Menu
Menu
Menu
Menu
Background: Insufficient reductions in maternal and neonatal deaths and stillbirths in the past decade are a deterrence to achieving the Sustainable Development Goal 3. The majority of deaths occur during the intrapartum and immediate postnatal period. Overcoming the knowledge-do-gap to ensure implementation of known evidence-based interventions during this period has the potential to avert at least 2.5 million deaths in mothers and their offspring annually. This paper describes a study protocol for implementing and evaluating a multi-faceted health care system intervention to strengthen the implementation of evidence-based interventions and responsive care during this crucial period. Methods: This is a cluster randomised stepped-wedge trial with a nested realist process evaluation across 16 hospitals in Benin, Malawi, Tanzania and Uganda. The ALERT intervention will include four main components: i) end-user participation through narratives of women, families and midwifery providers to ensure co-design of the intervention; ii) competency-based training; iii) quality improvement supported by data from a clinical perinatal e-registry and iv) empowerment and leadership mentoring of maternity unit leaders complemented by district based bi-annual coordination and accountability meetings. The trial’s primary outcome is in-facility perinatal (stillbirths and early neonatal) mortality, in which we expect a 25% reduction. A perinatal e-registry will be implemented to monitor the trial. Our nested realist process evaluation will help to understand what works, for whom, and under which conditions. We will apply a gender lens to explore constraints to the provision of evidence-based care by health workers providing maternity services. An economic evaluation will assess the scalability and cost-effectiveness of ALERT intervention. Discussion: There is evidence that each of the ALERT intervention components improves health providers’ practices and has modest to moderate effects. We aim to test if the innovative packaging, including addressing specific health systems constraints in these settings, will have a synergistic effect and produce more considerable perinatal mortality reductions. Trial registration: Pan African Clinical Trial Registry (www.pactr.org): PACTR202006793783148. Registered on 17th June 2020.
We will use a stepped-wedge cluster-randomised design with a nested process evaluation based on realist evaluation [18] to evaluate the process of implementation of ALERT to understand what works for whom and under what conditions [19]. We will also conduct a cost-effectiveness analysis to inform scalability. The cluster design was chosen as the intervention will be delivered at the hospital level. Our clusters are defined as a maternity ward of a hospital offering caesarean section and blood transfusion services with a minimum caseload of 2500 births per year. A stepped-wedge design was chosen to mirror scale-up for policy buy-in and for statistical efficiency as we expect larger cluster-level differences [20]. In addition, it enables the realist process evaluation and economic evaluation to take place in hospitals where we expect the intervention to be sufficiently mature in the way it is implemented. This protocol follows CONsolidated Standards of Reporting Trials (CONSORT) for stepped-wedge cluster randomised trial (SW-CRT) (Additional file 1) and the Standards for Reporting Implementation Studies (StaRI) (Additional file 2). ALERT will be implemented in four hospitals in Benin, Malawi, Tanzania and Uganda. These countries were purposely selected to allow for a range of health system characteristics and implementation challenges. While Malawi, Tanzania and Uganda share many health system characteristics (strong public health structures, nurse-midwifery and non-direct entry into midwifery education), there are also distinct differences (Table 1). For example, Malawi and Tanzania have strong task-shifting policies in maternity care whereby mostly non-physician clinicians perform caesarean sections [22]. In Uganda and Benin, in contrast, caesarean sections are performed exclusively by medical doctors. In Benin, direct entry into midwifery education is practised and maternity care is thus largely provided by midwives. Characteristics of study countries CS caesarean section # DHS StatCompiler and Survey reports for Demographic and Health Survey data, Benin; 2017–8; Malawi: 2015–16; Tanzania: 2015–16; Uganda: 2016 ## Additional analysis of Demographic and Health Survey data, Benin; 2017–8; Malawi: 2015–16; Tanzania: 2015–16; Uganda: 2016 ^WHO observer http://apps.who.int/gho/data/node.main.HWFGRP_0020?lang=en The trial will commence April 2021 for 30 months (Fig. 2). Trial hospitals were selected purposely to reflect the range of facilities and include typical hospitals currently caring for 30–50% of all births for the respective country [23]. In March 2020, we consulted with national Ministries of Health and prepared a list of all hospitals meeting the selection criteria of i) minimum caseload of 2500 births per year required based on trial sample size calculation; ii) caesarean section and blood transfusion services available; iii) preferably located in rural districts; and iv) consisting of a mix of typical public but also private-not-for-profit (faith-based) hospitals. We included public and private-not-for-profit hospitals to reflect the typical landscape of hospitals in sub-Saharan Africa and improve our results’ generalizability. We then selected four hospitals in each country (Fig. 3). ALERT intervention implementation schematic. Light green indicates the comparison cluster. Dark green indicates the cluster is receiving the intervention. BJ: Benin, MW: Malawi, TZ: Tanzania, UG: Uganda Map of the ALERT countries with key indicators for the selected study hospitals. CS: Caesarean section The intervention directly targets health care providers involved in intrapartum care and all women who give birth in the participating hospitals during the study period. In this study, the term ‘maternity care providers’ refers to nurses, nurse-midwives, midwives, auxiliary staff and medically trained staff such as obstetricians working in the maternity ward at one of the study facilities. Women will be eligible if they give birth to a newborn weighing ≥1000 g, which is a proxy for viable gestational age in settings with poor gestational age measurement. Women who gave birth in another location but receive care in the hospital after childbirth will not be included in the study as our intervention targets the intrapartum period. Intervention development was conceived in response to the SC1-BHC-19-2019 call from the European Commission to innovate and evaluate interventions to bridge the knowledge-do gap to improve health during the first 1000 days of life. Further, our intervention links to the 2030 Sustainable Development Goal agenda [24] and the Survive, Thrive, Transform aspirations of the United Nations [25]. The ALERT intervention focuses explicitly on the key elements of intrapartum care of i) admission, labour monitoring; ii) immediate maternal and newborn care; and iii) readiness and care for complications (Additional file 3). Thus, ALERT will cover all stages of labour, biologically effective interventions (such as appropriate admission, foetal monitoring, emergency preparedness like reducing time from decision to perform a caesarean section) and improving experience of care (such as promoting companionship and communication in the maternity wards). Our intervention is based on previous research in conceptualising and evaluating care QI and training interventions [14, 26, 27] and learning from the large Safe Childbirth Checklist trial in India [28]. Key intervention elements are continuous training and QI based on the assumption that the combination of these two is needed to address the underlying causes of inconsistent implementation of evidence-based practices. Further, intervention development and adaptation rely on end-user participation to consider women, families, and health providers’ perspectives [29]. The design pays attention to the experience of interaction between people and health systems. Understanding health systems responsiveness offers an opportunity to adapt care to changing clients/patients’ needs, promote women’s access to effective interventions and improve the quality of health services, ultimately leading to better health outcomes [30]. The intervention will include several training modules based on competency-based methodology and using the Laerdal Global health Mama Birthie low-cost models [31]. The training will be made available to maternity providers, similar to the successful Helping Mothers and Babies Survive modules [27]. Mentorship is increasingly recognised as an effective strategy to improve healthcare quality, either as part of QI bundles or as a stand-alone intervention [32, 33]. The ALERT mentoring and leadership training intervention component will use a cascade approach with i) in-facility clinical mentors linked to the QI approach and training; ii) mentorship from in-country ALERT staff for the head of the maternity unit; and iii) mentoring within the international ALERT team. Mentoring will address individual professional attitudes, inter-professional collaboration (teamwork), leadership strengthening for resource negotiations, and other aspects. The QI intervention aims to i) support the consistent implementation of the trainings provided; ii) address operational deficiencies identified during the formative research as part of the end-user participation strategies; and iii) support linkages between established maternal death review teams as well as other hospital improvement structures. We will use standard Plan-Do-Study-Act methodology. Data for follow-up will come from the perinatal e-registry or registers adapted to the type of data. The intervention will be delivered by maternity care providers in the study hospitals and supported by our research teams who are based in national universities and well-placed to deliver training and engage with supporting QI approaches. Local hospital-based training and management resources will be mobilized and integrated. To support the ALERT intervention’s institutionalisation and sustainability, there must be strong leadership from the districts and collaboration with the Ministries of Health, training institutions, and integration into existing QI structures in each country. We further linked our training approach to training resources within the countries, thus trainers of trainers as available at national and subnational level. In line with Juran’s trilogy and the WHO, we concur that promoting the combination of quality planning, control and improvement allows for more sustainable interventions [34]. The QI intervention will be informed by the collaborative QI approach [35] and will explicitly link to QI approaches already implemented in the facilities. To bolster knowledge, an adapted QI refresher training will be provided including the PDSA and problem-solving methods. Bottlenecks identified during the health facility assessment, operational deficiencies identified during the ALERT competency-based training sessions, and recommendations arising from the maternal death reviews will be the target of PDSA cycles addressed by the QI team. The hospital-based QI team will be supported by our research team and the head of the maternity unit to develop and implement feasible, small scale solutions. We recognise the barriers described to consistent implementation of QI particularly in resource-poor and understaffed settings [14, 36, 37]. PDSA cycles, although widely used, have been associated with limited effects [26, 38]. With this in mind, we plan to make adaptations to the collaborative QI approach in order to increase the effectiveness of the ALERT QI package (see Table 2 in additional file 3). By explicitly linking to the existing QI structures including perinatal audit and management, we aim to ease implementation and improve potential scalability [41, 42]. The mentoring approach linking to central national institutions is expected to improve accountability to support the structured and regular implementation of QI and thereby the needed control aspect as well as the link to the local management structures. The end-user participation element of the intervention design will allow the incorporation of quality planning which the WHO is now proposing as an essential component of QI [38]. Primary and secondary outcome indicators aFresh stillbirth is defined a stillbirth that happened during labour at the respective facility, thus where the foetal heartbeat was positive at admission; bThe cut-off level may be revised based on data from an ongoing study in Uganda and validation work; cSevere maternal morbidity will be defined using pragmatic criteria of major interventions (hysterectomy, laparotomy, blood transfusion, admission to intensive care unit or referral to higher level facility) This study includes three evaluations; 1) stepped-wedge trial; 2) realist process evaluation and 3) economic evaluation. The methods for each are described below. Our primary outcome is in-facility early perinatal mortality defined as in-facility (fresh) stillbirth and 24-h neonatal mortality. Selected secondary and process outcomes are listed in Table Table2.2. For a sub-sample of births, we will use lactate measurement using a simple point-of-care test (Nova Biomedical, StatStrip Xpress-I lactate) to obtain an objective measurement of hypoxic-ischaemic insults to be used in conjunction with the more subjective APGAR score due to interrater differences. It is suggested that lactate provides good predictive values on hypoxic-ischaemic insults as conventional pH measurement and base excess [43, 44]. Breastfeeding initiation will be assessed using information recorded in the perinatal e-registry and women’s reports at the time of discharge will be integrated into the exit interviews to determine responsiveness and experience of mistreatment. The primary and secondary outcome data will be collected through a perinatal e-registry and exit interviews with women being discharged following childbirth. The perinatal e-registry will include standard indicators of pregnancy risks and care received during the antenatal and perinatal period. The indicators were informed by similar clinical data collection in the European Union [45] and Tanzania [46]. We will support standardised admission and follow-up case notes to improve continuous documentation during care provision. After short training sessions facilitated by research staff, data will be entered continuously in the maternity ward by midwifery staff or data clerks (based on country preference). Exit interviews will be administered by research staff every six months during the implementation period (six time points) to 50 randomly selected women who had given birth in each hospital. To assess responsiveness and mistreatment, we will use a recently validated questionnaire with some adaptations [47]. The ALERT perinatal e-registry will provide primary and some secondary outcomes and will be implemented in all study hospitals. All women who meet the eligibility criteria will be included. The perinatal e-registry data will be entered on the maternity ward using the Research Electronic Data Capture (REDCap) platform available on tablets or computers [48]. The programme will have inbuilt ranges and branching logic programmed to improve data quality. Monthly data checking and feedback to providers will also be implemented. Weekly paper-based summary sheets will be used to check data completeness, and double-entry of data for 10% of the records will be done by an external facility supervisor. Supervision structures will include in-hospital supervision by an external resource to the maternity ward and by an ALERT research team data manager. The Hemming et al. formula for stepped-wedge trials was used to calculate the study’s power [49]. We used intra-cluster correlation coefficients for stillbirth and neonatal mortality from a study in Malawi [50] and maternal morbidity from a recent trial [51]. The inclusion of 16 hospitals, each with at least 2500 births per year, will give sufficient power (75–80%) to detect a 25% reduction among in-facility early perinatal mortality with baseline rates between 1.4 to 2.0% and 95%-confidence intervals. We also have sufficient power to assess several secondary outcomes, including maternal morbidity (Additional file 4). Randomisation will be stratified by country to ensure that hospitals are randomly selected and enrolled in six-monthly steps (four hospitals in four steps) in each country. Randomisation was performed by a statistician, independent from the implementation team, once the hospitals had consented to participate in the study. As with all training and QI interventions, we cannot blind participants (hospitals) to the intervention. However, women and families might not be aware of the exact step in the implementation of ALERT at the hospital where they give birth. The statistical analysis will be “intention-to-treat”, comparing ALERT intervention clusters (hospital maternity wards) with comparison clusters where care is provided according to national standards. We will define a “transition” period of two weeks during which the intervention is provided and adopted by the respective hospitals. We will use descriptive analysis to review the trends using interrupted time series analysis from the 30 months of data collection through the perinatal e-registry [52]. Seasonal variations will be described (e.g. birth weight and neonatal mortality) [53, 54]. While secular declines in stillbirths and early neonatal mortality have been slow in the past; we expect annual declines of at least 2% [55]. We will review secular trends over strata (countries) and clusters (hospitals) and estimate the heterogeneity of the effects across clusters as advised by Hemming et al. [56] Considering the limited number of clusters, we will use generalised estimating equations (GEE) adjusting for clusters and for the small sample [57]. We will adjust for clustering, time-trends and the sequence of inclusion of hospitals [56] and other methods to perform small-sample adjustments [58]. Additional sub-group analysis by stratification on select covariates such as birth weight, mode of delivery, time of delivery and type of outcomes will be conducted based on the power and sample size plausibility. We will assess how the intervention works by assessing how actors take up and implement the intervention components based on mechanisms that are triggered in specific contexts to generate the outcomes. We will analyse the differential effects of interventions in the ALERT settings: why is an intervention successful in one setting but perhaps without effect in another? The evaluation is structured along the realist research cycle [59], which starts with the development of an initial programme theory (Fig. 4). The realist research cycle [59] The initial programme theory will be developed based on the ALERT theory of change, a review of the most current literature and discussions with ALERT researchers. We will adopt a multiple embedded case study design to test the initial theory. In each country, we will select one hospital where the ALERT intervention is implemented at step 1 of the stepped wedge design and a second hospital involved in step 3. This phased recruitment of facilities will enable us to assess how the length of exposure to the intervention and changes over time influence observed outcomes. Selecting two hospitals per country will allow for cross-case comparison and identifying how mechanisms play out differently in different hospital-specific contexts. In each hospital, data will be collected on the implementation of the intervention, the context, the actors and the processes triggered by the intervention through a document review and interviews with (1) hospital managers, heads of maternity, midwives and district directors of health, and (2) mothers and families. We will also draw upon data from other work packages and participatory reflection sessions at ALERT consortium meetings. Audio-recordings will be done on devices which allow data encryption and in case this is not possible, recordings will be immediately transferred to a computer where data can be encrypted. All recordings will be verbatim transcribed and translated to English where needed. All transcripts will be entered in a NVIVO database and pseudo-anonymized to the highest degree possible using identifiers and codes for key variables. The ICAMO heuristic will be used in the data analysis [60]. The data will first be categorised using the intervention-actor-context-mechanism-outcome configuration. Next, a retroduction approach will be adopted, whereby explanations for the observed outcomes are identified by looking into the mechanisms, actual intervention modalities, actors and context elements. In-case and cross-case analysis will allow for the formulation of the ‘final’ programme theory, which will indicate what it is about the ALERT intervention that works for whom and in which circumstances. This will inform recommendations for scaling up the intervention and tailoring it to different contexts. Closely linked with the realist and effect evaluation will be an economic evaluation of the ALERT intervention. According to Drummond et al., economic evaluation is defined as “the comparative analysis of alternative courses of action in terms of both their costs and consequences” [61]. We propose to evaluate the economic impact of the ALERT intervention by conducting cost-effectiveness analysis, focusing on the mature intervention implemented in step three hospitals. The incremental cost-effectiveness ratio (ICER) will focus on the net costs per one reduction in stillbirth and in-facility perinatal mortality. We will utilise the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) to optimise the reporting of our evaluation [62]. We will follow the European Union’s open-access policy and strive to make all ALERT training modules, reports, and scientific articles publicly accessible. We will utilise the following dissemination channels: leaflets, a website (alert.ki.se), workshops and meetings at the local (district) and national level, conference presentations (local, national, and international), and peer-reviewed publications and reports. Key stakeholders, including the study participants (i.e., end-users), Ministry of Health policymakers, and other stakeholders interested in improving maternal and child health will be proactively sought out, and findings from the study shared.
