Menu
Menu
Menu
Menu
In 2012, Saving Mothers, Giving Life (SMGL), a multi-level health systems initiative, launched in Kalomo District, Zambia, to address persistent challenges in reducing maternal mortality. We assessed the impact of the programme from 2012 to 2013 using a quasi-experimental study with both household- and health facility-level data collected before and after implementation in both intervention and comparison areas. A total of 21,680 women and 75 non-hospital health centres were included in the study. Using the difference-in-differences method, multivariate logistic regression, and run charts, rates of facility-based birth (FBB) and delivery with a skilled birth provider were compared between intervention and comparison sites. Facility capacity to provide emergency obstetric and newborn care was also assessed before and during implementation in both study areas. There was a 45% increase in the odds of FBB after the programme was implemented in Kalomo relative to comparison districts, but there was a limited measurable change in supply-side indicators of intrapartum maternity care. Most facility-level changes related to an increase in capacity for newborn care. As SMGL and similar programmes are scaled-up and replicated, our results underscore the need to ensure that the health services supply is in balance with improved demand to achieve maximal reductions in maternal mortality.
We analysed data from the Zambia Chlorhexidine Application Trial (ZamCAT), a cluster-randomized controlled trial in which 39,797 pregnant women in Southern Province were enrolled and followed through 28 days post-delivery. The goal of ZamCAT was to evaluate the effectiveness of using chlorhexidine cord cleansing to reduce neonatal mortality (Hamer et al., 2015; Semrau et al., 2016). The trial operated in 6 of the 10 districts of Southern Province, including Kalomo District. Within each district, study investigators randomly assigned a total of 90 health facilities (clusters) to either the intervention or control group. Eligible health facilities provided routine antenatal services and had at least 160 births in their catchment area each year. Pregnant women living in the facilities’ catchment areas were identified either at their facility-based antenatal care visits or during community-based antenatal care outreach and offered enrolment. According to their study group, babies received clean dry cord care (control group) or topical application of a chlorhexidine solution once per day until three days after the baby’s cord fell off (intervention group). Field monitors made five home visits to all study participants throughout the course of the study – one antenatal, within two weeks of enrolment, and four postnatal (days 1, 4, 10, and 28 postpartum). All women in the trial, regardless of study group, received a standard package of services that included: a clean delivery kit, referral to a clinic in the presence danger signs for either the mother or baby, and messages about newborn health. Of the 42,356 women screened for the study, a total of 39,679 (90%) participated. Of these, 94% completed the study through the one month postpartum visit. The study team administered a series of questionnaires to the pregnant women before and after delivery that captured both demographics and health behaviours. The WHO has identified a set of medical interventions that address the direct causes of maternal death, with seven of these interventions defining Basic Emergency Obstetric Care (BEmOC) and an additional two defining Comprehensive Emergency Obstetric Care (CEmOC) (WHO, UNFPA, UNICEF, & AMDD, 2009). Assessment of these core signal functions at the facility level allows for measurement of a facility’s capacity to handle obstetric emergencies. As one of these interventions addresses resuscitation of the newborn, this group of seven interventions are also referred to as Basic Emergency Obstetric and Newborn Care (BEmONC) in the literature. For the purpose of our study, we kept the original BEmOC indicators and indicators of emergency newborn care separate. In ZamCAT, health facility assessments were conducted at facilities where the trial enrolled participants (n = 90) in the six districts at both baseline (September–October 2011) and endline (June–August 2013). The health facility assessment tool was based on ‘Monitoring Emergency Obstetric Care: A Handbook’ (WHO et al., 2009) and included an assessment of an expanded set of indicators that covered routine, basic emergency, and comprehensive emergency care for both mothers and newborns (Gabrysch et al., 2012). Kalomo District was selected by Zambia’s Ministry of Health as one of the four intervention districts for SMGL in Zambia and the only one in Southern Province. SMGL activities were launched in Kalomo in early February 2012 with a three-phase rollout, starting with those facilities with the highest volume of deliveries. SMGL was fully operational in all 34 Kalomo rural health facilities by September 2012, nearly 20 months after the ZamCAT was launched. SMGL activities continued to operate for another 13 months through the end of ZamCAT in October 2013. Study investigators collected all facility, household, and individual-level ante- and postpartum data between February 2011 and October 2013. To quantitatively assess the SMGL programme’s impact on FBB and delivery with a skilled provider we used a retrospective pre–post non-equivalent comparison group design (Grembowski, 2001). We compared two cohorts of pregnant women from ZamCAT: one that delivered before SMGL was implemented, and one that delivered after SMGL rollout. We treated the women living in Kalomo as the intervention group, while women from three adjacent and socio-demographically similar districts in Southern Province where SMGL was not implemented served as the comparison group. We excluded two of the ZamCAT study districts that we determined to be significantly different from Kalomo in terms of the socio-demographic characteristics of study women. To examine facility capacity for maternity and newborn care, we also analysed quantitative data from the health facility assessments conducted both before and after SMGL implementation. For the individual-level analysis, the final sample included all women with birth outcome data during the pre- and during-intervention periods in Kalomo (n = 6477) and the comparison districts of Choma, Monze, and Mazabuka (n = 15,203). For the facility-level analysis, we included all ZamCAT non-hospital health centres in Kalomo (n = 22) and the three selected comparison group districts (n = 52 at pre and n = 53 at post). The primary outcome of the individual-level analysis was FBB, expressed as a binary variable and reported by the mother in the household survey. The secondary outcome was attendance with a skilled birth provider, defined as a woman’s report on the household survey as delivered by a ‘nurse’ or ‘midwife’, and expressed as a binary variable. The main treatment or independent variable was living in the SMGL intervention district (Kalomo) either before or after the programme was fully operational. Several potential moderating variables were included in the multivariate analysis, including individual-level factors such as age, level of education, literacy level, parity, and number of household members, plus socio-economic status. We used asset ownership as a proxy for socio-economic status by generating scores using an asset index developed through a principal components analysis, modified from the Zambia Demographic and Health Survey wealth index (Rutstein & Johnson, 2004), and indexing the households by score into quartiles. To measure the facility’s capacity to provide emergency obstetric and newborn care services, we used an expanded list of indicators of emergency obstetric and newborn care (Gabrysch et al., 2012) from the health facility assessments. Of the 23 proposed indicators that span multiple dimensions of facility care, we were able to assess 17 as well as all 4 general requirements of the health facility (24/7 service availability, at least 1 skilled provider, communication tools and referral system, and reliable utilities). Table 2 illustrates the 17 indicators we were able to assess in our study, including: all three routine obstetric care indicators, all seven BEmOC signal functions, plus the two additional CEmOC functions, two of three routine newborn care functions, two of seven basic emergency newborn care functions (of which one, resuscitation of a non-breathing baby, is also classified as a BEmOC function), and both of the comprehensive emergency newborn care functions. The proposed indicators not measured in our study include: infection prevention including hygienic cord care (routine newborn care), and antibiotics to mother if the baby was preterm, corticosteroids in preterm labour, alternative feeding if baby unable to breastfeed, injectable antibiotics for neonatal sepsis, and prevention of mother-to-child transmission of HIV (pMTCT) if the mother was HIV positive (basic emergency newborn care). Notes: List adapted from proposed obstetric and newborn functions (Gabrysch et al., 2012). Existing EmOC signal functions are in italic bold (from the WHO/UN Handbook). aAt least one nurse, midwife, general doctor, or OBGYN at facility. bFunctioning communication equipment (landline, mobile, or radio). This does not include private cell phones unless the facility reimburses for cost of phone calls. cFacility has a functioning motorised vehicle with fuel that is routinely available and can be used for emergency transportation or access to a vehicle in near proximity that can be used for that purpose. dFacility routinely has electricity for lights and communication (at a minimum) from any power source during normal working hours; there has not been a break in power for more than two hours per day during the past seven days. eThe toilet/latrine is classified using criteria: Flush/pour flush to piped sewer system or septic tank or pit latrine; pit latrine (ventilated improved pit or other) with slab; composting toilet. fImproved water source include the following: Piped, public tap, standpipe, tubewell/borehole, protected dug well, protected spring, and rain water. gThermal protection: drying baby immediately after birth, skin-to-skin contact with mother, wrapping, no bath in first six hours (AMDD, n.d.). hNewborn intravenous fluid kit available in labour ward. iNewborn oxygen available in labour ward. To assess the SMGL programme’s overall impact in Kalomo on FBB and delivery with a skilled attendant, we calculated a difference-in-differences estimate for the two groups. To account for potential bias in this method we also used run charts to both examine the parallel trends assumption and to detect statistically significant shifts in rates of FBB and delivery with a skilled birth provider. Run charts are simple visual analytic tools that are used widely in quality improvement work, including most recently in health care settings, as a way to examine improvements over time (Perla, Provost, & Murray, 2011). The rule used for detecting a shift was at least six contiguous data points above or below the median. We used multi-level logistic regression to test the net effect of the SMGL intervention by comparing intervention and comparison sites while controlling for potential confounders. We regressed the individual-level outcomes (FBB and delivery with a skilled birth provider) against a dummy interaction variable that we created by taking the product of time (pre-intervention vs. during intervention) by group (SMGL vs. comparison sites). To account for differences in the Kalomo and comparison site populations, we also matched individuals in the intervention district with a sample of women from the comparison districts who had similar observable characteristics. To do this, we calculated the propensity score for each individual based on the estimated probability that this person might be in the SMGL group (D’Agostino, 1998). We matched women using the socio-demographic characteristics and other predictors that were both different between the two intervention groups and strongly associated with the outcome of FBB in the overall study population. These included: mother’s age, mother’s tribe, mother’s education, parity, distance to a health facility, HIV status, and household asset quartile. Individuals in the comparison areas without near matches were excluded. With that data in hand, we created a comparison group of individuals that did not have exposure to SMGL but shared the same characteristics as the SMGL-exposed group. To do the matching, we used the ‘greedy 5->1’ algorithm (Parsons, 2001). Next, we measured the average difference in the FBB outcome variable between the participants and the non-participants. We then ran the regression model again with the intervention and propensity score-matched comparison group to estimate programme effects. For the analysis of the facility-level data, we conducted chi-square tests of significance between the two groups on the proportion of facilities that met the minimum requirement for each of the facility capacity indicators. We used SAS Version 9.3 (SAS Institute, 2011) for all analyses. The Boston University School of Medicine Internal Review Board and the University of Zambia Research Ethics Committee approved of the protocol and informed consent forms for ZamCAT.
N/A
