Blood Availability and Maternal Mortality from Obstetric Hemorrhage: A Retrospective Case-Control and Cross-Sectional Study in Kenya

In 2017, an estimated 295,000 women died globally of maternity-related causes [1], with over half of these deaths in sub-Saharan Africa [2]. Obstetric hemorrhage remains a major contributor to maternal mortality worldwide, responsible for 34% of maternal deaths in Africa, with over 80% occurring in the postpartum period [3–5]. In Kenya, the estimated Maternal Mortality Rate (MMR) is 342 per 100,000 live births, which fails to reach the national target MMR of 147 [6,7]. Obstetric hemorrhage accounts for 40% of maternal deaths in Kenya [8,9], and 91% of these deaths are associated with suboptimal care, such as delays in timely intervention [8]. A recent multicenter study identified significant gaps in risk assessment, prevention, identification, and management of postpartum hemorrhage (PPH) following vaginal deliveries in Kenyan hospitals, including inadequate adherence to postpartum monitoring guidelines and delays in initiating treatment [10].

Rapid access to blood products is critical to reducing the risk of mortality from hemorrhagic shock, yet geographic and logistical challenges and blood shortages can hinder timely transfusion [11,12]. A 2018 multicenter study in 54 referral hospitals across Kenya identified that only 67% of obstetric patients with antepartum hemorrhage who had a clinical indication for a transfusion received the needed blood products [13]. These existing supply challenges predated more recent strains on Kenya’s blood programs. In 2019, the United States President’s Emergency Plan for AIDS Relief (PEPFAR) withdrew financial support from the Kenya Tissue and Transplant Authority, reducing the operational capacities of regional and satellite blood centers [14]. In the first year of the COVID-19 global pandemic, Kenya faced a significant donor shortage, meeting only 16% of blood demand in 2020 [15,16]. The cessation of World Bank funding for Kenya’s blood programs in 2025 threatens to further strain the blood system, with the Ministry of Health warning that national transfusion services could be severely disrupted without urgent intervention [17].

High maternal anemia rates in Kenya (26.2-40.3%) accentuate the importance of reliable blood systems, as anemia is correlated to blood loss, death, and near miss in labor and postpartum, and parturients with anemia are at a greater risk of PPH and adverse outcomes [18–21]. Additional risk factors associated with postpartum hemorrhage include grand multiparity, prior PPH, Cesarean delivery, prolonged labor, and non-administration of uterotonics[9].

While blood shortages have been linked to maternal mortality in sub-Saharan Africa, research on their impact on hemorrhagic deaths at the point of care in low-resource settings remains limited [11,22,23]. We examined blood availability and obstetric hemorrhage-related mortality at a national referral hospital in Kenya, aiming to (1) identify clinical and systemic factors associated with hemorrhagic versus non-hemorrhagic maternal deaths, and (2) assess whether implementation of a hospital-based blood bank affected blood availability and maternal mortality.

Study Setting

The study was conducted at Nakuru Level 6 Hospital, an 838-bed national referral facility in Kenya’s Rift Valley Province serving 2.1 million people and averaging over 12,000 live births annually. Blood products are supplied by the regional blood center and, since June 2020, by the hospital’s own blood bank. The study period was January 1, 2019, to June 30, 2022. We conducted two complementary analyses within the same dataset: a patient-level case-control study and a hospital-level repeated cross-sectional study, to capture both individual risk factors and system-wide patterns in blood supply and maternal mortality (Figure 1).

Data Sources

This study received ethical approval from the Egerton University Ethics Review Committee. With permissions from the hospital blood transfusion and performance improvement committees, we reviewed de-identified retrospective data from four sources: 1) Maternal Death Review forms (MDRs), 2) Maternal Death Notification forms (MDNFs), 3) District Health Information System (DHIS) electronic register, and 4) blood transfusion crossmatch register. Records were examined as they related to the blood continuum. Extracted variables from MDR and MDNFs included age, parity, referral status, parturition status on admission, transfusion status, cause of death, and contributory factors, including documentation of insufficient blood products. Blood transfusion registers provided monthly blood products dispatched to the maternity ward, and live birth counts were obtained from the DHIS register. All data were de-identified prior to analysis and handled confidentially. Access to raw data was restricted to study investigators, and files were stored securely in accordance with hospital and Ministry of Health confidentiality protocols.

Definitions

Maternal death was defined per WHO criteria as death from any cause related to or aggravated by pregnancy or its management (excluding accidental/incidental causes) during pregnancy or within 42 days of termination [24]. Obstetric deaths were classified as direct or indirect [24]. Hemorrhagic deaths included those from antepartum hemorrhage, intrapartum hemorrhage, postpartum hemorrhage, ectopic pregnancy, or uterine rupture, defined according to WHO guidelines [25,26]. Deaths from other causes or with no indication of hemorrhage were classified as non-hemorrhagic. Cases labeled as “indirect primary cause” or “unconfirmed non-hemorrhagic cause” included both WHO-defined indirect causes (e.g., anemia, HIV/AIDS, cardiovascular disease) and suspected but unconfirmed direct causes without documented hemorrhage (e.g., pulmonary embolism, amniotic fluid embolism). Contributory factors were abstracted from clinician-completed MDR and MDNF forms, based on clinical documentation and multidisciplinary review conclusions. These forms identify individual-, community-, and health system–level contributors to maternal death (Table 4).

Bias Mitigation

To minimize bias, we cross-verified information across multiple data sources, applied standardized definitions for exposures and outcomes, and restricted statistical analyses to variables with ≥80% completeness in maternal death records.

Case-Control Retrospective Study

We included all maternal deaths documented at Nakuru Level 6 Hospital between January 1, 2019, and June 30, 2022. Cases were deaths due to obstetric hemorrhage; controls were deaths from non-hemorrhagic causes. The analysis compared clinical and systemic factors associated with hemorrhagic death. Univariate logistic regression identified candidate variables (p<0.10), which were entered into a multiple logistic regression model adjusting for age and parity. Detectable effect sizes were calculated post-hoc using observed exposure proportions and logistic regression parameters, with α=0.05.

Repeated Cross-Sectional Study

We conducted a repeated cross-sectional analysis of monthly hospital-level data collected at three-month intervals from May 1, 2019, to May 31, 2022. All available records on maternal deaths, transfusions, live births, and blood product dispatches to the maternity ward were included. The primary outcome was the monthly count of hemorrhagic maternal deaths. Secondary outcomes included monthly blood units dispatched to the maternity ward, total maternal deaths, and live births. We compared pre- and post-implementation periods relative to the June 2020 launch of the hospital-based blood bank. Wilcoxon rank-sum exact tests were used for pre–post comparisons due to small sample size and non-normality. Poisson regression was used to model the association between monthly hemorrhagic deaths and blood units dispatched, adjusting for the number of live births. Post-hoc power analyses were conducted using observed means and standard deviations for key variables.

General Analytic Information

Analyses were performed in Stata/SE 16.1 (StataCorp, College Station, TX, USA). Continuous variables are reported as mean ± standard deviation (if normally distributed) or median [range] (if not). Categorical variables are reported as frequencies and percentages. Statistical significance was defined as p<0.05. This study followed the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) guidelines.

Case-Control Retrospective Study

Maternal Mortality

Between January 1, 2019, and June 30, 2022, there were 90 maternal deaths and 42,876 live births at or before arrival to Nakuru Level 6 Hospital, corresponding to a maternal mortality ratio of 210 ± 157 deaths per 100,000 live births. The mean age at death was 28.3 ± 5.8 years, and the median parity was 2 [range 0–8]. Over half of patients (47/88, 53.4%) were referred from another facility. The most common parturition status on admission was antenatal (33/77, 42.9%), followed by intrapartum (21/77, 27.3%) and postpartum (17/77, 22.1%); 7.8% (6/77) were admitted for abortion or ectopic pregnancy.

The three most common direct causes of death were obstetric hemorrhage (33/90, 36.7%), followed by eclampsia/preeclampsia (20/90, 22.2%) and infection or sepsis (17/90, 18.9%). Common indirect causes included anemia (18/90, 20.0%), cardiovascular disease (15/90, 16.7%), and respiratory disease (10/90, 11.1%). Deaths from indirect causes or unconfirmed direct causes without documented hemorrhage accounted for one-quarter of cases (23/90, 25.6%) and were classified as non-hemorrhagic.

Half of the cohort (39/80, 48.8%) received at least one blood transfusion during hospital admission. Contributory factors were documented in 85.6% (77/90) of cases; the most common were failure to recognize danger signs of pregnancy (28/77, 36.4%), resource-constrained care at Nakuru Hospital (22/77, 28.6%), and missed antenatal care (19/77, 24.6%) (Table 4).

Univariate Regression

In univariate analyses, receipt of a blood transfusion was associated with increased odds of hemorrhagic death (OR 4.15, 95% CI 1.57–10.96, p = 0.004) (Table 2). Lack of blood products, a clinician-documented contributory factor, increased the odds more than eightfold (OR 8.33, 95% CI 2.05–33.83, p = 0.003). Delays or miscommunication between referring facilities and Nakuru Hospital had a threefold increased odds of hemorrhagic death, though this did not reach statistical significance (OR 3.50, 95% CI 0.92–13.24, p = 0.065). Patient-level barriers, such as delays in seeking care or failure to recognize danger signs, were inversely associated with hemorrhagic death (OR 0.04, 95% CI 0.003–0.44, p = 0.009; OR 0.31, 95% CI 0.11–0.89, p = 0.03, respectively), indicating higher odds of non-hemorrhagic death in these groups.

No significant associations were found between hemorrhagic death and age, parity, referral status, anemia, missed antenatal care, resource constraints at the referring facility or Nakuru Hospital, lack of patient information, poverty, or cultural or religious barriers.

Multiple Logistic Regression

After adjusting for age, parity, and significant univariate variables, lack of blood products remained associated with hemorrhagic death (OR 19.91, 95% CI 2.23–177.97, p = 0.007). Delay in seeking healthcare was associated with higher odds of non-hemorrhagic death (OR 0.04, 95% CI 0.003–0.44, p = 0.009). A post-hoc power calculation using observed exposure proportions for lack of blood products indicated approximately 80% power to detect an odds ratio of 6.6 at α = 0.05; smaller effects were likely underpowered.

Repeated Cross-Sectional Study

Blood availability after implementation of the hospital-based blood bank in June 2020 did not show significant changes in monthly blood units dispatched to the maternity ward (88 ± 16 vs. 87 ± 16 units/month, p > 0.99) or in monthly hemorrhagic maternal deaths (1.0 ± 1.0 vs. 0.0 ± 1.0, p = 0.22) when comparing pre- and post-implementation periods (Table 3, Figure 2; full dataset in Table 5). No significant differences were observed in the mean number of obstetric patients receiving transfusions, total maternal deaths, or live births. Post-hoc power calculations indicated that the study had only 36% power to detect the observed difference in monthly hemorrhagic deaths and 5% power for monthly blood units, suggesting it was underpowered to identify modest pre-post changes.

In Poisson regression, however, each additional blood unit delivered to the maternity ward was associated with a 5% lower incidence rate of hemorrhagic death (IRR 0.95, p = 0.07), although this association did not reach statistical significance (Figure 3).

Our analyses at a Kenyan national referral hospital highlight key associations— most notably, that unavailability of blood products increased the odds of hemorrhagic deaths, and delays in seeking care increased the odds of non-hemorrhagic deaths. Blood bank data showed that lower blood availability correlated with higher hemorrhagic death rates, with fluctuations in supply before and after the hospital blood bank was established in June 2020. To our knowledge, this is the first study in a Kenyan healthcare setting to directly compare risk factors between hemorrhagic and non-hemorrhagic obstetric deaths. These findings build on a broader understanding of how systemic healthcare challenges contribute to Kenya’s persistently high maternal mortality rate.

We report a hospital-wide maternal mortality rate of 210 per 100,000 live births, aligning with national estimates which range from 36.2 to 510 [6,13,23,27–29]. Outcomes of obstetric complications are strongly affected by delays in accessing timely, life-saving care. According to the Three Delays Model, critical delays occur when seeking care, arriving at a health facility, and receiving adequate care [30]. The third delay— accessing quality care— is a key factor in maternal deaths in sub-Saharan Africa, often due to insufficient provider training, drug procurement challenges, staff shortages, inadequate equipment, and low staff motivation [9,11,22,27,31]. Prior studies in other countries in sub-Saharan Africa highlight the influence of third delay factors in postpartum hemorrhage deaths, including blood access, interhospital transfers, and lack of obstetric-gynecologic specialists [32].

Our findings further underscore the quantitative relationship between blood availability and hemorrhagic maternal deaths, reinforcing the critical role of robust hospital blood systems. After the hospital-based blood bank was established in June 2020, monthly blood units dispatched to the maternity ward did not substantially change, and there was no statistically significant reduction in hemorrhagic deaths. Post-hoc power calculations showed that the repeated cross-sectional analysis had only 36% power for the observed difference in monthly hemorrhagic deaths and 5% power for monthly blood units, indicating the study was underpowered to detect modest pre–post changes. Poisson regression analysis assessed monthly variability in blood availability across the full study period. This approach revealed that each additional blood unit was associated with a 5% lower incidence rate of hemorrhagic death, suggesting a possible dose-response relationship. While this trend did not reach statistical significance, it may still be clinically meaningful and hypothesis-generating. The persistently limited supply of blood products, the small number of monthly hemorrhagic deaths, and a post-implementation decline in live births likely contributed to the absence of a detectable change in the pre-post comparison. High transfusion rates among non-hemorrhagic deaths may reflect treatment of baseline anemia or unrecognized hemorrhage, particularly given the high prevalence of anemia in the study population. Prior studies in sub-Saharan Africa have identified multiple drivers of blood shortages, including donor scarcity, limited supplies and transport, inefficient recruitment, inappropriate transfusion, cultural barriers, stigma related to HIV status, and insufficient funding for blood services [11, 33, 34]. Future work could examine the longitudinal impact of decentralized hospital-based blood banks across multiple wards and endpoints, as well as contextual factors underlying blood availability trends.

Our secondary findings highlight that first- and second-delay barriers, including failure to recognize danger signs and delay in seeking healthcare, had stronger associations with non-hemorrhagic maternal deaths. This suggests gaps in patient knowledge and responses to complications like pre-eclampsia, infections, and indirect causes of death. Previous studies in Kenya have shown that only two-thirds of antenatal patients know even one danger sign, with most awareness centered on hemorrhagic symptoms rather than signs of other critical conditions (e.g., reduced fetal movement, abnormal vaginal discharge) [35]. Moreover, nearly half of patients lack a clear emergency plan [35]. A recent multicenter study in Nairobi similarly found that deaths from non-hemorrhagic causes (i.e., infection and eclampsia) were often linked to delayed care-seeking [36]. These findings emphasize the importance of educating patients about a broader range of danger signs and supporting them in establishing care plans.

This study has several limitations. As a single-center, retrospective review, it may not represent national trends, and data sources may contain redundant and missing information. Due to incomplete patient records, we could not collect data on clinical characteristics that might increase obstetric risks (e.g., anemia prevalence, estimated blood loss), nor on other key interventions in hemorrhage management (e.g., IV fluids, uterotonics). Furthermore, using the number of blood units dispatched to the maternity ward as a proxy for transfusions may underestimate actual transfusion events, and some deaths classified as non-hemorrhagic may have involved unrecognized hemorrhage. Finally, maternal deaths occurring outside the hospital or without completion of MDR/MDNF forms may be underreported, potentially underestimating the true burden of maternal mortality.

Our case-control analysis found that lack of blood products was significantly associated with hemorrhagic maternal death, while delays in seeking care were more often linked to non-hemorrhagic deaths. In the repeated cross-sectional analysis, implementation of a hospital-based blood bank did not significantly change blood supply or hemorrhagic mortality, likely due to minimal change in monthly blood units and low statistical power. These findings underscore the critical roles of reliable blood availability and timely care-seeking in improving maternal outcomes. Interventions to strengthen blood transfusion systems and expand patient education on obstetric emergency signs are essential in high-volume settings. Further research should evaluate locally adaptive blood bank models and strategies to reduce delays in emergency care access.

We thank our collaborators at Nakuru Level 6 Hospital for their approval of onsite study activities and for supporting data collection and analyses. We thank the Pathways for Innovative Transfusion Systems in Kenya (PITS- Kenya) “Bloodsafe” research program funded by the NHLBI/NIH (Strathmore IRB Approval # SU-IERC0992/21 and University of Pittsburgh IRB approval # STUDY21020040) for providing the infrastructure to conduct research activities. This international research collaboration initiative aims to identify and implement strategies to improve the availability and safety of blood transfusion services in Sub-Saharan Africa. (https://bloodsafe-research.org/)