Cluster-Randomized Controlled Study of SMS Text Messages for Prevention of Mother-to-Child Transmission of HIV in Rural Kenya

Background. Antiretroviral medications are key for prevention of mother-to-child transmission (PMTCT) of HIV, and transmission mitigation is affected by service delivery, adherence, and retention. Methods. We conducted a cluster-randomized controlled study in 26 facilities in Nyanza, Kenya, to determine the efficacy of SMS text messages on PMTCT outcomes. The relative risk and confidence intervals were estimated at the facility level using STATA. Results. 550 women were enrolled, from June 2012 to July 2013. The median age was 25.6 years, and 85.3% received ARVs. Maternal ARV use was similar between the intervention and control arms: 254/261 (97.3%) versus 241/242 (99.6%) at 34-36 weeks of gestation and 234/247 (94.7%) versus 229/229 (100%) at delivery. Among infants, 199/246 (80.9%) and 209/232 (90.1%) received ARVs (RR: 0.91; 95% CI: 0.77-1.14); 88% versus 88.6% were tested for HIV at 6 weeks, with 1/243 (0.4%) and 3/217 (1.4%) positive results in the intervention and control arms, respectively. Communication increased in both the intervention and control arms, with the mean number of 7.5 (SD: 5.70) compared with 6 (SD: 9.96), p<0.0001. Conclusions. We identified high ARV uptake and infant HIV testing, with very low HIV transmission. Increased communication may influence health-seeking behaviors irrespective of technology. The long-term effectiveness of facilitated communication on PMTCT outcomes needs to be tested. The study has been registered on ClinicalTrials.gov under the identifier NCT01645865.

This study was designed to test the hypothesis that reinforcement of key messages delivered via SMS would result in improvement in completion of key PMTCT cascade milestones. SMS text messages were designed based on our formative research conducted among HIV-positive women in the PMTCT program and community health workers [14]. Primary outcomes were antiretroviral uptake at 34–36 weeks of gestation and at delivery. Secondary outcomes included facility delivery, infant's uptake of HIV prophylaxis, infant HIV testing at 6–8 weeks after delivery, and HIV transmission. Women identified with HIV infection during antenatal care (ANC) receive either antiretroviral prophylaxis with zidovudine (AZT) starting in the second trimester of pregnancy or combination antiretroviral therapy (ART), with a CD4+ T-lymphocyte count less than 350 per cu mm used to determine the need for ART [15]. During the course of this study, guidelines evolved to expand HIV treatment to all HIV-positive pregnant women irrespective of CD4+ T-lymphocyte count, and this guidance was implemented in the high volume facilities included in this study [16]. All infants receive nevirapine prophylaxis from birth until 6 weeks postpartum, extended throughout breastfeeding if the mother is not receiving ART. Infants are tested for HIV using DNA PCR testing at 6 weeks postpartum. Community health workers (CHWs) serve as health extension workers in PMTCT programs in rural Kenya and are integral members of the communities in which they serve. CHWs ensure that women follow up in ANC, provide ongoing pregnancy and newborn health education to women, and encourage adherence to antiretrovirals. In this cluster-randomized study, 26 of the 54 health facilities within Homa Bay and Rachuonyo districts in Nyanza province were selected. Ten facilities serving <20 HIV-positive pregnant women yearly were excluded due to anticipated low enrollment. A volume-stratified sampling method was applied to ensure inclusion of a representative sample of types of health facilities among the remaining facilities. Two hospitals in Homa Bay were selected, and two of the four hospitals in Rachuonyo district were randomly selected and included. Eight health centers and fourteen dispensaries were randomly selected from the remaining forty health facilities. All health facilities were randomly allocated to be an intervention or control site, stratified by high volume (hospitals) and medium and low volumes (health centers and dispensaries). Consecutive eligible HIV-positive pregnant women presenting for ANC were invited to participate. Women were eligible to enroll in the study if they were less than 32 weeks of gestational age, were not currently receiving antiretroviral therapy, were planning to remain in the area for the duration of the study period, and agreed to follow-up of their infants until 6 weeks following delivery. Only women in the intervention arm were required to have access to a mobile phone (individual or shared) and be willing to receive SMS text messages. Male partners of women were also permitted to enroll in the study. An mHealth platform was designed and built in collaboration with DewCIS Solutions Ltd., a Kenya-based Information Technology company. A mobile web-based communications software system allowed for semiautomated delivery of preloaded messages. Participants were allocated to list-serves based on pregnancy stage to ensure delivery of appropriate messages. Participants were able to communicate with their assigned CHW by phone or text message to ask questions, report concerns, or provide notice of key transition points such as onset of labor or delivery. Participants were provided with airtime, equivalent of $3 US dollars, at each study visit to cover communication costs. Community health workers in both the intervention and control sites were retrained on PMTCT service delivery and on appropriate messaging to women during pregnancy and postpartum. Community health workers were provided with the equivalent of $6 US dollars for airtime expenses. SMS text messages were designed by the study team to address areas of need identified during our formative research [14]. SMS text messages were iteratively edited for content and tone and translated in collaboration with local study team members fluent in the local language. Participants received 3 to 6 SMS messages each week at a self-selected time of day and in their preferred language. The thematic areas covered four broad categories: PMTCT services, including appointment reminders and adherence support; motivational messages of hope and encouragement; male-partner involvement and engagement in delivery planning; and essential maternal child health messages including warning signs and nutrition. Women underwent structured interviews at four visits (enrollment; 36 weeks of gestation; delivery, 7 days postpartum; and 6–8 weeks postpartum) to record self-reported adherence to antiretrovirals in the past week, number and mode of communication between the participant and health workers, time and place of delivery, infant feeding practices, and any intervening clinical outcomes. The sample size for the study was estimated by inflating the sample size from a 2-arm simple randomized study using the design effect (DE), 1 + (n − 1)ρ, where n is the cluster sample size and ρ is the intracluster correlation. As there were no published data on the intracluster correlation coefficients for the selected outcomes, we estimated the coefficient of variation for a number of plausible intracluster correlation coefficients [17]. With a sample size of 286 women from 13 clusters per study arm, assuming an intracluster correlation <0.05 (range 0.01–0.51), the study had 80% power to detect an absolute effect size of 20% between intervention and control sites for the selected outcomes, including uptake of antiretrovirals, facility-based deliveries, and HIV infant testing at 6 weeks of age, after factoring in possible loss to follow-up of 10%. Baseline demographics and clinical characteristics were summarized using proportions for categorical variables and medians with associated interquartile ranges for continuous variables, stratified by intervention group. All outcomes were summarized at the individual and cluster level. For the individual level analysis, the overall proportion of women with the outcome of interest was estimated by intervention assignment. In the cluster level analysis, the proportion of women with the outcome of interest was estimated for each health facility. The study arm proportion was obtained as the mean of facility level proportions together with the standard deviation of the facility means. In our study, the intervention was randomly allocated to health facilities; hence, it is the unit of analysis. The effect of the mHealth intervention on each of the outcomes was estimated by comparing the two sets of health facility-specific proportions using two-sample t-tests. Differences between group proportions and associated confidence levels were estimated using t-distribution. Statistical significance tests were confirmed using the nonparametric Wilcoxon rank-sum tests. Since the number of health facilities per group was considered small, the comparison of facility level summaries was conducted using t-tests as this has been demonstrated to be the most robust approach compared to regression methods. The effect of the intervention was also estimated using relative risks and associated confidence intervals. We used a two-stage analytic approach to estimating the adjusted effect of the intervention. In the first stage, we estimated cluster level residuals for each health facility. For each outcome, we fitted a logistic regression model of the outcome and potential confounding variables including participant age, gestational age, whether the woman was newly diagnosed with HIV, and disclosure of HIV status to her partner and family. To measure the intervention effect as a risk difference, we estimated the residual for each health facility by taking the difference of the number of observed outcome events and the number of expected outcomes events. The adjusted risk difference was estimated by the difference between the mean facility-specific residuals in the intervention arm and the mean facility-specific residuals in the control. For the relative risk effect measure, the cluster-specific residual was estimated by the ratio of the observed outcome events to the expected outcome events for each health facility. The adjusted relative risk is estimated by ratio of the mean of the ratio residuals for clusters in the mHealth intervention arm to the mean of the ratio residuals for clusters in the control arm. Associated confidence intervals for each effect measure were also estimated. The protocol was approved by the Kenyatta National Hospital/University of Nairobi (KNH/UON) ethics review committee, the World Health Organization Institutional Review Board (IRB), and the Georgetown University IRB recognized KNH/UON as the IRB of record. All participants provided written informed consent.

N/A

• Dima Health