Quality and utilization patterns of maternity waiting homes at referral facilities in rural Zambia: A mixed-methods multiple case analysis of intervention and standard of care sites

Introduction Maternity waiting homes, defined as residential lodging near a health facility, are recommended by the WHO. An improved MWH model, responsive to community standards for functionality and comfort, was implemented at two purposively selected health facilities in rural Zambia providing comprehensive emergency obstetric and neonatal care (CEmONC) services (intervention MWHs), and compared to three existing standard-of-care MWHs (comparison MWHs) at other CEmONC sites in the same districts. Methods We used a mixed-methods time-series design for this analysis. Quantitative data including MWH quality, MWH utilization, and demographics of women utilizing MWHs were collected from September 2016 through May 2018 to capture pre-post intervention trends. Qualitative data were obtained from two focus group discussions conducted with pregnant women at intervention MWHs in August 2017 and May 2018. The primary outcomes were quality scoring of the MWHs and maternal utilization of the MWHs. Results MWH quality was similar at all sites during the pre-intervention time period, with a significant change in overall quality scores between intervention (mean score 83.8, SD 12) and comparison (mean score 43.1, SD 10.2) sites after the intervention (p <0.0001). Women utilizing intervention and comparison MWHs at all time points had very similar demographics. After implementation of the intervention, there were marked increases in MWH utilization at both intervention and comparison sites, with a greater percentage increase at one of two intervention sites. Conclusions An improved MWH model can result in measurably improved quality scores for MWHs, and can result in increased utilization of MWHs at rural CEmONC facilities. MWHs are part of the infrastructure that might be needed for health systems to provide high quality “right place” maternal care in rural settings.

Southern and Eastern Provinces, Zambia are primarily rural. This study was nested within a larger study evaluating the effectiveness of newly-constructed, community-informed MWHs to increase access to delivery services for women living furthest from care at BEmONC sites [17]. Additionally, five CEmONC facilities within two hours’ drive time from the 10 intervention BEmONC facilities were purposively selected to be included in this study, with intervention MWHs constructed at two of those five CEmONC facilities (Table 2). A Core Maternity Waiting Home Model (Core MWH Model) was implemented at two referral hospitals (intervention CEmONC sites) capable of conducting eight or more CEmONC signal functions in Southern and Eastern Provinces. The core pillars of the Core MWH Model for CEmONC sites, derived from original formative research [18–20], include: (1) infrastructure, equipment, and supplies to address the need for higher quality, safer MWHs where women can wait comfortably for delivery; and (2) health system linkages to ensure women receive appropriate antenatal or postnatal care while waiting. The first domain encompasses the construction of a quality cement structure without leaks; a lighting source; lockable doors and windows; a cooking area with utensils; bathing and laundry areas; latrines; beds, bedding, and mosquito nets; a lockable storage room for assets; dedicated space for postnatal women and newborns to stay; and access to water for drinking and hygiene [17]. The second domain requires being adjacent to a CEmONC facility and for CEmONC staff to regularly monitor waiting women and the condition of the MWH. Immediately after construction of the Core MWH Model at the intervention sites, ongoing maintenance of the MWH was assumed by the affiliated hospitals. ‘Policies, management and finance’ was a third pillar included in the main evaluation, but was not included in the CEMONC implementation plan. Upon completion of construction, the CEMONC sites assumed responsibility of the MWHs, with minimal guidance around management and governance. Three comparison CEmONC sites in Southern Province continued implementing the MWH ‘standard of care,’ which varied in quality (Table 2). A register system was instituted at all sites to capture MWH utilization (S1 File). A designated person who received a small stipend completed the registers at each site (Table 2). Intervention site Zimba opened in March 2017 and Intervention site Nyimba opened in April 2017. During construction of the MWHs in early 2017, Nyimba was re-designated an urban health center from a Level 1 Hospital. However, maternity services (including CEmONC signal functions) remained at Nyimba, and Nyimba remained the primary obstetric referral center for its district during the course of this evaluation. Similarly, Kalomo changed from a Level 1 Hospital to an urban health center in March 2018 and all CEmONC functions were transitioned to the new hospital a month later (Table 2). At Zimba, two shelters exist: the Core MWH Model and the prior existing MWH which accommodates any overflow. If a waiting woman transfers from the old MWH to the Core MWH Model, she is not re-registered, and the time spent at each MWH is not known. Data for Zimba thus include women waiting at the existing MWH and the Core MWH Model. Each woman is counted only once at initial registration (Table 2). We used an interrupted, two-group time-series design, systematically assessing the two intervention sites (Core MWH Model) and three comparison sites (standard of care) on a monthly basis between September 2016 and May 2018. We define the pre-intervention period to be from September 2016 through the opening of each MWH intervention site (March or April 2017) and a post-intervention period to be the 14 months following the opening of each site, through May 2018. We used mixed-methods to capture and triangulate data. First, quality assessment data were collected monthly from both intervention and comparison sites using a quantitative core model checklist (CMC), which was developed specifically for this project to measure quality, implementation fidelity and maintenance of quality after implementation (S2 File). This CMC evaluated nine core quality components of the MWHs identified during formative research [18–20]: infrastructure, safety, amenities, cleanliness, water, hygiene, sanitation, cooking and feedback. Second, MWH utilization data were extracted monthly from both intervention and comparison sites. Registers captured individual-level demographics and MWH arrival and discharge dates. We did not calculate a sample size for utilization a priori as these data were collected as part of routine monitoring. Local data collectors who underwent ethics training and training in all study instruments completed register data extraction and quality assessment data collection. Third, we conducted two focus group discussions (FGD) with sixteen pregnant women at each intervention MWH (S3 File) after the MWHs had opened. These four FGDs, facilitated by local data collectors trained in research ethics, qualitative interviewing techniques, the specific instruments, and fluent in the relevant local languages, captured perspectives on MWH quality, barriers and facilitators to MWH access and facility delivery, and reasons for MWH use. Women 15 years or older who had been staying the longest at the MWHs were recruited to participate in FGDs. Primary quantitative outcomes for this analysis are quality scoring of CEmONC facilities and utilization of MWHs at CEmONC facilities. To construct the composite quality score, the following domains were systematically assessed via the CMC: infrastructure, safety, amenities (including bedframes, mattresses, mosquito nets), cleanliness, water (access to potable water), hygiene (bathing area), sanitation (latrines), cooking (designated area and utensils), and feedback (system for receiving and addressing women’s comments/complaints). Each domain had between one to ten core components; the presence or absence of any individual component was scored as one or zero respectively. If present, additional points, if applicable, were added depending on material type (e.g. metal vs. thatched roof), functionality (e.g. absence of holes or leaks in roof), quantity of non-broken assets (e.g. bedframes), and condition (e.g. cleanliness scored as sufficient, needs improvement, or not clean). Scores under each domain were summed and standardized to 10. The domains were then summed to create a monthly composite quality score and scaled to 100. Indicators of utilization included mean number of women staying per month, average daily census (ADC), bed occupancy rate (BOR), and average length of stay (ALOS). Women were categorized as either staying for less than one night in the MWH or at least one night. The mean number of women utilizing an MWH per month was calculated by summing the total number of women who stayed at the MWH for any amount of time for any reason. Women who stayed less than one night, and/or had missing discharge and delivery dates were included in the mean number of women staying per month but excluded from calculations for average daily census (ADC), bed occupancy rate (BOR), and average length of stay (ALOS). ADC was calculated by summing total bed-days for all women who stayed at the MWH each month divided by the number of days in the month. Bed occupancy rate is ADC divided by the number of beds multiplied by one hundred. The ALOS was calculated by summing the bed-days for all women staying at the MWH divided by the number of women. For all variables, a woman’s contribution was counted for the calendar month in which she arrived at the MWH. Demographic characteristics include age, grade level completed, marital status, gravida, parity, pervious stillbirths, gestational age (EDD as reported on the ANC card on admission to the MWH), travel time from home, transportation mode, number of companions with the woman, and the companions’ relationship. Demographics are reported on all women who utilized any MWH for any length of time (including those without a discharge date). For the quality scores we calculated the mean and standard deviations. We used a difference-in-differences (DID) analysis to test for significance in quality between intervention and comparison sites during the pre- and post-intervention periods. The composite score and scores for individual domains are reported. We tested for significance in all quantitative data using first a t-test or chi-squared test for differences between intervention and comparison sites during the pre-intervention time period, and then a DID analyses for the post-intervention time period at the intervention sites as comparison sites lacked beds. For mean quality assessment scores, mean number of women per month, and ADC, the difference-in-difference estimates controlled for month due to the monthly nature of these variables. All analyses accounted for clustering. BOR was only calculated for the post-intervention period. Utilization data are presented as aggregate and stratified by intervention site; the two intervention sites had different pre-intervention utilization patterns. All quantitative analysis was done using SAS 9.4 (Cary, NC). P-values were considered significant at a level of alpha≤0.05. All qualitative analysis was conducted using NVivo 11 Software (QSR International, Doncaster, Australia). The FGDs were audio recorded, translated and transcribed verbatim into English. Some codes were created a priori based on the FGD guide; additional codes were created as themes emerged during the coding process. A content analysis was done for each time point and then the emerging themes were compared over time [21]. Qualitative data were triangulated with quantitative data to create a full picture of MWH quality and choice to utilize MWHs. Ethical approval was granted by the Boston University Institutional Review Board (protocol H-35321) and the ERES Converge IRB in Zambia (reference number 2016-June-023). Approval was also obtained from the Zambia National Health Research Authority and the Ministry of Health. The hospital administrator at each CEmONC site also granted approval for the evaluation. Written informed consent was obtained from each FGD participant. For FGD participants aged 15-17, assent was first obtained from the woman and consent was obtained from her guardian. If no guardian was available, the woman was ineligible for participation in the FGD. A waiver of consent was granted for data extracted from the registers. Quality assessment of the MWHs using the Core Model Checklist was non-human subjects research.

• bonawitz, rachael
• mcglasson, kathleen l.
• kaiser, jeanette l.
• ngoma, thandiwe
• fong, rachel m.
• biemba, godfrey
• bwalya, misheck
• hamer, davidson h.
• scott, nancy a.