Background: Mobile phones are increasingly used in health systems in developing countries and innovative technical solutions have great potential to overcome barriers of access to reproductive and child health care. However, despite widespread support for the use of mobile health technologies, evidence for its role in health care is sparse. Objective: We aimed to evaluate the association between a mobile phone intervention and perinatal mortality in a resource-limited setting. Methods: This study was a pragmatic, cluster-randomized, controlled trial with primary health care facilities in Zanzibar as the unit of randomization. At their first antenatal care visit, 2550 pregnant women (1311 interventions and 1239 controls) who attended antenatal care at selected primary health care facilities were included in this study and followed until 42 days after delivery. Twenty-four primary health care facilities in six districts were randomized to either mobile phone intervention or standard care. The intervention consisted of a mobile phone text message and voucher component. Secondary outcome measures included stillbirth, perinatal mortality, and death of a child within 42 days after birth as a proxy of neonatal mortality. Results: Within the first 42 days of life, 2482 children were born alive, 54 were stillborn, and 36 died. The overall perinatal mortality rate in the study was 27 per 1000 total births. The rate was lower in the intervention clusters, 19 per 1000 births, than in the control clusters, 36 per 1000 births. The intervention was associated with a significant reduction in perinatal mortality with an odds ratio (OR) of 0.50 (95% CI 0.27-0.93). Other secondary outcomes showed an insignificant reduction in stillbirth (OR 0.65, 95% CI 0.34-1.24) and an insignificant reduction in death within the first 42 days of life (OR 0.79, 95% CI 0.36-1.74). Conclusions: Mobile phone applications may contribute to improved health of the newborn and should be considered by policy makers in resource-limited settings.
The Wired Mothers study is a pragmatic, randomized, controlled trial with the primary health care facility as the unit of randomization. The study took place from 2009 to 2010 on the island of Unguja in Zanzibar, a semi-autonomous part of the United Republic of Tanzania. We followed the Consolidated Standards of Reporting Trials guidelines for reporting cluster-randomized trials [16]. The study comprised 24 primary health care facilities and pregnant women attending antenatal care at these facilities. Clusters eligible for randomization were the four primary health care facilities in each of the six districts of Unguja Island with the most antenatal care visits in the previous year and a midwife among the staff. There were no major differences between included facilities. They were all primary health care facilities staffed with 1 or 2 midwives and access to basic infrastructure and equipment. The distribution of facilities in relation to hospitals providing Emergency Obstetric and Neonatal care was the same in intervention and control clusters (Figure 1). The eligibility criteria for participants was pregnant women who attended their first antenatal care visit at 1 of the 24 primary health care facilities regardless of gestational age or mobile phone ownership. A total of 2550 women were included in the study (Figure 2). Twenty-two women miscarried and 82 women withdrew or were not contactable during follow-up. Of these, 15 were known to have travelled outside the study area and three were not pregnant. During the study period 5 women died as a result of direct obstetric complications. Research districts and location of intervention and control health facilities. Procedures for the selection of the study population. A Wired Mother with her child. The Research Council of Zanzibar approved the study protocol on January 27th, 2009. The trial is registered with ClinicalTrials.gov, {“type”:”clinical-trial”,”attrs”:{“text”:”NCT01821222″,”term_id”:”NCT01821222″}}NCT01821222. All women were informed about the nature and purposes of the study as summarized in the consent form written in the local language, Swahili. All women provided their consent either by signature or fingerprint prior to their inclusion in the study. Women were free to drop out of the study at any time without a change in the quality of care provided. All study results and completed questionnaires were kept confidential and were not accessible to people outside of the research team. The trial is registered after enrollment of participants, due to researchers not being aware of this International Committee of Medical Journal Editors publication demand for relatively benign interventions without individual randomization such as the Wired Mothers. The Wired Mothers mobile phone intervention was designed with the aim to link pregnant women to their primary health care provider throughout their pregnancy, childbirth, and postpartum period. The intervention was developed in Tanzania using simple technology and at low cost. It consists of two components: an automated short message service (SMS) system providing wired mothers with unidirectional text messaging and a mobile phone voucher system providing the possibility of direct two-way communication between wired mothers and their primary health care providers. Women in the intervention group were registered at their first antenatal care visit with date, a phone number, and gestational age. The phone number was either their own or an access phone number of a husband/relative/friend. A specially-designed software that creates an individual pregnancy timeline for each woman and automatically sends text messages to the registered phone number was developed. The content and the frequency of the messages varied throughout the pregnancy and were intensified to weekly messages during the 4 weeks before delivery. The content of the messages focused on health education on topics, such as danger signs in pregnancy and the importance of skilled delivery attendance as well as appointment reminders for the next antenatal care visit. A total of 29,000 SMS were sent during the intervention period. Because the wired mothers intervention was developed in the context of the Ministry of Health in Zanzibar prioritizing to reduce maternal mortality, a voucher system was added to improve access to emergency obstetric care and improve referral mechanisms. Each intervention woman received the phone number of her local midwife and a small voucher of 500 TSH (Tanzanian shilling) allowing to call her. The women were not provided with mobile phones but a referral link was created in the health system through provision of mobile phones to midwives in primary health care facilities, and to midwives/doctors/drivers at the hospital level. Twenty-four primary health care facilities were eligible for the study and the Ministry of Health agreed to let the facilities be included in the trial (cluster level consent). Meetings were held with staff in participating primary health care facilities to explain the nature and purpose of the trial. The enrolled primary health care facility staff also functioned as research assistants recording an inclusion questionnaire with demographic and covariate information, registering each contact with the women and completing an end-of-study questionnaire 6 weeks after delivery. Research assistants were assigned to the 3 hospitals providing emergency obstetric care and all contact with the enrolled women were similarly recorded. All pregnant women attending their first antenatal care visit in one of the participating primary health care facilities, if willing to participate, were included in the study. All enrolled women received an individual identity number and card. Pregnancy outcome was recorded at delivery for facility-based deliveries and for all included women in a follow-up interview 42 days after delivery. If the women did not return for the end-of-study interview, the research assistant contacted them either directly or by phone. Women attending the standard care received the protocols recommended in Zanzibar in the best format offered in these facilities. Double entry of data was performed in Epidata, transferred, and validated in SPSS. We evaluated the effect of a mobile phone intervention on the primary outcomes antenatal care (four or more visits) and skilled delivery attendance. These are presented in other papers [17,18]. Here, we present the interventions association with secondary outcomes stillbirth, perinatal death, and death of a child within the first 42 days of life. The intervention association with antenatal care and skilled delivery attendance is presented in other manuscripts. A perinatal death was defined as a composite of either a stillbirth or early neonatal death. We used the WHO agreed definition of stillbirth as any delivery in the third trimester (≥1000 g birth weight or ≥28 weeks of gestation) in which no signs of life (breathing, crying, heartbeat, movement) were evident [9]. An early neonatal death was defined as all babies born alive in the third trimester who die on or prior to day 7 after birth. The perinatal mortality rate is perinatal death per 1000 total births (live births and stillbirths). We included death of a child within 42 days (where end-of-study questionnaire was conducted) as a proxy of neonatal mortality. Power calculations were made on the primary outcomes skilled birth attendance and antenatal care attendance and did not take into account the clustering effect. Based on the number of antenatal care attendees from the previous year, the expected of size of the study population during a 3-month enrolment period was estimated to be 1100 women in the intervention group and 1375 women in the control group. Subsequently, a power calculation, based on data on antenatal care visits and skilled delivery attendance from the Tanzanian Demographic Health Survey (DHS 2005), was performed to document if the expected study population would be sufficient to document a true difference between the intervention and the control group [19]. To estimate whether this sample size was sufficient for detection of public health relevant effects of the intervention, we used data from the DHS 2005. For instance, with a 95% probability and a power of 90%, 894 women (447 in each group) were necessary for showing an increase of a relevant size (10% increase in the number of women delivering with a skilled birth attendant). Hence, according to our power calculations, our proposed sample size was sufficient to document an effect of our intervention on antenatal care and skilled delivery attendance. Primary health care facilities, stratified by district, were assigned by simple random allocation to either the mobile phone intervention or control group (Figure 1). Clusters and study participants were not masked due to the nature of the intervention requiring overt participation. Analyses were performed based on the “intention to treat” principle and all available data were included in the analysis. We adjusted for the clustering of our data using generalized estimating equations in all logistic regression analyses. We specified an exchangeable working correlation to allow for within cluster correlation and standard errors were based on the robust covariance matrix. We used the traditional logit link, which resulted in odds ratios (ORs) as an effect measure. However, for small values of the risk these can approximately be interpreted as relative risks. For our binary outcome measure, perinatal death yes/no, logistic multilevel analysis was used to analyze if there was a difference in perinatal deaths between the intervention and control groups. In this model, we included all socioeconomic and obstetric confounding variables and eliminated them using backward elimination (age, occupation, education, mobile phone status, residence, parity, previous caesarean section, multiple-gestation pregnancy). Variables with statistical significance were included in the final model. These were age and multiple-gestation pregnancy. Premature delivery, mode of delivery, four or more antenatal care visits, and delivery attendant were considered intermediate variables and not included in the model. We found no interaction between the intervention and explanatory variables. For other secondary outcomes we used a similar approach. Results were expressed as OR for perinatal deaths with 95% CI. Because perinatal mortality is a rare event this can be interpreted as a relative risk. For all models the criterion for significance was set at P<.05 and all analyses were performed using SPSS (version 20).