The study aims to test clinical effectiveness and cost-effectiveness of the implementation strategies to identify and reduce depression and suicide risk among adolescents living in urban slums. The study hypothesises that:
- A community-based anti-stigma campaign will lead to significant improvements in community behaviours toward adolescents with depression, other significant emotional or medically unexplained complaints and increased risk of self-harm/suicide; and
- A mobile device-based decision support system will improve the treatment of adolescents at high risk of depression, other significant emotional or medically unexplained complaints and increased risk of self-harm/suicide and lead to higher remission rates from depression and reduced suicide risk.
A cluster randomised, control trial (cRCT) involving UPHCs in urban slum clusters in the cities of Vijayawada (in the state of Andhra Pradesh) and New Delhi will be conducted. The adolescent population will be screened to identify two cohorts: (1) adolescents at ‘high-risk’ of depression, suicide/self-harm; and (2) a random sample of adolescents not at high-risk. These will comprise the evaluation cohorts and they will be followed up for 12 months. Detailed process and economic evaluations will also be conducted (refer to Figure 1 which outlines the study schema).
(Figure 1 about here)
The study will be conducted in two locations: slums clusters of Vijayawada and New Delhi. New Delhi is in the north of India, is the capital of the country and is the largest metropolitan city (population about 20 million) in India. Vijayawada is the second largest city in the south Indian state of Andhra Pradesh (population about 2 million). At each site, 30 slum clusters will be selected. For this study, a slum cluster is defined as slums within wards or geographical areas identified as slums/resettlement colonies. A ward is a local authority area, typically used for electoral purposes. In certain cities of India, such as Mumbai and New Delhi, a ward is an administrative unit of the city region; a city area is divided into zones, which in turn contains numerous wards. The slum clusters (constituting 1-5 slums) in Vijayawada, Andhra Pradesh and in New Delhi, under the Municipal Corporation of Delhi will be selected from slums located within a radius of 60 kms from the field office in both cities based on discussions with study collaborators and team members with prior experience working in these areas and will be selected to avoid contiguity. Each slum cluster will have a population of approximately 8000 and will be serviced by at least one UPHC. In Vijayawada, a few of the clusters selected lie in peri-urban areas but are informally considered as part Vijayawada both by local civil society organisations as well as by the Vijayawada municipality and are serviced by UPHCs and by the municipality.
High risk cohort: All consenting adolescents aged 10 to 19 years will be eligible for screening for depression, or suicide/self-harm risk. The screening will help to identify both high risk and non-high-risk cohorts. High risk is defined as presence of at least one of the two conditions:
- High risk of depression based on PHQ-9 score ≥10.
- Positive response (score ≥2) to the suicide risk question on the PHQ-9.
Field investigators will continue screening until they have identified about 990 adolescents per location at high risk of depression, or self-harm/suicide. This will form the ‘high-risk’ cohort. Because there will be some delay (up to 12 weeks) between screening and randomisation the high-risk adolescents will be rescreened to assess if they still meet the inclusion criteria prior to randomisation as literature suggests that around 25% of individuals initially identified at high risk will no longer meet the high-risk criteria at three months . Only those adolescents who continue to remain at high risk during re-screening will form part of the high-risk cohort.
Low risk general adolescent cohort: A second cohort of adolescents per slum cluster not at high risk for depression and/or self-harm/suicide will be identified by selecting a random sample from the remaining screened population matched by gender and age per slum. This will form the ‘low-risk’ cohort. The numbers per slum will be in proportion to the high risk per slum by gender and age.
The following adolescents will not be included in the study.
- Adolescents with severe physical or mental ill-health that would prevent them from giving consent or participating in the study
- Adolescents or their guardians who do not provide informed written consent
- Adolescents who are temporary residents of the slums such as visiting relatives/friends
The implementation team will comprise of researchers, project staff, and field investigators. Project staff will continuously monitor the trial implementation and support field staff, NPHW and UPHC doctors. Research staff will continuously monitor the data and provide feedback to the field staff about any deviations. In addition, 10% of data will be re-assessed by the designated field staff to ensure quality control, and corrective steps will be taken as needed. They will ensure maximum inclusivity of adolescents who manifest either physical or mental illness and field investigators will be provided training in lines with the adapted Guidelines on Disability Inclusion in Research  to ensure that as many adolescents as possible are included in the study.
Randomisation will be conducted at the level of the slum cluster, in a 1:1 ratio to intervention or control using a central allocation sequence. Natural remission variation between screening and rescreening may be used if deemed appropriate for stratification prior to randomization. The clusters will be stratified by location, average remission rate for depression between screened and re-screened population per slum cluster and median population of each slum cluster. The slum cluster selected under each intervention and control arms will be non-contiguous to avoid contamination. The random allocation will be performed using a software-based random number generator by a statistician not involved in study conduct or analysis. To maintain blinding, the unblinded statistician will generate the randomisation list and share it directly with field staff. The blinded study statistician as well as the principal investigator will remain blinded until data lock.
The two core intervention components of the ARTEMIS described above include the community anti-stigma campaign and electronic decision support system. Figure 2 outlines how the different components of the trial work together.
(Figure 2 about here)
The anti-stigma campaign will be implemented immediately post-randomisation and will continue throughout the intervention period, which will include different strategies of creating awareness using multimedia approaches, having a social contact approach (where people with lived experiences share their experiences through videos, in-person talks) and delivering these strategies.
The control arm slum cluster will receive usual care (currently the usual care practice involves very little support for mental disorders and only those with severe disorders are referred to a mental health professional) from their UPHC and general psychoeducation through brochures and pamphlets only. As part of good clinical practice, adolescents identified at high risk for depression or self-harm/suicide will be advised to consult a doctor or mental health professional for further diagnosis and treatment. If any adolescent is identified with severe depression, (a score of >=15 on PHQ9) and/or at high risk of suicide (suicide score >=2) during the screening, they and their parents/guardians will be advised to seek professional help immediately. A list of health facilities (with contact details) where such treatment is available will be provided to them. Further, the implementing team will request NPHW to follow up with the adolescent and their families more intensively and recommend them to visit a doctor/psychiatrist immediately. The implementation team will check to see whether the NPHW has made follow up visits to such adolescents.
For all UPHCs, that fall within the study sites, the implementing team will liaise with government and private pharmacies to ensure availability of appropriate medications. We will also explore collaborative strategies with the Ministry of Health and Family Welfare of both states (Andhra Pradesh and Delhi), to enhance the availability of common psychotropic medications for adolescents as per the Essential Drugs List in the UPHCs. Detailed contact information will be collected from all participants, such that follow-up can be done. The study participants will be constantly followed up for their wellbeing by the NPHWs and will also provide a great deal of motivation and encouragement to them to seek help for their mental illness on a regular basis.
Independent field investigators, blinded to intervention allocation, will be involved in data collection at each phase of the study. Quantitative data on stigma perceptions related to mental health, depression will be collected prior to randomisation (Time 0) and then subsequently at 3, 6, 12 months of the intervention. All data including clinical data of study participants will be captured on tablets, de-identified and saved in a secure server.
- Time - 0 month (pre-randomisation): The names of those adolescents at high-risk will be shared with trained interviewers who will then administer a detailed questionnaire that will enquire about socio-demographic characteristics, treatment history, history of any mental illness, family history of any mental illness, social support from friends and families, stressful events experienced in the last one year, resilience, history of any co-morbid major physical illnesses.. Questions related to costs incurred on treatment will be directed to the parents/guardians of the adolescents.
- Time – 3, 6, 12 months (intervention phase): Data will be collected from the study cohort from the intervention arm and the study cohort from the control arm (both low risk and high risk) across all the slum clusters by trained interviewers. Questionnaires administered will be the Patient Health Questionnaire (PHQ9), Knowledge, Attitude, and Behaviour (KAB) scale  and Barriers to Access to Care Evaluation- Treatment Stigma (BACE-TS) , Connor-Davidson Resilience Scale (CD-RISC - 10)  at these time points.
Tools used for data collection:
- PHQ9 - The PHQ9 consists of 9-items. The final score is calculated by assigning scores of 0, 1, 2, and 3, to the response categories of “not at all”, “several days”, “more than half the days”, and “nearly every day”, respectively. PHQ-9 total score for the nine items ranges from 0 to 27, and scores 5-9 indicate mild, 10-14 indicate moderate, 15-19 indicate moderately severe, and 20-27 indicate severe depression. This instrument will provide reliable estimate of depression in the communities and also proportion amongst the screen positive individuals who have accessed mental health services. A cut off score of ≥5 is indicative of mild depression needing further follow-up and assessment by the NPHWs, whereas a score of ≥10 needs clinical assessment by a PHC doctor.
- CD-RISC – 10 scale – This scale consists of 10 items measuring resilience. Responses are rated on a 5-point Likert scale, ranging from 0 (not true at all) to 4 (true nearly all the time). Each score has a minimum score of 0 and a maximum score of 4. The total scores are calculated by adding the scores of all 10 items. Higher scores suggest greater resilience and lower scores suggest less resilience, or more difficulty in bouncing back from adversity.
- KAB scale – KAB is used to measure the knowledge, attitude and behaviour related to mental health, and is a 16-item questionnaire. 12 of the items ascertain mental health knowledge, attitude and behaviours to understanding stigma. It has two subgroups – the behaviour subgroup and the knowledge and attitude subgroup. The behaviour subgroup is based on the Reported and Intended Behaviour Scale which uses a 5-point Likert scale (ranging from ‘agree strongly’ which scored 1, to ‘disagree strongly’ which scored 5). Higher scores suggest increased stigma, except for those questions which had a negative connotation. The knowledge and attitude subgroups were identified based on discussion with experts hence does not have the properties of a scale
- BACE – TS scale – This is a 12-item questionnaire with a four-point Likert scale, with scores ranging from 0 to 4, which asks questions on the stigma associated with seeking help for mental illnesses. Higher scores suggest higher stigma.
The SPIRIT 2013 statement, 33-item checklist and figure  is being used to schematically represent the study participants’ timeline of enrolment, eligibility screening, allocation, intervention and assessments at four timepoints (please see Fig. 3) and to guide the overall standards of a cRCT.
(Fig 3 about here)
1. Anti-stigma component (combined high and low risk cohorts)
- Change in mean behaviour scores of adolescents at 12 months using the KAB scale
- Change in mean knowledge and attitude scores and change in stigma perceptions of adolescents as assessed by Barriers to Access to Care Evaluation- Treatment Stigma (BACE – TS) Subscale. Change in attitude of parents demonstrated through qualitative interviews of parents selected through purposive sampling will also be captured. The qualitative tool development, data collection and analysis will be done by The George Institute’s qualitative researcher.
- Change from baseline in mean stigma scores of adolescents
2. mHealth component (high risk cohort)
- Proportion achieving remission (defined as: PHQ-9 <5, and/or suicide risk score <2) at 12 months.
- Mean difference in PHQ-9 scores of adolescents at 12 months.
- Proportion who have visited a doctor at least once between randomization and 12 months
For high-risk adolescents, assuming a 50 percent remission rate in the control arm at 12 months based on published data , and 30 slum clusters per intervention location, at least 27 subjects per slum cluster are needed to detect a 15 percent absolute improvement in the intervention arm (65 percent intervention vs 50 percent control) at 12 months with 90 percent power. This assumes a 2-sided significance level of 0.05 and an intra-class correlation coefficient (ICC) of 0.1 based on pilot data . To account for up to 20 percent of participants lost to follow-up, we will aim to enrol at least 33 participants per slum cluster for a total sample size of 1980 adolescents (60 clusters × 33 adolescents). Assuming a prevalence of 3% of adolescents with depression or increased risk of self-harm/suicide (47), we will therefore aim to screen about 1,100 adolescents in each slum cluster. Further assuming that 5% of eligible adolescents will refuse to participate, we aim to screen 1,160 adolescents in each slum cluster i.e. a total of 69,600 adolescents (60 slum clusters × 1,160 adolescents).
Primary analyses will be conducted at the participant level using either random-effect generalized linear mixed models or generalised estimating equations adjusted for clustering. For the primary outcome, remission at 12 months differences will be assessed using a log-binomial (log link function) or logistic regression (log link function) including a random cluster effect. If using generalised estimating equations, the effect of clustering will be accounted for using a repeated cluster effect with a compound-symmetry variance-covariance structure. The intervention effect will be estimated as odds ratio and corresponding 95 percent confidence interval.
Sub-group analyses will be conducted according to UPHC-level characteristics (size, location and health service characteristics) and patient-level characteristics (demographic factors and clinical factors e.g. depression severity at baseline). The primary outcome will be assessed at 12 months (and reported as secondary outcomes at 3, and 6 months).
Continuous outcomes such as knowledge and behaviour scores of adolescents will be analysed in a similar manner but using generalized linear mixed model with identity link function in place of logit or log link function and with the baseline value of the score included as a covariate in the model. A pre-specified analysis plan including sensitivity analyses, potential covariate adjustments, analyses for secondary endpoints and detailed assumptions (e.g., missing data handling) will be developed prior to unblinding. To ensure that the analysis is not influenced by potential results, all analyses will initially be programmed using randomly scrambled groups. The real group allocation will be used once the statistical analysis plan is finalised, the database locked, and all analyses programmed.
A trial-based and a modelled evaluation of long-term costs and outcomes will be conducted from a health system perspective. Intervention costs will include salaries, training, and equipment. A within trial comparison of costs (e.g., visits, medications) will enable estimation of potential cost-offsets associated with the intervention. Average differences in utility observed between treatment arms will be determined using the Short Form 6-Dimension (SF6D), which will be calculated using the Brazier algorithm  from the Short Form 36 (SF36) questionnaire. The SF6D is a utility instrument that is derived from the SF36 which is a standard tool for measuring health-related quality of life in economic evaluations . It includes a specific domain for assessing mental well-being, has routinely been used in adolescent mental health research and is validated for use in India . To capture costs and outcomes beyond the trial, a decision-analytic model will be developed to enable long term morbidity and quality of life to be simulated. Incremental costs per Quality Adjusted Life Years gained will be determined. Sensitivity analyses will determine robustness of the assumptions of the model.
Processes will be monitored continuously throughout the intervention phase. A formal evaluation will be carried out at the end of the intervention using Michie’s Behaviour Change Theory  as the over-arching framework to guide our understanding of the intervention implementation and impact. This theory assesses the capability, opportunity, and motivation of adolescents, PHC workers and doctors to engage with the intervention. The researchers will take a case study approach in which a purposive sample of clusters will be selected. In addition to Behaviour Change Theory, the MRC framework for evaluating complex interventions  and the RE-AIM (Reach, Effectiveness, Adoption, Implementation, and Maintenance) framework  will be used to evaluate the implementation. Qualitative data about the experiences of the adolescents, community participants, UPHC doctors, NPHW, and other key stakeholders will be gathered by researchers through focus groups and in-depth interviews until thematic saturation is reached. For participants in the mHealth intervention arm, the mHealth system will capture health services usage analytic data to assess intervention fidelity. A mixed methods analysis will be conducted informed by the Medical Research Council UK (MRC) guidelines . The process evaluation will assist in the interpretation of the ARTEMIS study results, to understand barriers and facilitators in implementation of the intervention components, explain what aspects work or need modification in future trials, and evaluate whether the intervention was implemented appropriately, and as planned.
Protocol version – V7, 20 Dec 2021
Protocol submitted to: Clinical Trial Registry of India (CTRI)
Trial status: Study approved by CTRI (Reference Number CTRI/2022/02/040307, http://ctri.nic.in/Clinicaltrials/pmaindet2.php?trialid=47111&EncHid=&userName=ARTEMIS-%20CTRI/2022/02/040307).
Trial registration date: 18 February 2022
Recruitment start date: tentatively after 2nd May 2022
Recruitment end date: tentatively 1st May 2022 (one year following recruitment start date)
The study is approved by the George Institute for Global Health India.. Approval of the ARTEMIS project from the Health Ministry’s Screening Committee (HMSC), Indian Council for Medical Research (ICMR) has also been received (ID 2020-9770). Any protocol amendment will be shared with the ethics committee approval. If there are protocol deviations, those will be reported to the trial registry too. Any protocol amendment following start of the trial will also be reported to the trial registry.
Preliminary discussions about the study and interventions with each ward members, PHC doctors and NPHWs were conducted during the initial slum cluster mapping and household listing exercises. Informed written consent will be obtained from all participants who are aged 18 and 19 years and from parents/guardians of under aged adolescents (10 to 17 years) along with assent from these underage adolescents, by trained interviewers at the time of screening for high-risk populations and prior to administering the detailed questionnaire at baseline and for focused group discussions and interviews. Any individual identified at imminent risk of suicide at baseline will be advised to seek professional help from a doctor/psychiatrist immediately whose contact details would be provided. All data collection and reporting will be compliant with ICMR guidelines . Independent study monitoring will occur for a subset of participants. Data will be de-identified, stored on George Institute India servers, and held in strict compliance with Good Clinical Practice guidelines using our standard operating procedures for data security, confidentiality, backup, and audit trails. As required, all raw data and any derived datasets will be stored at the George Institute India Hyderabad office for at least 10 years from study completion.
Adverse event reporting
Serious adverse events will be recorded after the start of the intervention phase. A serious adverse event is defined as death due to any cause in the intervention or control arm, hospitalisation due to psychiatric disorders, or a history of self-harm or attempted suicide during the intervention period. These events will be captured using a standardised case report form and reported to an independent Data Safety Monitoring Committee Chairperson on a monthly basis and Ethics Committee Chairperson within 24 hours of the implementing team being aware of the incident. This data will be analysed during the Data Safety Monitoring Committee meetings that will be held on a quarterly basis.
In order to address safety concerns, at least one interim analysis will be conducted with results reviewed by the Data Safety and Monitoring Committee (DSMC). If the interim analysis shows a benefit from the intervention the study will proceed as planned as the sustained effect at end of study will remain an ongoing question. A recommendation to discontinue the study prematurely will be based upon there being clear evidence that the intervention provides harm. Although no formal/ binding stopping rules are suggested, an increase in the proportion of participants with a suicide risk score >2 at 3 and 6 months in the intervention arm of 3 standard deviations (a 2-sided p-value < 0.0027) or more would be regarded as strong evidence of early harm. The DSMC will reveal the unblinded results to the Trial Steering Committee if, taking into account both statistical and clinical issues and exercising their best clinical and statistical judgement, the unblinded results provide sufficient evidence that the trial treatment is on balance harmful.
Ethical approval, site approvals and modification of IEC materials will be completed 15 months after the commencement of the study. Screening of eligible adolescent population and baseline interviews of study participants will be completed by 20 months. Randomisation of PHCs into intervention and control arm will be done in May 2022. Implementing the intervention (anti-stigma campaign and EDSS decision making), follow up and endline data collection will be completed by 34 months. Publications will be done to highlight various aspects of the project throughout the project period and specially to share the final results. Findings will also be disseminated to various stakeholders including policy makers at a policy symposium held before the end of the study. The final trial report will be written by the researchers of ARTEMIS guided by the PI and CoIs. Professional writers will not be hired to write the trial report.