Characteristics of maternity waiting homes and the women who use them: Findings from a baseline cross-sectional household survey among SMGL-supported districts in Zambia

Objective Maternity waiting homes (MWHs) have been identified as one solution to decrease maternal morbidity and mortality by bringing women living in hard-to-reach areas closer to a hospital or health center that provides emergency obstetric care. The objective of this study was to obtain data on current MWH characteristics and the women who use them as well as women’s perceptions and experiences with MWHs among seven Saving Mothers Giving Life (SMGL) supported districts in Zambia. Methods A cross-sectional household survey design was used to collect data from 2381 mothers who delivered a child in the past 13 months from catchment areas associated with 40 health care facilities in seven districts. Multi-stage random sampling procedures were employed with probability proportionate to population size randomly selected. Logistic regression models, Chi-square, and independent t-tests were used to analyze the data. Results Women who lived 15–24 km from a health care facility were more likely to use a MWH when compared to women who lived 9.5–9.9 km from the nearest facility (AOR: 1.722, 95% CI: 1.450, 2.045) as were women who lived 25 km or more (AOR: 2.098, 95% CI: 1.176, 3.722.881). Women who were not married had lower odds of utilizing a MWH when compared to married women (AOR: 0.590, 95% CI: 0.369, 0.941). Over half of mothers using a MWH prior to delivery reported problems at the MWH related to boredom (42.4%), management oversight (33.3%), safety (33.4%), and quality (43.7%). While the study employs a robust design, it is limited by its focus in Saving Mothers Giving Life districts. Conclusion MWHs, which currently take many forms in Zambia, are being used by over a third of women delivering at a health facility in our study. Although over half of women using the existing MWHs noted crowdedness and nearly a third reported problems with the physical quality of the building as well as with their interaction with staff, these MWHs appear to be bridging the distance barrier for women who live greater than 9.5 km from a health care facility. A cross-sectional household survey was used to collect data from the 40 study cluster catchment areas involved in the study (20 health care facilities identified to receive a minimum core model MWH and 20 comparison facilities). A team of local residents, hired as research assistants, literate in the appropriate local languages and English, and with previous experience collecting quantitative data for research studies, were trained in human subjects’ protection and qualitative and quantitative data collection methods during a 5-day training. Data were captured electronically using SurveyCTO Collect Software installed on encrypted tablets. The survey was pre-tested among 50 respondents (women who had delivered a baby within the past year) recruited from a local clinic. All four languages were represented (Bembe, Nyanja Tonga, and Tumbuka). Adjustments were made in response to the pre-test, mainly changing more formal translations into the vernacular. No major changes were required. When possible, questions on the survey were drawn or adapted from existing instruments [14]. Zambia is a land locked country in sub-Saharan Africa. At the time of the survey, there were 10 provinces and 74 districts. According to WHO, the current population is 16 million with a maternal mortality ratio of 224 per 100,000 live births [14]. While the country has had marked improvements in maternal health, a woman’s lifetime risk of maternal death remains high at 1 in 79. It is for these reasons that Zambia was chosen as one of two original countries in sub-Saharan Africa to take part in a five-year public-private partnership aimed at accelerating reduction in maternal and newborn mortality [15]. Launched in 2012, SMGL takes a health systems approach to improve access to clean, safe childbirth services and timely emergency care for pregnant women [16]. Seven SMGL districts (Choma, Kalomo, Lundazi, Mansa, Nyimba, Pemba, and Chembe) in three provinces (Eastern, Luapula, and Southern) are targeted for this study. Although they are now part of SMGL districts, at the time of data collection, SMGL activities had not commenced in Choma or Pemba. To ensure facilities included in the study are resourced appropriately to adequately manage obstetric complications, 40 rural health facilities, located within 2 hours of travel time to a CEmONC referral facility were selected from a list of eligible facilities that met the following inclusion criteria: (i) capable of performing a minimum of 5 of the 7 BEmONC signal functions and (ii) providing intrapartum care to a minimum of 150 women per year; or (i) staffed with at least one skilled birth attendant on staff, (ii) routinely providing active management of third stage of labor, and (iii) having no stock outs of oxytocin or magnesium sulfate in the last 12 months. We chose 12 months to ensure facilities had stability in commodities and human resources. In 4 districts health facilities were randomly chosen, while in the remaining 3 districts health facilities meeting criteria were purposively sampled from eligible facilities with input from district health teams. Selection and assignment of study clusters is described in detail elsewhere [17]. The survey was conducted in the 40 study cluster catchment areas in the seven SMGL districts over three weeks in March 2016. These data will later become part of a robust evaluation study of 20 sites receiving the minimum core model MWH and 20 comparison sites. For the survey, multi-stage random sampling procedures were employed in the seven districts with probability proportionate to population size randomly selected. The sample frame of clusters included villages located more than 9.5 km from the health care facility within their catchment area along the most direct route, identified through geo-coding. Details of the sampling frame and protocol for this study are reported elsewhere [17]. In the second stage of sampling, all households within the selected villages were listed and then randomly selected. The sample consisted of women who met the following inclusion criteria: (i) had delivered in the last 13 months (to obtain recent delivery data and reduce recall bias), (ii) 15 years of age or older, and (iii) lived in a village that was 9.5 km or farther from one of the health care facilities included in our sample. To ensure a representative sample of the target population, a multi-stage random sampling procedure was used: (i) villages 9.5 km or greater from the health care facility along the most direct route were identified within each of the seven districts, (ii) households within each village were randomly ordered and approached to contact an eligible respondent (i.e., mother who has recently delivered), and (iii) if more than one eligible respondent was in the household, one of these respondents was randomly sampled. Participants from eligible households were recruited, consented, and enrolled in the study. The research assistant recorded the geo-location of the village center to determine distance to the nearest health care facility. Eligible participants provided written informed consent, which was documented in writing or with a fingerprint and witness signature prior to beginning the survey. If participants were under the age of 18 years, child assent and guardian or husband (if over the age of 18 years) was obtained. Each household survey took approximately 45 minutes. The final sample included 2381 mothers who had delivered a child in the last 13 months for a response rate of 86.9%. Of the women who were eligible but did not respond, 280 mothers were unavailable, 60 refused participation, and 20 mothers withdrew after beginning the survey or had incomplete surveys and were dropped from the study. Participants received a small token of appreciation (chitenge, a local fabric) in acknowledgment of their time. Ethical approval was obtained from Boston University Institutional Review Board (IRB), University of Michigan IRB, and the ERES Converge Research Ethics Committee in Zambia. The analytic strategy for the current study was to: (i) provide descriptive statistics for the study sample, (ii) examine key characteristics of mothers and households who used a MWH for their most recent delivery, (iii) examine the prevalence of MWH utilization for women delivering in the year prior to the start of the study, and (iv) examine perceived characteristics of MWHs among women who delivered at these locations across the seven districts included in this study. For the analyses, STATA 14.0 was used to estimate the models outlined above [18]. All logistic regression models provide adjusted odds ratios (AOR) and 95% confidence intervals (95% CI) and accounted for the multi-stage sampling procedure. Moreover, Chi-square and independent means t-tests were used to assess differences between individual districts (when compared to the combined group of respondents in remaining districts to maintain a large enough sample to make meaningful comparisons). Missing data were handled using listwise deletion.