Mental Health Care Support in Rural India

Key Points

Question

Can an antistigma campaign and a mobile technology–based electronic decision support system result in reduced stigma and improved mental health for adults at high risk of common mental disorders at the primary health care level?

Findings

This cluster randomized clinical trial included 44 primary health center clusters with 9928 eligible participants (3365 in the high-risk cohort). There was a significant difference in mean depression scores between intervention vs control groups at 12 months, with lower scores in high-risk cohort.

Meaning

A multifaceted primary health center intervention with high implementation fidelity may be effective in reducing depression risk.

This cluster randomized clinical trial evaluates the use of a digital mental health intervention and community-based antistigma campaign in reducing risk of depression and mental health–related stigma.

Abstract

Importance

More than 150 million people in India need mental health care but few have access to affordable care, especially in rural areas.

Objective

To determine whether a multifaceted intervention involving a digital health care model along with a community-based antistigma campaign leads to reduced depression risk and lower mental health–related stigma among adults residing in rural India.

Design, Setting, and Participants

This parallel, cluster randomized, usual care–controlled trial was conducted from September 2020 to December 2021 with blinded follow-up assessments at 3, 6, and 12 months at 44 rural primary health centers across 3 districts in Haryana and Andhra Pradesh states in India. Adults aged 18 years and older at high risk of depression or self-harm defined by either a Patient Health Questionnaire–9 item (PHQ-9) score of 10 or greater, a Generalized Anxiety Disorder–7 item (GAD-7) score of 10 or greater, or a score of 2 or greater on the self-harm/suicide risk question on the PHQ-9. A second cohort of adults not at high risk were selected randomly from the remaining screened population. Data were cleaned and analyzed from April 2022 to February 2023.

Interventions

The 12-month intervention included a community-based antistigma campaign involving all participants and a digital mental health intervention involving only participants at high risk. Primary health care workers were trained to identify and manage participants at high risk using the Mental Health Gap Action Programme guidelines from the World Health Organization.

Main Outcomes and Measures

The 2 coprimary outcomes assessed at 12 months were mean PHQ-9 scores in the high-risk cohort and mean behavior scores in the combined high-risk and non–high-risk cohorts using the Mental Health Knowledge, Attitude, and Behavior scale.

Results

Altogether, 9928 participants were recruited (3365 at high risk and 6563 not at high risk; 5638 [57%] female and 4290 [43%] male; mean [SD] age, 43 [16] years) with 9057 (91.2%) followed up at 12 months. Mean PHQ-9 scores at 12 months for the high-risk cohort were lower in the intervention vs control groups (2.77 vs 4.48; mean difference, −1.71; 95% CI, −2.53 to −0.89; P < .001). The remission rate in the high-risk cohort (PHQ-9 and GAD-7 scores <5 and no risk of self-harm) was higher in the intervention vs control group (74.7% vs 50.6%; odds ratio [OR], 2.88; 95% CI, 1.53 to 5.42; P = .001). Across both cohorts, there was no difference in 12-month behavior scores in the intervention vs control group (17.39 vs 17.74; mean difference, −0.35; 95% CI, −1.11 to 0.41; P = .36).

Conclusions and Relevance

A multifaceted intervention was effective in reducing depression risk but did not improve intended help-seeking behaviors for mental illness.

Trial Registration

Clinical Trial Registry India: CTRI/2018/08/015355.

Introduction

Mental illness, behavioral disorders, and self-harm account for about 16% of all disability adjusted life-years (DALYs) worldwide.¹ In India, the National Mental Health Survey estimated the lifetime prevalence of any mental disorder among adults is 15%, with nearly 150 million people in need of treatment² and a doubling of disease burden since 1990.³ Despite the burden, the treatment gap for managing mental disorders is 75% to 95% in low- and middle-income countries (LMICs)⁴ with only around 4% of individuals with depression able to access minimally adequate treatment in LMICs, such as India,⁵ with multiple and supply side barriers that prevent care.^6,7

The Comprehensive Mental Health Action Plan 2013-30 from the World Health Organization (WHO)⁸ proposes rights-based, primary care focused approach for reducing treatment gap using evidence-based digital solutions. Given limited trained mental health professionals, alternative solutions that involve delivery of mental health care by upskilling primary care physicians and community health workers holds promise.⁷ The Indian government contracts village-based community health workers known as Accredited Social Health Activists (ASHAs), with one ASHA servicing a population of about 1000 individuals. ASHAs are generally educated to grade 8 to 10, or lower in some remote areas. ASHAs are provided basic training in health services delivery supporting the government maternal and child health programs on a part-time (2 to 3 hours per day) contractual basis. They can provide additional services as per need, which requires scalable strategies for training and provision of support tools. The potential benefits of an electronic decision support system in managing different health conditions, including mental health, have been demonstrated mainly in high-income countries⁹; however, few studies demonstrated improvements in clinical outcomes.¹⁰

The Systematic Medical Appraisal Referral and Treatment (SMART) mental health project¹¹ was conceived as an implementation platform for India’s national strategies to improve mental health services.¹² It was underpinned by the WHO’s Mental Health Gap Action Programme (mhGAP)¹³ with a goal of improving evidence-based management and prevention of stress, depression, and self-harm and suicide (referred to as common mental disorders [CMDs]). It builds on a pilot project¹¹ that tested an electronic decision support system and a community-based antistigma campaign to deliver mental health services in rural primary health care settings.¹⁴

In this article, we report the results of a trial that aimed to evaluate the feasibility, clinical effectiveness, and cost-effectiveness of the SMART mental health intervention. We hypothesized that (1) an electronic decision support system will improve the treatment of adults at high risk of CMDs and lead to a reduction in symptoms and (2) a community-based antistigma campaign will address barriers to accessing mental health care and lead to significant improvements in community behaviors toward people with mental disorders.

Methods

Ethical Considerations

This study was approved by the ethics committees of George Institute for Global Health India and All India Institute of Medical Sciences, New Delhi, India, and the Health Ministry’s Screening Committee, Indian Council of Medical Research. Written informed consent was obtained from all participants. All data management processes were compliant with national privacy law. Data were securely stored and analyzed on servers located at the George Institute India office. Serious adverse events were recorded, and further details are available in eAppendix 4 in Supplement 2.

Design

The intervention was evaluated using a parallel group cluster randomized usual care–controlled trial with the primary health care center (PHC) as the cluster of interest. The study protocol has previously been published¹⁵ and is presented in Supplement 1.

Study Setting

Forty-four PHCs from 3 districts in 2 Indian states—Faridabad and Palwal Districts of Haryana and West Godavari District of Andhra Pradesh—were included. A median (IQR) of 3 villages (2-4) per PHC participated.

Eligibility Criteria

Eligibility was based on the PHC physician’s consent to participate, noncontiguous villages with a total population of at least 6300, and a minimum of 1 ASHA per 1000 population. Consenting adults aged 18 years and older were assessed for risk of CMDs. High-risk criteria included high risk of depression based on a Patient Health Questionnaire–9 item (PHQ-9) score of 10 or greater, high risk of anxiety based on a Generalized Anxiety Disorder–7 item (GAD-7) score of 10 or greater, or a positive response (score ≥2) to the suicide risk question on the PHQ-9. Participants with severe mental or physical illness that would prevent regular follow-up were excluded.

ASHAs were trained to conduct village-based assessments of the adult population for depression, anxiety, or self-harm and suicide risk. Two cohorts were then generated from this screening process. In the high-risk cohort, ASHAs continued screening until they identified at least 150 individuals per PHC at high risk of CMDs. Because of the time required to screen the large eligible population (up to 12 weeks), participants with a positive response at initial screening were rescreened prior to randomization to assess if they still met the inclusion criteria. We expected around 25% of individuals to have natural remission and consequently the final high-risk cohort was estimated to be at least 110 individuals per PHC. In the non–high-risk general adult cohort, 150 adults per PHC not at high risk of CMDs were identified by selecting a random sample from the remaining screened population.

Randomization and Blinding

Randomization was conducted at the level of the PHC using a computer-based 1:1 allocation stratified by geographic region, median remission rate, and median population size. An unblinded statistician not involved in the study, generated the randomization list to share with field staff. Independent outcome data collection was conducted by blinded field staff. Statistical analyses were independently performed and validated by 2 statisticians blinded to the intervention allocation.

Intervention and Implementation

A detailed description of the intervention components are outlined in the published protocol and summarized here (eFigure 1 in Supplement 2).¹⁵ Both the antistigma campaign and digital health interventions were rolled out simultaneously in the trial. In Haryana, the intervention was implemented from September 21, 2020, to September 20, 2021, and in Andhra Pradesh from December 4, 2020, to December 3, 2021. Data were cleaned and analyzed from April 2022 to February 2023.

Antistigma Campaign

The antistigma campaign comprised pamphlets, posters, and brochures; videos of people with mental disorders and their caretakers talking about their illnesses and a local celebrity emphasizing treatment for mental disorders; and a drama about mental health and treatment benefits staged live in some villages and video recordings screened in other villages.¹⁴ Resources were adapted to local context and language (Hindi for Haryana sites and Telugu for Andhra Pradesh sites).¹⁶

Digital Health Intervention

ASHAs were trained to use an Android application (app) on a 7-inch tablet to identify and refer individuals at high risk of CMDs. PHC physicians were trained in identification and clinical management of CMDs using an app, based on recommendations from WHO’s mhGAP guidelines.¹³ The physicians used psychoeducation, brief counseling (talk therapy), and, where indicated, medication treatment with commonly used antidepressants, such as sertraline and fluoxetine. Patients were referred to a mental health specialist if the electronic decision support system algorithm suggested this was recommended. Assessments were done at the PHCs or via health camps in the villages. The latter approach mitigated the challenge of needing to travel long distances from villages to seek care at the PHCs. The ASHA and physician apps contained a recall and reminder system that supported 2-way sharing of essential clinical information between the physician and ASHA. Participants at high risk were prioritized for ASHA follow-up in the village to support the treatment plan initiated by the physician. The ASHAs assessed treatment adherence and whether follow-up visits with physicians occurred. They also assessed reasons for treatment delay, identified current stressors, and provided supportive counseling on ways to cope with these stressors. An interactive voice-recorded system provided personalized reminders for follow-up care to participants, ASHAs, and physicians. Data were encrypted on the tablet device and securely transmitted using 128-bit SSL to a server based at the research institute’s offices in Hyderabad (eAppendix 1 in Supplement 2).

Staff Training and Remuneration

ASHAs in the intervention arm were employed on a part-time basis. They attended a 3-day hybrid in-person and online workshop that increased their knowledge about mental illness and provided training in the use of the app. The online delivery was implemented during COVID-19 mobility restriction periods. In addition, in-field support was provided to ASHAs for the initial 2 to 3 weeks until they were proficient in using the app. Booster workshops were conducted at 6 and 9 months, and ongoing in-field training was provided using a virtual conferencing platform.

Intervention arm PHC physicians received a professional development course on mental illness and guidance in the use of the app. They were trained in the interpretation of the electronic data collected by the ASHA and diagnosis and interpretation of the decision support output based on mhGAP-IG¹⁷ for management decisions. The focus was on depression, suicide or self-harm, and other serious emotional and medically unexplained concerns as outlined in the mhGAP-IG.¹⁷ They were provided regular in-person support at the beginning of the intervention and as needed during the intervention. The PHC physicians received booster training at 6 and 9 months.

ASHAs and PHC physicians were remunerated for participating in training and intervention delivery at levels aligned with government sector payments. In villages where ASHAs were unavailable, residents from those villages were trained to use the intervention and were reimbursed at similar levels as the ASHAs (eAppendix 1 in Supplement 2).

Control Group

Participants in the control arm received enhanced usual care provided by PHC facilities. Those identified as at high risk were informed about their risk and advised by ASHAs to seek care from a PHC physician or psychiatrist. ASHAs were provided a 2-day in-person training to improve their understanding of mental illness and the need to seek care. PHC physicians were informed that patients may present for assessment and care and were encouraged to use their clinical judgement to provide care at the PHCs or refer to specialists. Any participant with elevated risk of self-harm or other symptoms that potentially reflect serious mental illness were referred to the nearest psychiatrist. For PHCs in both control and intervention arms, we liaised with government and private pharmacies to enhance the availability of antidepressants (fluoxetine and sertraline) as per the PHC essential drugs list.

Data Collection

In the first data collection period (the screening visit), only PHQ-9 and GAD-7 scores were collected to determine risk status. In the second data collection period, all people initially identified as at high risk were rescreened and a random selection of individuals not at high risk were selected to construct the high-risk and non–high-risk cohorts and complete baseline data were collected. An electronic data collection app was used for quantitative data collection allowing for real-time queries on missing, nonlogical, and outlier data. A more detailed mixed methods process evaluation outlining other data sources will be reported separately.¹⁸

Study Outcomes

The study outcomes are detailed here, and more information is available in eAppendix 2 in Supplement 2. Further details on the instrument scales used are available in eAppendix 3 in Supplement 2. Primary outcomes included PHQ-9 scores at 12 months in the high-risk cohort and behavior scores at 12 months in the combined high-risk and non–high-risk cohort on the Mental Health Knowledge, Attitude, and Behavior scale.^19,20

Secondary outcomes included remission (defined as all 3 of the following: PHQ-9 score <5, GAD-7 score <5, and suicide risk score <2) in the high-risk cohort at 12 months; GAD-7 scores in the high-risk cohort at 6 and 12 months; PHQ-9 scores in the high-risk cohort at 6 months; proportion of people at high risk of CMDs at the end of the study who visited a physician at least once in the previous 12 months; and stigma, knowledge, and attitude and stigma perceptions scores at 12 months in the combined non–high-risk and high-risk cohort using the Barriers to Access to Care Evaluation–treatment stigma subscale²¹ and the Mental Health Knowledge, Attitude, and Behavior and Stigma scale.

Implementation outcomes included the proportion at each PHC in high- and non–high-risk cohorts who attended live or recorded drama and were exposed to all antistigma video components, the proportion followed up by ASHAs at least once and seen by physician at least once, the proportion followed up by ASHAs at least 6 times and seen by physician, mean visits by an ASHA, the setting in which PHC physician care was accessed (village health camps, teleconsultation, PHC visit), and the proportion who visited a specialist (among those referred).

Statistical Analysis

The prespecified statistical analysis plan is available online at https://osf.io/96w5c/ and in Supplement 1. All analyses were done on an intention-to-treat basis and at the study participant level while adjusting for clustering at the PHC level. Prespecified subgroups included PHC size (less than vs greater than or equal to the median population), PHC location (Andhra Pradesh vs Haryana), sex (male and female), age group (18-44 years, 45-59 years, and ≥60 years), and illness severity at baseline (suicide score ≥2, PHQ-9 or GAD-7 score ≥15). For continuous variables, the primary analysis was conducted using repeated measure linear mixed models. Models included the intervention, visit (3, 6, or 12 months), the interaction between visit and intervention, and the baseline measure of the outcome as fixed effects. A random PHC effect was used to model within-cluster correlations and a repeated visit effect with unstructured covariance to model within-individual correlations. Binary outcomes at 12 months were analyzed using mixed-effects logistic regression with the intervention as a fixed effect and a random PHC effect. The primary analyses used all available data with no imputation. Controlled multiple imputations as well as covariate-adjusted analyses and per-protocol analyses were performed as sensitivity analyses.²² All tests were 2-sided with a nominal level of α set at 5%. All analyses were performed with R version 9.4 or above (R Foundation) and SAS version 7.15 (SAS Institute).

Results

In total, 44 PHC clusters were randomized to intervention or control with a total of 9928 eligible participants (3365 [33.9%] at high risk and 6563 [66.1%] not at high risk) (Figure 1). Of the included participants, 5638 (57%) were female and 4290 (43%) were male. The mean (SD) age was 43 (16) years. A total of 9057 (91.2%) participants were followed up at 12 months. Due to COVID-19 restrictions, prerandomization data collection took considerably longer than planned, particularly in Haryana. Postrandomization follow-up occurred as per protocol (eFigure 2 in Supplement 2). Baseline characteristics were well balanced between randomized groups (Table 1). The high-risk cohort was older, had a higher proportion of females, and a higher level of mental health comorbidities compared with the non–high-risk cohort. PHC facility characteristics by randomization group demonstrate reasonable balance between the 2 groups (eTable 1 in Supplement 2).

Figure 1. Trial Profile.

PHC indicates primary health care center.

^aOne participant was misclassified as non–high risk. From those with a negative response at initial screening (n = 97 587) in 22 PHCs, 150 from each PHC were randomly selected to constitute the non–high risk trial cohort.

^bSeven participants were misclassified as non–high risk. From those with a negative response at initial screening (n = 64 854) in 22 PHCs, 150 from each PHC were randomly selected to constitute the non–high-risk trial cohort.

Table 1. Baseline Demographic and Clinical Characteristics by Treatment Arms.

Characteristic	No. (%)
	High-risk cohort		Non–high-risk cohort
	Intervention (n = 1697)	Control (n = 1668)	Intervention (n = 3279)	Control (n = 3284)
Age, mean (SD), y	47 (15)	46 (15)	41 (16)	41 (16)
Sex
Female	1139 (67)	1156 (69)	1694 (52)	1649 (50)
Male	558 (33)	512 (31)	1585 (48)	1635 (50)
Marital status
Never married	68 (4.0)	78 (4.7)	400 (12)	420 (13)
Currently married	1425 (84)	1374 (82)	2632 (80)	2650 (81)
Separated/divorced/widowed	204 (12)	216 (13)	247 (7.5)	214 (6.5)
Education
No schooling	934 (55)	777 (47)	955 (29)	777 (24)
Primary school	489 (29)	564 (34)	1146 (35)	1100 (33)
Secondary school	181 (11)	197 (12)	587 (18)	697 (21)
Higher secondary or more	93 (5.5)	130 (7.8)	591 (18)	710 (22)
Occupation
Household duties	1015 (60)	1017 (61)	1317 (40)	1358 (41)
Retired
Organized sectora	76 (4.5)	94 (5.6)	321 (9.8)	379 (12)
Informal sectorb	526 (31)	477 (29)	1335 (41)	1247 (38)
Student/other occupations	80 (4.7)	80 (4.8)	306 (9.3)	300 (9.1)
Comorbidities
Cardiovascular disease/diabetes	246 (14)	240 (14)	286 (8.7)	308 (9.4)
Mental disorder	88 (5.2)	134 (8.0)	15 (0.5)	25 (0.4)
Cancer	10 (0.6)	20 (1.2)	4 (0.1)	11 (0.2)
Taking allopathic medicines regularlyc	543 (32)	576 (35)	698 (21)	753 (23)
Taking any medicines for mental disorders regularly	58 (3.4)	63 (3.8)	5 (0.2)	5 (0.2)
Family history of a mental disorder	62 (3.7)	59 (3.5)	68 (2.1)	55 (1.7)
Substance use (tobacco, alcohol, or other substances)	716 (42)	698 (42)	1141 (35)	1123 (34)
Experienced a stressful event in previous 12 mo	1272 (75)	1224 (73)	1835 (56)	1678 (51)
Trial outcomes, mean (SD)
Behavior scored	14.8 (4.1)	15.8 (4.1)	15.9 (3.9)	16.3 (4.0)
Knowledge scored	12.82 (2.33)	13.00 (2.44)	12.82 (2.51)	12.82 (2.31)
Attitude scored	12.37 (2.24)	12.55 (2.26)	12.68 (2.32)	12.53 (2.30)
Stigma scoree	0.53 (0.69)	0.42 (0.56)	0.35 (0.57)	0.28 (0.47)
PHQ-9 scoree	13.3 (4.4)	13.0 (4.6)	1.6 (2.2)f	1.4 (2.1)f
GAD-7 scoree	10.9 (4.1)	11.5 (3.9)	1.5 (2.0)f	1.4 (2.0)f
Suicide risk score ≥2e	386 (23)	322 (19)	0 (0)f	0 (0)f

Abbreviations: GAD-7, Generalized Anxiety Disorder–7 item; PHQ-9, Patient Health Questionnaire–9 item.

^a Organized sector includes government employment and private employment.

^b Informal sector includes agricultural labor, manual labor, farming, skilled work, and business/trading.

^c Regular medicine use was based on self-report and verified by looking at the medicine prescription.

^d A higher score indicates better mental health knowledge, attitude, and behaviors (eAppendix 3 in Supplement 2).

^e A lower score is associated with lower stigma or lower severity of symptoms (eAppendix 3 in Supplement 2). The suicide risk score is derived solely from item 9 of the PHQ-9.

^f For the non–high-risk cohort, scores are derived from the initial screening visit.

Implementation Outcomes

Fidelity to the intervention components was generally high. For the digital care model, there was generally a high rate of adoption of the intervention for those identified at high risk (Table 2). The absolute number of people referred to a specialist was low (n = 224) and of those referred, 53 (23.7%) visited a specialist. For the antistigma campaign, the median (IQR) percentage of people in the high- and non–high-risk cohorts who attended the live or recorded drama and were exposed to all the antistigma video components was 84.0% (65.7-95.9) and 85.4% (64.2-94.9), respectively (further details available in eTable 2 in Supplement 2).

Table 2. Intervention Fidelity—Digital Health Care Model for the High-Risk Cohort.

Characteristic	Median (IQR)
% Followed up by ASHAs at least once	98.0 (96.6-100.0)
% Seen by a physician at least once	93.2 (91.9-95.9)
% Followed up by ASHAs at least 6 times and seen by physician	87.6 (51.9-93.7)
Visits by an ASHA	11.5 (6.2-19.0)
No. seen by a PHC physician	56 (39.0-97.5)
Setting in which PHC doctor care was accessed, No. (%)
Village-based health camps	837 (54.9)
Teleconsultation	93 (6.1)
In-person visit to the PHC clinic	595 (39.0)

Abbreviations: ASHA, Accredited Social Health Activists; PHC, primary health care center.

Primary and Secondary Outcomes

For the high-risk cohort, mean PHQ-9 scores were lower in the intervention arm (2.77 vs 4.48; mean difference, −1.71; 95% CI, −2.53 to −0.89; P < .001) at 12 months (Table 3). The intervention was also associated with improved remission rates at 12 months (74.7% vs 50.6%; odds ratio [OR], 2.88; 95% CI, 1.53 to 5.42; P = .001; effect size, 0.6). There were no differences in PHQ-9 and GAD-7 scores at 6 months (eFigure 3 in Supplement 2).

Table 3. Primary and Secondary Outcomes.

Outcomesa	Mean (SD)b		Mean difference (continuous outcomes) or OR (binary outcomes) (95% CI)b	P value	ICC at 12 moc
	Intervention	Control			Intervention	Control
High-risk cohort at 12 mo unless specified
Coprimary outcome
PHQ-9 score	2.77 (0.29)	4.48 (0.29)	−1.71 (−2.53 to −0.89)	<.001	0.12	0.13
Secondary outcomes
GAD-7 score	2.33 (0.29)	3.90 (0.29)	−1.57 (−2.39 to −0.74)	<.001	0.15	0.20
Achieved remission, No. (%)	1118 (74.7)	748 (50.6)	2.88 (1.53 to 5.42)	.001	0.27	0.20
Self-reported receipt of mental health treatment, No. (%)	1427 (94.5)	98 (2.5)	677.95 (244.32 to 1881.16)	<.001	0.25	0.48
GAD-7 score at 6 mo	4.91 (0.29)	5.02 (0.29)	−0.11 (−0.94 to 0.72)	.79	0.26	0.37
PHQ-9 score at 6 mo	5.65 (0.29)	5.54 (0.29)	0.10 (−0.73 to 0.93)	.81	0.25	0.32
Combined non–high-risk and high-risk cohort at 12 mo
Coprimary outcome
Behavior score	17.39 (0.26)	17.75 (0.26)	−0.36 (−1.11 to 0.39)	.34	0.29	0.26
Secondary outcomes
Knowledge score	14.45 (0.25)	13.11 (0.25)	1.34 (0.62 to 2.06)	.001	0.26	0.29
Attitude score	14.01 (0.19)	12.88 (0.19)	1.14 (0.58 to 1.69)	<.001	0.22	0.25
Stigma score	0.21 (0.07)	0.43 (0.07)	−0.22 (−0.42 to −0.03)	.02	0.37	0.64

Abbreviations: GAD-7, Generalized Anxiety Disorder–7 item; ICC, intraclass correlation; PHQ-9, Patient Health Questionnaire–9 item.

^a For PHQ-9, GAD-7, and stigma, lower scores are better; for behavior, knowledge, and attitude, higher scores are better.

^b All estimates are derived from repeated-measure linear mixed model with scores at 3, 6, and 12 months as the dependent variable. The model includes treatment arm, visit, their interaction, and the baseline score as fixed effects, random cluster effect, and repeated visit effect. Odds ratio estimates are from mixed-effects logistic regression with the outcome at 12 months as the dependent variable. The model includes intervention fixed effects and primary health care center random effects.

^c ICC for scores were calculated as ratio of between- and within-variance estimates from the linear mixed-effects null model with scores at 12 months as the dependent variable and only primary health care center random effects. The ICC for binary outcomes was calculated as the ratio of variance estimate, derived from logistic mixed-effects model with primary health care center random effect and sum of variance estimate and π² / 3. Models were fit separately to intervention and control arms.

For the combined non–high-risk and high-risk cohort, there was no difference in mean behavior scores at 12 months (17.39 vs 17.74; mean difference, −0.35; 95% CI, −1.11 to 0.41; P = .36); however, there were improvements in all 3 secondary outcomes, including increased knowledge and attitude scores and reduced stigma scores in the intervention arm compared with the control arm. The intervention was associated with improved (higher) mean behavior scores; however, this difference became nonsignificant at 6 months and 12 months. There were moderate improvements in community knowledge, attitude, and stigma scores at 12 months and in behavior scores at 6 months (standardized mean difference, 0.42-0.47) (eFigure 4 in Supplement 2).

Subgroup analyses showed no significant heterogeneity across any of the prespecified subgroups (Figure 2) for PHQ-9. For the behavior score, there was a statistically significant heterogeneity in treatment effect despite overlapping 95% CIs for the subgroups by sex, cohort, and depression severity.

Figure 2. Subgroup Analyses for the 2 Primary Outcomes.

NA indicates not applicable; PHC, primary health care center; PHQ-9, Patient Health Questionnaire–9 item.

Sensitivity analyses using covariate-adjusted analysis and imputed analysis regarding the primary and secondary outcomes were consistent with the primary models (eTables 3 and 4 in Supplement 2). Per-protocol analysis of behavior and PHQ-9 scores also showed similar results to intention-to-treat analysis, mainly due to high implementation fidelity. In a post hoc analysis, self-reported mental health medication use was higher in the intervention arm compared to control (10.1% vs 2.2%; OR, 5.11; 95% CI, 2.17 to 12.0; P < .001).

Discussion

In this randomized clinical trial, both the community antistigma campaign and the ASHA-led digital health interventions achieved a high level of implementation fidelity. For those at high risk of CMDs, the intervention was associated with lower depression scores compared to usual care and improvements in almost all secondary outcomes. The intervention had a major impact on access to mental health treatment with almost everyone reporting having seen a physician in the previous 12 months (compared to less than 3% in the control arm). There were large reductions in risk of depression, anxiety, and self-harm risk with a 3-fold increase in odds of remission (effect size 0.6). With only a modest increase in mental health medication use (10.1% of intervention arm participants), the benefits appear to be mainly due to nonpharmacological management. The improvements in the control arm align with the findings from Whiteford et al²³ in a high-income country setting where they found 53% of people with depression have natural remission at 12 months, and the intervention effect size of 0.6 is of similar magnitude to much smaller studies of individual-focused (rather than health service–focused) interventions previously conducted in LMIC settings.²⁴ For the high-risk cohort, there was no heterogeneity by any of the prespecified subgroups, suggesting there is potential to spread and scale to other populations and regions of India. Although there were significant improvements in PHQ-9 and GAD-7 scores at 12 months, there were no differences at 3 and 6 months, which suggests that implementation programs need a longer time frame for effects to be evident in line with recommendations in mhGAP-IG.²⁵ The length of time to observe a treatment effect may have been due to COVID-19–related delays in implementation combined with the likely need for sustained contact and follow-up by community health workers.

For the total adult population, the intervention did not demonstrate improvements in behavior scores at 12 months. However, it was associated with moderate improvements in community knowledge, attitude, and stigma scores at 12 months and behavior scores at 6 months, suggesting that more intensive efforts are needed to achieve sustained community-level change.²⁶ The inability to sustain behavior change on health care seeking at 12 months aligns with a previously published review of high-income country interventions²⁴ and the recent Lancet Commission Report on Stigma and Discrimination.⁶ Our trial design cannot assess whether the dual approach (antistigma campaign and digital care delivery) is needed to improve health outcomes; however, there are likely synergistic effects from combining approaches.^6,11,14

The 2022, the WHO and the Lancet Commission on Mental Health advocated for innovative strategies to reduce the burden of mental disorders globally, particularly using task-sharing workforce models and digital health solutions.^7,8,27 For other health conditions, digital health technologies have health care access and quality²⁸; however, in mental health, there are few large, robustly evaluated interventions.²⁹ Given the lack of evidence from LMICs, this study makes a novel contribution to the evidence base with potential for integration with India’s National Mental Health Programme.^12,30

Strengths and Limitations

The key strength of this study is high implementation fidelity in 2 diverse Indian states with a high follow-up rate. ASHAs achieved high follow-up rates despite their other work responsibilities. Because ASHAs are residents in the community, they have intimate knowledge of how best to reach people opportunistically at their homes. The digital health application also generated prioritization lists, which improved workflow efficiency. Similarly, the physicians effectively organized their work routines both through allotted times to receive ASHA referrals and home visits as part of dedicated health camps that were coordinated by ASHAs. Despite these strengths, given India’s heterogeneity, the intervention strategy would likely need adaptation to local context in other regions or abroad. Another limitation is that elevated PHQ-9 and GAD-7 scores are indicative of increased risk of depression and anxiety, respectively, and a confirmed clinical diagnosis is needed. However, the standard threshold scores for these instruments are commonly used for outcome assessment in research and public health programmes.²⁷ Because of known high rates of natural remission, we rescreened participants to define the high-risk cohort. This prioritized people with persistent symptoms, and therefore the findings may not be generalizable to a lower-risk cohort.

Additionally, the COVID-19 pandemic impacted trial monitoring and timelines, especially with prerandomization data collection. It also impacted delivery of face-to-face antistigma campaign elements, and often the campaign was either stopped or shifted to an online modality due to lockdowns and mobility restrictions. Although teleconsultations were introduced during COVID-19, access to such services is grossly limited and it is not likely to have had a substantive impact on mental health service provision. The impact of these issues on study outcomes is difficult to assess; however, despite the disruptions, PHQ-9 and GAD-7 scores demonstrated a consistent downward trend in the intervention arm.

Conclusions

While the evidence for mental health digital interventions is growing globally, few studies have focused on LMICs and tend to be small scale with short follow-up duration. The SMART mental health strategy, which combined a digital intervention with a multimedia antistigma campaign, was well adopted, effective in lowering depression severity, and increased remission rates from depression, anxiety, and self-harm.

Supplement 1.

Trial protocol and statistical analysis plan

Supplement 2.

eTable 1. PHC facility characteristics

eTable 2. Intervention fidelity - exposure to anti-stigma campaign components

eTable 3. Primary and secondary outcomes – covariate-adjusted analyses

eTable 4. Primary outcomes, GAD-7 score and remission – controlled multiple imputation analyses

eFigure 1. SMART Mental Health Intervention

eFigure 2. Study timelines

eFigure 3. Depression and anxiety primary and secondary outcomes at 3, 6, and 12 months (high-risk cohort)

eFigure 4. Knowledge, attitude, behaviour outcomes at 3, 6, and 12 months (combined high-risk and non-high-risk cohort)

eAppendix 1. Training and clinical visit activities provided by ASHAs and PHC doctors

eAppendix 2. Study Implementation Outcomes

eAppendix 3. Study outcome survey instruments

eAppendix 4. Serious adverse events definition

Supplement 3.

Data Sharing Statement