therefore designed to enhance the CAP's acceptability through delivery of better care. The role of the PHCP is to help the CAP to provide safe, acceptable care to patients, including promotion of and respect for human rights and avoidance of harmful practices. The PHCP will also assist the CAP in accepting the provision of evidence-based care to patients, when the PHCP determines that it is indi-cated. This collaboration should promote mutual respect, trust, support, referral and advice among the CAPs and PHCPs. Evidence-based medications will be prescribed only by trained health care person-nel. CAPs in the intervention arm commit themselves to implementing enhanced patient care through the collaborative arrangement with PHCPs. This commitment is explicitly obtained from each CAP prior to their inclusion in the trial. In this regard, they agree and commit themselves to utilizing the arrange-ment to avoid practices that are potentially harmful or abusive and to engaging with PHCPs for the overall care of patients referred to them. The frequent visits of the research team will also be used to encourage and reinforce their commitment. As described in Section 7.2.4, CAPs in the intervention arm who refuse to implement the collaborative arrangement or persist in the use of harmful/abusive ap-proaches will be excluded from further participation in the trial.

The entry point to treatment for patients will be from the traditional healers or faith healers (i.e., CAPs). The CAPs take responsibility for the patient and commence treatment. The contract to care for the pa-tient therefore lies with the CAP. The discretion about when to seek clinical support from or refer to the PHCP lies with the CAP. The PHCP in turn provides the necessary support and assistance needed. In providing clinical support to the CAP, the PHCP may engage with the CAP, the patient, and caregivers of the patient. Beyond specific referrals and consultations, the PHCP will maintain regular weekly scheduled personal contacts with the CAPs as part of the collaboration, using such contacts to reinforce information on best clinical practice, provide information on patient rehabilitation, and attend to other clinical issues that the CAPs may bring to the attention of the PHCP.

Main components of the collaborative shared care

The main components of the collaborative shared care can be grouped into two:

1. Clinical support to respond to the medical need of psychotic patients: This involves the provision of medical care to patients, especially those in conditions of acute psychotic disturbance, by PHCP following the request from CAP.
2. Clinical support to improve service on a continuous basis: This involves engagement and inter-actions between PHCP and CAP during which the former provides the latter with specific in-formation to improve service, organize rehabilitation for their patients, and generally provide better care. This interaction will involve the provision of specific information by the PHCP to CAP on scheduled and regular visits and with full documentation.

These two activities, as now further described, will form the basis of the evaluation of the collaborative care both in regard to the fidelity of its implementation as well as its impact or effectiveness.

Clinical support to respond to the medical need of psychotic patients Intervention will be delivered collaboratively by providers in the intervention arm (CAPs and PHCPs) The intervention is presumed to start from the time TP consents to participate in the study until discharged from treatment (by either the CAP or PHCP). (Patients who enter the study are monitored and evaluated for outcome until the end of the primary outcome at six months irrespective of whether they have been discharged from treatment earlier). Intervention is initiated at the CAP facility but may subsequently involve the PHCP facility or other medical facility, depending on whether the patient is referred to any such other facility or not and based on the agreement between the CAP and the PHCP. It will consist of the treatment modalities as determined by the CAP and by the PHCP when called by the CAP or during routine engagement visits to the CAP. That is, all eligible and consenting patients who are recruited to the intervention arm of COSIMPO will have their treatments initiated by the CAP to whom they have presented. Some will continue with such treatments if the CAP does not seek the input of the PHCP in the care of the patients. Some other patients for whom the CAP seeks the input of the PHCP will receive treatments as prescribed by the PHCP in addition to whatever treatments the CAP is also offering to them.

When a PHCP is involved in the care of a particular patient, the PHCP assesses the patient at the CAP facility, provides feedback to the CAP, and initiates treatment according to the Mental Health Gap Action Programme Intervention Guide (mhGAP-IG). The PHCP continues to monitor the patient by regular visit to be determined by clinical need of patient (for example, administration of specific medications may require dai-ly or more visits to monitor their effects). If the patient is on herbal medication, PHCP takes this into con-sideration in administering medication. PHCP may refer patient to see a physician or specialist, if judged necessary, and in consultation with the CAP.

Specified reasons for consultations of PHCPs by CAPs

Based on information derived from our formative studies, COSIMPO will specify the following specific reasons for the healer to consult the PHCP:

• Very disturbed, agitated or aggressive patients
• Patients at risk of absconding
• Patients at risk of self-harm or harm to others
• Patients who have side effects of medication
• Patients who are not getting better despite efforts by the healer
• Patients who have other comorbid physical conditions, such as fever, injuries, breathing dif-ficulties (breathlessness), severe throbbing headaches, urinating excessively either during the day or at night, drinking water excessively, previously diagnosed hypertensive and or diabetics or any other condition that requires medical attention.

Care by the PHCP Once a PHCP is consulted by a CAP for assistance in the care of a patient, the PHCP will follow the guidelines for the management of patients with psychosis as laid down in the mhGAP-Intervention Guide in figures 11 and 12 below. The provider will conduct assessment and offer treatment according to the procedures in the Guide and as adapted for COSIMPO. Referral to other levels of care, when needed and dictated by the Guide, will be decided by the PHCP but implemented with the approval of the CAP. The PHCP will keep a record of all interactions and treatments in patients' records note books provided by COSIMPO research team to be kept at the PHC facility where the PHCP is based. The records will be in form of patient clinical narrative entries as typically collected by the PHCPs, but supplemented with specific forms and checklists that capture the activities included in the collaborative care program, designed for COSIMPO, and in the form that can be abstracted into the study data base. The checklist will include de-tails including date, time, and duration of visit; psychosocial advice or intervention provided; counseling on harm reduction given; perceived receptiveness and cooperativeness of the CAPs. Space will be provided for writing detailed notes on the experience of the visit including suggestions to improve the collaboration.

Duration of treatment period and follow-up This research project is not imposing a specific treatment duration period. PHC workers and CAPs are free to select the treatment and duration in line with usual standard practice, and may or may not be influenced by the guidelines in the collaborative shared care package. Thus, all clients attending primary care clinics and CAPs will receive usual treatments, in line with routine standard practice, and may or may not be influenced by the guidelines in the collaborative shared care, or whatever duration is prescribed by the PHCP worker or the CAP. However, primary outcome will be assessed at 6 months following entry into the study, irrespective of whether the patient is still in treatment or not.

Instruments

The following instruments will be used for baseline and outcome assessments:

The Structured Clinical Interview for DSM-IV Axis disorder Positive and Negative Syndrome Scale (PANSS) World Health Organization's Disability Assessment Schedule 2.0 (WHODAS 2.0 direct patient version and PROXY VERSION) Internalized Stigma of Mental Illness Life Chart Schedule (LCS) Service Utilization Questionnaire

Checklist of harmful practices by CAPs:

Knowledge about Schizophrenia Interview (KASI)

Checklist of victimization and abuse by caregivers:

Family Burden Interview Schedule

Blinding: TPs will not be blinded to the patients' treatment arms (as it is impossible to conceal the fact of collaboration or lack of it from the patients). There shall be two groups of trained RAs: those conducting the baseline assessment and those conducting the outcome evaluations. While the former cannot be blind to patient trial arm (given that RA's will conduct the assessment at the recruitment facility and will get to know about the fact of collaboration or lack of it), the latter will be kept blind to trial allocations. In view of the possibility that some patients may still be on admission at the facility (in which recruitment has occurred) at the 3-month assessment point, it may be difficult to conduct this particular assessment blind by the RAs. However, the main outcome measures (at 6 month) will be administered by outcome assessors (RAs) independent of the intervention and blind to the allocation of treatment. TPs will be instructed not to disclose details of care being received to the outcome assessors. The primary outcome measure (PANSS) will always be completed first in order to minimize the risk of bias in the event of unmasking during the assessments and, if it occurs, the point of unmasking will be recorded. Sensitivity analyses will be carried out to assess the effect of unmasking on the primary outcomes. The intervention and outcome assessment teams will not have any interactions during the trial and will have separate physical location and administrative management. Table 13 shows a summary of blinding and treatment allocation of the COSIMPO team member.

Risks and potential discomforts in COSIMPO Human Subjects Protection Procedures for human subjects' protection in this research will follow guidelines and policies of the Institutional Reviews Boards (IRBs)-- of the sponsoring organization, the University College Hospital/University of Ibadan, ethical review board and the College of Health Science, Nkrumah, Ghana and the NIMH DSMB. The site investigators have completed Human Participant Protections Education for Research Teams sponsored by National Institutes of Health, U.S. Department of Health and Human Services.

Definition of Adverse Events (AE) Even though the possibility of adverse events, which are directly related to the intervention is low, a critical incident register will be maintained during the study to record specific serious adverse events (SAEs) and Other Adverse Events (OAEs).

SAEs include:

Death, due to any cause Any suicide attempt Any violence towards others Harmful or abusive treatment practice Victimization (violence against the TP) Hospital admission due to psychiatric problem Hospital admission due to serious medical emergency Incarceration Events that may require extraordinary response, such as high risk of suicide attempt, or rapid and severe clinical deterioration Stigmatization (due to the intervention) Sexual abuse, including rape An event or finding with the potential to become an international incident Other Adverse Events (OAEs)

Other adverse events (AE), which are any other medical occurrence trial patient which does not necessarily have a causal relationship with this treatment, will be identified. This may include any of the following:

1. Clinically significant problems not requiring hospitalization or extraordinary response
2. Onset or worsening of medical problems not requiring hospitalization
3. Other events that do not meet criteria for serious adverse events, for example, a report of suicidal ideation without risk of a suicide attempt Other foreseeable risks/discomforts and mitigation strategies This research study has little risks and discomforts. However foreseeable risks associated with the COSIMPO trial will be described to every prospective subject as part of standard information before consent is provided. Also steps to mitigate such risks for every subject category by the research team would be described to the different category of subjects as applicable.

Response to SAEs When a RA detects SAEs, the required information below will be collected systematically on tablets (electronic data collection) during the outcome assessments. In addition such will be immediately reported to the research supervisor who will in turn report to the trial management, who will them immediately inform the PI all within 24 hours. Also the information on the SAE will be uploaded daily to the database and the DM will complete the first part of the SAE report and bring such information to the attention of the trial manager.

Information required for reporting adverse events

Data collection: Interviews will be conducted at baseline, 3,and 6 months. The questionnaires will be administered by trained research assistants (RA). Data will be collected directly using tablet computers and uploaded to a remote server by the data manager every two days.

Three types of data will be collected:

1. Baseline data
2. Follow-up outcome data
3. Process data For all data, range and consistency checks will be performed at weekly intervals separately for each data source and, where relevant, consistency checks will be carried out. Any queries identified will be resolved promptly by the data management team, and the database will be updated, maintaining the audit trial. All data will be kept in separate databases and only merged into a master database after data collection is completed and each individual database is 'locked.' Access to data will be password protected at multiple levels and no member of the trial team apart from the data manager will have access to these passwords. The database will be backed up daily on a password protected computer after upload.

Data quality and security of data The site data manager will protect the confidentiality of personal data principally through procedures to separate study data and participant identifiable data. These identifier data will be stored electronically in a password-protected folder and hard copies of the consent form will be stored in a locked cupboard with access only to the data manager and the site PI. All subsequent data will be collected and collated using only the trial ID number. Regarding the qualitative data set that will be collected during process evaluation, participant audio files will be stored in a password-protected folder and transcribed data will be anonymized with only participant ID information.

Data Coordinating Centre (DCC) An anonymized and password protected dataset will be transferred from Ghana to UI to the off site server via the internet. Only trial managers in both sites would have access to the data during the period of data collection. All DCC procedures, including reporting to the DSMB, will be overseen by DCC coordinator and Hub Administrator. Study statistician, supported by the statistical consultant, will conduct statistical analyses of data and will provide any other statistical advice as requested.

Data and safety monitoring

The NIMH Global DSMB will ensure the safety of TPs. The board will review and approve the study protocol, informed consent and all relevant documents. The main responsibilities of the DSMB are, but not limited to, the following:

Review of protocols, consent procedures, consent forms, and safety plans prior to the initiation of the study; Monitoring of the progress of the study, including recruitment and retention of participants, adverse events, serious adverse events (SAEs), reasons for participant withdrawal, adherence to the timeline of the study, quality of data, and protocol violations; Make definitive and authoritative decisions about the continuation, modification, or termination of the study, based on the balance of adverse events and beneficial outcomes. The data and safety monitoring plan will involve the continuous evaluation of safety, data quality and data timeliness. The PI will be expected to continuously review data and patient safety which will be reported at TMC and TSC meetings and all discussions will be documented. The PI will also submit twice yearly progress report to the NIMH DSMB.

-All paper data files (quantitative and qualitative) will be stored in lock and key cabinets and the computerized data will be password protected. Any back-ups made of the computerized data will be put on separate hard drives which will also be password protected. Only named research staff (AR, SS and the data manager) will have access to the data onsite.

Contamination Contamination will be minimized through the geographical spread of the randomized clusters which will make the possibility of patients moving from a control to an intervention CAP facility unlikely and also by not publicizing the availability of the collaboration in the intervention arm to other CAPs.

Statistical analyses Descriptive analyses Initial analyses will compare baseline characteristics of individuals who consented and did not consent and participants who could and could not complete baseline assessments, and a comparison of the distribution of potential confounding factors. Findings will be reported as per CONSORT guidelines for cluster randomized controlled trials (Campbell, Marion K., Diana R. Elbourne, and Douglas G. Altman, 2004) including a trial flow chart. This will include the flow of clusters and participants through each stage of the trial, including the number eligible, randomly assigned, receiving the intended treatment, completing the study protocol and analyzed for the primary outcome.

The outcome measures will be summarized at baseline, 3 month, and 6 month follow-ups by intervention arm and overall. These will be summarized by means (SD), medians (IQR) or numbers and proportions as appropriate (and including age, gender, baseline outcome score), adjusting for cluster. For continuous outcomes, histograms within each arm will be plotted in order to assess how closely the scales follow a normal distribution to determine how to describe the outcomes and how to properly do the inferential analysis. The clusters will be described in terms of gender, education, country site, and profile of the CAPs (e.g., small vs. large).

Outcome analyses The primary analyses will be intention-to-treat at the 6-month follow-up visit adjusted for baseline measures. That is, after randomization, patients will be analyzed according to study arm allocation irrespective of the treatment being actually received. These analyses will also disregard adherence to the intervention or withdrawal from the trial. Random-effects logistic regression will be used to analyze binary outcomes, adjusting for country site and cluster size as random-effects variables. Analyses of continuous outcomes will use random-effects linear regression, additionally adjusting for the baseline value of that outcome. Adjustments will also be made for any of a-priori defined set of potential confounding variables for which randomization did not achieve balance between the two arms at baseline.

Missing data The number (%) with complete data will be reported. In instances where the methods for dealing with missing data have been recommended in the scales, these will be used. When no such recommendations are available, the investigators will employ multiple imputations and a fully conditional model based on Markov Chain Monte Carlo (MCMC). Imputed data will be used in secondary analysis.

Inclusion Criteria:

1. Patients aged 18 and over and who speak the study language of Yoruba (Nigeria) or Kwi (Ghana). Only patients who speak the languages of the study will be included.
2. All patients presenting to the CAPs in the selected clusters with a confirmed diagnosis of non-organic psychosis as assessed by research interviewers using the Structured Clinical Interview for DSM (SCID).
3. Patients must be on admission at the CAP facility and must be symptomatic at the time of recruitment as indicated by a minimum total PANSS score of 60.
4. Patients must have a caregiver who is willing to give consent to participate in the study.

Exclusion Criteria:

1. Those with serious physical illness in need of urgent medical attention (this could be a serious infection, injury, etc.).
2. Serious cognitive impairment that may interfere with assessment by the research team.
3. Those who will not be in the study area for at least six month following recruitment.
4. Women who are pregnant or will attempt to become pregnant during the study period.