SaferBirths bundle of care protocol: a stepped-wedge cluster implementation project in 30 public health-facilities in five regions, Tanzania

Background: The burden of stillbirth, neonatal and maternal deaths are unacceptably high in low- and middle-income countries, especially around the time of birth. There are scarce resources and/or support implementation of evidence-based training programs. SaferBirths Bundle of Care is a well-proven package of innovative tools coupled with data-driven on-the-job training aimed at reducing perinatal and maternal deaths. The aim of this project is to determine the effect of scaling up the bundle on improving quality of intrapartum care and perinatal survival. Methods: The project will follow a stepped-wedge cluster implementation design with well-established infrastructures for data collection, management, and analysis in 30 public health facilities in regions in Tanzania. Healthcare workers from selected health facilities will be trained in basic neonatal resuscitation, essential newborn care and essential maternal care. Foetal heart rate monitors (Moyo), neonatal heart rate monitors (NeoBeat) and skills trainers (NeoNatalie Live) will be introduced in the health facilities to facilitate timely identification of foetal distress during labour and improve neonatal resuscitation, respectively. Heart rate signal-data will be automatically collected by Moyo and NeoBeat, and newborn resuscitation training by NeoNatalie Live. Given an average of 4000 baby-mother pairs per year per health facility giving an estimate of 240,000 baby-mother pairs for a 2-years duration, 25% reduction in perinatal mortality at a two-sided significance level of 5%, intracluster correlation coefficient (ICC) to be 0.0013, the study power stands at 0.99. Discussion: Previous reports from small-scale Safer Births Bundle implementation studies show satisfactory uptake of interventions with significant improvements in quality of care and lives saved. Better equipped and trained birth attendants are more confident and skilled in providing care. Additionally, local data-driven feedback has shown to drive continuous quality of care improvement initiatives, which is essential to increase perinatal and maternal survival. Strengths of this research project include integration of innovative tools with existing national guidelines, local data-driven decision-making and training. Limitations include the stepwise cluster implementation design that may lead to contamination of the intervention, and/or inability to address the shortage of healthcare workers and medical supplies beyond the project scope. Trial registration: Name of Trial Registry: ISRCTN Registry. Trial registration number: ISRCTN30541755. Date of Registration: 12/10/2020. Type of registration: Prospectively Registered.

This is a continuous quality improvement (CQI) project, which will be implemented in a stratified stepped-wedge cluster (regions) randomisation approach (Fig. 1). Randomization was done using simple random sampling i.e. the first region, Manyara, was purposively selected for logistical and strategic reasons, whereas subsequent regions were selected randomly. Clusters will receive the intervention at different time points based on the randomization [45]. Prospective observational data will be collected before and after introduction of the intervention periods at each cluster. The before implementation data, collected for three to 12 months (Fig. ​(Fig.1),1), will serve as baseline/control data. Random and sequential crossover of clusters from control to intervention will be conducted until all clusters are exposed. More clusters will be exposed to the intervention towards the end of the study. Stepped-wedge cluster randomized trial in 30 health facilities in 5 regions This design is preferred because, firstly, it provides a robust scientific evaluation of health service delivery of interventions, and it does not involve individual patient recruitment, rather a cluster (i.e. health facilities in one region). Secondly, for ethical purposes it ensures that all clusters receive the proven beneficial interventions by the end of the project. Thirdly, not all health facilities will implement the bundle at the same time, but in a stepwise manner, providing a learning opportunity for more optimal implementation at the next facilities. Data collected at each facility and context-specific issues will inform local training models and use of effective plan-do-study-act (P-D-S-A) cycles as study implementation continues [46, 47]. The project will be implemented in 30 health facilities in 5 clusters (regions) in mainland Tanzania: Manyara, Tabora, Geita, Shinyanga, and Mwanza as described in Fig. ​Fig.1.1. These health facilities were selected in consultation with the Tanzania Ministry of Health Community Development, Gender, Elderly and Children (MoHCDGEC) and the President’s Office- Regional Administration and Local Government (PO-RALG) based on the following criteria: 1) high burden of maternal and perinatal mortality, 2) high volume of deliveries, and 3) alignment with the ministry’s strategic priorities for maternal and newborn interventions in the country. Table 1 shows total births, fresh stillbirth rates, macerated stillbirth rates, 7-days neonatal mortality and maternal mortality rates in 2019 for the 30 health facilities. The data was obtained from respective health facility records of the health management information system. All of the health facilities provide Comprehensive Emergency Obstetric and Basic Newborn Care services and have separate labour units and operating theatres. Total birth and mortality rate in the study hospitals (2019) NA Data was not available for that particular health facility, DH District Hospital, RRH Regional Referral Hospital, HC Health Centre The SBBC is a quality improvement combination of innovative clinical and training tools (Fig. 2) coupled with LDHF on-the-job training aimed at empowering healthcare workers to improve care around labour and delivery. Description of the innovative tools in the SaferBirths Bundle of Care package (Laerdal Global Health, Stavanger, Norway) The interventions will target three main areas: labour management, newborn resuscitation (HBB), and management of bleeding after birth (Fig. ​(Fig.2).2). SBBC will provide a combination of training and clinical solutions for these situations including a data management system that provides rapid local feedback (Fig. 3). The bundle is the result of 10 years of a multidisciplinary collaboration between international institutions from within and outside Tanzania. All the innovative tools were co-created with midwives and doctors working in maternity departments at Haydom Lutheran Facility (rural setting), Muhimbili National Facility (urban), Temeke Referral Regional Hospital (urban), researchers from Stavanger University Hospital/SAFER and engineers from Laerdal Global Health (saferbirths.com/publications). Interventions that will be implemented during the SaferBirths Bundle project period The SBBC interventions will be coupled with semi-automatic data capturing components, whereby clinical data through the use of Moyo, NeoBeat and the Liveborn App, and training data through the use ofNeoNatalie Live will be uploaded to a secure access-controlled research server. Data will be automatically processed, and basic statistics andanalyses will be easily available on local dashboards (Fig. ​(Fig.3).3). This data can be shared with local healthcare workers, used to facilitate PDSA cycles (Fig. ​(Fig.2)2) and for benchmarking between facilities. Such availability of data will enable a system with rapid and objective feedback on training, clinical quality of care, and patient outcomes (Figs. 4) to guide new efforts (PDSA-cycles) and potentially increase motivation for CQI. Specifically, each health facility will utilise the objective data-driven feedback to adjust ongoing LDHF on-site simulation training, targeting the identified gaps in clinical care. Integration of innovations, data capturing and feedback mechanisms SBBC is coordinated by Haydom Lutheran Hospital in collaboration with the two Tanzanian sectoral Ministries in Health (MoHCDGEC and PO-RALG), the Tanzanian Midwifery Association (TAMA) and the Paediatric Associations of Tanzania (PAT) to scale up and ensure that implementation is rooted in the local health organizations. Project investigators, in collaboration with experts from SAFER (a simulation and implementation center based in Stavanger, Norway), will conduct courses in newborn care, labour management and a basic simulation-based train-the-trainer course (SimBegin) of 15 selected national candidates from TAMA and PAT. These 15 participants will become national facilitators and conduct cascade trainings of facility champions on newborn care, labour management and simulation-based training in the five regions. Figure 5 describes the training cascade that will be implemented. Building this capacity within the professional associations means the value can be applied not only to the SBBC project, but also to other training initiatives around the country where simulation-based learning andragogy can be useful. Illustration of the planned training cascade and implementation strategy Two facility champions from each facility will be trained to become site facilitators (Fig. 4). In total, there will be 60 facility champions (2 × 30 facilities) trained by the national facilitators for 6 days in 5 batches (stepwise for each region) with focus on 1) the use of the innovative tools, 2) how to facilitate simulation based LDHF on-the-job trainings, and 3) collection and use of local data for rapid feedback and PDSA CQI. Regional coordinators (n = 5) will be oriented on the SBBC interventions and how to coordinate technical and administrative issues in each region. Standardised methods will be used to ensure that participants are sufficiently trained, including those outlined in WHO Standards for Improving Quality of Maternal and Newborn Care in health facilities [48]. Following the six-day training of facility champions, a five-day training will be conducted at each facility, targeting all healthcare workers and service providers at the facility maternity and newborn care units. The training will be divided into a theoretical educational part (2 days) and a clinical practicing part including simulations (3 days) (Fig. ​(Fig.4).4). The two facility champions at each site will assist 2 national facilitators to conduct these trainings. The 30 health facilities will be provided the innovative training, clinical and data management tools, which will remain at the facility beyond the project implementation period to sustain improvement efforts. With support from the project’s regional coordinators, the facility champions will be responsible for following up day-to-day duties in relation to the bundle and facilitating LDHF on-the-job trainings related to newborn resuscitation and management of various labour complications. Additionally, the facility champions will utilize the captured data to conduct scheduled debrief meetings to enable healthcare workers to reflect on the care they provide. To ensure that facility champions are well-supported in their new role, national facilitators, in collaboration with the regional coordinators, will conduct scheduled supportive supervision to provide in-house training and support on a regular basis. A mentorship program will be developed and implemented in close collaboration with simulation experts at SAFER to support especially the national facilitators (faculty development) (Fig. ​(Fig.44). Prior to introduction of the SBBC interventions at the facility level, situational analysis meetings will be conducted with clinical staff at the different sites in consultation with regional, council and facility health management teams. The aim will be to assess and identify bottlenecks related to readiness, availability, and quality of provided intrapartum care. The results of these assessments will be used to cater implementation of the bundle to meetthe needs of the respective health facilities. The study will include parturient women, their offspring, and healthcare workers in the 30 selected health facilities. Pregnant women at gestational age 28 weeks and above with a live foetus at labour admission will be enrolled in the study. All healthcare workers in the maternity ward (prenatal, labour, postnatal and neonatal wards) in the selected health facilities will be involved. We used Stata function stepped-wedge to calculate the power [50] with the following input parameters. An average of 4000 baby-mother pairs per year per health facility giving an estimate of 240,000 baby-mother pairs for a 2-years duration (i.e., an average of 1 year pre-implementation and 1 year post implementation), 25% reduction in perinatal mortality from 2 to 1.5% at a two-sided significance level of 5%. Assuming intracluster correlation coefficient (ICC) to be 0.0013 [51], the study power stands at 0.99. Purposive sampling will be used in the qualitative components of the study. Maximum variation sampling will guide the selection of participants for in-depth interviews and focus group discussion with healthcare workers, facility management team as well as mothers receiving care. This approach secures a wide variety of people of interest and consequently a broad range of perspectives to better understand contextual factors influencing implementation. Qualitative data will be collected until saturation is achieved. Research assistants (at least two at each site) dedicated for data collection and regional coordinators (one for each region) will be trained on data collection and quality control procedures, including correction of errors and safe transfer of data to the central research server. The clinical data (process and patient outcomes) collected at each facility will be entered into an electronic data collection system i.e. Open Data Kit (ODK) developed specifically for this project. Refresher training of research assistants will be conducted by the study investigators and data manager as needed. Data will prospectively be collected at all sites for a period of 3 years. Clinical process and patient data will not include personal information: instead a unique identification (ID) number will be provided. The regional coordinators will oversee the data collection processes at the regional level. Several datasets will be collected during both baseline and implementation phases. We will monitor implementation level data (e.g. facilitators and barriers for scale-up), process level data (e.g. the adoption of training, use and experiences with the SBBC tools), and outcome/output level data (patient outcomes, clinical actions, and biomedical signal data). A facility readiness and service availability assessment tool will be used to conduct bottleneck analyses of each health facility readiness for good quality intrapartum care. A quality improvement tool will be used to develop a specific plan based on the bottleneck analysis. Knowledge about CQI process implementation at SBBC sites will be collected using Logbooks and PDSA diaries. Systems will be established to document the dissemination of trainings and the frequency and quality (performance) of training among healthcare workers in each site over the 3 years. On-site simulation training data will be semi-automatically collected by the NeoNatalie Live simulator and uploaded automatically to the central server at Haydom. Healthcare workers’ knowledge and skills on intrapartum care before and after the initial 5-day training will be collected using multiple-choice questions and OSCEs. Their periodic skills retention tests will be done using OSCEs. The adoption of the SBBC will be further evaluated through in-depth interviews and focus group discussions with healthcare workers, caregivers, and facility leaders/management. Healthcare workers will be asked to ascertain acceptability and barriers of the new interventions. The perceptions of women about intrapartum care will be assessed with semi-structured interviews. Patient morbidity, mortality and clinical events/actions during intrapartum care data will be collected from patient case notes, labour and delivery registers. The use of clinical SBBC tools (Moyo, NeoBeat and Upright bag mask) will be recorded on separate data collection forms Biomedical heart rate signal data will be automatically collected by Moyo (foetal heart rate) and NeoBeat (newborn heart rate). In each facility, the research assistants will assess the quality and completeness of the data before uploading via the electronic data collection system into the central secure research server/database at Haydom Lutheran Hospital. Signal data collected by Moyo and NeoBeat will automatically upload to the research server. The regional coordinators will perform quality data check on a daily basis. A data manager, located centrally at Haydom, will be responsible for overseeing the data management process including quality control. The data manager will run queries and perform final quality control before saving the data on the secure research server. Any querry identified will be sent back to the regional coordinator who will communicate with the research assistants for resolution. Training data will be collected partly automatically by the NeoNatalie Live simulator and others will be manually entered into the database immediately after training. De-identified qualitative data will be collected using digital recorders. They will be transcript, processed and stored in password protected research computers with controlled access. After data correction and cleaning, the data manager will prepare summary tables of selected variables identified for rapid evaluation of study progress. The summary tables will be shared with the regional coordinators, health management team in the regions and health facilities, and the national and site facilitators. The feedback will then be shared with healthcare workers and utilized to assess the strengths and identify procedure-practice gaps which need improvement, in line with the PDSA model and CQI. Some facilities will have direct access to their own “dashboard” (Fig. ​(Fig.3)3) and can follow their daily statistics. A final data analysis plan will be established in collaboration with the respective ministries, principal investigators and statistical partners in the consortium to answer all the specific project objectives. For binary correlated data (e.g. before and after training practices for health care workers), the one-sided McNemar’s test will be considered. Differences in training attendance and knowledge and skills acquisition rates between health-facilities will be compared using both bivariate and logistic regressions. The training scores will be modelled as longitudinal data and compared over time using time series analyses and mixed model effects. Statistical process control methods will also be utilised. The clinical care indicators and perinatal/maternal outcomes will be analysed and compared both within and between clusters/facilities (and regions) by random intercept models factoring in the stepped-wedge design. Poisson regression models will be used to estimate the differences in rates of morbidity and mortality. Qualitative thematic content data analysis methods will be used to analyse qualitative data [52]. Qualitative data will be processed and analysed using COREQ guidelines [53].

N/A

• Dima Health