The effects of a locally developed mHealth intervention on delivery and postnatal care utilization; A prospective controlled evaluation among health centres in Ethiopia

Background: Although there are studies showing that mobile phone solutions can improve health service delivery outcomes in the developed world, there is little empirical evidence that demonstrates the impact of mHealth interventions on key maternal health outcomes in low income settings. Methods: A non-randomized controlled study was conducted in the Amhara region, Ethiopia in 10 health facilities (5 intervention, 5 control) together serving around 250,000 people. Health workers in the intervention group received an android phone (3 phones per facility) loaded with an application that sends reminders for scheduled visits during antenatal care (ANC), delivery and postnatal care (PNC), and educational messages on dangers signs and common complaints during pregnancy. The intervention was developed at Addis Ababa University in Ethiopia. Primary outcomes were the percentage of women who had at least 4 ANC visits, institutional delivery and PNC visits at the health center after 12 months of implementation of the intervention. Findings: Overall 933 and 1037 women were included in the cross-sectional surveys at baseline and at follow-up respectively. In addition, the medical records of 1224 women who had at least one antenatal care visit were followed in the longitudinal study. Women who had their ANC visit in the intervention health centers were significantly more likely to deliver their baby in the same health center compared to the control group (43.1% versus 28.4%; Adjusted Odds Ratio (AOR): 1.98 (95%CI 1.53-2.55)). A significantly higher percentage of women who had ANC in the intervention group had PNC in the same health center compared to the control health centers (41.2% versus 21.1%: AOR: 2.77 (95%CI 2.12-3.61)). Conclusions: Our findings demonstrated that a locally customized mHealth application during ANC can significantly improve delivery and postnatal care service utilization possibly through positively influencing the behavior of health workers and their clients. The study was conducted in Semen Shewa Zone, Amhara region, central Ethiopia in a total of 10 health facilities (5 intervention, 5 control) serving around 250,000 people. Health facilities within 10 km of the main road connecting Addis Ababa to North East Ethiopia were included for both intervention and control arms of the study. The main purpose of limiting the study sites within the above distance was to ensure that all study participants had sufficient mobile phone network connectivity and similar physical access to health facilities. Health centers in the non-bordering districts (more than 50 km away from the nearest intervention health center) were considered as control to minimize intervention contamination (the possibility of mothers in one of the intervention groups moving to the other health centers or vice versa because of perceived improved care in the intervention areas). These health centers were also within 10 km of the main road. The study was an interventional study by design with intervention and control health facilities with comparable endpoints. The study used two approaches to gather data from women visiting the health facilities for obstetric care; It is important to note that women followed in the longitudinal study are not necessarily the same women who participated in the cross-sectional surveys. The sampling procedure and follow-up schedule is shown in Fig 1. The health workers (nurses/health officers—A nurse and health officer are health professionals with at least 2 and 4 years of training respectively and providing maternal and child health care is one their key activities. -) in the intervention group received an android phone loaded with an application that had four major features described below. First health workers were requested to register all clients who came to their health facility for ANC, delivery care or PNC using the same forms they would otherwise fill in using the paper based registry book. The application generated a unique identification code based on the names and other background information of each pregnant woman (Kebele which is comparable to ‘village’ and is the smallest administrative unit in Ethiopia where approximately 5,000 people live, House number) allowing longitudinal follow up of each pregnant woman and sending a reminder to the health worker for each follow up visit. A scanned copy of the obstetric form is shown in S1 supporting information. The project provided an air-time worth 10 US dollars per month for each health worker to cover the cost for submitting the forms and calling pregnant women ahead of their scheduled visits. Health workers were able to fill in the forms off-line and needed the cellular connection only to submit the forms and receive the reminders/educational messages and generate report on service utilization. The software application (an open-source package called Open Data Kit (ODK—available on https://opendatakit.org/)) was adapted locally by a team of IT professionals working at Addis Ababa University and was continuously modified based on the actual needs of the health workers. Complete data were transmitted to the central server at Addis Ababa University using the cellular network of the phone. The application can be shared with interested users upon request. The effectiveness of the intervention was evaluated by comparing the differences in the following 3 indicators in the intervention and control health centers after 12 months of implementation of the intervention; Information regarding the 3 outcome measures was obtained through a facility record review using a standard checklist. To achieve the required sample size maternity service utilization data for 6 months (July to December 2014) were compared. Intervention started in February 2014. Sample size was determined based on the following assumptions; Proportion of pregnant women who had institutional delivery at birth among those who had at least one ANC visit—p1 = 21% [25] as control and taking a 10% difference in proportion between the intervention and control areas to signify public health importance (p2 = 31%); α error (two sided)– 0.05; 90% power; n2/n1 (ratio between the intervention and control areas)– 1:1. With a 15% upward adjustment for possible non-response, the sample size was computed approximately as n1 = n2 = 500. A total of 15 Health workers (three per health center) received a two-day initial training and then quarterly (every 3 months) a refresher training (2 days each) on the mobile phone application, its potential use to improve data management and maternity care as well as the central role of client-centered care in ensuring quality care. The training focused on three main areas of maternity care: antenatal care (including birth preparedness and complication readiness), delivery care and postnatal care. The principal investigator and the IT team (2–3 people) gave the trainings. Considering the median duration of pregnancy at first ANC visit for urban Ethiopia (at 4.4 months) [4], it was necessary to follow the cohort of pregnant women who had their first visit when the study began for at least 5–6 months to continue the observation until delivery and immediate postpartum. The IT staff provided solutions in situations where health workers accidentally deleted electronic forms or faced technical challenges to transmit data to the central server. First we compared the baseline and follow-up characteristics of antenatal care clients in the intervention and control health centers using X2 tests. We ran logistic regression models with intervention status as primary independent variable and the primary outcomes (having 4 ANC visits, Institutional delivery and PNC visit within 6 hours of delivery) as dependent variables, with age, residence and parity as co-variates. To examine the possibility of ‘dose-response’ relationship, differences in the primary outcomes (Institutional delivery and PNC visit) between women who had one and four ANC visits were compared using the 95%CI for the difference of proportions. The criterion for significance was set at α = 0.05. All analyses were conducted with STATA version 12. Respondents gave informed verbal consent after they were informed about the purpose of the study as described below. Data collectors read the information sheet and consent form which explained the objectives and benefits of the research, that their participation should be voluntary, confidentiality of information collected, their right to withdraw from the study without any consequence, and the minimum risk involved by participating in the study. Study participants were also given the contact addresses (phone numbers and email addresses) of the principal investigator and IRB chairperson in case they have questions to clarify. Once potential respondents decided to participate in the study, data collectors signed on the consent form and proceeded with the interview. The research protocol was approved by the Institutional Review Board(IRB) of the College of Health Sciences at Addis Ababa University (Protocol number 040/12/SPH). The study was financially supported by United Nations Population Fund (UNFPA) Ethiopia Country Office.