Mobile -Solutions Aiding Knowledge for Health Improvement

1. Study design: Using a community-based, cluster randomized controlled trial design (cRCT) the investigators will examine the impact of an integrated "m-health package - (M- SAKHI) for mothers on appropriate infant feeding practices, starting in the first / second trimester of pregnancy to one year after delivery, to improve child feeding practices and child growth, and reduce the prevalence of undernutrition in their children. This will result in two treatment groups (see diagram). The clusters for the trial will be villages under the administrative area of ASHAs. Villages with a population of 1000- 2000 (ranging from 751 to 2000) with single ASHA will be randomized to receive either this intervention (intervention villages) or continue with existing delivery of care (control villages). Selecting a single village per ASHA criteria will reduce the chances of contamination of study groups. The data will be collected using a longitudinal design because the investigators want to study the impact of intervention on the mothers starting at pregnancy until her infant is 24 month old. This is preferred to a cross sectional design to reduce the impact due to migration of mothers in or out of the clusters, who may not receive the intervention as planned. Cross sectional surveys are better suited if a before and after intervention design is used. Lastly on-going government programs can contaminate the intervention in a before-after cross section design. However in a CRCT, the government programs will be present in both study and control clusters, so the additional impact of the intervention can be evaluated. Additionally, growth velocity and development of an infant whose mother is receiving the intervention can also be evaluated using a longitudinal design.

This design will control for potential confounding factors (Observed and unobserved) because an adequate number of clusters (292 villages) will be randomly allocated to the treatment groups. This random allocation of treatments by villages will reduce selection bias that would arise if better performing PHCs with favorable local conditions were purposely selected. Contamination of intervention will be constrained by the administrative and geographic separation of the villages, and by buffer areas.

The outcome assessments will be made by field research officers (FROs) on a cohort of 2,728 mother-infant dyads (1,364 in each treatment group) measured at baseline and at follow up visits. There will be follow-up assessments every trimester during pregnancy and every month from birth until the children are 12 months of age, and then every three or 6 months till 24 months. Three of these assessments (when infants are 15,18 \& 24 months) will be made after the education intervention has ceased in order to assess the sustained impact on the prevalence of child stunting and infant feeding practices (see evaluation plan or details). The investigators expect a likely high correlation between baseline and follow up outcome measures, and low loss to follow up, thus making this approach the most efficient study design.

In this cRCT the interventions will be allocated at a community level, but the outcome assessments will be at the individual level. This is the best comparative design for the proposed interventions, which if delivered to individual mothers would most likely lead to contamination of the intervention in the densely populated rural village communities.

The proposed field area will be the centrally located villages in the catchment areas of Indian Government, PHC in 3 districts (Nagpur, Bhandara and Wardha) of eastern Maharashtra (see Map).There are around 28 PHCs.Each Primary Health Centre (PHC) covers an average population of 30,000 with an average of 30 villages. The administrative area under an ASHA is villages with population of 1000-200 (range of 751 to 2000).

As per the investigators' experience in the Maternal and Neonatal Health (MNH) Registry, the Crude Birth Rate (CBR) of the study area is 16.4 per 1000 population, thus expected or projected number of ante-natal cases (ANC) per thousand populations would be 18 per year considering 10 per cent pregnancy wastage.

At any given point of time, 60 per cent ANCs (11) will be available per thousand population, of which 6 will be up to 20 weeks of gestation. An additional 6 can be registered during next 3 months of enrollment period.

Using standard formulae, the sample size required would be 2,728 mother-infant pairs (1,364 per group) from 292 clusters .This sample would provide 90% power to detect a 15% relative reduction in stunting. Based on prior experience the investigators estimate that after accounting for probable cases of loss to follow-up and delivery outside the study area, about 70 percent of ANCs would be available for inclusion in the study. Thus 8 ANCs per 1000 population would be available for inclusion in the study and a population of 341000 needs to be covered to get required number (2728) subjects.

The unit of randomization will be the ASHA who provides her services to a village with a average population of 1000-2000 (ranging from 751 - 2000) persons. Using a computerized random number generator the investigators will randomize 146 ASHA (1 ASHA per village) to deliver services to pregnant women till her infant is 12 months using the M-SAKHI intervention in addition to the exiting services (intervention group), and, 146 to control group i.e routine provision of services or existing level of service delivery (uniform allocation), across each geographic area. The random allocation sequence will be generated using Stata® software.

All efforts will be made to enroll the women in their first / second trimester (up to 20 weeks of gestation). In both arms, to encourage women to enroll as early as possible, the eligible couples will be identified by ASHA and incentivized by home pregnancy test kits, so that they could report as soon they missed their periods. The average expected number of births for the population covered by 292 villages in the PHCs over 12 months would be 3000, which would be more than required for trial recruitment, thus confirming the feasibility of the sampling scheme.