A Comprehensive Approach to Improving Emergency Obstetric and Newborn Care in Kigoma, Tanzania

Introduction: To address high levels of maternal mortality in Kigoma, Tanzania, stakeholders increased women’s access to high-quality comprehensive emergency obstetric and newborn care (EmONC) by decentralizing services from hospitals to health centers where EmONC was delivered mostly by associate clinicians and nurses. To ensure that women used services, implementers worked to continuously improve and sustain quality of care while creating demand. Methods: Program evaluation included periodic health facility assessments, pregnancy outcome monitoring, and enhanced maternal mortality detection region-wide in program- and nonprogramsupported health facilities. Results: Between 2013 and 2018, the average number of lifesaving interventions performed per facility increased from 2.8 to 4.7. The increase was higher in program-supported than nonprogramsupported health centers and dispensaries. The institutional delivery rate increased from 49% to 85%; the greatest increase occurred through using health centers (15% to 25%) and dispensaries (21% to 46%). The number of cesarean deliveries almost doubled, and the population cesarean delivery rate increased from 2.6% to 4.5%. Met need for emergency obstetric care increased from 44% to 61% while the direct obstetric case fatality rate declined from 1.8% to 1.4%. The institutional maternal mortality ratio across all health facilities declined from 303 to 174 deaths per 100,000 live births. The total stillbirth rate declined from 26.7 to 12.8 per 1,000 births. The predischarge neonatal mortality rate declined from 10.7 to 7.6 per 1,000 live births. Changes in case fatality rate and maternal mortality were driven by project-supported facilities. Changes in neonatal mortality varied depending on facility type and program support status. Conclusion: Decentralizing high-quality comprehensive EmONC delivered mostly by associate clinicians and nurses led to significant improvements in the availability and utilization of lifesaving care at birth in Kigoma. Dedicated efforts to sustain high-quality EmONC along with supplemental programmatic components contributed to the reduction of maternal and perinatal mortality. The outcomes and impact of the Program were evaluated from 2013 to 2019 with objectives to (1) assess capacity, functionality, effective coverage, and quality of routine and EmONC care; and (2) measure maternal and newborn outcomes. In 2013, the Program established a comprehensive data system for monitoring and evaluation throughout the region, complementing the aggregated routine health management data reported to the Ministry of Health. The system included periodic data collection from all public and private health facilities that conducted at least 90 deliveries per year, regardless of program-support status. Between 2013 and 2019, the number of facilities providing delivery care included in the assessment increased substantially, as a reflection of increased facility-based care at birth in the region. Originally, 127 health facilities were included in the 2013 evaluation and were revisited in subsequent years. The latter evaluations also included facilities that had recently started to provide delivery care and thus were not captured in 2013 (47 added in 2016 and an additional 23 in the 2018 and 2019 evaluations). This approach allowed the assessment of the system-wide increase in capacity and functionality of maternal care services and its relationship with pregnancy outcomes. When compared to the routine health management and information system in Kigoma, facilities included in the endline evaluation provided care for 95.5% of deliveries in the region.42 From 2013–2019, the number of facilities providing delivery care included in the assessment increased substantially, as a reflection of increased facility-based care at birth in the region. The health facility assessments (HFAs) evaluated facility infrastructure, availability of equipment and supplies, essential drug stocks, staffing, ability to provide routine obstetric and newborn care, capacity to collect routine maternal and child health data, and performance of EmONC interventions.43 The assessments were first conducted in 2013, and all HFAs were conducted by CDC personnel and Tanzanian data collectors. Subsequent assessments took place in January 2016, January 2018, and February 2019.43 Availability of services and items that are essential for obstetric and newborn care were assessed by observation and direct verification. EmONC functionality was assessed based on the performance of the EmONC signal functions44 in the 3 months before the data collection. A detailed description of the assessment methodology can be found elsewhere.43 The health assessment questionnaire is included in Supplement 1. In 2013, CDC developed a system and tools (Supplement 2) for periodic data collection of all births that occurred in health facilities providing maternity care in the region.43 The approach was designed to collect individual observations on all women who delivered in the facilities where the HFAs were conducted. Information on each delivery included maternal characteristics, the obstetric diagnosis, delivery type and outcome, newborn characteristics and care, obstetric surgeries, postpartum pregnancy complications, and status of mother and baby at discharge. Data were extracted from all available inpatient logbooks, patient records, audits, and morgue registers. In hospitals and health centers, individual patient data were triangulated across various sources—such as labor and delivery, postpartum, female ward, surgical, admission/discharge registers, and hospital morgues—to ensure completeness of information. In dispensaries, where only the maternity service statistics register documents delivery and postpartum events, individual data did not need triangulation; however, in a few instances, maternal or perinatal deaths that occurred in dispensaries were identified in audit data performed at the hospital or district level; these outcomes were linked to the dispensary of occurrence. Data were collected retrospectively for 12–30 months at a time using specially designed inventory and data extraction and triangulation tools. Information on women with direct and indirect obstetric complications was classified according to ICD-10 standards. Women with multiple complications were classified according to their most severe direct obstetric complication; if they only suffered indirect obstetric complications, they were classified according to the most severe indirect complication.45 Facility-based maternal deaths were identified among all female deaths that occurred in the facility using all registers available at admission and discharge, all relevant wards, operating rooms, and the morgue. The number of data sources, their quality, and completeness improved over time. Additionally, in 2015, the Ministry of Health introduced facility death registers in all hospitals and health centers. Death registers were kept on each ward and captured individuals’ information related to age, sex, and cause of death. Tallied monthly, the quality of maternal and neonatal mortality data gradually improved. The percentage of maternal deaths with unspecified cause of death, for example, declined from 10% to 2% between 2013 and 2018. All information related to maternal deaths collected from service statistics sources were reviewed by 2 CDC obstetricians and 1 Tanzanian medical doctor not associated with the implementation activities. Perinatal deaths were originally extracted from registers maintained in labor and delivery, postnatal and neonatal wards, and the morgue. Data inventory and tools were expanded in 2018 and 2019 to capture additional registers: the death registers, which also captured perinatal deaths; registers in newly opened neonatal intensive care units and Kangaroo Mother Care corners; and, a separate morgue stillbirth and newborn death register. Thus, the enumeration of perinatal deaths was likely to be more complete for 2016–2018 than in earlier periods. We describe the evaluation results from data collected in 2013, 2016, and 2019. Data collected in 2013 refer to EmONC performance and outcomes in 2013, those collected in 2016 refer to performance and outcomes in 2016, and those collected in 2019 refer to performance and facility pregnancy outcomes in 2018. Using the total population figures for Kigoma region from the 2012 Population and Housing Census and the region-wide growth coefficient, we estimated the total population for 2013 and 2018.46 To estimate the annual number of births in 2013 and 2018, we multiplied the annual population with the crude birth rate derived from the Kigoma reproductive health surveys conducted by CDC in 2014 and 2018.47 For our analyses, we focused on measures related to the availability of routine and emergency obstetric care and indicators recommended to assess coverage, utilization, and quality of EmONC services.44 We also assessed essential obstetric and newborn care capacity using: (1) general facility infrastructure (in terms of availability of uninterrupted power supply, clean and safe water supply, communication and emergency transport availability); (2) availability of trained staff (at least 1 staff member trained in EmONC in the previous 1–3 years); (3) availability of supplies and essential medicines; (4) performance of routine maternal care (services available 24/7, use of partographs, and use of active management of the third stage of labor); (5) performance of essential newborn care (initiation of immediate breastfeeding, skin-to-skin, and promotion of kangaroo mother care); and (6) availability of protocols, guidelines, and forms needed for conducting service delivery. EmONC services are defined by a set of lifesaving interventions, or “signal functions,” recommended by the World Health Organization (WHO) to treat the major direct obstetric complications.44 BEmONC interventions include administration of parenteral antibiotics, uterotonics, and parenteral anticonvulsants; manual removal of placenta (MRP); removal of retained products; assisted vaginal delivery (AVD); and basic neonatal resuscitation. CEmONC includes 2 additional services: performance of obstetric surgeries (e.g., cesarean delivery) and performance of blood transfusion. Facilities were classified based on whether they had, within the previous 3 months, performed each of these signal functions. Because AVD—using either forceps or vacuum extractor—is relatively uncommon in Tanzania, some facilities were classified as CEmONC or BEmONC even if they did not perform AVDs within the past 3 months (i.e., CEmONC-1 and BEmONC-1). Facility assessors looked at evidence of whether each of the EmONC signal functions had been used in the 3 months before the study, if the required drugs and/or equipment were present, and if health providers at the facility had the training to perform the service. Evidence of all these elements was required to record that a signal function was performed/available. We examined indicators of EmONC utilization, as recommended by WHO,44 (institutional delivery rate, population cesarean delivery rate, met need for obstetric care, and direct obstetric case fatality rate [CFR]) and other key outcome indicators (facility maternal and perinatal mortality, neonatal and stillbirth rates) in 2013 and 2018 and compared them by facility type and program support status. Program support status was defined based on the receipt of various interventions (Table 1). To evaluate the efficacy of the Program in Kigoma region, we assessed health facility and pregnancy outcomes in public PS facilities and public and private non-program-supported (NPS) facilities at 3 points in time: 2013 (before the program interventions were scaled up), 2016, (after interventions were scaled up to additional centers), and 2018 (after dispensaries were added to the Program in 2016 and before closing the Program). Before-and-after comparisons can only be performed for health facilities that initiated support after 2013. A baseline for facilities that started their interventions before 2013 cannot be established. We present descriptive results as percentages, means, rates, and ratios by facility type, program support, and ownership status. All analyses were performed using SAS v. 9.6 software. Statistical tests were computed for rates and ratios only, using z-statistics.