Mothers and their babies are at the highest risk of death during delivery and during the first week postpartum [1]. Even though 59% of births in sub-Saharan African countries now occur in health facilities [2], in 2018, approximately 2.476 million babies died during the first 28 days of life, with approximately 1 million dying during the first day of life and almost another 1 million within the first week [3,4]. In Mali, an estimated 67 stillbirths [5] and an additional 75 newborn deaths [6] occur on average every day; the predominant causes are birth asphyxia, prematurity, and sepsis [7]. The maternal mortality ratio in Mali is 562 per 100,000 live births [3], almost 2.7-fold the global rate [3]. Most of these deaths are because of factors related to the place of delivery and quality of care [4,8]. In particular, the lack of appropriately trained staff in facilities, delays in referrals, and inadequate supplies and equipment are evident in many resource-poor settings [9]. This suggests that high-quality care, especially during late pregnancy, childbirth, and the early newborn period, are essential [10], and simple evidence-based interventions are likely to meaningfully reduce the number of maternal deaths, stillbirths, and early neonatal deaths [11,12].

Traditional capacity-building models for training health care providers, delivered through off-site classroom-style lectures or directed reading, are still the norm in low- and middle-income countries [8,13,14]. However, to disrupt the delivery of services, these approaches have been found to have limited (if any) impact on learning outcomes, and any knowledge and skills gained are neither transferred to other coworkers at the facility nor translated into improved provider practice and performance [13,15]. Moreover, there is no evidence that these methods contribute to knowledge retention after training, and limited conclusions can be drawn about their long-term impact on measures of performance in terms of service quality and maternal and newborn outcomes. A review of intervention studies aimed at improving the performance of health workers in low- and middle-income countries has suggested that problem-based learning and competence-based learning [15], including supervision and audits with feedback [13,14], along with brief team learning sessions within the workplace (health facility), using simulation, followed by simulation-based practice and feedback, can improve learning outcomes and provider performance compared with training approaches that do not include subsequent practice [13,16].

Over the last few years, an increasing number of African countries have implemented competency-based training of health care providers with the goal of improving maternal and child health outcomes and reducing perinatal mortality [17-19]. Low-dose, high-frequency (LDHF) training is a type of competency-based training that promotes maximal retention of clinical knowledge, skills, and attitudes through short, targeted, simulation-based learning activities, delivered during several in-service sessions and reinforced with structured practice sessions [20].

In that vein, in response to a World Health Organization report on health systems and policy indicators documenting that Mali did not have a policy on maternal, perinatal, and neonatal death audit (World Health Organization 2014), Mali’s Ministry of Public Health and Hygiene made it mandatory to report maternal, perinatal, and neonatal deaths and adverse events within 48 hours of their occurrence, through the Integrated Disease Surveillance and Response Network, and to convene an audit of the events within 15 days of the notification.

Mali’s national directive to implement facility-based maternal and perinatal morbidity and mortality audits offers an opportunity to leverage it and target training on skills and competency gaps identified during audits. However, to the best of our knowledge, the potential of using facility-based maternal and perinatal morbidity and mortality audits to identify areas that require attention for subsequent LDHF hands-on training has not yet been exploited. Moreover, even though LDHF hands-on training has demonstrated a significant positive impact on maternal and early neonatal outcomes, including reduced rates of fresh stillbirth and first-day mortality [17,21], retained placenta, and postpartum hemorrhage [19]. Limited conclusions can be drawn about the impact of LDHF hands-on training on maternal and neonatal care practices during intrapartum and postpartum periods and other perinatal outcomes, especially 7-day mortality and morbidity. There is very little known about the cost-effectiveness and the processes of implementation of LDHF training [22,23].

This study proposes a strategy to leverage maternal and perinatal morbidity and mortality audits being implemented in Mali, as a basis for identifying gap areas that can be addressed by a case-based focused LDHF on-site training strategy within primary health care–level facilities (Centre de Santé Communautaire [CSCOM]). These trainings will also target community-based causes of mortality and morbidity in health facilities. The trial will assess the effectiveness of LDHF hands-on training informed by maternal and neonatal mortality audits in reducing perinatal mortality and in increasing adoption of key maternal and newborn care practices during the delivery and the postpartum period. Secondary objectives include assessment of the implementation process and the cost-effectiveness of the intervention.

The study will be conducted in 4 districts in the Koulikoro region, Mali. The study area was selected as it is part of a larger project implemented by the Canadian Red Cross in collaboration with the Malian Red Cross (Croix-Rouge Malienne). This project is an integrated 4-year (2016-2020) maternal neonatal and child health (MNCH) intervention that targets 6 districts across the Koulikoro and Sikasso regions and is funded by the Government of Canada. The project supports the primary and secondary health facilities (CSCOMs and Centre de Santé de Référence [CSREF]—referral health centers) to improve the quality of their MNCH services and strengthen the health management information system. At the community level, the Malian Red Cross has a strong presence because of the training and support of local community health workers and volunteers for MNCH-related activities at remote integrated management of childhood illness sites.

The Koulikoro region is the second-most-western region of Mali. It covers an area of 90,120 square kilometers. With an estimated population in 2017 of 3.1 million and a density of 26.8 people per square kilometer. The Koulikoro region is separated into 7 districts: Kati, Kangaba, Koulikoro, Kolokani, Nara, Banamba, and Dioïla. The Canadian Red Cross and Malian Red Cross program encompasses the latter five. Of these, 4 districts (Koulikoro, Kolokani, Banamba, and Dioïla) will be included in this study.

This pragmatic (effectiveness) cluster randomized controlled trial will be implemented in 84 community health facilities (CSCOM) in the Koulikoro region, Mali. Facilities will be randomly allocated into either of two intervention arms or the control arm.

Community health facilities, including community health workers serving the catchment area, are the units of randomization (clusters). Although there are some differences in the services provided at these facilities according to the baseline facility assessment conducted in December 2016, only those health facilities providing basic obstetric care with at least one health care provider trained in the active management of the third stage of labor (AMTSL), essential newborn care (ENC), perinatal care (PNC), and those who had not yet started implementing the audits are considered for inclusion in the study. All facility deliveries (both mother and newborn) within study clusters and live home births referred to a facility within the perinatal period are included in the trial, subject to informed consent. In that respect, a master list of 84 facilities meeting the inclusion criteria within the 4 districts will be generated.

Health facilities were randomly allocated into one of the three study arms, resulting in 28 facilities per study arm. The randomization of clusters was performed by an independent epidemiologist not directly involved in the research project using a pseudorandom number sequence generated using Excel (Microsoft). Owing to the nature of the intervention, blinding of study participants and data collectors is not possible; however, outcome data will be collected by independent data collectors not involved in the intervention delivery. The study setting (Figure 1), recruitment strategies, random allocation sequence, intervention implementation, and data collection points (Figure 2) are summarized below.

The group of facility-based health workers and community health workers within the 4 health districts of the Koulikoro region within each facility (unity of randomization) are eligible to enter the trial. Eligible facilities are those that provide basic obstetric care and have at least one facility-based health care provider trained in AMTSL, ENC, and PNC. The eligibility assessment of facilities was made possible through the data collected during the baseline facility assessment conducted in December 2016.

Facilities within the 4 health districts of the Koulikoro region that did not provide basic obstetric care or those who did not have any health workers who received training in maternal and neonatal health care providers in AMTSL, ENC, and PNC have been excluded. Rural maternities were not eligible, as they are not included in the national audit policy.

The intervention uses a problem-based learning approach that uses the information emerging from the implementation of maternal and newborn audits to improve health worker skills training. This training is delivered by a study-trained facilitator, employed by Mali’s Ministry of Public Health and Hygiene, that attends and provides feedback during the audit meeting, followed by an immediate (next day) one-day in-service case-based focused training [8,13,14]. Specifically, each training cycle starts with the implementation of a maternal and neonatal morbidity and mortality audit, during which the trainer identifies gaps in knowledge and skills. This is followed by on-site LDHF hands-on training of maternal and neonatal health care providers on the skills and competency gaps highlighted by the audits. The training sessions also include direct observation of care, supervision, and use of simulations. In addition, community health workers are also trained when issues identified during the audit are related to their practice, such as in the identification of danger signs in the community or timely referral to the facility.

We hypothesize that the provision of on-site case-based LDHF hands-on training of maternal and neonatal health care providers, focused on gaps identified through clinical audits, will improve quality of care, thus decreasing adverse events during labor and the immediate postpartum and perinatal period in facility births, in comparison to traditional off-site generic training. The training of community health workers will improve community management via identification of danger signs and timely referral during pregnancy, labor, and the postnatal period (Figure 3). Ultimately, the aim of the intervention is to address key barriers to improving maternal and perinatal health by focusing on increasing the capacity of maternal and neonatal health care providers along with improving the recognition of danger signs and referral during the natal and postnatal period by community health workers.

This cluster randomized trial aims to assess whether LDHF hands-on training, informed by maternal and neonatal mortality audits in routine clinical settings, reduces perinatal mortality. Secondary objectives include estimating the effectiveness of the intervention in improving maternal and neonatal care practices among health workers and the comparative cost-effectiveness of the case-based focused LDHF hands-on training, compared with traditional training.

The perinatal mortality rate per 1000 live births occurring in each facility and the corresponding catchment area for the births that had taken place in the facility will be used to assess the effectiveness of the intervention. It is defined as the sum of fetal deaths (stillbirth), defined as the death of a fetus weighing 500 g or more, or of 22 weeks of gestation or more, occurring in health facilities, and early neonatal death, namely, the death of a live born within the first 7 days of life occurring in the facility or in the community. As health facility studies are not an appropriate source of data for calculating perinatal mortality incidence [24] unless all babies are born and stayed in a health facility for 7 days, the data on perinatal mortality will be extracted from both facility-level data and household survey questionnaires.

The secondary outcomes are as follows:

A sample size of 600 births per cluster was calculated to have 80% power to detect a reduction in the perinatal mortality rate from 30% to 27% live births, which corresponds to a 10% reduction in the perinatal mortality rate between intervention and control arms (α=5%, intracluster correlation of 0.003, and coefficient of variation of cluster size 0.9 and a 0.96 correlation of baseline measures with outcome). The number of clusters available per arm was 28. The expected average number of deliveries per month in each cluster was approximately 38, resulting in 16 months of follow-up. Two additional months of data collection are planned to account for losses and refusals and allow for adjustments in the analysis, resulting in a final average cluster size of approximately 675 births over 18 months.

This operational research study aims to improve the practical knowledge of health workers to improve the quality of management of childbirth and PNC to reduce perinatal mortality. It is part of Mali’s ongoing efforts to achieve target 3 of the Sustainable Development Goals and whose results would be useful for potential replication or scale-up. We will valorize the findings by adopting strategies for dissemination and knowledge transfer. For this purpose, the results will be disseminated to the Ministry of Health and Health Care providers. We will also hold a deliberative workshop with project users and stakeholders to discuss the results and lessons learned from the implementation and effects of the intervention to provide guidance from the perspective of replication or scaling-up. In addition, we will disseminate the results among the scientific and political communities through publications in scientific journals and presentations at international conferences.

The distinctive feature of this intervention compared with the other LDHF interventions is that it adopts problem-based learning, including supervision and audits with feedback [14]. Thus, the audit component will identify knowledge gaps related to the maternal and neonatal care provided in the facilities and then identify the appropriate training modalities to correct deficiencies in the management of maternal and neonatal care around the delivery and in the postpartum period.

The analytical approaches will provide evidence about the benefits and costs of this pragmatic intervention to improve the quality of maternal and newborn care in Mali with the aim of reducing perinatal mortality. The results will provide policy makers and practitioners with crucial information to facilitate the translation of best practices in capacity strengthening for health care providers in the routine practice of maternal and child care services. Indeed, although LDHF hands-on training has been implemented and evaluated in some resource-limited settings with the aim of reducing maternal and perinatal mortality, previous studies have analyzed only a few outcomes [17,21].

Another distinctive feature of this trial is that it adopts a pragmatic and holistic approach, and the analysis will contain both qualitative and quantitative data collected over a period of 14 months. The trial will provide a better understanding of how LDHF hands-on training of health care providers and community health workers could complement routine care to improve maternal and child health outcomes in resource-limited settings. In addition, this trial occurs during routine health service delivery and will demonstrate the potential of routinely collecting health management information system data to analyze the effectiveness of a major public health intervention.

Despite these strengths, some limitations of this study need to be considered. The facility-level data collection instrument was developed based on existing consultation and delivery registries, as well as partograph forms. As data are not being collected by direct observation of health care providers but rather from the partographs and registries they fill out, we are also concerned with the completeness, reliability, and accuracy of data to be collected. To overcome this important limitation, refresher training on the use of partographs will be held in all health facilities that are part of the trial. This will strengthen the health care providers’ capacity to use the partograph and aim to improve data completeness and quality for this study.

Some data will be lost to follow-up if the newborn’s mothers are referred to higher-level health facilities. However, we estimate that this will be a minimal disruption to the trial, and we will report the number of these cases in our analyses. In addition, underreporting of infant deaths is usually greater for deaths that occur very early in infancy in low-income countries [24]. As the Ministry of Health is a stakeholder in this project, a strategy will be developed in cooperation with the study team to motivate health workers to improve data completeness, reliability, and accuracy. We will also triangulate mortality data collected in the health facility with the mortality data collected in the household surveys to assess the level of completeness and accuracy and to reduce the likelihood of underreporting of perinatal and maternal deaths in the facility. The data collector will also conduct a follow-up home visit, aided by community relays and volunteers, for all facility births to minimize losses to follow-up.

Finally, completeness and accuracy of recall, including age at death, may deteriorate with time and is also related to the skill and cultural sensitivity of the person carrying out the interview [24]. However, as the recall period was relatively short in this trial (<21 days), the risk of recall bias will be reduced.