Menu
Menu
Menu
Menu
Background: Residential accommodation for expectant mothers adjacent to health facilities, known as maternity waiting homes (MWH), is an intervention designed to improve access to skilled deliveries in low-income countries like Zambia where the maternal mortality ratio is estimated at 398 deaths per 100,000 live births. Our study aimed to assess the relationship between MWH quality and the likelihood of facility delivery in Kalomo and Choma Districts in Southern Province, Zambia. Methods: We systematically assessed and inventoried the functional capacity of all existing MWH using a quantitative facility survey and photographs of the structures. We calculated a composite score and used multivariate regression to quantify MWH quality and its association with the likelihood of facility delivery using household survey data collected on delivery location in Kalomo and Choma Districts from 2011-2013. Results: MWH were generally in poor condition and composite scores varied widely, with a median score of 28.0 and ranging from 12 to 66 out of a possible 75 points. Of the 17,200 total deliveries captured from 2011-2013 in 40 study catchment area facilities, a higher proportion occurred in facilities where there was either a MWH or the health facility provided space for pregnant waiting mothers compared to those with no accommodations (60.7% versus 55.9%, p <0.001). After controlling for confounders including implementation of Saving Mothers Giving Life, a large-scale maternal health systems strengthening program, among women whose catchment area facilities had an MWH, those women with MWHs in their catchment area that were rated medium or high quality had a 95% increase in the odds of facility delivery than those whose catchment area MWHs were of poor quality (OR: 1.95, 95% CI 1.76, 2.16). Conclusions: Improving both the availability and the quality of MWH represents a potentially useful strategy to increasing facility delivery in rural Zambia. Trial registration: The Zambia Chlorhexidine Application Trial is registered at Clinical Trials.gov (identifier: NCT01241318)
Our study utilized data collected from two separate studies conducted in two contiguous districts in Southern Province, Kalomo and Choma Districts, between 2011 and 2013: 1) a formative evaluation of MWH conducted in 2013 on the physical quality of MWHs [19]; and 2) a cluster-randomized controlled trial conducted in Southern Province between 2011 and 2013 [20]. At the time of data collection, Kalomo had a primarily rural population (93%) of 258,570 [21] and 35 health facilities, including 27 health centers (HC), six health posts and two referral hospitals [22]. Choma District also had a mostly rural (76%) population of 247,860 [21] with 29 HCs, six health posts and two referral hospitals. We would most likely expect to find an MWH at the HCs and hospitals, but not health posts, as they do not provide delivery services. Of those hospitals and HCs that offer delivery services, at the time of our survey 25 of Kalomo facilities had MWHs (86%), whereas only six in Choma had MWHs (19%) [22]. In Kalomo District during the time of our study, there was an ongoing initiative to improve maternal health by addressing the three delays through a public-private partnership called Saving Mothers, Giving Life (SMGL) [23–25]. The package of interventions implemented through SMGL since 2012 have included, among others, community mobilization and sensitization activities, improvements in referral systems, mentoring health staff, and investments in supply chain and equipment at facilities. There is evidence that SMGL had a significant impact on rates of facility delivery in Kalomo [26]. Therefore, the effect of this package of interventions is addressed in our analysis as a potential confounder. Data on facility delivery were obtained from the Zambia Chlorhexidine Application Trial (ZamCAT), a cluster Randomized Controlled Trial (RCT) in which 39,797 pregnant women in six districts of Southern Province, including Choma and Kalomo, were enrolled at their first antenatal care (ANC) visit and followed through 28 days post-delivery [20, 27]. The goal of ZamCAT was to evaluate the effectiveness of using chlorhexidine cord cleansing to reduce neonatal mortality. The woman-level data included background characteristics, collected during the initial enrollment survey, and location of their delivery, collected during a survey after delivery at the 1 and 4 day postpartum household visits. Data for facility capacity for emergency obstetric care were obtained from a health facility assessment (HFA) tool conducted as part of ZamCAT between June and August 2013. Not all facilities in each of the two districts were included in the ZamCAT study. Facility-level criteria for inclusion in the study were: (1) an estimated 160 births per year in the catchment area, (2) routine provision of ANC services, and (3) willingness to participate. In Kalomo District, 22 facilities, all HC, were selected, representing 81% of all HC in the district and in Choma 18 facilities (HC) were selected, representing 62% of all HC in the district. The HFA tool captured basic indictors of capacity to perform maternal and newborn health signal functions and other indicators of routine maternity and newborn care. Signal functions are a set of medical interventions that address the direct causes of maternal death [28]. Full details of the ZamCAT trial are described elsewhere [20, 27]. In this analysis, the main outcome indicator was delivery at any facility, determined by location of birth (health facility or hospital) reported by the mother at the ZamCAT household postpartum visit. Indicators of woman’s socio-economic status (household wealth, education level), maternal demographics (age, marital status), and pregnancy characteristics (parity, ANC) were obtained from the ZamCAT enrollment questionnaire. The indicator for facility capacity for emergency obstetric care was a continuous score, calculated as the sum of the basic emergency obstetric and newborn care (EmONC) signal functions (maximum of 7) [28] and 1 point for each of the following: electricity, water, 24-h care, and availability of a skilled provider (defined by the WHO as someone trained to manage normal pregnancies and to identify, manage and refer complications) [3], for a total 11 possible points. Signal functions were assessed by asking the facility in-charge interviewee whether or not the function had been performed within the last 3 months at that facility. The primary independent variable was a composite quality score for the MWH. As part of a formative evaluation [19], the study team systematically assessed and inventoried the functional capacity of all existing MWHs in the selected districts (n = 31; 25 in Kalomo and 6 in Choma). The health facilities with existing MWHs were identified in advance by the district health offices. The composite quality score was created by utilizing both quantitative and qualitative data collected at each MWH, inclusive of a series of questions asked of a member of affiliated clinic staff as well as photos that captured the state of the physical structure/space, availability of a water source within 200 m, availability/state of a toilet and bedding, and availability/state of a gathering and cooking area. These items emerged from our literature review of the physical factors that may be important determinants of quality in an MWH. In seven cases, facilities did not have a separate MWH structure at the time of assessment, but instead utilized a clinic area as a designated space for waiting women. We evaluated these spaces using the same criteria. The scoring system for each criterion ranged from 0 (not available) to 5 (present and fully functional). There were a total possible 75 points if all criteria for each of the 15 components was rated highest. Based on their composite score, MWHs were also categorized into tertiles and labeled as “low”, “medium” and “high” quality. We limited our analysis to those catchment areas for which we had health facility and woman-level data from ZamCAT (n = 40 sites; 18 of these had an affiliated MWH, three had a designated area for waiting mothers, often the clinic wards, and 19 of these had no structure nor designated area). Our sample of sites excluded seven health facilities in Kalomo and three in Choma that had an MWH but for which we did not have ZamCAT woman-level data. Of these 10 excluded, three facilities were hospitals, as hospital catchment areas defined differently than for HCs, and thus were not assigned for randomization for ZamCAT. The other non-hospital health facilities/HCs were not included in the ZamCAT study because they did not meet the health facility inclusion criteria. We conducted bivariate analyses examining associations between background characteristics of the women in our sample and our outcome of facility delivery to determine what factors to control for in the regression analyses. We used the Pearson chi-square test for categorical variables and t-tests for continuous variables if the data were normally distributed or non-parametric Wilcoxon rank-sum tests if non-normally distributed. Any characteristics associated with outcome variables with p-value <0.20 were included in the adjusted logistic regression model. We used multiple logistic regression to assess the likelihood of facility delivery based first on the facility capacity score and then on the composite MWH quality score. We also regressed the primary individual-level outcome (facility delivery) against the category of the MWH quality (low, medium, high), adjusting for covariates that may have moderated the effect, such as sociodemographic characteristics and distance to the facility, as well as for facility capacity score and level of SMGL program implementation. SMGL level of implementation was defined by using three time periods: data on women who 1) delivered before or during January 2012, the time at which the SMGL rollout started; 2) women who delivered between February and August 2012 and may have had some exposure to the SMGL program; and 3) women who delivered from September 2012 to the end of the ZamCAT data collection in October 2013 and were most likely exposed to some level of SMGL activities. In Choma the SMGL implementation level was always zero as SMGL never operated in that district during those time periods. Quantitative data were analysed in SAS version 9.3 [29].
N/A
