Implementation of Maternal and Newborn Health Mobile Phone E-Cohorts to Track Longitudinal Care Quality in Low- and Middle-Income Countries

Country-Specific Implementation Approaches

The MNH eCohort teams in each of the 4 implementing countries tailored the eCohort to that country’s context. Table 2 summarizes the country-level adaptations to the methodology regarding data collection strategies, enumerators, languages, incentives, and mobile phone distribution. Country teams reviewed all modules to ensure adherence to national guidelines and other contextual factors (using local language and phrases, standard treatment guidelines, and locally available medications).

Selection of Implementation Sites

In each of the 4 countries, 2 sentinel sites were selected using a slightly different rationale (Table 3). The eCohort was not intended to be nationally representative. Instead, it aimed to provide actionable evidence for health system improvement in selected sites that represent a typology of health system challenges throughout the country. We initially aimed for 1 site to be predominantly urban and the other predominantly rural. In Kenya and South Africa, sites were selected based on poor MNH outcomes. In Ethiopia, health outcomes and prior relationships with health system actors, as well as availability of electricity and network coverage, determined the site selection. India chose 2 sites, including 1 with relatively better outcomes, to assess if care competence could explain differences in outcomes. Both sites in India include a mix of urban and rural.

Sampling and Recruitment

We aimed to recruit women who were representative of health system users in the selected sites. This was achieved through the selection of facilities included as enrollment sites, as well as the decisions (discussed in later sections) regarding the distribution of mobile phones to women participating in the eCohort. A total sample of approximately 1,000 women was recruited per country, with 500 women enrolled in each of the 2 sites. Women were approached for enrollment in the MNH eCohort if they presented for their first ANC visit at an enrolling health facility during the data collection enrollment period. Eligibility criteria included: (1) being at least age 15 years in Ethiopia and Kenya, 18 years in India, or 15 years in South Africa with parental consent to participate); (2) being pregnant and attending their first ANC visit; (3) planning to continue to receive maternal health care in the same subnational area; and (4) being willing to be contacted on the phone every month for the duration of the study. Exclusion criteria included being at the health facility for a follow-up/subsequent ANC visit; being at the health facility for another health concern or to accompany someone else; and any physical, mental, or cognitive disability that would prohibit the ability to provide informed consent. Sample size justification is provided elsewhere.

Within each site, a master facility list—including private facilities—was obtained, and facilities were stratified by type, including public primary facilities, public secondary (hospitals), private primary facilities (e.g., clinics, maternal and child health [MCH] specialty centers or MCH specialty clinics), and private hospitals. In each site, we also obtained data on the volume of ANC visits by facility type (using DHS or health management information system data). Care-seeking patterns across the 4 strata (i.e., proportion of women who use public primary, public secondary, private primary, or private secondary facilities for ANC) dictated the sampling strategy. We aimed to include health facilities that would be representative of ANC utilization patterns at the site. The number and types of facilities selected for enrollment are described in Table 3.

In India and South Africa, private-sector facilities were excluded due to refusal to participate in the study. In Ethiopia, 2 private facilities, 1 MCH center, and 1 hospital were dropped due to a lack of ANC first clients for the first week of data collection. Enrolling women at the secondary level was also limited, given the fact that few women initiate care for ANC at the secondary level but may be referred to higher-level care throughout their pregnancy.

Collaboration with subnational authorities and the facility in-charge was an important step to ensure participation of health facilities as enrollment sites. In all sites, inception visits with the subnational authorities and at selected health facilities built trust and fostered collaboration.

Data Collection

The mixed-mode nature of the eCohort requires 2 distinct types of data collection: in-person data collection at baseline and endline and computer-assisted telephone interviewing for the rest of the survey.¹⁰ The number and background of enumerators employed for the in-person vs. phone data collection varied across the 4 countries (Table 2). All data collectors had a minimum of secondary-level education, had previous data collection experience, and were fluent in all necessary local languages. In Ethiopia and India, data collectors had a degree in health, which includes nursing or master’s degrees, though such advanced degrees were not a requirement. In Kenya, in-person data collectors had nursing degrees, and phone-based data collectors were required to have a bachelor’s degree. In South Africa, data collectors for in-person surveys resided in the selected sites.

In-Person Recruitment

Mainstage data collection began in May 2023 in Ethiopia, June 2023 in Kenya, May 2023 in South Africa, and October 2023 in India. Data collectors were assigned to participating health facilities to enroll women and complete the baseline survey. During the enrollment period, they were posted to a health facility and enrolled eligible women until the desired number of women at each health facility (based on ANC volume and DHS utilization estimates) was reached.

In-person data collection for the baseline survey was needed to identify participants, collect data on the health facilities where women were recruited, and establish a relationship between the study and the participants. The in-person recruitment also included health assessments (described later) and a review of health cards. During this initial contact, data collectors explained and answered questions about the study, obtained informed consent, and distributed phones to women.

A central component of the in-person data collection was to gather sufficient information to allow tracing of enrolled women during the phone follow-up. As part of the baseline interview, data collectors gathered as many phone numbers as possible from each woman, as well as information about her residence. On average, women provided 1.6 phone numbers in Ethiopia and 2.7 in Kenya. Additional phone numbers may have been a secondary phone that she owned, a partner’s phone, or a friend/neighbor’s phone. During the in-person baseline, data collectors also established a preferred time of day for follow-up phone calls.

Mobile Phone Follow-Up and Call Tracing

Between the in-person baseline and endline surveys, women were followed up by phone calls made at approximately monthly intervals. These interim surveys aimed to ask participants about additional care that they had received, the content of that care, the women’s perception of that care, and their health status. For the purposes of the MNH eCohort, computer-assisted telephone interviews were the preferred data collection approach because it helped retain women in the cohort by sustaining rapport and relationships, facilitated participation by all women (including women who were illiterate, compared to text message responses), and ensured the accuracy of data collection (compared to interactive voice response). Building rapport with participants could also be achieved by having the same data collector call participants over the course of the study, which also appeared to have improved retention—a strategy employed by Kenya, Ethiopia, and South Africa to date. At the end of each call, data collectors confirmed the preferred phone number and time for the follow-up call. The preferred time of call for women changed throughout the duration of this study. In the calls during her pregnancy, she may be working, studying, or carrying out other duties, but after the birth of the baby, the preferred time of call was likely to shift. Data collectors were responsive to these changes in the demands of the participants’ time and schedule and conducted calls at a convenient time for participants.

Data collectors followed clear procedures during phone data collection (Box). Previous experience with phone surveys indicates collecting multiple phone numbers is a best practice. In the experience of the MNH eCohort, women were typically reached on their primary phone number, which the woman owned, and frequent calling of additional numbers was seen as a nuisance. However, in the unfortunate event of a maternal death, which occurred 3 times so far, those were identified by calling additional phone numbers (of husbands or sisters).

If a woman was reached on a number that was not noted as her primary phone number, data collectors confirmed if the preferred number and time of day had changed. If no contact was made with the woman in a particular month, she was noted as missed for that month, and data collectors attempted to call again 1 month from the date of the first call made. Women were considered lost-to-follow-up in the eCohort if they went 2 months without any contact, and data collectors stopped attempting to call that woman.

The phone-based data collection was an important component of the eCohort and required substantial infrastructure to carry out the many calls required in this study. In India and Ethiopia, the eCohort used an existing call center within the research team. In South Africa, phone data collection infrastructure was established for the first time. In Kenya, phone data collection was contracted out to an external firm. We found that having prior experience with phone-based data collection was very helpful for smooth implementation of the eCohort. Phone-based data collection required significant investment in infrastructure (phone, computers, headsets, staff, integrated call management systems, and Internet), quality control systems, and staff capacity. Call center operations also relied on strong power and network connections, and if those failed, interruptions to study calls occurred. In the event of these interruptions, alternative methods for calls (i.e., using mobile phones) were used.

Attrition Management and Incentives

To ensure ongoing participation in the study, it was critical for participants to have sufficient airtime to keep their phone lines open and able to receive calls.¹⁰ Different incentive approaches were applied by each country regarding airtime (Table 2), but generally, each country credited each woman with a small amount of airtime upon completion of each phone interview. In Kenya, if a woman did not respond to the calls, a text message reminding her of the study interview and airtime top-up was sent, which often resulted in a completed survey.

Ethiopia completed the eCohort in February 2024, and Kenya completed the eCohort in July 2024. In Ethiopia, of 1,000 women, 92% completed at least 1 instance of module 2 (additional ANC care during the antenatal period), 89% completed module 3 (delivery care and delivery outcomes), and 83% of women completed the endline survey that concludes participation in the eCohort (module 5). In Kenya, of 1,010 women, 90% completed at least 1 instance of module 2, 88% of women were followed throughout the pregnancy (module 3), and 86% of women completed the endline survey (Figure). Data collection in India and South Africa is ongoing.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE

Percentage of Women Who Completed Modules of the MNH eCohort in Ethiopia and Kenya

Women who have dropped out of the study cite fatigue and that the calls take up too much time as reasons for dropping out. To date, refusals and loss-to-follow-up are not outside of the expected range.¹⁴ Calculating loss-to-follow-up with the study still ongoing is difficult but will be carefully analyzed upon completion of the cohorts and in all subsequent analyses.

Throughout the eCohort, pregnancy loss, neonatal death, and infant death were inevitable outcomes experienced. The eCohort defined stillbirths as fetal deaths that occurred at 28 weeks or more of gestation and miscarriages as fetal deaths that occurred before 28 weeks of gestation. Women who experienced these outcomes were asked to respond to a series of relevant questions (e.g., in case of stillbirth, questions focused on the content and quality of intrapartum care; in case of neonatal or infant death, questions covered additional care-seeking and cause of death). After experiencing any of these outcomes, women completed their study participation and were no longer contacted for further follow-up interviews. Retention rates reported here include these women who concluded their survey participation early. About 5% of women experience these outcomes; 54 women in Ethiopia and 55 women in Kenya ended their participation during rounds of module 2 or after module 3 after a miscarriage, stillbirth, or neonatal and infant death.

In-Person Endline Survey

The in-person endline survey facilitated a second round of health assessments, including weight and head circumference measurements for the babies and repeats of the health assessments taken at baseline for the women. The in-person endline also concluded the women’s participation in the survey and provided an opportunity to distribute incentives. Endline surveys are currently ongoing, but contextual factors appear important to consider when planning the endline. For example, in Kenya, women typically returned to their maternal village after delivery and would be out of the study area during the endline data collection. In Ethiopia, confinement periods may have impacted the ability to carry out in-person endline data collection. Where baseline interviews took place at health facilities, endline surveys could take place in a more central location, not necessarily at the health facility, and on a more flexible schedule. To accommodate women who were out of the study period, endline surveys could also be conducted on the phone to gather a few final key measures and forego the health assessments.

Procurement and Budget Implications

An understanding of the procurement and budgetary implications of this study is important. Knowing the cost drivers in these first 4 countries can help future implementers consider how the eCohort might be adapted or implemented in varying contexts with varying budgets and resources. In each of the 4 countries, the MNH eCohorts were completed with a budget of US$200,000. Funds were distributed from QuEST Harvard to our network partners in the study countries. Teams allocated the funds as they saw fit for their context. In Kenya, a data collection firm with experience administering phone-based surveys was hired to carry out all data collection except the baseline and endline surveys. Costs associated with analysis and dissemination are not included here.

Table 4 shows the budget breakdown across key categories, including staff to oversee study implementation (including programmers to set up data collection instruments and databases), data collector salaries, costs associated with in-person data collection, procurement (phones, tablets, data collection equipment, and incentives provided to women), costs associated with administering phone surveys (establishing a call center or contracting to a firm), phones/airtime, and other study requirements.