Figure 1

Conceptual framework of intervention context, Kushal Maa intervention targets, outcomes and anticipated long-term impacts. Kushal Maa intervention targets include knowledge, social support and postnatal care. Primary behavioural and health outcomes are postpartum depression, exclusive breastfeeding and needed postpartum contraceptive uptake.

Table 2

Perinatal care, breastfeeding and contraceptive indicators overall and by rurality, across study states

Table 3

Study site characteristics

Inclusion and exclusion criteria

Women eligible for study enrolment must meet the following inclusion criteria: 18+years of age at the time of enrolment, in the last trimester of pregnancy (30–33 weeks of gestation), any parity, single or multiple gestation (1–2), knowledge of site-specific local language, access to own phone (either feature phone with WhatsApp ability or smartphone) and if the potential participant does not have their own smartphone, access to household-level smartphone for participating in both her marital and natal home.

Women who meet any of the following exclusion criteria will be excluded from participation in this study: incapable of providing informed consent and any impairment that prevents participation in the study intervention (eg, hearing or speech impairment).

Participant recruitment and enrolment

Participant flow through the study is detailed in figure 4 and online supplemental table S1. Recruitment will be carried out simultaneously across study sites following the same protocol. First, within one geographical area, we will select a sampling frame of primary health centres, where we will be recruiting from. With the assistance of the community health providers, auxiliary nurse midwives, ASHAs, Community Health Officers and any other administrative staff, study research assistants (RAs) will review antenatal care registration data and list potential study participants.

As soon as the listed women meet the gestational age eligibility criteria (ie, 30–33 weeks of gestation), the study team will travel to the household, preferably accompanied by the relevant ASHA worker for screening and enrolment activities. If a woman is identified as eligible for the study, the RA will carry out an informed consent process including a detailed description of study procedures, possible risks and benefits to participation and participant rights in research. The potential participant and their family members will have the opportunity to ask questions. Concurrence will be requested from household heads as appropriate (eg, mothers-in-law and husbands) for women’s participation in the study, and each woman will be requested to provide written (or thumbprint with witness, if illiterate) confirmation of informed consent per Indian guidelines.55 A copy of the informed consent form and the study information sheet will be provided to the women for their records (see example under online supplemental material).

Randomisation

Informed by the group nature of the intervention, limited gestational age range eligibility criteria and site-specific birth rates, each sequential group of 15 study participants per site will be block randomised to intervention or control arm using fixed-size blocking to ensure equal number of blocks for each arm and consistent intervention group sizes. Randomisation is conducted within REDCap’s randomisation module that is programmed to provide fixed-sized blocks (n=15 participants each) to ensure equal allocation in each block. Blocks thus formed are then randomly assigned to either intervention or control arms.

Blinding

Due to the nature of the behavioural intervention delivered at a group level, it is not possible to blind study participants or intervention moderators. However, to increase study robustness, our study will maintain blinding for study investigators, research coordinators, RAs conducting outcome assessments and data analysts.

Data collection and management

On participant enrolment, the RAs will collect detailed participant contact information and administer the baseline survey (table 1) in person in a private location at the participant’s residence or another convenient location. Follow-up surveys will be administered at 6 weeks, 3 months and 6 months post partum by the research team. Follow-up surveys at 6 weeks and 3 months post partum may be conducted over the phone, only if the participant is out of reach and has gone to natal home after delivery, whereas the 6-month endline survey will be conducted in person (table 1). For consistency, follow-up data collection points are defined by estimated due date due date with a window period of ±2 weeks from the scheduled date. At roughly 6 months post partum, participants will have had the opportunity to participate in 28 weekly Kushal Maa sessions. Self-reported health outcome data will be validated as possible against women’s health cards and CHW-maintained registers. Reasons for study withdrawal will be captured for all participants. All participant data will be collected on encrypted and password-protected tablets using REDCap secure data collection software, uploaded to project servers and regularly checked for quality. However, data will be gathered using paper-based forms as a backup in the event that tablets are not working or REDCap is experiencing technical difficulties. The site principal investigators are in charge of the safe storage of these paper forms. The data are transcribed into the REDCap secure data collecting programme, checked for accuracy and uploaded to the project server.

Backend data on intervention engagement will include weekly moderator tracking of each participant’s attendance and level of participation in group interactive sessions and use of the chat platform (posts, frequency, topic). Deidentified transcripts of group interactive sessions and text chats will be used to further understand health educational needs and interests for postnatal women.

Orientation to intervention digital platforms: Zoom and WhatsApp

Once participant eligibility for study participation is confirmed through screening and informed consent received, RAs will briefly orient all study participants on the use of the intervention-related applications (ie, Zoom, WhatsApp), confirming participants that they could or could not be assigned to the intervention arm, including facilitating their download on participant phones and providing them with an instructional information sheet. Further orientation on intervention-related applications will be conducted by intervention moderators for study participants randomised to the intervention group.

Participant retention

To maximise participant retention, data collection and intervention implementation activities will be closely tracked. Loss to follow-up will be reduced through collecting complete participant contact information at study recruitment, including contact information of two family members or friends (one each from marital and natal homes), at two inperson data collection points and following up in person with participants who were unable to be reached via phone. For intervention participants, loss to follow-up will be reduced through use of local moderators for intervention activities, follow-up to understand any non-attendance and inperson enrolment and early data collection activities.

A participant will be considered lost to follow-up if she fails to attend at least four group call sessions in sequence and study staff are unable to reach the participant after at least four contact attempts. End-of-study is defined as when the last participant collects the endline data collection.

Effectiveness and intervention mechanism analyses

Primary analyses will be conducted first as intention-to-treat including all randomised participants, regardless of their level of engagement in the intervention. Secondary analyses will be done per-protocol, with intervention participants categorised into high (75% of group calls), medium (50% of group calls) or low attendance who attended a medium level of calls (at least four group calls defined as per-protocol). The quantitative analysis will be done using the Consolidated Standards of Reporting Trials guidelines for individual RCTs.56

Study participant sociodemographic characteristics and obstetric history will be described overall and by intervention group using appropriate distributional statistics (means, SD, medians, IQRs). Baseline equivalence of sociodemographic characteristics across the final analytic sample will be assessed using χ² or t-tests, as per variable distribution.

We will estimate the effectiveness (Aim 1) of the Kushal Maa intervention following an intent-to-treat approach using mixed-effects logistic regression analyses to accommodate for any clustering induced by the group-oriented delivery of our intervention for primary outcomes at 6 months: exclusive breastfeeding, unmet need for postpartum contraceptives and postpartum depression. Secondary analyses of other maternal and child health behaviours and outcomes will follow a similar approach, employing logistic or linear mixed-effects regression for binary and continuous outcomes and survival analysis for time-to-event outcomes (eg, length of exclusive breastfeeding). Perprotocol analyses will be conducted as secondary.

Analyses of the intervention mechanism (Aim 2) will evaluate whether maternal and child health knowledge and social support mediate the relationship between intervention participation and the primary outcomes. To maximise rigour, these analyses will be conducted using modern principles of structural equation modelling (SEM) and causal inference methods. SEM readily allows for the inclusion of multiple mediators simultaneously and the creation of latent variables which represent shared variation among similar measures that are likely to be correlated. As part of our analyses, we will investigate whether the various candidate mediators are sufficiently correlated to be treated as measures of one or more latent variables. The resulting latent variables may be used as mediators, reducing the effects of measurement error and improving statistical power for tests of mediation. Causal inference methods extend SEM-based mediation approaches to yield optimal estimates of indirect effects for non-continuous outcomes and/or mediator, an important benefit for this study given that its outcomes are binary. We will use Mplus (Los Angeles, California, USA) to perform analyses for Aim 2 because it unites SEM and latent variables with causal inference-based mediation methods in the same software platform and because it can adjust SEs for the clustering of participants within intervention groups. To extend the rigour of these analyses, we will prespecify and control for the covariates that could influence the mediator-outcome relationship (eg, age, region) and perform sensitivity analyses to investigate the robustness of the results to potential confounding of mediator-outcome relationships by any remaining unobserved confounders. Technology assessment analysis (backend data) will contribute to this aim. We will analyse quantitative backend data to gain more insight into women’s intervention participation and engagement, including attendance and level of participation in group interactive sessions and use of the chat platform (posts, frequency, topic, etc) using frequencies and means/medians. WhatsApp group discussion data will be translated and analysed qualitatively for content to further understand educational needs and interests for postnatal women.

Process evaluation activities will explore study implementation through review of monitoring indicators collected from back-end data sources such as women’s weekly attendance and engagement in group calls, interactions in the text chat group, etc, as well as from descriptive analysis of weekly reports submitted by intervention moderators regarding their perspectives of the group. Further evaluation will occur through review of monitoring data, study documents and interviews with study personnel.

Cost and cost-effectiveness analyses

Hypothesising and anticipating that the intervention will yield effectiveness, we will conduct cost analysis and cost-effectiveness analyses (Aim 3) from the individual, health system (inclusive of programme and health system costs) and societal perspectives (inclusive of participant, programme and health system costs) (table 3). We will value and calculate costs from the individual perspective and health system perspective, following published guidelines on costing approaches performed in LMICs. The health system costs will be derived from the study budget and accounting records and relevant estimations. Start-up costs, which include office furniture and equipment, employee recruitment costs and other resource costs incurred for the purpose of initiating implementation, will be calculated and presented separately from the implementation cost. We will annualise the start-up cost but not discount the value of the start-up capital items.

The implementation costs include personnel salaries (eg, administrative staff, CHWs), training, CHW travel costs, supplies (eg, office supplies), equipment and buildings (eg, rent, utilities) and other recurrent costs (eg, telephone and internet). The costs will be collected during the trial period in Indian rupees and will be inflated to the year of analysis and presented in both Indian rupees and US$.

Costs from the individual perspective include healthcare-related out-of-pocket payments and will be measured through repeated individual surveys, asking participants the time it took them to seek care (opportunity cost) and the amount they paid for travel, medications, food and accommodation and other expenses. One central objective to the cost analysis and cost-effectiveness analysis for this study is to evaluate the value of the support group. We hypothesise that time spent in the programme, specifically engaging with a support group, will increase a sense of connectedness and lead to the formation of community, both of which will contribute to better maternal health in general and mental health in particular. Each participant’s social connectedness assessment will be represented as a social network indicator building from betweenness, degree and eigenvector centrality of individual communication to generate a measure of centrality.

We will estimate the incremental cost-effectiveness ration (ICER) of Kushal Maa as compared with the standard of care. Incremental costs will be divided by incremental health effects to generate a deterministic estimate of ICERs, expressed as a cost per: (1) exclusive breastfeeding added, (2) postpartum depression averted and (3) met need for postpartum contraceptives. Time horizons of 1 year and 10 years will both be considered to cover all costs and effects of the intervention on postpartum depression. To understand the short-term cost-effectiveness of Kushal Maa, a 1- year time horizon that encompasses the trial period will be used. In the analysis for short-term cost-effectiveness, no modelling or simulation will be employed. We will use the data collected from this RCT to calculate ICER using the following equation:

To evaluate the long-term cost-effectiveness of Kushal Ma, we will use a 10-year time horizon with a cycle length of 1 year. We will use TreeAge Pro V.2023 to construct a decision tree and conduct the analysis. We will test for uncertainty by conducting one-way and probabilistic sensitivity analyses.

We will use the calculated social network indicator to stratify the cost-effectiveness analysis (CEA) and examine the ICER of Kushal Maa under varying levels of social connectedness. This study will measure social connectedness and associated opportunity cost to gain the connectedness through capturing the frequency, duration and sum of unique contacts with whom each participant engaged during the trial period. We will ask participants to report their contacts (table 3). The information will be used to construct matrices, enabling us to calculate degree centrality and betweenness centrality for each participant. The information will enable us to evaluate how the cost-effectiveness of Kushal Maa in improving maternal and neonatal health outcomes may vary by social connectedness. The findings will be critical to informing potential replication and programme expansion costs as well as scaling and implementation.