Public health problem
Suicide is a public health challenge worldwide, 1 particularly in India. In 2020, India’s suicide rate increased by 10% and in 2021 by 7.2%, reaching 12.4 per 100 000 in 2022, its highest ever recorded rate. 2–4 India has the highest number of suicide deaths in the world, Reference Vijayakumar, Chandra, Kumar, Pathare, Banerjee and Goswami5 with age-standardised rates among males and females of 11.1 and 14.7, respectively, compared wth global rates of 5.4 and 12.6. 6 India recently released its first National Suicide Prevention Strategy, Reference Vijayakumar, Chandra, Kumar, Pathare, Banerjee and Goswami5 but its realisation is impeded by the lack of data on the real-world implementation of evidence-based interventions to prevent suicides.
Need for integrated intervention
There is consensus that stand-alone interventions are unlikely to achieve reductions in suicide rates. Reference van der Feltz-cornelis, Sarchiapone, Postuvan, Volker, Roskar and Grum7 Nonetheless, individual suicide prevention strategies, shown to be effective globally, can inform the design of integrated interventions to be tested at scale. Reference Gask, Dixon, Morriss, Appleby and Green8,Reference Kawanishi, Aruga, Ishizuka, Yonemoto, Otsuka and Kamijo9 Priority areas, informed by epidemiological data about suicide attempts and deaths, also need to be considered. In India, suicide rates are highest among youth aged 15–29 years, Reference Phillips and Cheng10,Reference Dandona, Kumar, Dhaliwal, Naghavi, Vos and Shukla11 requiring action at the school and community level. India’s youth suicide rates are among the highest globally. Reference Balaji, Mandhare, Nikhare, Shah, Kanhere and Panse12 Pesticide self-poisoning contributes to 14.9% of suicides in India, 4 necessitating strategies to reduce access to pesticides as recommended by the World Health Organization (WHO). 13 In rural settings in low- and middle-income countries (LMICs) like India, primary, community and lay health workers are a critical part of the health workforce and can help identify and manage persons at risk for suicide if trained.
Our solution
A suicide prevention intervention combining multiple strategies for their synergistic impacts has not yet been implemented at scale and tested for its effectiveness in any LMIC context. Addressing this evidence gap, the Suicide Prevention Implementation and Research Initiative (SPIRIT) integrated strategies including promotion of mental health in schools, reduction of access to lethal means and internationally recognised suicide prevention strategies, such as promoting mental health and well-being in schools and enhancing community capacity to identify and manage suicidal behaviour. 13 SPIRIT conducted a cluster randomised controlled trial Reference Pathare, Shields-Zeeman, Vijayakumar, Pandit, Nardodkar and Chatterjee14 of an integrated suicide prevention intervention in a rural district of India, which is identified by the government as a priority region for suicide prevention, given its high rates of suicide (particularly among women and farmers) and poisoning (including by pesticides). 15 The integrated intervention included strategies for adolescents in public schools, central pesticide storage and community health worker (CHW) training for identification and referral of people at risk, as well as management and follow-up support.
A cluster randomised controlled trial was chosen for its ability to yield robust conclusions, unlike traditional randomised controlled trials, which are unsuitable for village-level interventions. This trial aimed to test the integrated package’s effectiveness by comparing intervention and control villages on deaths by suicide and suicide attempts. Given that no previous study has implemented these interventions at such a large scale, we also assessed the extent to which this package could be feasibly implemented and the effectiveness of each of the three sub-interventions on outcomes identified a priori.
Our wider objective was to generate quality, high-impact evidence on processes involved in, and outcomes of, integrated, multilevel suicide prevention interventions to guide policymakers and researchers looking to implement similar interventions at scale in India and other underserved contexts.
Method
Trial design
We conducted a pragmatic cluster randomised controlled trial in Mehsana district in the state of Gujarat in western India, with villages as the primary sampling unit/clusters. Mehsana is a rural district with a 1.52 million population spread over ten subdistricts. 15 The trial followed Consolidated Standards of Reporting Trials (CONSORT) guidelines for cluster randomised controlled trials Reference Schulz, Altman and Moher16 and was delivered according to protocol, Reference Pathare, Shields-Zeeman, Vijayakumar, Pandit, Nardodkar and Chatterjee14 with timeline modifications due to COVID-19. All procedures contributing to this work comply with the ethical standards of national and institutional committees on human experimentation, and with the Helsinki Declaration of 1975 as revised in 2013. Ethical approval has been obtained from the Institutional Review Board of the Indian Law Society (no. ILS/36/2017) for all procedures involving human subjects. Written informed consent was obtained from all participants prior to the study’s commencement. All adverse events protocols were maintained (Supplementary File 1 available at https://doi.org/10.1192/bjp.2025.10431). As seen in Supplementary File 1, the study had 3 phases: the baseline period of 12 months without any interventions in villages (November 2018–October 2019); the intervention phase (November 2019–January 2022), when all 3 interventions were rolled out in the villages in a staggered manner; and a post-intervention follow-up phase of 12 months following rolling out of the final intervention out in the last village (February 2022–January 2023). For the control arm, the same time frames were used. This approach was chosen to account for the staggered roll-out period (different start and end dates for all 54 intervention villages). The study protocol was registered on 7 April 2017 in India’s Clinical Trial Registry (no. CTRI/2017/04/008313).
Participants
Villages were eligible according to the following criteria: the village council agreed; a high school with ≥35 students; a population of ≤6000; and farming as the primary occupation.
For sub-intervention 1, data were collected from adolescents in grade 9 (13–17 years of age) from public high schools in the villages included in the trial, and for sub-intervention 3, from CHWs working in study villages.
Randomisation
Of 273 villages in Mehsana with government high schools, 255 schools consented, of which 161 had a population of ≤6000 and were potential clusters for randomisation. Villages were randomly allocated in a 1:1 ratio to either the intervention or control arm by an independent statistician using a code designed for Stata version 14.2 for Windows (Copyright 1985-2015 StataCorp LLC. Windows based, 4905 Lakeway Drive College Station, Texas 77845 USA, 800-STATA-PC, http://www.stata.com).
Intervention villages
The intervention arm was exposed to all three interventions, detailed in the protocol. Reference Pathare, Shields-Zeeman, Vijayakumar, Pandit, Nardodkar and Chatterjee14
Sub-intervention 1: school-based intervention
A locally adapted version of the evidence-based Youth Aware of Mental Health (YAM), a universal mental health promotion intervention effective in multiple national contexts. Reference Wasserman, Hoven, Wasserman, Wall, Eisenberg and Hadlaczky17 Adaption involved translation and back-translation, contextual tailoring and piloting. Reference Pathare, Shields-Zeeman, Vijayakumar, Pandit, Nardodkar and Chatterjee14 The intervention aims to prevent depression, reduce suicidal ideation and promote mental well-being among adolescents. It is a 5 h programme consisting of 3 h of role-play sessions and 2 h of interactive lectures and discussions about mental health, in addition to educational materials. YAM was delivered within the classrooms by trained community volunteers.
Sub-intervention 2: lockers for pesticides in central storage facilities (CSF)
Previous feasibility studies have shown that storage facilities for pesticides reduced suicide rates in Tamil Nadu, South India. Reference Vijayakumar, Jeyaseelan, Kumar, Mohanraj, Devika and Manikandan18 A storage facility was installed in each village at a central location identified by the village council, with lockers resourced through study funding. Each farming household was offered a locker free of charge to store pesticides. Facility spaces were either provided free by the village council or were rented from local community members. A trained facility manager, appointed following the facility’s opening, monitored the checking in and out of pesticides.
Sub-intervention 3: training CHWs in the identification and management of suicidal risk
In intervention villages, CHWs (accredited social health activists, Aanganwadi [rural child care centre] workers, auxillary nurse midwives, multipurpose health workers and community health officers) employed by the Department of Health and Family Welfare and the Ministry of Women and Child Development in Gujarat were trained in the identification, management (including referral to specialised care) and follow-up of persons at risk of suicide and/or self-harm. Adapting the self-harm/suicide module of the WHO Mental Health Gap Action Programme intervention guide 19 used in 100+ countries, 20 the training covered an overview of suicide; risk and protective factors; suicide risk assessment; psychosocial interventions such as psychoeducation, behavioural activation, stress management and problem-solving; and introduction to, and networking with, mental health services. Using a train-the-trainer model, ten instructors were first selected and trained, who then trained other CHWs in the local language, Gujarati, supervised by the local research team. Training involved two-day workshops comprising interactive lectures, group discussions, role-play and skills practice, followed by a reminder application and resource materials.
Throughout the intervention period, promotional strategies were used including street plays/movies, posters, jingles, fortnightly meetings, printouts of a religious calendar with information on resources where help could be sought, etc.
Control villages
Residents in the control villages received enhanced usual care, which involved the provision of brochures with information on mental health services and other resources for seeking help, such as emergency helpline numbers and contact details of public and non-governmental healthcare services.
Procedures
Data were collected in intervention and control villages from July 2018 to February 2023. Due to limitations in official suicide data (including underreporting by the National Crime Report Bureau [the official government agency in charge of collecting data about suicides]), lack of a coroner-based health inquiry system, social stigma and legal consequences, Reference Arya, Page, Armstrong, Kumar and Dandona21 a surveillance system was developed using community informants and record reviews. Specifically, data on suicides and attempted suicides were collected from community-based informants (e.g. CHWs, local religious and political leaders etc.), and we performed manual examination of the records of all medico-legal cases from hospitals and the police. The development and methodology of the surveillance system are published elsewhere, Reference Vijayakumar, Pathare, Jain, Nardodkar, Pandit and Krishnamoorthy22 and the manual used is now a WHO manual serving as a model for other LMICs. 23
Outcomes
Primary outcomes
For the overall integrated intervention, the primary outcome was a reduction in the number of suicides and suicide attempts (combined) in intervention villages compared with controls, 12 months following implementation of the 3 interventions in intervention villages.
For sub-intervention 1, the primary outcome was a reduction in self-reported depression and suicidal ideation among school-going adolescents in intervention villages compared with controls, at 3 and 12 months after receiving the school-based intervention. Depressive symptoms were assessed using Patient Health Questionnaire 9 (PHQ-9), Reference Ganguly, Samanta, Roy, Chatterjee, Kaplan and Basu24 comprising 9 items rated 0–3 for possible totals of 0–27, with higher scores indicating greater depression. We used the clinical cut-off (score of ≥10 indicating moderate or higher levels of depression) and the cut-off for community studies (score of ≥5 indicating mild or higher levels). Reference Ganguly, Samanta, Roy, Chatterjee, Kaplan and Basu24,Reference Mohta, Malhotra, Gupta, Mani, Patra and Nongkynrih25 Suicidal ideation was assessed using the Suicidal Ideation Attributes Scale (SIDAS). Reference van Spijker, Batterham, Calear, Farrer, Christensen and Reynolds26 We used the first item on SIDAS, where a score of 0 indicates no suicidal ideation and ≥1 indicates some level of suicidal ideation in the past month.
For sub-intervention 2, there was no additional outcome other than the primary outcome.
For sub-intervention 3, the primary outcome was the number of people with suicidal ideation detected and referred by trained CHWs to mental health professionals for support and care provision in intervention villages.
Secondary/process outcomes
For each sub-intervention, we examined reach and uptake. For sub-intervention 1, process outcomes were the proportion of schools agreeing to implement, and which implemented, the intervention, and the proportion of adolescents in participating schools who received the intervention. For sub-intervention 2, process outcomes included the proportion of villages that agreed to establish a CSF and provided space for its construction, and the frequency of use of the lockers. For sub-intervention 3, process outcomes were the proportion of CHWs trained in intervention villages and their knowledge, confidence, skills and attitudes 6 and 12 months following training compared with CHWs in control villages.
Sample size and power calculations (details in protocol Reference Pathare, Shields-Zeeman, Vijayakumar, Pandit, Nardodkar and Chatterjee14 )
To detect an absolute difference of at least 20% in the incidence of suicide and attempted suicide following the intervention, a sample size of 60 901 was needed (significance level 0.05, power 80%). Accounting for design effect, clustering effect and potential attrition (20%), the required sample size was 146 184 individuals (73 092 in control, 73 092 in intervention), spread across 124 villages (62 per arm). Reference Phillips and Cheng10 Sample sizes were also estimated to test sub-intervention 1 (N = 4557 across arms) and sub-intervention 3 (N = 836 across arms).
Statistical analysis
An intention-to-treat analysis was performed for the primary outcome. A random effect Poisson regression model was used to allow for variation between the clusters in primary outcome incidence as a gamma distribution. Because a negative binomial regression model resulted in similar estimates, we report a random effect Poisson regression model with robust standard error correction accounting for overdispersion. This analysis was adjusted for independent effects of allocation to intervention or control arm, period and the interaction between arm and period. Estimates are presented as an incidence rate ratio. Incidence is calculated per 100 000 person-years (defined as the number of years that the person has contributed to the data). For primary and process outcomes pertinent to the sub-interventions (change in depression and change in knowledge, skills and attitude levels), differences between continuous and categorical outcome variables were calculated using an appropriate multilevel linear and logit effect model, adjusting for clustering by villages as a random effect. Estimates are presented as adjusted mean differences for continuous and adjusted odds ratios for categorical outcomes. All hypothesis testing was carried out at the 5% (two-sided) significance level; Bonferroni correction was further used for adjusting the statistical level. Missing data were assumed to be randomly missing, and no imputation was performed. Analyses were performed using Stata version 14.2 for Windows.
Results
Please refer to Fig. 1 for the CONSORT diagram. Of 161 eligible villages, 62 were randomised to each arm. All 62 control villages consented to participate in the trial and collected data. Of the 62 villages allocated to the intervention, 8 refused participations, resulting in 54 intervention villages. There were no significant differences in baseline characteristics between villages that refused versus those that agreed to participate in the trial. Primary outcome data (from the surveillance system) were collected for 116 villages (62 control, 54 intervention). Baseline demographic data, including age, gender distribution, population size and occupation, were not significantly different between intervention and control villages. Details on the number of students (sub-intervention 1), the number of villages with CSFs (sub-intervention 2) and the number of CHWs (sub-intervention 3) are shown in Fig. 1.
Trial profile/Consolidated Standards of Reporting Trials (CONSORT) diagram. YAM, Youth Aware of Mental Health; CSF, central store facility; mhGAP, Mental Health Gap Action Programme; CHWs, community health workers.
Table 1 summarises the results for the overall primary outcome (suicide attempts and deaths combined) and its two components. There was no statistically significant difference in the overall combined outcome between the two arms. During the 12 month follow-up period there were fewer deaths by suicide in the intervention arm (n = 44) than in the control arm (n = 66), and more suicide attempts in the intervention arm (n = 92) than in the control arm (n = 81). The incidence rate ratio for death by suicide was 0.57 (95% CI 0.32–1.02, P = 0.061), indicating a 43% lower risk of death by suicide in the intervention arm compared with the control, with a marginally significant difference between the arms. Regarding suicide attempts, there was no significant difference between the arms.
Absolute number of events (suicides and suicide attempts) and calculated suicide and attempted suicide rates based on the population size of control and intervention villages
Pre-intervention period is November 2018 to October 2019; post-intervention period is February 2022 to January 2023; robust s.e. 0.9.
a. The estimated rate ratio is adjusted for the random effect of clusters/villages and the fixed effect of the period (baseline rate of pre-intervention period). Surveillance data collection was a continuous process from July 2018 to March 2023. A rate ratio of 0.57 (1.00 – 0.57 = 0.43) denotes a relative risk difference of 43% for a reduction in suicides between control and intervention arms.
*Marginally significant.
Among school-going adolescents (n = 2560 in intervention, n = 4364 in control), there were significant differences between the two arms at baseline in the proportions of those with mild or higher levels (≥5) of depression (31% in control versus 24% in intervention), and of those with suicidal ideation (7.9% in control versus 5.2% in intervention). At the 3 month follow-up (Table 2) both arms had improved, with the intervention arm having significantly lower proportions of those with mild or higher levels of depressive symptoms (odds ratio 1.59, 95% CI 1.04–2.41, P = 0.031) and suicidal ideation (odds ratio 3.7, 95% CI 1.6–8.9; P = 0.003). While there were further reductions in these proportions in both arms at the 12 month follow-up, the difference between the arms was no longer significant (Table 2).
Regression estimates for depression and suicidal ideation in adolescents at 3 and 12 month follow-up
Regression estimates are presented as adjusted mean differences for Patient Health Questionnaire 9 (PHQ-9), a continuous variable, and adjusted odds ratios for PHQ-9 and the Suicidal Ideation Attributes Scale (SIDAS) when presented as categorical variables.
Q1, first quartile; Q3, third quartile.
*Statistically significant. Bonferroni correction was applied, and significance level was set at 0.01.
Regarding the primary outcome for sub-intervention 3, CHWs identified 108 individuals at risk for suicide in the intervention villages over the 12 month follow-up period. Of these, 88 (81.48%) had a history of suicide attempts and 20 (18.51%) had suicidal thoughts when meeting with the CHW. CHWs referred 78 (72.2%) individuals of the 108 to mental health professionals, and 63 (58.33%) were followed for at least 6 months. Thirty individuals were lost to follow-up.
Implementation process/secondary outcomes
School-based sub-intervention 1
Regarding reach, a total of 47 (87.03%) of the 54 schools that were approached (each village in the intervention arm had one school) agreed to roll out YAM. The seven schools that declined stated that the academic schedule was too hectic to accommodate a new intervention, primarily because their academic activities had already been disrupted by two consecutive waves of the pandemic. Regarding uptake, in 11% of schools (5 of 47), more than 90% of YAM participants received the full intervention. In 88% of schools, 51–90% of participants completed the full intervention; only one school had less than 50% of students completing all 5 sessions; 72.81% of participants (1864 of the 2560 who consented) attended an average of 4 of 5 YAM sessions.
Central storage for pesticides, sub-intervention 2
Regarding reach, of the 54 villages that consented to participate and that were part of the intervention arm, 52 (96.2%) village councils provided space for community storage facilities for pesticides. Based on formative research, we estimated that approximately 50% of households engaged in agricultural practices that required pesticides. Accordingly, 8370 lockers were made available across these 52 CSFs. Of the lockers provided, 88.83% were used at least once and 82.53% were used frequently (more than twice in 12 months) to store pesticides.
Training of CHWs, sub-intervention 3
There were no significant differences between the control and intervention arms in demographic variables or CHWs’ knowledge, confidence, attitudes and skills at baseline (n = 436 CHWs in intervention, n = 436 CHWs in control). CHWs in the intervention group showed significantly higher levels of knowledge and confidence at the 6 month follow-up (P = 0.001), and higher levels of knowledge, confidence, attitudes and skills at the 12 month follow-up (P < 0.05; Table 3).
Regression estimates of knowledge, attitude and skills among community health workers (CHWs) at the end of 6 and 12 months
Regression estimates are presented as adjusted mean differences and 95% confidence interval.
ATSP, attitude towards suicide prevention; Q1, first quartile; Q3, third quartile.
*Statistically significant.
Discussion
Our trial evaluated the effectiveness of an integrated suicide prevention programme comprising three evidence-based interventions in rural India. India contributes significantly to global suicide deaths, characterised by high rates among youth, women and rural communities. Reference Patel, Ramasundarahettige, Vijayakumar, Thakur, Gajalakshmi and Gururaj27 Implementing community-based interventions in India is challenging due to sociocultural stigma, limited rural health infrastructure, fragmented governance and logistical complexities. Reference Kaur, Kallakuri, Mukherjee, Wahid, Kohrt and Thornicroft28,Reference Rappai, Cherian, Lukose and Vijayakumar29 Understanding these factors is key to evaluating and scaling suicide prevention strategies. Reference Rappai, Cherian, Lukose and Vijayakumar29
The integrated intervention did not have a statistically significant effect on the combined outcome of suicide attempts and deaths. However, when analysed separately, the intervention villages had a 43% lower risk of suicide deaths at 12 months. This reduction is noteworthy in LMICs such as India, where suicide rates are high and rising – up by 11.6% in Gujarat and 15.3% nationwide from 2020 to 2021. 2,3 Suicide attempts increased in both arms, slightly more so in the intervention arm. This may be due to improved surveillance capturing data for the first time and greater help-seeking behaviours due to awareness-raising activities. Enhanced CHW capacity may have led to the identification of more cases. It is also possible that the intervention reduced fatal outcomes, resulting in more non-fatal attempts being reported.
Comparisons with existing evidence
Few large-scale randomised trials of multilevel suicide prevention interventions exist in LMICs. Reference van der Feltz-cornelis, Sarchiapone, Postuvan, Volker, Roskar and Grum7,Reference Hofstra, van Nieuwenhuizen, Bakker, Özgül, Elfeddali and de Jong30 Our findings add to mixed results globally: some trials showed reduced deaths or attempts while others did not. Differences in interventions, populations, designs and contexts make direct comparisons difficult. Notably, none included means restriction as in our study.
Results of the few published trials of multilevel interventions have been mixed, with one having reduced suicide deaths among police officers in Canada; Reference Mishara and Martin31 one having reduced attempts among ninth-graders in the USA; Reference Schilling, Aseltine and James32 one having reduced a similar composite outcome of deaths by suicide and suicide attempts in rural areas, but not in metropolises, in Japan; Reference Ono, Awata, Iida, Ishida, Ishizuka and Iwasa33 and one reporting no impact on such a composite outcome in New Zealand. Reference Collings, Jenkin, Stanley, McKenzie and Hatcher34 These trials differ in various ways – the interventions were integrated (e.g. unlike our study, none of the four trials restricted access to means of suicide); the populations were targeted; design (e.g. unlike our study, the regions were not randomly assigned to intervention and control arms in the Japanese study Reference Ono, Awata, Iida, Ishida, Ishizuka and Iwasa33 ); the health service and cultural and historical contexts in which the trial was based; the details reported in their publications; and their duration. Sri Lankan studies on safe pesticide storage showed limited effectiveness, Reference Pearson, Metcalfe, Jayamanne, Gunnell, Weerasinghe and Pieris35,Reference Konradsen, Pieris, Weerasinghe, van der Hoek, Eddleston and Dawson36 prompting a national ban – a feasible approach due to centralised governance and limited local manufacturing. However, similar bans face significant hurdles in India due to extensive pesticide production, federal governance and bureaucratic constraints. SPIRIT therefore introduced centrally managed pesticide lockers, supervised by trained facility managers, aligned with local governance structures and supported by prior research in Tamil Nadu. Reference Vijayakumar, Jeyaseelan, Kumar, Mohanraj, Devika and Manikandan18 While bans remain a critical long-term goal, community storage offers a feasible short-term strategy in India while advocacy for regulation continues.
More trials across contexts with consistent, detailed reporting are needed to advance definitive conclusions about the effectiveness of multilevel interventions, including dimensions such as components tested, populations, context etc. Multilevel suicide prevention interventions may require longer follow-up to show effects. Our 12 month window may have been too short. In contrast, trials in Hungary and Japan found positive outcomes after 2–3.5 years. Reference Ono, Awata, Iida, Ishida, Ishizuka and Iwasa33
Sub-intervention outcomes
All three sub-interventions were implemented as planned, with high reach and uptake – despite differing levels (schools versus communities) and the disruption resulting from the pandemic. This is notable given our study’s almost unprecedented large scale (students in 47 out of a possible 54 schools; 52 of 54 villages established central storage facilities to house 8370 lockers, with nearly 90% of these in use; 436 CHWs in all 54 villages trained); and because these interventions were at different levels (schools in sub-intervention 1 versus entire communities in sub-interventions 2 and 3).
Few multilevel interventions have targeted youth in LMICs, despite growing concerns about their mental ill-health and suicide rates. Reference Hofstra, van Nieuwenhuizen, Bakker, Özgül, Elfeddali and de Jong30,Reference Doty, Bass, Ryan, Zhang and Wilcox37,Reference van der Feltz-Cornelis, Sarchiapone, Postuvan, Volker, Roskar and Grum38 Our results are promising in terms of the acceptability of a preventative intervention among schools, students and parents who consented to their children receiving the intervention and data collection. The intervention was also effective in reducing depression and suicidal ideation at the 3 month follow-up; however, this effect was not sustained 12 months later. These findings are consistent with studies showing that school-based universal mental health promotion interventions often show effects but are not sustained in the long term. Reference Wells, Barlow and Stewart-Brown39 There was a higher number of students in the control arm compared with the intervention arm, primarily because recruitment and data collection for the control group were completed before COVID-19 restrictions disrupted intervention delivery and participation. Future studies should explore booster sessions or dose–response effects. Training of CHWs led to sustained (1 year) improvements in knowledge, confidence, skills and practices, and they identified, provided support and followed up with over 100 individuals at risk.
Limitations and strengths
COVID-19 delayed implementation of the interventions, which contributed to variation in exposure duration, a common finding for all non-COVID trials conducted during the pandemic. The trial interventions were planned for roll-out in a staggered manner over 6 months from November 2019, but roll-out took place over 26 months, being completed by January 2022. The 3 interventions were therefore not all introduced in intervention villages at the same time or within the same 6 month period, and villages had different periods of exposure to each of the 3 interventions. Another limitation is that we are unable to evaluate the individual effects of each of the three interventions, because we looked at combined effects at the village level. Although there was a clear trend suggesting that the integrated intervention reduced deaths by suicide, the trial was not powered to detect statistical differences in suicide deaths because it focused on suicide attempts and deaths by suicide combined as an aggregate indicator. Fourth, there are no reliable state- or national-level data on suicide attempts in India; our community-based surveillance, later endorsed by WHO and an important contribution of this trial to the field, helped fill this gap but may have missed some cases. Reference Vijayakumar, Pathare, Jain, Nardodkar, Pandit and Krishnamoorthy22,23
Despite challenges, the trial demonstrates the feasibility of delivering complex interventions on a scale in low-resource settings. The randomisation of regions enhanced internal validity. Thoughtful, sustained engagement of leaders, the health and educational sectors and other key stakeholders in participating villages increased the acceptability of the randomisation, the implemented interventions and the surveillance system. Our trial adds to the scarce evidence base on multilevel suicide prevention interventions implemented in LMICs, and its findings can inform future suicide prevention efforts in underserved areas of LMICs.
Supplementary material
The supplementary material is available online at https://doi.org/10.1192/bjp.2025.10431
Data availability
Following the data-sharing policies of the institute, deidentified data will be made available on request to one of the principal investigators, S.P., and to the institutional head, V. Joshi, via the corresponding author.
Acknowledgements
We thank Dr Anant Bhan, Prof. Dr Jacquleen Joseph, Prof. Dr Lakshmanan Jeyaseelanan, Dr Santosh K. Chaturvedi and Dr Thara Rangaswamy, members of the Data Safety Monitoring Board (DSMB), for their invaluable contributions to this trial. We thank Dr Camilla Wasserman and Dr Vladimir Carli for helping with adaptation of the YAM intervention in the Indian context. We also thank Mohammad Shahjahan (BCCP), Dr Holly Cambell-Rosen (US National Institute of Mental Health, NIMH) and the other dedicated members of the Advisory Committee, along with Prof. Dr Michel Wensing, Dr Bethany Hipple Walters, Dr Soumya Swaminathan, Raj Mariwala, the late Keshav Desiraju, N. Ram, Akhileshwar Sahay, Dr Nikhil Gupte and Dr Wouter den Hollander; and Dr Frederiek Schouten, who played an instrumental role in the successful completion of this research trial. Finally, we thank Prof. Srividya Iyer for her detailed and helpful comments and suggestions for improving this manuscript.
Author contributions
S.P., L.V. and L.S.-Z. have full access to all data in the trial, and take responsibility for the integrity of the data and the accuracy of data analysis. Concept and design: L.S.-Z., L.V. and S.P. Acquisition, analysis and interpretation of data: L.S.-Z., L.V., S.P., N.J., D.P. and I.L. Drafting of the manuscript: L.S.-Z., L.V., S.P., N.J., D.P. and I.L. Critical revision of the manuscript for important intellectual content: all authors. Statistical analysis: D.P. and N.J. Funding acquisition: L.S.-Z., L.V. and S.P. Administrative, technical and material support: N.J., I.L. and S.P. Supervision: L.S.-Z., L.V. and S.P.
Funding
The SPIRIT Project is funded by NIMH (grant no. 5U19MH113174- 03 Revised), to S.P., L.V. and L.S.-Z. (principal investigators).
Declaration of interest
None.
Analytic code availability
Access to analytical code will be provided at reasonable request for the purpose of reproducing results or replicating procedures.
Open access
eLetters
No eLetters have been published for this article.