The study seeks to provide access to healthcare information and maternal and infant care services to rural women, a population subset known for low utilization of antenatal care and postnatal care, suboptimal hygiene and nutritional practices, susceptibility to cultural taboos affecting maternal and infant health and preference for traditional remedies of low efficacy over available treatment of ailments affecting their and their infant’s health. Technology will serve as two bridges for rural women. First, it will work as a bridge across low literacy barriers and provide healthcare information in a timely manner to rural women. Voice messages will be readily accessible to women in areas where female literacy is very low. Second, access to healthcare solutions and rural health workers over phone will bridge the barriers to access healthcare, including transport and cultural constraints. In order to study the efficacy of the solution for wider adoption in the public health system in Pakistan, the research will use a randomized controlled trial of application of voice message nudges, cellphone delivered healthcare messages to spur timely action to improve mother and infant health outcomes and document uptake of maternal health services.
The study chooses Pakistan as it has several attributes where such a behavioral economics intervention can be useful. In rural Pakistan, pregnancy-specific healthcare utilization from skilled professionals and facility delivery rates are pretty low. According to the most recent Pakistan Demographic Health Survey 2012-2013, prior to our intervention, more than two-thirds of pregnant women (73%) received at least one antenatal care during their pregnancy; however, only 36% of pregnant women had the recommended at least four antenatal care visits with dramatic differences between urban and rural areas, 62% and 26%, respectively. On the other hand, less than half of births (48%) took place in a health facility, with a wide gap between urban and rural 68% vs. 40%. Likewise, only 44% of rural births were delivered at a hospital or health clinic, whereas the ratio was 71% in urban births. Lastly, while 44% of rural women did not have any postnatal visits in the first two days after birth, the corresponding ratio was only 23% among urban women.
Care-seeking behaviors and pregnancy outcomes among rural Pakistani women are similar to those observed in other developing countries, making our findings immediately relevant to large parts of the world. For example, according to Demographic Health Surveys from 28 African countries,75% of pregnant women received at least one antenatal care, on average; and antenatal care utilization was even less than 50% in some countries such as Zimbabwe, Burkina Faso, and Ethiopia. Moreover, 38% of pregnant women received the recommended four or more antenatal care visits in Africa. Likewise, only 43% of births took place in a facility in sub-Saharan countries, where neonatal mortality rates are the highest in the world Finally, in low-income countries worldwide, only 37% of women received postnatal care within the first two days after birth.
Delayed recognition of pregnancy complications, inadequate antenatal and postnatal care as well as eschewing facility deliveries are among the major factors leading to high maternal and neonatal mortality ratios in South Asia and sub-Saharan Africa However, these deaths could largely be prevented through behavioral modifications to health-seeking behaviors, including antenatal and postnatal care use from skilled professionals and facility deliveries. In this line, our intervention aims to advance pregnancy-related healthcare utilization via informational voice nudges among pregnant women in rural Pakistan with potential applicability to similar populations elsewhere.
Our design aims to test the efficacy of informational nudges in improving maternal health knowledge and care uptake using voice messages as the medium of communication. We manipulate call frequency, message timing, and the provision of small financial incentives. Participants received voice messages for up to 26 weeks, depending on the time of the recruitment. All treatments are implemented at the cluster level to minimize spillover effects across treatment arms. In this village-level cluster randomized controlled trial, we compare the outcomes of the four treatment arms, A, B, C, and D, to the control arm, E. We summarize the experimental design as follows:
• Treatment Arm A: High-Frequency Informational Nudges Timed to Gestational Age. This group received two weekly messages timed to gestational age.
• Treatment Arm B: Low-Frequency Informational Nudges Timed to Gestational Age. This group received one weekly message timed to gestational age.
• Treatment Arm C: Low-Frequency Informational Nudges + Small Cash Incentives. This group received one weekly message timed to gestational age. In addition, at the end of the weekly voice calls pertaining to the maternal healthcare literacy messages, the participants were offered a small financial incentive (20 Rupees) if they agreed to listen to a general health message by pressing ‘1’ on the keypad, displayed in Message Content Appendix B. If they accessed the additional information, it would give them a phone balance transfer of 20 Rupees (~USD 0.20). Therefore, Arm C assesses the value of financial incentives in boosting the efficacy of informational voice nudges.
• Treatment Arm D: Random Order Informational Messages. This group received one message per week in a random order, i.e., the message content was not synchronized to the gestational age.
Our experimental design aims to answer several important questions not explored by large-scale RCTs conducted exclusively in the countryside of poor nations. The key question we investigate is whether and to what extent mobile phone-based informational nudges can help address behavioral impediments to pregnant women's maternity-specific care-seeking in such localities. Relatedly, arms A and B allow us to examine whether high- versus low-frequency informational messages have differential impacts on knowledge and behavioral outcomes.
We hired rural female community health workers, known as Lady Health Workers (LHW), to administer the recruitment protocol to the rural women in the study villages. We deployed them to visit every household in their locality to identify and register pregnant women. In each village (cluster) of the sample, we targeted to reach all the pregnant women who were in the first trimester of pregnancy.
Participation was voluntary and obtained by the use of approved and standard informed consent procedures. Accordingly, a printed informed consent form was read and explained in the local language to the women. Individuals who granted consent were registered. Given that LHWs have been institutionalized and reasonably acceptable in rural Pakistan (Douthwaite and Ward, 2005), we did not face any resistance to participation in our study. Initially, 1556 women were recruited in 403 villages in Pakistan's Chakwal and Swabi districts. Following this, the villages (clusters) were randomly assigned to five arms, four treatment (A, B, C, D) arms, and a control group.
To eliminate the possibility of sample selection bias, we provided free mobile phones to 300 participants who did not have one because our interventions required access to a mobile phone. If any participant had access to another family member’s cellphone, that number was recorded for women who reported that they could receive calls on it. In addition to the phone number of the participants, their preferred times, and days to receive calls throughout the week were recorded to increase the likelihood that women, including those having partial access to shared phones, could listen to the messages.
We managed to register nearly all pregnant women in the first trimester of their pregnancy at the time of recruitment and provided a mobile phone to those who did not have one. Thus, we are confident that our sample is representative of the study population at hand and is scalable to hundreds of millions of women in similar conditions in South Asia and Sub-Saharan Africa.
A baseline survey, aiming to elicit participant demographic information, household characteristics, and maternal history, was administered by LHWs in person to each of the 1556 subjects at the time of recruitment.
Then, to administer data collection at the endline, we contracted Lady Health Visitors (LHVs), trained as skilled birth attendants, to ensure the accuracy of the endline information as more specific information on delivery and conditions around it was needed. The LHVs collected data using an Android cellphone application developed to ensure the accurate and standardized administration of the endline survey. It was successfully administered to 1,399 participants, gathering pregnancy-specific health knowledge and care utilization, such as prenatal care, facility delivery, and postnatal care.
Through our work, we aim to demonstrate that beyond creating the choice of modern maternal care in the close vicinities of pregnant women, women in rural areas need to be equipped with health literacy, without which the available healthcare choices may not be appropriately legible to them. Health literacy is likely to change malformed opinions and manifests healthcare choices in bold relief, allowing women to adopt health behaviors beneficial to them and their newborns. In some ways, health literacy can allow for accurate reading of the choice architecture available to pregnant mothers, otherwise seen through the translucency of free-flowing opinions.