ABSTRACT
Introduction: Many pregnancy-related deaths can be avoided if health care workers reliably provide prompt, evidence-based care during the final stages of pregnancy and delivery. We report the impact on the institutional maternal mortality ratio (iMMR) of a quality improvement (QI) initiative that focused on establishing timely and reliable obstetric care in public Brazilian hospitals for the primary causes of maternal deaths.
Methods: The QI initiative used a quasi-experimental time-series design implemented in 19 Brazilian maternity hospitals comparing 3 periods: baseline (January 2018 to November 2019), implementation (December 2019 to March 2021), and post-implementation (April 2021 to September 2021). We used a sequential approach, referred to as the “4Rs” (Recognize, Rescue, Reassess, and Refer), with the Modified Early Obstetric Warning Score (MEOWS) as a key tool, to identify clinical deterioration and implement care bundles directed at the management of 3 principal life-threatening conditions around the time of birth: postpartum hemorrhage, sepsis, and hypertensive disorders of pregnancy (HDPs). Clinical staff received tailored training and support within a structured learning system that brought multiple teams together to rapidly test and implement evidence-based changes.
Results: Comparing the baseline with the implementation period, run charts detected an overall reduction of 34.2% in all causes of maternal deaths (from 83.7 to 55 deaths per 100,000 live births). Additionally, the iMMR due to the 3 analyzed life-threatening conditions decreased by 60.9% between baseline and the combined implementation and post-implementation periods (from 44.1 to 17.2 deaths per 100,000 live births); postpartum hemorrhage-related deaths were reduced by 72.9% (from 11.7 to 3.17 deaths per 100,000 live births) and sepsis-related deaths were reduced by 100% (from 20.4 to 0 deaths per 100,000 liv births). No changes by HDP-related deaths were observed. There was a 178% increase in iMMR by all causes (mainly attributed to COVID-19) during the post-implementation period, but deaths related to life-threatening conditions either remained stable or decreased even further.
Conclusion: Optimization of the steps in a care sequence and reliable implementation of care bundles directed at the principal life-threatening conditions around the time of birth, using QI, tailored training, and the MEOWS tool, seem to be feasible and promising approaches for reducing iMMR.
BACKGROUND
Circumstances leading to maternal death are complex and multifactorial, but up to 80% of peripartum deaths are thought to be preventable.1 Across the globe, the leading causes of mothers’ death are postpartum hemorrhage (27%), hypertensive disorders of pregnancy (HDPs) (14%), sepsis (11%), complications of abortion (8%), embolism (3%), and other direct causes such as obstructed labor and complications of delivery (10%).2 These are the same leading causes observed in Brazil and other low- and middle-income countries (LMICs).3,4 HDPs, including eclampsia and preeclampsia, are a significant contributor to mortality in Latin America and the Caribbean, contributing to 22% of all maternal deaths in the region.4
Quality improvement (QI) is a systematic, data-driven approach to enhance health care processes and outcomes through iterative testing and refinement. QI methods can help to strengthen the delivery of reliable interventions that are effective, sustainable, and cost-effective.5,6 When used to support clinical interventions, QI methods have successfully improved maternal health outcomes in both high- and low-income nations.7–9
A care bundle is a set of evidence-based practices that, when implemented together, improve patient outcomes more effectively than when applied individually. Care bundles used to manage an emergency are well-recognized strategies to improve adherence to clinical standards to manage life-threatening conditions, reducing the time between recognition and treatment.10,11
Despite numerous advances in reducing maternal death, LMICs are far from reaching the United Nations Sustainable Development Goals related to maternal and newborn outcomes. To address this challenge in Brazil, the Institute for Healthcare Improvement, in collaboration with the Albert Einstein Hospital and MSD Brazil, carried out a QI initiative to reduce the institutional maternal mortality ratio (iMMR) using a framework developed by the Institute called the Breakthrough Series Collaborative (BTS). BTS is a structured, short-term learning system that brings together multiple teams to rapidly test and implement evidence-based changes, closing the gap between best practices and daily operations to achieve breakthrough improvements in quality and outcomes. This article presents the learnings and results of this QI initiative.
METHODS
Context
Between 1990 and 2013, the national maternal mortality ratio (MMR) in Brazil dropped by 55%. Since then, however, MMR in Brazil has been unchanged, ranging generally between 55 and 60 maternal deaths per 100,000 live births,3,12 with the exception of during the height of the COVID-19 pandemic (March 2020 to May 2021), when maternal mortality increased by more than 70%.12,13
In Brazil, maternal death is primarily associated with the delay in recognizing clinical deterioration at the health facility level and the reliability of providing effective and standardized care.14,15 Well-designed processes to recognize clinical deterioration and perform immediate clinical actions are crucial to prevent pregnancy-related mortality.14,15
The circumstances that lead to maternal death are complex and multifactorial, including factors related to inequity and racism. For example, by 2018, more than a half of all maternal deaths occurred among Black women, who are twice as likely to die during pregnancy and childbirth than White women.16,17 Regrettably, during the pandemic, these inequities were even more evident. Black women accounted for 1,095 of the pregnant and postpartum women who died from COVID-19, representing 54% of this group by March 2022.12
Implementation Approach
Our QI initiative utilized the Model for Improvement,5 a widely recognized framework developed by the Institute for Healthcare Improvement, that guides teams to set specific improvement aims, identifies processes and outcomes to measure, and develops change ideas to test, using rapid Plan-Do-Study-Act (PDSA) cycles. Teams were supported to identify system inefficiencies, prioritize interventions, and foster a culture of continuous improvement through collaboration and shared learning. The approach is grounded in evidence-based practice and integrates both quantitative and qualitative data to ensure a comprehensive understanding of system performance and outcomes. The teams were connected through a series of virtual and face-to-face meetings using the BTS methodology, which provided a platform for the teams to learn, collaborate, and rapidly disseminate best practices.18
Before launching, the BTS governance team convened a meeting of implementers, clinical experts, and hospital leaders to create consensus about the initiative’s aim, clinical intervention, and implementation strategy. The outputs of the meeting included (i) an aim to reduce the iMMR by 36%, with focus on the 3 most prevalent life-threatening conditions, (ii) definition of the clinical intervention as the “4Rs” sequence, described below, and care bundles for the life-threatening conditions, and (iii) agreement on the implementation approach that addressed a set of system enablers, using QI methods and tailored training of the clinical staff.19
At the first learning session, we engaged hospital teams to identify the local environmental conditions and capabilities that support the teams to implement the clinical changes. These primary drivers of improved survival and associated ideas for improvement are summarized in Table 1.
Primary Drivers of Improved Women’s Survival Around Birth and Associated Changes the Participating Hospitals Tested
The implementation intervention started with a 2-day face-to-face learning session in November 2019. Due to the COVID-19 pandemic, all subsequent quarterly face-to-face 2-day sessions between December 2019 and March 2021 were replaced with 1-hour monthly virtual sessions. During the in-person and virtual sessions, QI teams received training on QI methods; learned how to design initial tests of change using PDSA cycles to increase the percentage of patients with a calculated Modified Early Obstetric Warning Score (MEOWS) and the use of the care bundles; and shared experiences of successes and challenges during implementation. QI teams comprised an executive leader, obstetrician, and obstetric nurse.
Between the 1-hour monthly virtual learning sessions, we offered one-to-one support to the hospitals upon request. We asked each hospital to send a monthly report including a description of the tests of changes, changes implemented, and updates of the process and outcomes measures. Improvement coaches provided QI teams with feedback mentoring based on the data and information included in these reports. To coach leaders to support and enable change in their systems, the leadership of the hospitals participated in 2 separate virtual learning sessions explicitly designed to convey the crucial role of leadership (e.g., support for the initiative, understanding of the methods, removal of barriers, and presence at the front line, among other topics),20 and to advance equitable care.
Job Instruction
Training Within Industry,21 a proven methodology for standardizing and improving work processes, has gained traction in health care settings after demonstrating its effectiveness across various industries. Training Within Industry provides a structured framework for training individuals to perform tasks correctly, safely, and efficiently.
In this initiative, we utilized Job Instruction,22 one of the core components of Training Within Industry, to enhance the implementation of care bundles for maternal health. Job Instruction involves breaking down complex tasks into key elements, including important steps, key points, and their underlying rationale. By providing this structured approach, we supported health care providers to practice these elements in a standardized way, fostering consistency and quality in performance across all maternal care processes.
Job Instruction requires the teams to create Job Breakdown Sheets of the most important processes containing the important steps, key points (related to safety and best performance), and the reasons of each key point.23 With the Job Breakdown Sheets ready, the instructor will train the front line, one-by-one, using the 4 steps of training:
Prepare the worker
Present the process
Try out performance
Follow up
Clinical skills require a competency-based approach with a strong emphasis on skill acquisition. While Job Instruction is valuable for hands-on tasks, its effectiveness is maximized through individualized instruction and practical application. Ideally, this involves one-on-one training, allowing for personalized feedback and repetition. Furthermore, conducting the training in a simulated environment or the actual clinical setting where the procedure takes place enhances skill transfer and confidence.
In our case, the BTS team trained virtually 1–2 health care workers per hospital, providing a foundation in the Job Instruction methodology. The subsequent face-to-face training at each hospital offered by the health care professionals was crucial for ensuring competency in hands-on clinical skills. This approach allowed for direct observation, individualized feedback, and practice in the actual clinical environment, ultimately leading to improved skills to care for patients.
Study Design
We undertook a quasi-experimental time-series study to determine the effectiveness of the QI initiative to improve standard clinical surveillance and care and the impact of this on iMMR in public maternity hospitals across Brazil. To evaluate changes in the iMMR, we analyzed data collected from the 19 participating organizations during 3 periods:
Baseline (January 2018 to November 2019)
Implementation (December 2019 to March 2021)
Post-implementation (April 2021 to September 2021)
This article follows the structure of Standards for Quality Improvement Reporting Excellence (SQUIRE 2.0).24
Settings and Participants
All 28 public hospitals that had participated in another Brazilian QI Collaborative to reduce high cesarean delivery rates were invited to join the BTS initiative.25 Of these, 20 institutions volunteered to join the new initiative. In addition, 1 hospital with a high iMMR in the north of Brazil that had not participated in the other QI Collaborative asked to join and was also included in the BTS.
Clinical Intervention
This initiative focused on the 3 most common causes of maternal death: postpartum hemorrhage, sepsis, and HDPs. “Failure to rescue” is the failure or delay in recognizing and responding to a hospitalized patient experiencing complications from a disease process or medical intervention,26 and the concept has been broadly applied to obstetric settings where an estimated 40% of obstetric deaths have been attributed to this breakdown in clinical care.27
The 4Rs
We structured the clinical activities according to a sequence of 4 steps that recognizes, rescues, reassesses, and refers (“4Rs”) patients to the most appropriate level of care after clinical deterioration, applied to every woman admitted to each of the hospitals for obstetric care:
Recognize: For sepsis and HDPs, we used MEOWS28 for timely detection of clinical deterioration. This score was used at the time of first contact with the patients in each transition of care and to reassess the patient’s clinical condition after the second R (Rescue). MEOWS has been used extensively, including in Brazilian studies, to support efforts to reduce maternal mortality.29–33 For postpartum hemorrhage, we used blood loss ≥500 mL for a vaginal birth and ≥ 1,000 mL for cesarean delivery as the primary trigger to initiate a rescue response,34 as this preceded clinical deterioration.
Rescue: A response to early clinical deterioration. In this initiative, we used care bundles and tailor-made clinical interventions to manage hemorrhage. Additionally, we used care bundles specifically designed to manage any suspicion of HDPs. The bundle for treating suspected sepsis adheres to the standards recommended by the 2018 Surviving Sepsis Campaign guidelines.35 For postpartum hemorrhage and HDPs, our team of clinical experts selected evidence-based interventions after a comprehensive literature review and the BTS team developed the bundles using the Model for Improvement.
Reassess: Reassess the effects of the rescue using MEOWS.
Refer: Escalate the care to the most appropriate level within or outside the hospital.
The 4Rs approach was a modification and expansion of a sequence of clinical actions originally described for the management of obstetric-associated hemorrhage.36 Before this initiative, none of the hospitals had a standardized way to manage the 3 life-threatening conditions with the interventions based solely on the decision of the physician during the event.
Care Bundles
The care bundles were based on the concept originally proposed by the Institute for Healthcare Improvement11 and widely adopted in obstetric practice36 that a small set of clinical interventions applied together, directed at a specific clinical problem, is a more effective clinical strategy than piecemeal application of those care elements. During this BTS, these evidence-based clinical bundles were introduced to hospitals that did not previously have standardized evidence-based protocols to manage the 3 life-threatening conditions. The ultimate use of the bundle elements was left to the discretion of the attending physician during the event, depending on the patient’s unique needs. The sequential approach to diagnosis and treatment was modified from a previously reported approach to management of obstetric-associated hemorrhage.36
The care bundles for postpartum hemorrhage, sepsis, and HDPs are shown in Table 2. Compliance with the bundle was monitored through regular audits and results were used to guide iterative refinements and support team learning throughout the improvement process.
Care Bundles for Postpartum Hemorrhage, Sepsis, and Hypertensive Diseases of Pregnancy
In addition to care bundles, we also proposed 3 actions to prevent postpartum hemorrhage: assess hemorrhage risk, quantify blood loss during delivery, and adminster 10 IU oxytocin after delivery. Blood was collected using a drape during vaginal delivery or a suction canister during a cesarean delivery. Blood-soaked items like pads and sponges were weighed, and their dry weight was subtracted to calculate how much blood they had absorbed. The total blood loss was then determined by adding the collected blood to the amount calculated from the soaked items.
For blood loss >1,000 mL in a vaginal birth or >1,500 mL in a cesarean delivery, traditional care bundles are not suitable. The expert panel recommended that obstetricians and gynecologists utilize a more individualized approach to invasive procedures, tailoring them to each patient’s specific needs and clinical circumstances. These procedures may include a hysterectomy, the use of an intrauterine balloon, compressive sutures, and vascular embolization. The integration between the “4Rs” and care bundles’ triggers is summarized in Figure 1.
The “4 Rs” Flowchart for Sepsis and Hypertensive Disorders of Pregnancy (A) and for Hemorrhage (B)
Abbreviations: C-section: cesarean delivery; MEOWS: Modified Early Obstetric Warning Score; QBL, quantity blood loss.
All participating hospitals were required to implement at least 1 care bundle for 1 of the 3 life-threatening conditions. The choice of life-threatening condition was determined by the preferences of the QI teams at each hospital; however, all hospitals had access to all care bundles. Regardless of the chosen life-threatening condition to focus on, each hospital was required to implement the proposed actions to prevent postpartum hemorrhage.
The care bundle for postpartum hemorrhage was divided into 2 bundles designed to be applied sequentially. The first response care bundle was applied when blood loss was ≥500 mL and ≤1000 mL for vaginal birth and ≥1000 mL and ≤1500 mL for cesarean delivery. The second response care bundle was applied if the patient continued to bleed after the application of the first response care bundle. If bleeding persisted despite the application of both care bundles, the treatment was determined by the clinical judgment of the care team.
Equity and Anti-Racism Training
Recognizing the profound impact of racial disparities on maternal health outcomes, we incorporated anti-racism training as a critical component of this QI initiative. Evidence demonstrates a disproportionately high rate of maternal mortality among Black women in Brazil, highlighting the urgent need to address systemic inequities within the health care system.12 To mitigate the influence of implicit bias and promote equitable care for all patients, we conducted virtual anti-racism training sessions for hospital leaders between October 2020 and February 2021. These sessions aimed to raise awareness of racial disparities in maternal health and empower leaders to implement strategies that counter race-based inequities in their respective institutions. A detailed description of this intervention and its impact will be published in a separate manuscript.
Enabling the Environment for Change
Meetings between experts and the hospital-based leadership team addressed the system enablers (all necessary drugs, materials, time allocated, and equipment) required for the frontline team to carry out the new standard of care (the “4Rs,” MEOWS, and care bundles). They guided leadership behaviors (e.g., walk rounds and devolved leadership approaches). Leadership buy-in was sought to allow more autonomy for the nurses to trigger and apply the care bundles previously approved by the obstetricians.
Measurement
QI teams collected the following data from hospital records: the number of births, maternal mortality by cause, the use of MEOWS, the number of women diagnosed with potential life-threatening conditions, and the number of live births.
QI teams provided baseline data on the number of live births and deaths 12 months prior to the intervention. For all other measures, baseline data was established by analyzing hospital data for common and special cause variation during the intervention according to Shewhart.37 Except for maternal mortality, all other measures were newly implemented by the hospitals as part of this QI initiative.
To be compliant with the care bundles, the clinical teams had to have applied all elements of the bundle (“all or nothing” concept),11 but we also measured the percentage of patients receiving each element of the bundle.
For process measures, we created a spreadsheet for each hospital with instructions on collecting and reporting data. We recommended a random sample of at least 20 patients monthly for each process measure. The way the QI teams collected process measures varied. Some collected the process measures by reviewing patients’ charts. Other hospitals collected process measures while the care was happening. In both cases, the QI teams used the tool created by our team to collect data. The hospital QI team assigned 1 person to collect data and enter it into the online platform. During the implementation period, the QI team at each site shared their de-identified aggregate monthly measures using online data entry (Simple QI®). All measures’ formulas are described in the Supplement Table.
For this initiative, we used aggregated data from the units to track changes in iMMR over time. This approach was adopted because every hospital (despite focusing on a specific bundle) had access to the complete set of bundles. Subgroup analysis of hospitals applying improvement to individual bundles was not done because of sample size limitations.
Analysis
We used monthly run charts analysis38 to detect changes from the baseline for all causes of death and the life-threatening condition. To detect significant changes in performance from the baseline, we extended the median baseline values into the implementation and post-implementation period and applied statistical process control rules to detect special cause variation. The causes of maternal deaths, including COVID-19 deaths, are presented in a table, and we used a scatter plot chart to correlate the percentage of patients with MEOWS score calculated and the iMMR. Minitab® v18 (USA) was used for all statistical analyses.
Ethical Considerations
For this article, access to the BTS database was approved by the Research Ethical Board of Albert Einstein Hospital. The database has QI process measures, aggregated and de-identified, and does not exhibit any data referring to or mentioning the patients involved, compliant with local data protection laws. The present article does not present any data that could identify the facilities.
RESULTS
Of the 21 hospitals that joined the BTS at the start of this initiative, 1 hospital never provided data and was excluded from the study, and another hospital dropped out.
We included the remaining 19 maternity hospitals participating in the initiative and contributing data for the analysis. All 19 maternity hospitals provided data on iMMR consistently for the entire baseline, implementation, and post-implementation period. Most institutions were in the southeastern region of Brazil (63%), and 7 belonged to Brazil’s north/northeastern region (37%). Table 3 describes the characteristics of the participating institutions.
Characteristics of the Participating Institutions, by the Life-Threatening Condition Chosen for Improvement
During the implementation period of the intervention, efforts were made to enhance clinical skills and redesign the care model in the participating maternity hospitals. As demonstrated in Table 4, there was a notable change in care practices and the obstetric care environment favorable for iMMR reduction.
Changes Documented in the Hospitals Before and After the BTS, by the “4Rs” Clinical Steps
Across the 19 hospitals, the implementation of quantifying blood loss rose from 0% to 95%, and the administration of 10 IU of oxytocin after delivery rose from 78% to 98% (data not displayed). In the case of sepsis, the percentage of patients for whom health care providers reliably used the care bundle increased significantly from 42% to 82%.
After the intervention began, hospitals were instructed to track and report the percentage of patients with MEOWS scores recorded. Just after the hospitals started testing and implementing MEOWS, 61% of patients had scores documented. This percentage steadily increased throughout the implementation period, ultimately stabilizing at 98%. We saw an inverse relationship over time between the reduction of iMMR and the increased adoption of MEOWS (Figure 2).
Correlation Between Implementation of MEOWS and the iMMR From Life-Threatening Conditions, June 2019 to March 2021
Abbreviations: iMMR, institutional maternal mortality ratio; LTC, life-threatening condition; MEOWS, Modified Early Obstetric Warning Score.
We estimate that between December 2019 and September 2021, the QI interventions prevented approximately 33 maternal deaths due to life-threatening conditions and other causes, except COVID-19 (Table 5). However, the number of live births was higher during the baseline period than the implementation and post-implementation periods. This could lead to an argument that the exposure to risk was lower during and after implementation, which would naturally result in fewer deaths. A closer examination of the data, however, reveals a different story. When we calculate the iMMR by dividing the number of deaths during the baseline period by the number of live births, we arrive at an iMMR of 76.4 (102/133,471 X 100,000 live births). In contrast, when we perform the same calculation for the implementation and post-implementation periods, the iMMR is significantly lower at 27.3 (33/120,916 X 100,000 live births). Therefore, the observed reduction in the number of deaths cannot be attributed to the difference in live births during the baseline period compared to the implementation and post-implementation periods.
Number of Deaths per Period and the Estimated Number of Deaths Prevented, by Cause of Death
Comparing the baseline to the implementation period, analysis of run charts detected a 34.2% reduction in the all-cause iMMR (from 83.7 to 55 deaths per 100,000 live births). However, there was a sharp increase of 178% between the implementation and post-implementation period, from 55 to 152.9 deaths per 100,000 live births, with 70% of the deaths attributed to COVID-19 (Figure 3).
Run Chart of the Aggregated iMMR From All Causes of Death in the 19 Participating Hospitals, January 2018 to September 2021a
a December 2019 to March 2021 represents the implementation period of the quality improvement initiative.
Abbreviations: iMMR, institutional maternal mortality ratio.
Comparing the baseline to the combined implementation and post-implementation period, the iMMR due to life-threatening conditions decreased by 60.9%, from 44.1 to 17.2 deaths per 100,000 live births (Figure 4). The iMMR due to postpartum hemorrhage decreased by 72.9%, from 11.7 to 3.17 per 100,000 live births, while the iMMR from sepsis decreased by 100%, from 20.4 to 0 deaths per 100,000 live births (data not displayed). The number of deaths did not decrease for HDPs.
Run Chart of the Aggregated iMMR Due to Life-Threatening Conditionsa in the 19 Participating Hospitals, January 2018 to September 2021b
a Life-threatening conditions comprised postpartum hemorrhage, sepsis, and hypertensive disorders of pregnancy.
b December 2019 to March 2021 represents the implementation period of the quality improvement initiative while April to September 2021included the post-implementation period.
Abbreviations: iMMR, institutional maternal mortality ratio.
Table 6 shows the variation in percentage points in the care bundle practices suggested by the QI initiative. Data for the postpartum hemorrhage second response care bundle are not included because none of the hospitals collected the related data.
Variations in Care Bundle Practices Among Participating Hospitals, May 2019 to October 2021
DISCUSSION
This BTS initiative presents a feasible QI strategy for decreasing iMMR in public hospital settings in a developing country. This BTS addressed the 3 leading maternal causes of death found by the WHO: postpartum hemorrhage, sepsis, and HDPs. After the QI intervention, we observed a reduction in deaths related to postpartum hemorrhage and sepsis, but not HDPs, resulting in a lower aggregate iMMR. The durability of the reduction in deaths from hemorrhage and infection was proved despite the disruptions associated with the COVID-19 pandemic. These findings reinforce the impact and sustainability of the redesign of obstetric care in hospitals that were delivered using a combination of QI and targeted trainings.
Despite efforts to reduce maternal deaths in Brazil, MMR has remained high and relatively unchanged since 2013.3,12 Our results support the calls for expanding traditional strategies that focus on clinical training, infrastructure, medical equipment, medications, and human resources to include more engaging training methods for clinical bundles39 combined with QI approaches40 to implement evidence-based care more reliably.
The core achievement of our initiative was the successful implementation of the clinical intervention to reduce iMMR from the major life-threatening conditions using the “4Rs,” which included an adaptation of the “failure to rescue” concept26 applied to the resource-constrained setting of a middle-income country. Although MEOWS is a well-known tool to recognize clinical deterioration in pregnant women,41 its application in Brazilian public maternity hospitals was novel for many hospitals.31 During the COVID-19 pandemic, the Brazilian Ministry of Health recommended using the tool in all public hospitals assisting pregnant women.42
Our intervention introduced novel combinations and applications of existing QI tools and approaches. While some hospitals had previously utilized the MEOWS, our initiative innovated by linking MEOWS with care bundles and mandating MEOWS calculation during all transitions of care and following bundle implementation. This comprehensive integration of MEOWS represented a unique approach to enhancing patient safety and standardizing care practices. This BTS design linked a process that reliably identified patients at risk of life-threatening conditions to a reliable mechanism to deliver necessary care quickly. The “4Rs” system enabled all participants, not just the physicians, to respond when the clinical team identified a woman at risk; this meant that nursing teams in the participating hospitals could trigger and implement the bundles—an unusual role in the Brazilian setting.
Although many hospitals have postpartum hemorrhage protocols, evidence suggests that the mere presence of these protocols does not reduce bleeding-related deaths.43 In our initiative, the use of QI and training to improve the reliability of the application of actions for preventing, detecting, and implementing bundles to treat hemorrhage-related deaths was associated with a 72.9% reduction in iMMR. A similar impact on maternal hemorrhage was reported recently by Gallos et al.,44 who saw a significant decrease in postpartum hemorrhage in an intervention group of African hospitals that received a clinical bundle implemented with a package of activities that included training, local champions, and audit and feedback; mortality was not reported. Other studies showed a direct correlation between adherence to bundles and improved maternal outcomes and provider performance.36
The early identification of sepsis, even before a definitive diagnosis, is crucial in preventing maternal mortality,45 mainly as maternal sepsis definitions are controversial.46–48 While sepsis is reported as the third leading cause of death worldwide,2 infection is responsible for nearly half (46%) of all facility-based maternal deaths in Brazil.49 This BTS showed that early detection through the MEOWS, together with the introduction of the suspicion of sepsis bundle, was associated with a 100% decrease in sepsis-associated maternal deaths, reinforcing the recommendation of early detection and treatment50 (i.e., application of the sepsis bundle within an hour of suspicion of sepsis), rather than waiting for sepsis to be confirmed.
After the QI intervention, iMMR due to HDPs did not change from the baseline values despite adaptive changes to the implementation design (e.g., additional clinical sessions, reinforcements in quality training, adjustments, and customization of the HDP bundle for each maternity unit setting). We also did not observe improvement in bundle compliance for HDP. Failure to reduce maternal mortality due to HDP include poor implementation of the HDP bundle, the inadequacy of the HDP-specific bundle, the clinical condition of patients in this group (age, comorbidities, cardiovascular risk, and others), structural issues (lack of required equipment and medications and different concentrations of magnesium sulfate among hospitals making standardization more difficult), and poor antenatal care (patients arriving at the hospitals with advanced HDP). While the California Pregnancy-Associated Mortality Review Committee in the United Sates attributed 65% of HDP-related deaths to missed diagnoses and ineffective care,51 our findings suggest that existing clinical care bundles and QI interventions and training to deploy those bundles were insufficient to rescue women presenting with HDP. Other QI interventions using the toolkit approach have successfully reduced eclampsia and morbidity due to HDP,52 but the impact of this approach on mortality was not reported. Improving prenatal care, standardization of medication doses among hospitals, and incorporating new implementation strategies or complementary clinical activities may be necessary to decrease HDP-related deaths in Brazil.
Furthermore, the absence of significant changes in maternal mortality due to HDP despite reductions in deaths from postpartum hemorrhage and sepsis warrants further analysis. While our intervention focused on improving recognition and management through care bundles, excluding the onset of labor and mode of delivery as a component of the HDP care bundle may have limited its effectiveness. However, decisions regarding the mode and timing of delivery are complex, often influenced by a combination of clinical, provider, and patient preferences, as well as institutional practices. In our setting, the intervention did not address these factors directly, which may have contributed to the lack of observed improvement in HDP-related mortality. This highlights the need for future iterations of care bundles to incorporate interventions, such as early delivery protocols, that directly address the clinical progression of HDP.
Finally, we highlight that we undertook this QI initiative during the COVID-19 pandemic. Despite the adverse context, the initiative achieved sufficient engagement of the obstetric teams to continue actively participating in the adapted QI initiative. Teams continued to attend virtual learning sessions and undertake activities in their maternity units. The improved outcomes for hemorrhage and sepsis were sustained despite an increase in overall iMMR attributed to COVID-19 infection.12,13 The second wave of COVID-19 coincided with the final months of this BTS; after March 2021 COVID-19 became the leading cause of maternal death, obscuring the impact of the QI work performed by this BTS. The adjustments to deliver the BTS virtually during the pandemic provided a potential model for remote QI implementation and training.
Limitations
Although our work showed a significant decrease in iMMR for postpartum hemorrhage and sepsis, the results should be interpreted cautiously. The hospitals selected for the study had considerable variations in baseline performance and significant variations in the number and values of iMMR. Variations in baseline iMMR and low actual numbers of maternal deaths per month in each hospital did not allow for rigorous subgroup analysis. The hospitals that actively participated in bundle deployment likely contributed disproportionately to reported improvements in mortality for postpartum hemorrhage and sepsis in the aggregated results. In addition, some hospitals already had high-quality maternity programs with low or no maternal deaths, providing a minimal opportunity for contribution to overall improved iMMR. However, these hospitals provided significant opportunities for a study of positive deviance, supporting ideas for improvement across the learning collaborative.
A more rigorous selection of intervention hospitals and inclusion of a matched group of hospitals that did not receive the intervention would have increased our confidence that the reductions in mortality from postpartum hemorrhage and sepsis were directly attributable to the QI and training strategies that we deployed. However, we know of no other environmental or policy interventions that could be attributed to our observed results.
While we focused on iMMR as an aggregate outcome, the lack of granular data on presenting cause-specific cases limits our ability to generate additional insights. This underscores the importance of future QI initiatives adopting more comprehensive data collection frameworks, including systematic collection and analysis of case tracking, clinical intervention timing, and associated outcomes. These data would enable more robust evaluations of intervention effectiveness and help identify critical areas for further refinement. Additionally, while the impact of race and ethnicity are essential factors in determining maternal outcomes in Brazil,12 as it is worldwide, we were not able to discern differences in maternal outcomes based on race in this set of hospitals.
The absence of data on perinatal outcomes, such as fetal loss and newborn mortality, also represents a limitation of this study, avoiding having a broader impact of these conditions on the maternal-newborn dyad and restricting the scope of our findings. Future QI initiatives should prioritize the systematic collection of perinatal data to provide a more comprehensive evaluation of QI interventions.
CONCLUSION
A combination of implementation approaches consisting of QI, Job Instruction, equity and antiracism training, care bundles, and MEOWS was delivered using a Breakthrough Series Collaborative in a group of public hospitals in Brazil to improve the reliability of evidence-based practices that are known to reduce pregnancy-related maternal deaths. We used a QI approach along the peripartum patient journey to deliver a sequential set of clinical surveillance and response activities (the “4Rs”) and ensure reliable implementation of evidence-based care bundles. These activities were linked to a significant decrease in aggregate mortality for 2 of 3 targeted life-threatening conditions in participating public hospitals.
Furthermore, during the implementation and post-implementation periods, deaths related to life-threatening conditions either remained stable or decreased even further despite the excess mortality associated with the COVID-19 pandemic. The scalability of our approach must be further tested using robust evaluation designs in greatly expanded settings and contexts. If replicated in these contexts, this QI implementation and clinical management approach may inform expanded national and global policies and strategies to reduce maternal mortality in LMICs.
Acknowledgments
We thank all the health care professionals of the participating institutions; with their empowerment and motivation, this initiative was successful. We also appreciate the technical and administrative teams supporting this initiative’s development.
Funding
The activities in this publication were supported by funding from MSD, through its MSD for Mothers initiative. This grant was awarded to Sociedade Brasileira Israelita Albert Einstein for the project ‘Reduction of Maternal Mortality and Appropriate Birth in Brazil: Adoption of safe practices for the Protection of Mothers and Babies’ during 2019–2021. MSD had no role in the design, collection, analysis, and interpretation of data, in writing of the manuscript, or in the decision to submit the manuscript for publication. The content of this publication is solely the responsibility of the authors and does not represent the official views of MSD. MSD for Mothers is an initiative of Merck & Co., Inc., Rahway, NJ, USA.
Author contributions
All authors contributed to the data curation, formal analysis, investigation, method, validation, visualization, and writing. Additionally, PBo, PBa, and CGB contributed to the conceptualization, project administration, and supervision. All authors reviewed and approved the final version of this article.
Competing interests
None declared.
Footnotes
↵† Members of the group: Adriana Cristina D’Ornelas, Adrieli Cristina Menezes, Aline Cristhina Martins Barros do Prado, Aline Pires de Freitas Belchior, Ana Carolina Barbosa PorDeus, Andrea de Almeida Vasconcelos Nogueira, Andrea Maria Novaes Machado, Andréa Millen Penedo, Angela Maria Machado, Angélica Cristina Rodrigues, Carla Cristiane de França Araújo, Caroline Macedo de Camargo, Clélia Andrade Salustrino, Cleuma Reis Rodrigues de Freitas, Douglas Vieira Leite Benevides, Dulce Mara Carlos Batista dos Reis, Edna Cristiane da Silva Pimenta, Fernanda Carvalho Raimundo, Karine Moraes de Andrade, Hosana Silvia Almeida Correa Zenardi, Leila Marques da Silva Santos, Luana Rodrigues Ferreira Silva, Marcelo Tetisuo Ikegami, Maria Benita Alves da Silva Spinelli, Mary Angela Meneses Sanches, Michelle Ribeiro Achcar, Nathalia Moura de Mello e Silva, Norma Suely de Carvalho Fonseca Assunção, Renata Gondim Carvalho Cunha, Roberta Carneiro Mendonça, Robson Braguetto, Rudiane Daniela Vicentine Sivieri, Sheyla Guimarães Oliveira, and Tatiana Nobre e Souza.
Notes
Peer Reviewed
Cite this article as: Borem P, Gushken AKF, Gushken AP, et al.; on behalf of the Abraço de Mãe Collaborative group. A comprehensive strategy to mitigate institutional maternal mortality: lessons from a quality improvement initiative in Brazilian maternity hospitals. Glob Health Sci Pract. 2025;13(2):e2400130. https://doi.org/10.9745/GHSP-D-24-00130
- Received: April 29, 2024.
- Accepted: January 20, 2025.
- Published: December 31, 2025.
- © Borem et al.
This is an open-access article distributed under the terms of the Creative Commons Attribution 4.0 International License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly cited. To view a copy of the license, visit https://creativecommons.org/licenses/by/4.0/. When linking to this article, please use the following permanent link: https://doi.org/10.9745/GHSP-D-24-00130
