Background: There are high expectations that mobile health (mHealth) strategies will increase uptake of health care services, especially in resource strained settings. Our study aimed to evaluate effects of an mHealth intervention on uptake of maternal health services. Methods: This was an intervention cohort study conducted at six public antenatal and postnatal care clinics in inner-city Johannesburg, South Africa. The intervention consisted of twice-weekly informative and pregnancy stage-based maternal health information text messages sent to women during pregnancy until their child was one year of age. The intervention arm of 87 mother-infant pairs was compared to a control arm of 90 pairs. Univariate and multivariate analyses were used to compare the probability of the outcome between the two groups. Results: Intervention participants had higher odds of attending all government-recommended antenatal and postnatal visits, all recommended first year vaccinations (OR: 3.2, 95% CI 1.63–6.31) and had higher odds of attending at least the recommended four antenatal visits (OR: 3.21, 95% CI 1.73–5.98). Conclusion: We show an improvement in achieving complete maternal-infant continuum of care, providing evidence of a positive impact of informative maternal mHealth messages sent to pregnant women and new mothers. Trial registration ISRCTN, ISRCTN41772986. Registered 13 February 2019—Retrospectively registered, https://www.isrctn.com/ISRCTN41772986.
This multi-centre cohort intervention study ran from May 2014 to June 2015 and included maternal-infant pairs recruited from six participating public healthcare facilities offering ANC and PNC/EPI services in the inner city of Johannesburg and Hillbrow, urban neighbourhoods within Johannesburg’s inner-city. These neighbouring areas have a high population density, are predominantly low-income and have high rates of alcohol abuse, gender-based violence, unemployment (estimated at 23% in 2013) and HIV (27% HIV positivity among pregnant women in 2013) [22, 23]. There are 16 public healthcare facilities offering ANC and PNC/EPI services in inner city Johannesburg, and Mobile Alliance for Maternal Action (MAMA) was offered in six of these. Among these six, three were selected as intervention facilities, and another three were selected as control facilities from the 10 not offering the MAMA intervention. All sites provided standard ANC and PNC services to study participants and were purposively selected based on client similarity and proximity to each other. The intervention consisted of free one-way maternal health SMSes sent twice weekly throughout pregnancy and for one year postnatally. The SMSes, which contained supportive and informative information timed to the stage of pregnancy and age of the child, sent as part of the Mobile Alliance for Maternal Action (MAMA) South Africa project [24]. The SMS content was initially drafted by BabyCentre UK and then customised for the South African context by a team of local maternal and infant health professionals. The SMSes covered a range of maternal and infant health topics such as healthy eating, reminders to go for ANC/PNC appointments, psycho-social support, PMTCT support messages (if HIV-related messages were requested) and delivery planning (for examples, see Appendix). The intervention was offered to all pregnant women receiving ANC care at the intervention sites and supplemented the clinical standard of care offered. In this setting 98.4% of households owned a mobile phone [25] and almost all the study women had their own phone that they received the intervention messages on. Intervention participants joined the SMS intervention between their 11th and 39th week of pregnancy, thereby receiving between two and 28 intervention SMSes before delivery. An additional 104 messages were sent postnatally, and included reminders for each vaccination during the first year. Study recruitment was initiated two years after the SMSes were first offered. Intervention participants were identified from a list of the SMS recipients who had received the full year of postnatal messages. All women for whom the telephone number was listed, were contacted by phone and invited to participate. Each woman was called up to five times on separate days if there was no answer to the phone call. Control arm participants were identified while they were receiving PNC services at a control recruitment site, screened for eligibility and invited to participate. All women within the mother–infant pairs in both the intervention and control groups were required to be over the age of 18 at recruitment, to have received ANC and PNC services at one the participating ANC/PNC sites between July 2012 and June 2014, to have delivered with a skilled birth attendant at one of two participating delivery sites, and to have had regular access to a cellular phone. All participants were also required to attend a face-to-face interview and provide their infant’s Road to Health (RTH) monitoring booklet. The primary outcome was the proportion of mother–child pairs who would receive comprehensive maternal, neonatal and infant care. The aim was to include as many as possible of the women who had signed up for the intervention. Due to the low number of women who could be included, we made a post hoc sample size calculation. The sample size was based on complete EPI coverage at one year of age as no previous data for the composite score could be found. No reliable local data were available, so we used the 2013 WHO data on South Africa’s measles vaccination coverage rate at one year of age, which was 66%, as a baseline [26]. To identify an increase in coverage from 66 to 86%, the minimum required for herd immunity from most childhood vaccines [27], at 80% power and 95% confidence, a sample size of 68 individuals per arm was identified [28]. Socio-demographic data were collected during participant interviews. ANC attendance data were collected from clinical ANC records and EPI coverage data were collected from infant RTH booklets. All study data were digitised and stored using Research Electronic Data Capture (REDCap), hosted at the University of Witwatersrand. REDCap is a secure, web-based application designed to support data capture for research studies [29]. Across the three intervention sites, 1770 women signed up to receive the SMSes. Of those, 379 (21.4%) could be reached by phone (for the rest there was no telephone number or the person did not respond to the phone call) and were invited to participate in the study of which 181 (47.8%) showed up for the interview. In the control arm, 290 participants were identified and invited to participate while 175 (60.3%) attended the interview. Just over half of the 356 women interviewed (n = 179, 50.3%) had missing ANC records and were excluded from the analysis (see Fig. 1). Complete data for all outcomes was available for a total of 177 individuals; 87 in the intervention arm and 90 in the control. Participant flow diagram Continuum of care is typically defined by the data available in individual studies [30, 31]. The study team defined the primary outcome as a binary composite of two maternal and infant indicators; having had four or more ANC visits (indicator 1), and receiving all first-year infant vaccinations at one year of age (indicator 2). These indicators are based on recommendations from the South African National Department of Health maternal health and EPI guidelines [32, 33]. Delivery with a skilled birth attendant, a common indicator within other maternal health continuum of care work [31], was not included as it was part of the study inclusion criteria. Notably, rates of facility-based births in South African urban settings are very high; recently estimated to be 99% in Gauteng province, where Johannesburg is located [34]. Secondary outcomes included attendance to at least two, three, four and five ANC visits, the mean number of ANC visits attended, and mean vaccination coverage of the two groups. The study aim was not to focus on HIV-related outcomes in part due to data on a related HIV-positive cohort having been published elsewhere [35]. Chi-square tests were conducted on binary variables, Student’s t-tests were used for continuous variables and univariate and multivariate analyses were used to compare the probability of the outcome between the two groups. Pre-existing socio-demographic cohort differences were identified and adjusted for in the model, as noted. All data analysis was conducted using Stata version 13 [36] and statistical significance was considered at p < 0.05.
N/A