Article Text

The impact of a direct to beneficiary mobile communication program on reproductive and child health outcomes: a randomised controlled trial in India
  1. Amnesty Elizabeth LeFevre1,2,
  2. Neha Shah2,
  3. Kerry Scott2,
  4. Sara Chamberlain3,4,
  5. Osama Ummer4,5,
  6. Jean Juste Harrisson Bashingwa6,7,
  7. Arpita Chakraborty8,
  8. Anna Godfrey3,
  9. Priyanka Dutt3,4,
  10. Rajani Ved9,10,
  11. Diwakar Mohan2
  12. On behalf of the Kilkari Impact Evaluation Team
    1. 1University of Cape Town, School of Public Health and Family Medicine, Cape Town, South Africa
    2. 2International Health, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland, USA
    3. 3BBC Media Action, London, UK
    4. 4BBC Media Action, New Delhi, Delhi, India
    5. 5Oxford Policy Management, New Delhi, Delhi, India
    6. 6MRC/Wits-Agincourt Unit, University of the Witwatersrand, School of Public Health, Johannesburg, Gauteng, South Africa
    7. 7Computational Biology Division, Department of Integrative Biomedical Sciences, Institute of Infectious Disease and Molecular Medicine, University of Cape Town, Faculty of Health Sciences, Cape Town, South Africa
    8. 8Research and Evidence Practice, Oxford Policy Management, New Delhi, Delhi, India
    9. 9National Health Systems Resource Centre, New Delhi, Delhi, India
    10. 10Bill and Melinda Gates Foundation, Delhi, India
    1. Correspondence to Dr Amnesty Elizabeth LeFevre; aelefevre{at}gmail.com

    Abstract

    Background Direct-to-beneficiary communication mobile programmes are among the few examples of digital health programmes to have scaled widely in low-resource settings. Yet, evidence on their impact at scale is limited. This study aims to assess whether exposure to mobile health information calls during pregnancy and postpartum improved infant feeding and family planning practices.

    Methods We conducted an individually randomised controlled trial in four districts of Madhya Pradesh, India. Study participants included Hindi speaking women 4–7 months pregnant (n=5095) with access to a mobile phone and their husbands (n=3842). Women were randomised to either an intervention group where they received up to 72 Kilkari messages or a control group where they received none. Intention-to-treat (ITT) and instrumental variable (IV) analyses are presented.

    Results An average of 65% of the 2695 women randomised to receive Kilkari listened to ≥50% of the cumulative content of calls answered. Kilkari was not observed to have a significant impact on the primary outcome of exclusive breast feeding (ITT, relative risk (RR): 1.04, 95% CI 0.88 to 1.23, p=0.64; IV, RR: 1.10, 95% CI 0.67 to 1.81, p=0.71). Across study arms, Kilkari was associated with a 3.7% higher use of modern reversible contraceptives (RR: 1.12, 95% CI 1.03 to 1.21, p=0.007), and a 2.0% lower proportion of men or women sterilised since the birth of the child (RR: 0.85, 95% CI 0.74 to 0.97, p=0.016). Higher reversible method use was driven by increases in condom use and greatest among those women exposed to Kilkari with any male child (9.9% increase), in the poorest socioeconomic strata (15.8% increase), and in disadvantaged castes (12.0% increase). Immunisation at 10 weeks was higher among the children of Kilkari listeners (2.8% higher; RR: 1.03, 95% CI 1.00 to 1.06, p=0.048). Significant differences were not observed for other maternal, newborn and child health outcomes assessed.

    Conclusion Study findings provide evidence to date on the effectiveness of the largest mobile health messaging programme in the world.

    Trial registration number Trial registration clinicaltrials.gov; ID 90075552, NCT03576157.

    • child health
    • immunisation
    • nutrition
    • individual randomized trial

    Data availability statement

    Data are available upon reasonable request. Anonymised data are available upon reasonable request from the study Principal Investigator Dr. Amnesty LeFevre aelefevre@gmail.com.

    https://creativecommons.org/licenses/by/4.0/

    This is an open access article distributed in accordance with the Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits others to copy, redistribute, remix, transform and build upon this work for any purpose, provided the original work is properly cited, a link to the licence is given, and indication of whether changes were made. See: https://creativecommons.org/licenses/by/4.0/.

    Statistics from Altmetric.com

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

    What is already known on this topic

    • Direct-to-beneficiary mobile communication programmes offer great potential to disseminate health information at a population level; reaching a large number of beneficiaries rapidly and at low cost. As a result, they are among the few examples of digital health solutions to scale widely in a range of settings globally. Despite their potential, no evidence exists of the impact of these programmes at scale in low-income and middle-income countries where the majority of maternal and child deaths occur each year. The evidence that does exist is limited to smaller scale pilots; key programme components of which often change in the process of scaling up.

    What this study adds

    • Kilkari is the largest direct-to-beneficiary mobile communication programme in the world and has reached over 10 million women and their families across 13 states in India. Our study is the first randomised controlled trial conducted to date of a beneficiary mobile communication programme at scale. Exposure to Kilkari was significantly associated with improvements in a few important health practices, including the use of reversible contraceptive methods, but not others, including exclusive breast feeding. Subgroup analyses highlight the differential impact among key population segments, including the poorest.

    How this study might affect research, practice or policy

    • Study findings provide the most conclusive evidence to date on the effectiveness of the largest beneficiary mobile communication programme in the world. The differential targeting of key population segments could serve to deepen the impact. Further research is needed to assess the impact of shorter duration campaigns less susceptible to SIM change, network disruptions and population migration.

    Background

    Direct-to-beneficiary mobile health (mHealth) programmes which provide stage-based health information to new and expectant mothers have proliferated rapidly throughout the last decade.1–3 In low-income and middle-income countries where the majority of maternal and child deaths occur, mHealth messaging programmes are among the few examples of digital health programmes to successfully scale. Five programmes globally have scaled to reach over a million subscribers in Bangladesh,4 5 India,6 7 South Africa1 and Tanzania.

    Evidence linking scaled mHealth messaging programmes to changes in health outcomes is limited7–12; a factor which may impede sustainability and further expansion.13 The scant evidence available suggests varied impacts on a range of outcomes including improvements in birth weight7; infant feeding practices7 8; utilisation of early and complete antenatal care (ANC),8 9 facility-based skilled birth attendance,10 12 maternal and childhood immunisation services7; and reductions in perinatal mortality.11 However, efforts to compare and generalise these findings have been hampered by wide variations in study contexts and programme components including the content, modality, frequency of messages and supporting programmatic activities.14 The small scale of these programmes at the time that they were evaluated also limits the generalisability of the findings observed. As programmes scale they often undergo significant changes in design, implementation and resource availability, which can lead to a ‘voltage drop’ in a range of factors including reach, exposure and impact on beneficiary practices and health behaviours.2 15 These challenges underscore the need for evidence at each stage of an mHealth programme’s maturity from pilot to delivery of services at scale, to a large number of beneficiaries.

    Kilkari is the world’s largest mHealth messaging programme of its kind. Kilkari is an outbound service that makes weekly, stage-based, prerecorded calls about reproductive, maternal, neonatal and child health (RMNCH) directly to families’ mobile phones, starting from the second trimester of pregnancy until the child is 1 year old. BBC Media Action designed and piloted Kilkari in the Indian state of Bihar in 2012–2013, and then redesigned and scaled it in collaboration with the Ministry of Health and Family Welfare (MOHFW) between 2015 and 2019. Prior to its transition to MoHFW in April 2019, Kilkari had been scaled to 13 states across India and reached over 10 million subscribers.16

    From 2018 to early 2020, an individually randomised controlled trial (RCT) was conducted in four districts of Madhya Pradesh to determine the effectiveness of Kilkari. This study sought to assess whether exposure to mHealth information calls during pregnancy and postpartum improved infant feeding and family planning practices. This is the first study to evaluate the impact of a maternal mHealth messaging programme at scale; additional details on the evaluation protocol have been published elsewhere.17

    Methods

    Trial design

    The study is an individually RCT with a parallel and unblinded design.

    Participants

    At the time of randomisation in late 2018, women enrolled to the study (n=5095) were 12–34 weeks of gestation, more than 18 years of age, could speak and understand Hindi, and owned or had access to a mobile phone during the day when Kilkari calls were likely to come. Women were excluded who did not consent or who were mobile subscribers of Bharat Sanchar Nigam Limited (the state-owned telecommunications company) due to poor network coverage in the RCT districts in Madhya Pradesh. Given the shared nature of mobile phones, the husbands of women enrolled to the study were additionally interviewed as part of endline survey activities.

    Study setting

    Women in India have limited access to and use of mobile phones. Despite near universal household level phone ownership (92.8%), only 47.8% of Indian women have access to a mobile phone (41.6% in rural areas and 62.7% in urban).18 In the central Indian state of Madhya Pradesh (population 75 million) which is home to an estimated 20% of India’s total population, 19.1% of women rural areas and 49.5% in urban had access to a phone that they themselves could use in 2015.19 Beyond limitations in women’s phone access, Madhya Pradesh’s population health status falls below national level averages for most RMNCH indicators. In 2015, only 49.6% of women reported using any modern method for family planning and 12.1% reported having an unmet need for family planning.19 While over half (53.0%) of pregnant women attended ANC clinics in the first trimester, only 35.7% received the recommended four ANC visits.19 Despite high rates of institutional delivery (80.8%), only 54.9% of women and newborns received a postnatal health check within 2 days following birth.19 Among children, 34.4% were breastfed within 1 hour of birth and 58.2% were exclusively breastfed until 6 months of age.19 One in four children (25.8%) were wasted (weight-for-height) and 42.0% were stunted (height-for-age).19 Among children 12–23 months of age, 53.6% were fully immunised (Bacille Calmette-Guerin (BCG), measles and three doses each of polio and Diphtheria, Pertussis, and Tetanus (DPT)).19

    Intervention description

    Kilkari is comprised of 90 min of RMNCH content sent via 72 once weekly voice calls: 24 during pregnancy, 24 within the first 6 months postpartum and 24 from 7 to 12 months postpartum. Individual calls span an average of 77 s in duration and are framed as coming from ‘Dr Anita’. Across health content areas, 18% of cumulative call content is on family planning (benefits of family planning, modern reversible methods, sterilisation, pregnancy tests); 13% on child immunisations (diseases covered, doses); 13% on nutrition (malnutrition, growth monitoring, anaemia); 12% on infant feeding (quality of food, breast feeding, complementary feeding, anaemia); 10% on pregnancy care (ANC, institutional delivery, rest, nutrition, tetanus toxoid, emergency services); 7% on entitlements; 7% on diarrhoea; 7% on postnatal care (newborn danger signs, cord care, hypothermia); and the remainder on a range of topics including intrapartum care, water and sanitation (WASH) and early childhood development (figure 1).

    Figure 1

    Summary of Kilkari content during pregnancy and up to 1 year post partum. WASH, water and sanitation.

    The mobile numbers provided by women randomised to receive Kilkari began receiving calls no later than 8 months (34 weeks) after conception, making them eligible for at least 57 of the 72 messages. The timing of RCT enrolment was intended to mirror the timing of enrolment to Kilkari elsewhere in India. Once subscribed, Kilkari used an algorithm to retry mobile numbers up to nine times—three times in each week—three times on the first day, and twice then twice each day for the next 3 days—to reach a subscriber. Respondents were given the option to unsubscribe from the service at any time.

    Outcomes

    The primary outcome of this study was the reported practice of exclusive breast feeding for infants 0–6 months of age. The secondary outcome was the use of modern reversible contraceptive methods (including intrauterine contraceptive device (IUCD), injectables, oral contraceptive pills, emergency contraceptive pills, condoms) at 1 year post partum. The latter outcome was added after the trial registration but included in the published protocol. Three reasons underpin its inclusion. First, family planning messages constitute the largest overall message share (figure 1). Second, family planning calls occur throughout the extended postpartum window allowing for a large window of exposure to the intended messages. Third, the complexity of phone sharing practices in this population means that men may be the actual listeners of some of the Kilkari calls.20 Family planning represents includes behaviours which depend on both men and women both of whom may have been exposed to Kilkari calls. The outcome of immediate/early breast feeding noted in the trial registration was not emphasised as a primary or secondary outcome in our protocol. The study team felt that changes in these outcomes might be challenging to observe for two reasons. First, immediate/early breast feeding can be supply side dependent in the case of facility based and/or skilled attendant deliveries (which is why we reported it for normal deliveries and all deliveries) and thus, information provided to pregnant women is unlikely to move the indicator. Second, there are only two Kilkari messages (message 17: pregnancy month 8, week 1 and message 23: pregnancy month 9, week 3) which include content on early and immediate breast feeding.

    To assess the impact of Kilkari exposure on RMNCH outcomes, endline surveys were administered to enrolled women and their husbands after 12 months postpartum. The women’s endline survey included modules on RMNCH knowledge, practice, decision-making and demand for information and supply side services for the following health areas: pregnancy and delivery care, newborn care, child health, infant and young child feeding (IYCF), family planning, media consumption and reported demand for and receipt of health services from frontline health workers. The men’s survey sought to measure RMNCH knowledge; mobile phone ownership, use and literacy; and women’s access to phones. In addition to relying on respondent recall to answer survey questions, investigators also collected data from the participant’s Mother and Child Health card, which is issued to every pregnant woman at the local health facility and updated each visit and includes the child’s immunisation record.

    Sample size

    To detect a 5% difference in the reported practices of exclusive breast feeding and reversible modern contraceptive use assuming an alpha of 0.05, 80% power, the estimated sample size for a two-sample proportions test would be 3200 women. After adjusting for a design effect (variance inflation factor) of 1.25 due to clustering at the level of the Community Development Block (India’s subdistrict level administrative units), a 20% loss to follow-up from enrolment to endline, and a potential loss of 35% of women due to poor reporting of phone numbers and/or changes to original phone numbers provided, a total of 5100 women were targeted for enrolment in the study. We assumed that 40% of women in the reproductive age would have access to a mobile phone in 2018 at the start of the study, and that 12.5% of households would have a pregnant woman between 4 and 7 months, and thus anticipated needing to visit up to 750 000 households across the four study districts.

    Enrolment of participants

    Study participants were identified through a household listing survey carried out from July to October 2018 which included an intensive mapping of households and their residents and screened participants for eligibility. Women identified during the listing as meeting the eligibility criteria (Hindi speaking, 4–7 months pregnant, with access to a mobile phone during the day) were screened again by a separate baseline survey team. Those confirmed to meet eligibility criteria were administered a 1.5 hour baseline survey inclusive of modules on sociodemographic characteristics; phone ownership, use and digital competency; RMNCH knowledge and practices. Following the baseline survey, study participants who completed the face to face survey and consented to receive Kilkari messages were randomised to receive Kilkari calls (intervention) or not receive messages (control). Exposure to Kilkari spanned through the child’s first birthday. Study participants were interviewed at 12 months post partum as part of endline survey activities carried out from November 2020 to March 2021. No harms are anticipated for study participants enrolled to either study arm.

    Randomisation and masking

    The participants were randomised after stratification on a range of variables potentially associated with exposure (listening to messages) and likely to influence outcomes, including gestational age, parity, age of woman and ownership of phone. Stratification sought to ensure a balance of covariates between the intervention and control groups. The individual randomisation procedure was done on a block-by-block basis (block randomisation) after enrolment was completed for each block. This ensured similar number of intervention and control subjects for each community development block. Randomisation was performed by picking a random sample of participants for each Block using the sample command in Stata with the use of the above listed variables as stratifiers. The randomisation and allocation processes were carried out by the study investigators in late 2018 following the completion of baseline survey activities. The random sample picked was allocated to receive the intervention immediately after randomisation. The participants could not be blinded due to the nature of the intervention. The data collectors administering surveys were not aware of the allocation.

    Statistical methods

    To assess exposure to Kilkari content, call data records from the interactive voice response (IVR) system were linked to baseline and endline survey data. Listening patterns were assessed for each subscriber by call, for the duration of the their subscription to Kilkari. The content of the Kilkari calls was mapped to key health outcomes including knowledge and practice. Exposure was defined at a listening threshold of 50% or more of the cumulative duration of the calls mapped to the outcome. For example, 7 Kilkari calls (week 20, 24, 29, 34, 65, 66, 68) included content on reversible modern contraceptive methods, for a total of 10.5 min of audio content on this topic (online supplemental table 1). To be ‘exposed’ subscribers would have had to listen to 50% or more of the cumulative message total of these messages, for example, 5 min or more of the contents.

    Supplemental material

    Primary analyses of outcomes were done with modified intention-to-treat (ITT) analyses at the individual level so that outcomes were analysed regardless of the degree of listening to Kilkari. The term modified refers to the requirement of being able to determine the outcomes beyond 1 year of the postpartum period (ie, only those who were reached by the endline survey were included in the analysis). Relative risk (RR) for primary and secondary outcomes was calculated for the intervention group compared with the control group using log-binomial regression with general estimating equations to account for correlation within villages.

    To assess the impact of exposure on outcomes, compliance adjusted treatment effects were additionally generated using the instrumental variable (IV) methodology. The study arm based on randomisation was considered the instrument variable. As noted above, exposure was defined as listening to at least 50% of the cumulative duration of the calls mapped to the outcome. RR estimates were calculated using log binomial models for each outcome with adjustment for clustering at the village level. The IV analysis was carried out using the IVREG package in R.21 Frequencies and proportions are used to describe differences across intervention and control groups (ITT) as well as those exposed to 50% or more of Kilkari content.

    Role of the funding source

    The Bill and Melinda Gates Foundation had no role in the study design; collection, analysis and interpretation of data; in the writing of the report; or in the decision to submit the paper for publication. All authors confirm that they had full access to all the data in the study and accept responsibility for the publication submitted.

    Patient and public involvement

    Patients were first engaged on identification in their households as part of a household listing carried out in mid/late 2018. Those meeting eligibility criteria were interviewed as part of the baseline survey, and ultimately randomised to the intervention and control arms. Prior to the administration of the baseline, a small number of patients were involved in the refinement of survey tools through qualitative interviews, including cognitive interviews, which were carried out to optimise survey questions, including the language and translation used. Finalised tools were administered to patients at baseline and endline, and for a subsample of the study population, additional interviews carried out over the phone and via qualitative interviews between the baseline and endline surveys. Unfortunately because of COVID-19 and associated travel restrictions, patients could not be involved in the dissemination of study findings. However, public dissemination of the results has occurred through a number of global level presentations.

    Results

    Sample characteristics

    Among the 5095 women enrolled at baseline, 2695 were randomised to the intervention arm and 2400 to the control arm (figure 2). An estimated 95% (n=2559) of the mobile numbers provided by women randomised to the intervention arm received at least 1 Kilkari call that was answered. Table 1 shows the characteristics for the ITT population of women (n=5095) enrolled at baseline and their husbands (n=3842) interviewed at endline.

    Table 1

    Sample characteristics for women and men’s surveys in four district of Madhya Pradesh

    Figure 2

    Consolidated Standards of Reporting Trials flow diagram.

    Exposure to Kilkari calls

    The mobile numbers provided by those randomised to the intervention arm began receiving Kilkari calls no later than the first week of the eighth month of their pregnancy (Kilkari weekly call 17). Successful calls are those that reach the handset and are answered. We cannot identify who answered the call. Figure 3 depicts the number and duration of successful calls for each of the Kilkari calls among the mobile numbers provided by those randomised to the intervention arm. As denoted by the orange bars, the number of successful calls peaked at Kilkari call 22 (pregnancy month 9, week 2) and declined gradually from call 25 (baby month 1, week 1) to 38 (baby month 4, week 2). Declines in successful calls were attributed primarily to deactivations as a result of adverse events (miscarriage, death, stillbirths), SIM change and increases in call non-delivery rates. From the timing of the baseline survey to 12 months postpartum when the endline survey occurred, 44% of women enrolled into RCT changed their SIM cards and thus were no longer receiving Kilkari calls. Women who retained the same SIM card from baseline to endline tended to be more educated, within the higher socioeconomic strata, and advantaged caste groups. Call non-delivery rates increased throughout 2018 and were underpinned by the bankruptcy of Reliance communications, one of India’s leading mobile network providers.

    Figure 3

    The number of successful* Kilkari calls and per cent of listening by call over time for those randomised to the intervention arm in four districts of Madhya Pradesh, India. *Successful calls are those which are delivered to the handset and answered for 1 s or more.

    High listenership, depicted by the black and blue horizontal lines in figure 3, is defined as the proportion of subscribers who listened to at least 50% and 75%, respectively, of the total content of an individual call. Across individual Kilkari calls, high listenership rates were stable over time, ranging from 45% to 65% among successful calls. An average of 65% of subscribers listened to more than 50% of the cumulative total of successful calls while 31% listened to more than 75% of the total content of calls. The odds of listening to 50% or more of the cumulative content were greater among those enrolled earlier in pregnancy, with the same phone number from baseline to endline surveys, and with double digit numeracy (online supplemental figure 1).

    The Kilkari programme’s algorithm that keeps calling subscribers who do not answer up to nine times each week increased the proportion of successful calls, especially among subscribers in the most marginalised demographic groups (figure 4). On average, seven call attempts were needed to reach subscribers in the poorest socioeconomic strata as compared with four attempts needed to reach those in the higher (middle, richer and richest) socioeconomic strata. Across ethnic groups, seven call attempts were needed on average to reach scheduled caste subscribers versus four in the more advantaged other backward castes and general castes. Similar patterns were observed for those without education and with two or more children. Efforts to assess the timing of calls per day are reported elsewhere22 and broadly suggest that call answer rates were lowest earlier in the day and overall for those in the most marginalised sociodemographic groups.

    Figure 4

    Number of call attempts needed to reach subscribers by sociodemographic characteristic. Dark blue bars denote the more marginalised subscribers as compared with those in the lighter blue. OBC, other backward castes.

    Primary outcome: exclusive breast feeding

    Kilkari was not observed to have a significant impact on exclusive breast feeding (ITT, RR: 1.04, 95% CI 0.88 to 1.23, p=0.64; IV, RR: 1.10, 95% CI 0.67 to 1.81, p=0.71). However, the proportion of children born through a normal delivery who were immediately breastfed was 3.0% higher in the control arm as compared with the intervention (ITT, RR: 0.96, 95% CI 0.93 to 1.00, p=0.027). Significant differences were not observed for other IYCF outcomes assessed, including complementary feeding and the provision of iron folate (table 2).

    Table 2

    Infant and young child feeding and family planning practices in four districts of Madhya Pradesh

    Secondary outcome: modern reversible contraceptive methods

    Kilkari was associated with significantly higher reported use of modern reversible contraceptive methods (condoms, oral contraceptive pills, emergency contraceptive pills, injectables, IUCD) using both ITT (RR: 1.12, 95% CI 1.03 to 1.21, p=0.007) and IV (RR: 1.72, 95% CI 1.06 to 2.77, p=0.028) analyses, corresponding to an absolute difference of 3.7% across study arms and 7.4% higher among exposed versus not exposed groups (table 2). Differences in the use of reversible methods were driven by significantly higher use of condoms (ITT, RR: 1.12, 95% CI 1.01 to 1.23, p=0.024) and a marginally significant increase in the use of oral contraceptive pills (ITT, RR: 1.28, 95% CI 0.96 to 1.69, p=0.088). Subgroup analyses suggest that absolute differences in reversible method use between those exposed to Kilkari and those not exposed were highest for those with any male child (9.9% difference), in the lowest three socioeconomic strata (difference of 15.8% difference in the poorest, 8.8% in poorer and 6.7% in poor), and in scheduled castes (12.0% difference) (figure 5, online supplemental table 2A–C).

    Figure 5

    Proportion of those exposed versus not exposed using reversible modern contraceptive methods by sociodemographic characteristics. OBC, other backward castes.

    In addition to differences in the use of modern reversible contraceptive methods, the proportion of men or women sterilised since the birth of the child was 2.0% lower in the intervention arm (12.9%) than reported in the control arm (14.9%). Findings were significant across study arms (ITT, RR: 0.85, 95% CI 0.74 to 0.97, p=0.016) and among those exposed to Kilkari versus those not exposed (IV, RR: 0.56, 95% CI 0.37 to 0.85, p=0.006).

    Effects on other RMNCH practices

    Table 3 presents impact findings for a range of other RMNCH practices, including facility delivery, essential newborn care, WASH and childhood immunisations. Findings suggest that the proportion of children immunised at 10 weeks was 2.8% higher in the intervention arm as compared with the control arm (ITT, RR: 1.04, 95% CI 1.00 to 1.07, p=0.048). Significant differences were not observed for other RMNCH practices assessed. Subgroup analysis found that differences in the proportion of children fully immunised at 12 months were higher among women exposed to Kilkari who had no schooling (difference of 15.2% across exposed vs not exposed groups) or primary school completed (difference of 9.9%) as compared with those with secondary school education or higher (difference of 1.4%) (online supplemental tables 2A–C). A similar trend was observed for immunisations at birth, 10 weeks, and 14 weeks.

    Table 3

    Delivery, essential newborn care, WASH and immunisation practices in four districts of Madhya Pradesh

    Effects on knowledge, decision-making, discussion, demand, supply

    Tables 4–6 and online supplemental tables 3A,B and 4A,B present data on differences in women and men’s knowledge, decision-making, reported discussion, demand for and receipt of health information for RMNCH outcomes assessed. ITT findings suggest that the proportion of women (RR: 1.04, 95% CI 1.01 to 1.07, p<0.001) and men (RR: 1.05, 95% CI 1.00 to 1.09, p=0.032) who know to ask a health provider questions about family planning use was higher among those in the intervention arm compared with those in the control arm (table 2). This corresponds to a 3.4% and 3.0% difference across study arms for women and men, respectively. Similarly, both ITT (RR: 1.05, 95% CI 1.00 to 1.11, p=0.032) and IV (RR: 1.24, 95% CI: 1.01 to 1.53, p=0.044) findings suggest that women subscribed to Kilkari were more likely to know that they could obtain a pregnancy test from Accredited Social Health Activities, the Government of India’s community health worker cadre (table 5). Among a range of outcomes assessed on decision-making, discussion, demand and receipt of information, Kilkari was associated with an increased likelihood of women’s involvement in the decision-making on when to start giving their children complementary food (ITT RR: 1.03, 95% CI 1.00 to 1.06, p=0.041; IV RR: 1.11, 95% CI 1.00 to 1.24, p=0.042) (table 3). This corresponds to a 2.8% difference across study arms for women. Significant differences were not observed for other intermediate outcomes assessed.

    Table 4

    Women’s involvement in decision-making, discussion and demand for information in four districts of Madhya Pradesh

    Table 5

    Family planning knowledge among women in four districts of Madhya Pradesh

    Table 6

    Family planning knowledge among men in four districts of Madhya Pradesh

    Discussion

    Study findings provide evidence on the exposure and impact of mHealth information messages across a range of RMNCH outcomes. Kilkari was not significantly associated with significant improvements in infant feeding practices, including the primary outcome of exclusive breast feeding. A significant increase was observed in the use of modern reversible contraceptive methods, particularly among those women exposed to Kilkari with any male child, in the poorest socioeconomic strata, and in disadvantaged ethnic groups. Increases in reversible method use were driven principally by increases in condom and oral contraceptive use, and occurred concurrently with a significant decrease in the proportion of men or women sterilised since the birth of the child. Increases in the proportion of children immunised at 10 weeks and the proportion of women involved in decision-making on when to introduce complementary foods were also observed. Significant improvements in other RMNCH outcomes were not observed. Findings suggest that an average of 65% of subscribers listened to more than 50% of the cumulative total of successful calls while 31% listened to more than 75% of the total content of calls. High listenership rates were stable over time; a likely indication that if calls reach the mobile device and are answered, subscribers tend to the listen to them. Analyses of the programme’s algorithm that attempts to call subscribers up to nine times to yield a successful call found that a greater number of call attempts are needed to reach the most marginalised.

    Kilkari includes health information content across more than 11 health areas. The theory of change underpinning the evaluation is published elsewhere.17 Infant feeding outcomes were selected as the priority outcomes from the outset because they represent a sizeable share of message content (12%) and the practice of these behaviours is not supply side dependent, although immediate breast feeding can be. In the case of exclusive breast feeding, the low (58%) pretrial prevalence of this behaviour additionally left much potential for improvement.19 In practice, significant improvements were not observed in exclusive breast feeding. Qualitative research findings published elsewhere help to shed light on the factors underpinning these results.23 Kilkari messages that told women to give only breastmilk were interpreted to mean ‘give only breastmilk as the main source of nutrition’—since janam ghutti (a local term referring to homemade or indigenous herbal tonic), water and the provision of small quantities of sweets were not a source of nutrition, giving them to babies was not considered by caregivers to be incompatible with following the message to exclusively breastfeed.23 When Kilkari did specify not to give ghutti or other food items, women often did not seem to absorb this from the message.23 Broadly, women did not see valid reasons to stop giving substances such as ghutti which they thought helped to support the child’s health and in instances of illness could prevent worse illness. While quantitative survey findings did indicate a significant increase in the proportion of the mothers involved in decision-making over when to introduce complementary foods, mothers were also not the sole decision-makers and families too felt strongly that ghutti, water and other substances were helpful for babies. Women also reported appreciating and trusting the advice of elders who encouraged giving ghutti and other substances. Specific to the practice of breast feeding, most women in rural Madhya Pradesh reported having very limited support to exclusively breastfeed, including refrigeration, consistent electricity supply, pumping and bottle sterilisation, as well as support to solve physiological issues that may interfere with breast feeding. In sum, complex behaviours that have competing social norms and are currently seen to work quite well are challenging to change and receiving a small number of Kilkari voice calls advocating that one adopt a particular practice did not catalyse a change in behaviour.

    From a measurement perspective, qualitative findings affirmed those from cognitive interviews done to support the design of survey tools.24 25 While the National Family Health Survey-4 findings from 2015 to 16 suggest that the reported practice of exclusive breast feeding was 58% in Madhya Pradesh, our findings suggest a much lower prevalence (intervention: 10.8%; control: 10.4%) when additional probing questions are incorporated in the survey tool. Discrepancies lie the wording of survey questions and indicator framing. Our estimates of exclusive breast feeding are based on two questions in contrast to other surveys: (1) women who reported practicing exclusive breast feeding in the first 6 months of their child’s life and (2) women who report never having fed their child janam ghutti, water, honey, animal milk, formula milk, jaifal or tea in the first 6 months of life. Across study arms, only 10.8% of women in the intervention arm and 10.4% of women in the control arm met this standard. Arguably this estimate is a more reliable indication of the true prevalence of exclusive breast feeding in this population.

    In contrast to infant feeding, Kilkari had a significant impact on the secondary outcome of modern reversible contraceptive use. Family planning messages constituted the largest overall content area (18%) and were provided throughout the extended postpartum window. The impact on modern reversible contraceptive use was driven by increases in condom use and oral contraceptives, and as noted above, occurred concurrently with a decrease in sterilisation. A multitude of factors may explain these results. First, 49% of women in our sample reported sharing mobile phones and among women who did not own their own phones (25%), it was often their husbands who listened to Kilkari calls.23 Exposure to Kilkari calls among men may thus have catalysed some of the increase observed. Second, the study population was characterised by high economic migration among men; a factor which too may have influenced women’s decision-making on method preferences and the observed shift from permanent to reversible methods. Lastly, increases were significantly higher among subgroups including those with any male children. Given social norms around gender preferences for children, Kilkari may have served as a ‘tipping point’ to prompt behaviour change among those who had already had a male child. Indeed, qualitative findings suggest that men and women retained and appreciated Kilkari messages that aligned with their pre-existing worldviews, social norms and existing practices but overlooked or de-emphasised content that did not.23

    Comparison with findings elsewhere

    Our study is the largest effectiveness trial conducted globally to date of a mHealth messaging programme operating at scale. Comparable RCTs of mHealth messaging programmes in other low resource settings were completed as part of pilot studies or smaller deployments in Africa.8–11 The changes in programme features and design which frequently occur as programmes transition from small scale pilots to large scale deployments can lead to a voltage drop in their effectiveness at scale and have implications for the generalisibiltiy of results.2 15 By starting with a programme already firmly established and scaled, we have sought to improve emerging evidence on the generalisability of such interventions. By recruiting study participants directly from their communities, we have sought to test the intervention under ‘imperfect’ real world conditions ensuring that findings are generalisable to and representative of the population at large with access to mobile phones. Further efforts to power the sampling frame to detect a 5% difference across study arms in the primary outcomes of reversible contraceptive method use and exclusive breast feeding sought to ensure that modest differences in outcomes would be detected. While challenges with exposure exceeded expectations (and are discussed below), subgroup analyses on key population subgroups were particularly critical to understanding the differential effect programme activities may have had. Elsewhere we continue to explore equity considerations in greater depth.22 Analyses of call data records made possible by MOHFW and with the support of technical partners including BBC Media Action, Beehyv and IMI Mobile were integral to our efforts to determine exposure to specific subsets of Kilkari messages by outcome and explore linkages between dosage of listening and reported health outcomes. Ours is the first evaluation of a mHealth messaging programme to make this link.

    The RCT presented here is part of a large mixed methods evaluation of Kilkari.17 Findings add to the limited evidence available on the effectiveness of large-scale health information messaging programmes4 7 12 and smaller scale initiatives in Malawi8 and Zanzibar.9–11 However, comparisons with other programmes are challenging given the fundamental differences in programme components, implementation partners, study contexts, scale of implementation and evaluation design. Kilkari was designed as an integrated component of a suite of mHealth services that involved training of frontline health workers (Mobile Academy), equipping them with job aids (Mobile Kunji), and reinforcing health information provided to beneficiaries through contacts with the public health system (Kilkari).2 16 However, the programme was ultimately scaled as a standalone service with key features, including the timing of calls, modified based on what was feasible to implement at scale.26 By comparison, programmes in Bangladesh, Malawi, Zanzibar and elsewhere in India, all feature mHealth information messaging as part of a larger suite of services, often enabled by frontline health workers who provide face-to-face communication services, supported by an implementing non-government organisation with donor funding. The sole exception is South African National Department of Health’s MomConnect programme which does include a HelpDesk27 component but principally is centred around the delivery of 145 twice weekly text messages to new and expectant mothers.28

    Globally, the future of mHealth messaging programmes remains uncertain. While several programmes have scaled to reach large numbers of subscribers,4–7 28 sustainability remains a continued challenge. While a number of factors underpin this, key drivers include dependency on donor funding, limited government capacity for oversight and stewardship, and lack of evidence on impact and value for money.13 The emergence of COVID-19 has further increased demands on already finite health resources. Despite the uncertainties, programmes like Kilkari, which have the infrastructure in place to send health information content out to millions of beneficiaries, offer much potential for delivering other kinds of health information, including information related to COVID-19. More broadly, our findings underscore the potential impact mHealth messaging programmes may have in providing women and their families with access to health information. By not restricting the framing of Kilkari messages to women, the programme may have tapped into a demand among men for health information. Further, efforts to ensure that Kilkari content was accessible regardless of subscribers’ education, caste or socioeconomic status, may underpin pro-poor findings observed for reversible contraceptive method use. Long-standing social norms, particularly those around son preference, are challenging to change. However, efforts to differentially target key population segments whom we know services are likely to impact the most may be one future avenue to explore.

    Limitations

    Evaluation activities have several notable limitations. When compared with pregnant women at a population, our sample of women with access to a mobile phone during the day is likely more advantaged; a structural bias most mHealth initiatives face. However, when compared against women enrolled to the Kilkari programme elsewhere in India, the RCT sample may be different for two reasons. First, our RCT sample was identified via a household listing and screening survey where an estimated 50% of the pregnant women at a population level were eligible since they had access to a phone during the day. This exceeds the average population level coverage attained by Kilkari across 13 states in India (21%) and elsewhere in Madhya Pradesh (18%) which identifies subscribers passively through government tracking registries. Our RCT thus included a higher proportion of the population by including women who would not necessarily have been registered in the government’s tracking registry.29 The RCT faced challenges in ensuring continuous exposure to Kilkari messages among those in the intervention arm as a result of high rates of SIM change (44%), male population migration (potentially with their phones; 25% of men could not be interviewed as part of the endline survey), and increases in call non-delivery rates attributed to the bankruptcy of the telecommunications provider for the Kilkari programme, in mid-2018. All of these factors exceeded initial expectations and resulted in a loss of power due to lower compliance and programme exposure. A third limitation of the study lies in our reliance on self-reported outcomes for all outcomes except immunisations, many of which may be subject to social desirability bias. A fourth limitation lies in the information and recall biases associated with collecting data on practices occurring over a long window at a single time point (12 months postpartum). Additionally, we were not able to measure outcomes related to self-efficacy to practice RMNCH behaviours. Lastly, there may be unmeasured confounders but we assume the randomisation process distributed them evenly across the two arms.

    Conclusions

    Study findings provide the most conclusive evidence to date on the effectiveness of a mHealth messaging initiative at scale in India and globally. The moderate impact observed on modern reversible contraceptive method use, particularly among the most marginalised, is promising while the lack of any measurable changes in infant feeding behaviours is disappointing. Further research is needed to understand whether the impact might be deepened through microtargeting-specific subsets of the beneficiary population with tailored programme content; by changing the number of messages and the duration of the service; and by complementing Kilkari with supplementary face to face communication to create an enabling environment more supportive of behaviour change.

    Supplemental material

    Data availability statement

    Data are available upon reasonable request. Anonymised data are available upon reasonable request from the study Principal Investigator Dr. Amnesty LeFevre aelefevre@gmail.com.

    Ethics statements

    Patient consent for publication

    Ethics approval

    This study involves human participants and was approved by Institutional Review Boards from the Johns Hopkins Bloomberg School of Public Health in Baltimore, Maryland USA and Sigma Research and Consulting in Delhi, India provided ethical clearance for study activities. Verbal informed consent was obtained from all study participants.

    References

    Supplementary materials

    • Supplementary Data

      This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Footnotes

    • Handling editor Sanni Yaya

    • Twitter @jbashingwa, @ac1112, @priydee

    • Collaborators The Kilkari Impact Evaluation Team* (listed in Alphabetical order) Smisha Agarwal, Salil Arora, Jean JH Bashingwa, Aarushi Bhatanagar, Sara Chamberlain, Rakesh Chandra, Arpita Chakraborty, Neha Dumke, Priyanka Dutt, Anna Godfrey, Suresh Gopalakrishnan, Nayan Kumar, Simone Honikman, Alain Labrique, Amnesty LeFevre, Jai Mendiratta, Molly Miller, Radharani Mitra, Diwakar Mohan, Deshen Moodley, Nicola Mulder, Angela Ng, Dilip Parida, Nehru Penugonda, Sai Rahul, Shiv Rajput, Neha Shah, Kerry Scott, Aashaka Shinde, Aaditya Singh, Nicki Tiffin, Osama Ummer, Rajani Ved, Falyn Weiss, Sonia Whitehead.

    • Contributors AEL is the overall study PI and guarantor. She accepts full responsibility for the work and conduct of the study, had access to the data, and controlled the decision to publish. She additionally helped to secure the funding, led the design of the study tools, supported oversight of field work and analysis and wrote the first draft of the manuscript. NS inputted on the study tools, supported data monitoring of the field work, conducted the data analysis, helped to draft the data tables and edited the manuscript. KS helped to secure the funding, design the study tools, support the field work, input on analyses and edit the manuscript. SC helped to secure the funding, draft and review study tools, interpret data analyses and study findings, and edit the manuscript. OU help to revise study tools, support field work including training and monitoring survey data quality, input on analyses and edit the manuscript. JJHB help to support monitoring survey data quality, input on analyses and edit the manuscript. AC helped to draft and review study tools; support data collection, including training, monitoring; interpret data analyses and study findings, and edit the manuscript. AG, PD and RV helped to draft and review study tools, interpret data analyses and study findings, and edit the manuscript. DM helped to secure funding, helmed the study design including sampling and randomisation, helped draft study tools, led the data analysis and edited the manuscript.

    • Funding Bill and Melinda Gates Foundation (OPP1179252).

    • Competing interests All authors have completed the Unified Competing Interest form (available on request from the corresponding author) and declare that the research reported was funded by the Bill and Melinda Gates Foundation. AG, SC, PD are employed by BBC Media Action; one of the entities supporting program implementation. The authors do not have other relationships and are not engaged in activities that could appear to have influenced the submitted work.

    • Patient and public involvement Patients and/or the public were involved in the design, or conduct, or reporting, or dissemination plans of this research. Refer to the Methods section for further details.

    • Provenance and peer review Not commissioned; externally peer reviewed.

    • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.