https://orcid.org/0009-0006-6930-3132
Figures
Abstract
Maternal and neonatal mortality rates remain high in sub-Saharan Africa despite increased facility-based childbirth, reflecting persistent gaps in quality of care. We evaluated the impact of MENTORS, a facility based nurse mentorship and training program, on provider knowledge and childbirth care quality. In an unblinded parallel-arm cluster randomized controlled trial conducted between November 2021, 40 health facilities in Kenya were randomized to receive continuing medical education lectures, simulation-based drills, and in-facility mentorship to strengthen obstetric and newborn care. Baseline data included facility assessments and provider surveys; endline data included provider surveys, a standardized neonatal resuscitation simulation, and childbirth observations. Primary outcomes included provider knowledge, quality of routine maternity care, and quality of neonatal resuscitation management. All outcomes were assessed at endline. The trial was registered at ClinicalTrials.gov (NCT05110521) and the AEA RCT Registry (0008449). At baseline (November-December 2021), 357 providers were interviewed; at endline (May-August 2022), 376 providers were interviewed and 4,377 childbirths were observed. Routine maternity care quality was low across facilities, with providers performing 58.6% of essential actions in intervention facilities and 57.0% in control facilities at endline. There was no significant difference between the intervention and control facilities in provider knowledge (unadjusted domain effect 0.006, 95% CI: -0.038, 0.051) nor the quality of routine maternity care (0.018, 95% CI: -0.029, 0.065). However, providers in intervention-exposed facilities demonstrated improved quality of neonatal resuscitation in simulation-based assessments (0.094, 95% CI: 0.000, 0.188), with suggestive evidence of improved performance in standardized childbirth observations as well (0.041, 95% CI: -0.001, 0.083). Routine maternity care quality was suboptimal across both intervention and control facilities. While the facility-based mentorship program did not significantly improve provider knowledge or the quality of routine childbirth care, it showed evidence of improvements in neonatal resuscitation, although these findings were not robust across all sensitivity analyses.
Citation: Zhang L, Vatsa R, Chang W, Akinyi S, Little S, Gakii C, et al. (2026) Impact of a long-term mentorship program for childbirth providers on quality of care: A cluster randomized controlled trial. PLOS Glob Public Health 6(9): e0006073. https://doi.org/10.1371/journal.pgph.0006073
Editor: Patience A. Afulani, University of California San Francisco, UNITED STATES OF AMERICA
Received: February 11, 2026; Accepted: August 4, 2026; Published: September 2, 2026
Copyright: © 2026 Zhang et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: Data cannot be shared publicly because of potentially sensitive patient information, and the ethical approval under which the data were collected indicated the data were to be accessible to only the study team and would not be transmitted elsewhere. For further information, please contact the Amref Health Africa Ethics and Scientific Review Committee (esrc.kenya@amref.org) or the Harvard Chan School Office of Regulatory Affairs and Research Compliance (orarc@hsph.harvard.edu).
Funding: This study was funded by the United States Agency for International Development (USAID)’s Development Innovation Ventures program (FAIN 7200AA20FA00025) and the Bill & Melinda Gates Foundation (INV-018440 and INV-043251). Jacaranda Health contributed some independent funds to support the study as well. JC and MM received a salary from USAID and the Bill & Melinda Gates Foundation. RV was supported by award Number T32GM007753 and T32GM144273 from the National Institute of General Medical Sciences (NIGMS) for his MD-PhD Training. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institute of General Medical Sciences or the National Institutes of Health (NIH). USAID, the Bill & Melinda Gates Foundation, NIGMS, and NIH had no role in study design, data collection and analysis, decision to publish, or preparation of the study manuscript.
Competing interests: I have read the journal’s policy and the authors of this manuscript have the following competing interests: AW, CK, SA, SL, and SR are employees of Jacaranda Health.
Introduction
Globally, maternal mortality rates reveal stark disparities: 10 per 100,000 live births in high-income countries compared to 346 per 100,000 in low-income countries, with sub-Saharan Africa accounting for 70% of global maternal deaths [1]. Like much of sub-Saharan Africa, Kenya continues to face high maternal mortality, estimated at 379 maternal deaths per 100,000 live births [1]. These figures reflect persistent challenges including inadequate access to emergency obstetric care and problems with the quality of maternal health services [2,3].
While significant strides have been made over the past few decades in increasing access to routine maternal and newborn care and the rate of facility-based childbirth in sub-Saharan Africa, [4] several studies show that higher institutional childbirth rates have not consistently translated into better outcomes [5,6]. Women continue to experience both “too little, too late”, care that is delayed, unavailable, or delivered with inadequate resources, and “too much, too soon”, the overuse of non-evidence-based or unnecessary interventions [7]. In Kenya, quality of maternity care is highly variable and often lacking, characterized by unnecessary interventions such as C-sections, disrespectful treatment, and non-adherence to routine clinical guidelines [8,9].
Recent efforts to improve maternal and newborn health have largely focused on training healthcare providers. Global initiatives have emphasized “skills and drills”, simulation exercises, and short courses typically 1–5 days in length, designed to strengthen clinical performance and emergency response [10,11]. Evaluations across low-and-middle income countries have documented improved adherence to clinical protocols, and greater provider knowledge, but effects on maternal and newborn health outcomes have been mixed [10,12,13]. A potential reason for the modest effect of these programs on health outcomes is the limited nature of the training. Without repeated practice, reinforcement, or sustained supervision, providers may struggle to integrate newly learned practices into routine clinical care [14].
Mentorship programs have been proposed as an alternative. These multi-month longitudinal programs typically involve more experienced healthcare providers working with less experienced mentees to improve their performance, support their professional development, and enable skills transfers [15]. Evidence to date suggests that mentorship programs can improve knowledge and competency, particularly in simulated settings [16–18]. However, cluster-randomized and stepped-wedge trials of training and coaching interventions have shown that improvements in knowledge and adherence to clinical practices do not consistently translate into better teamwork, clinical performance, or maternal and neonatal outcomes [19–21]. Further evidence is needed to determine whether these improvements translate into improved quality of care and health outcomes. Thus, there is a need to rigorously evaluate the impact of clinical training and mentorship programs on quality of maternity care, especially when scaled up in sub-Saharan African settings like Kenya [10,16].
In this paper, we conduct a randomized evaluation of MENTORS, a nurse mentorship and emergency obstetric and newborn care training program developed by Jacaranda Health, a maternal and child health organization in Kenya. Unlike most training programs that rely on short-term or one-off sessions, [10] MENTORS is designed to provide longitudinal, facility-embedded support for maternity care. We use direct childbirth observations and provider surveys from facilities across Kenya to measure quality of childbirth care, maternal and neonatal health outcomes, and provider knowledge.
Methods
Ethics statement
This study was licensed by the National Commission for Science, Technology, and Innovation in Kenya (NACOSTI/P/21/13369) and approved by the institutional review board of the Harvard T.H. Chan School of Public Health (IRB21–1013) and the ethics and scientific review committee of Amref Health Africa (P1047/2021). The trial was implemented in the 40 study facilities following approval of the study protocol by respective facility in-charges and county health directors. Following the study endline, control arm facilities also began receiving the intervention package.
All data were collected by enumerators who received extensive training in research ethics, proper conduct in maternity wards, and study procedures by the local implementation partner. Written, electronic, or thumbprint-based informed consent was obtained from all facility in-charges, health workers, and patients involved in the study. Only women who were stable and able to focus on the consent process were approached for participation. An emphasis on voluntary participation, participant confidentiality, and freedom to withdraw consent at any time was upheld throughout. Women under 18 years of age, but at least 15 years of age, were allowed to consent and participate in the study on their own volition, given pregnant individuals are considered mature minors in Kenya. To prevent undue harm to any patient, psychological supports and abridged protocols were offered in cases of stillbirth, miscarriage, or infant loss at the time of childbirth.
Study design
We conducted an unblinded parallel arm cluster randomized controlled trial (RCT) between November 2021 and August 2022 in 40 public and faith-based health facilities across eight counties in Kenya (Fig A in S1 Appendix). The intervention, designed and implemented by Jacaranda Health, contained two parts: i) a nurse mentorship and emergency obstetric and newborn care training program (“MENTORS”) and ii) a digital health platform supporting pregnant and postpartum women through informational messages, appointment reminders, and a clinical helpdesk (“PROMPTS”). In this manuscript, we restrict our focus to the MENTORS program. Previous research from our team separately analyzed the impact of the PROMPTS program, finding the platform improved mothers’ antenatal and postpartum knowledge, preparedness, and health behaviors, but did not affect care seeking behavior for danger signs.[22]
Facilities selected for inclusion in the study were either government-owned or faith-based, had an average of 50–400 normal vaginal births per month (based on records from the Kenya Health Management Information System), and were in counties in which Jacaranda Health had a plan to expand. To ensure no overlap with concurrent interventions, eligible facilities were also required to have no ongoing mobile health or quality improvement research initiative. The birth volume criterion was set to enable sufficient statistical power for outcomes derived from standardized childbirth observations and because the MENTORS program primarily targets facilities of this size, namely mid- to high-volume facilities.
To ensure sufficient distance between study facilities to minimize the risk of spillovers, candidate facility locations were reviewed by the study team and in the case where two facilities were near each other (i.e., in the same sub-county), the lower-level or lower-volume facility was excluded. Block 1:1 randomization was done at the facility level and stratified based on tertile of normal vaginal birth volume. Following randomization, the minimum distance between facilities in different study arms was approximately 10 kilometers.
Intervention
MENTORS is a facility-based provider-facing program aimed at increasing and sustaining skills in basic emergency obstetric and newborn care, which includes the timely recognition and management of common complications during childbirth and the immediate postpartum period. As of 2024, MENTORS had trained over 10,000 nurses in basic and emergency obstetric and newborn care across 400 + health facilities. To enable low-cost scale up, the program uses a “training of trainers” model (Fig B in S1 Appendix). Lead mentors underwent a structured training program conducted by Jacaranda Health, consisting of six weeks of full-time training that included both clinical content in basic and emergency obstetric and newborn care and supervised shadowing with experienced mentors. Lead mentors were assessed during the training period on clinical knowledge and skills, including performance on standardized knowledge assessments and neonatal resuscitation examinations, prior to deployment. Lead mentors helped facilitate mentorship activities within their respective counties and trained other providers to become in-facility mentors. In-facility mentors were selected from existing maternity staff in participating facilities, including nurse midwives, clinical officers, and medical officers, based on clinical experience and recommendations from county health leadership. These in-facility mentors received one week of full-time training with knowledge-based pretests and posttests and skill-based simulations. They subsequently led longitudinal training and mentorship for facility-based peers, including continuing medical education (CME) lectures, simulation-based drills, and childbirth debriefs.
Each participating facility is expected to hold a minimum of two CME lectures and one simulation-based drill every month offering hands-on opportunity to bridge classroom learning with clinical practice. The CME lectures consist of onsite and virtual sessions approximately 45 minutes in length that cover a wide range of topics including normal labor and childbirth, postpartum hemorrhage, neonatal resuscitation, shoulder dystocia, respectful maternity care, hypertensive disorders of pregnancy, complications of abortion, and infection prevention. Meanwhile, the drills consist of 1.5–2-hour long simulation-based trainings covering topics such as normal childbirth, shoulder dystocia, eclampsia, breech birth, and assisted vaginal birth. Mentors select topics based on their identification of facility needs and training equipment availability.
Data
Data were collected by trained enumerators at multiple points throughout the study period (Fig C in S1 Appendix). Baseline data collection, which occurred from November 5th 2021 to December 7th 2021, consisted of a facility assessment and provider survey. Endline data collection occurred from May 2nd 2022 to August 3rd 2022, and consisted of a provider survey, neonatal resuscitation simulation, and childbirth observations. Endline data collection was split into two waves based on geographic location, with 20 facilities per wave, and lasted for one to two months in each facility depending on the number of childbirths observed. 90 enumerators conducted childbirth observations across the two waves, with three to four enumerators collecting data per facility. Enumerators observed births at least four days per week, including on weekends and overnight.
Facility Assessment: Enumerators conducted a baseline facility assessment with facility staff collecting information on facilities’ infrastructure, staffing, available procedures, medical supplies and equipment, referral management, treatment protocols, and patient-borne costs associated with maternity services, using tools adapted from prior studies [23].
Provider Survey: Maternity ward providers were eligible for the provider baseline survey if they had no plans to move to a different facility within the following 6 months. Providers were interviewed about their clinical education, experience with different maternity services, knowledge of recommended maternal and newborn health care practices, recent in-service training in obstetrics, and workplace motivation. The endline provider surveys included the same elements as the baseline survey and added a standardized simulation experience, previously validated in Tanzania, to assess the newborn resuscitation skills of birth attendants [24].
Childbirth Observations: Childbirth observations began approximately 5 months after the intervention began. Enumerators – who themselves were required to have pursued or graduated with an accredited medical, nursing, or midwifery degree and to have experience in maternal and newborn health – received training on the observation tools under the supervision of OBGYNs and midwives. They were trained to minimize disruption to care processes and were instructed to intervene only in the case of an avertable life-threatening emergency. The childbirth observations involved a detailed checklist of procedures in routine and emergency obstetric and newborn care, adapted from a checklist previously used in Kenya [25]. Emergency obstetric care was recorded for four specific types of complications: neonatal resuscitation, postpartum hemorrhage, shoulder dystocia, and severe pre-eclampsia/eclampsia. Maternal data (e.g., age, marital status, parity, gravidity) and newborn details (e.g., birthweight, sex) were obtained from the Ministry of Health maternity register following review of patient files after discharge.
Outcomes
The outcomes for this study were pre-registered on the United States Clinical Trials Registry (NCT05110521; registered 10/25/2021) and American Economic Association RCT Registry (AEARCTR-0008449; registered 10/28/2021). Primary and secondary outcomes were organized within pre-specified domains to account for multiple comparisons and reduce the risk of false positives.
Pre-registered primary outcomes for the MENTORS component of the study included provider knowledge (of obstetric and newborn care), quality of routine maternity care, and quality of neonatal resuscitation management. Provider knowledge was assessed in the endline survey as the proportion of correct responses to ten knowledge questions on normal labor, infection prevention and control, pre-eclampsia/eclampsia, postpartum hemorrhage, newborn care, and shoulder dystocia. Quality of routine maternity care was measured from childbirth observations as the share of essential actions that providers performed, based on an equally weighted 20-item index published by Tripathi et al [26]. The Tripathi index has been previously validated in Kenya, with higher scores denoting better quality of care. In exploratory analyses, we disaggregated care quality across the four stages of the Tripathi index (initial client assessment and examination, first stage of labor, second and third stage of labor, and immediate newborn and postpartum care). Finally, quality of neonatal resuscitation management was measured using a structured, scenario-based simulation based on the Helping Babies Breathe curriculum [27]. Providers were presented with a standardized clinical scenario of a non-breathing newborn following a complicated delivery and were asked to demonstrate resuscitation using a mannequin and standard equipment. Enumerators guided the simulation by providing clinical information (e.g., heart rate) in response to provider assessments. Performance was evaluated using a standardized checklist of essential resuscitation steps, and providers were assessed on the fraction of equally weighted steps performed correctly. We also assessed, as an exploratory outcome, whether providers achieved a perfect score on the neonatal resuscitation simulation as all elements of the simulation need to occur for a resuscitation to be successful in practice.
Pre-registered secondary outcomes included an index of provider motivation and satisfaction, measured via a scale validated in Kenya by Mbindyo et al. (2009); [28] an index of quality of management of maternal and neonatal complications (i.e., neonatal resuscitation, postpartum hemorrhage, shoulder dystocia, severe pre-eclampsia/eclampsia); an index of quality of infection prevention; an index of quality of partograph completion; an index of provider treatment of patients during labor and childbirth, assessed using an interpersonal abuse and unsupportive birth environment measure developed by Berger et al. (2022); [29] a composite index of severe maternal and neonatal morbidity (i.e., postpartum hemorrhage, severe pre-eclampsia, eclampsia, maternal sepsis, neonatal sepsis or systemic infection, uterine rupture, neonatal asphyxia, maternal mortality, and perinatal mortality); and timeliness of care delivery (e.g., between facility arrival and first contact with a provider). In exploratory analyses, we evaluated process- and outcome-based measures for individual complications observed by enumerators with sufficient frequency during childbirth observations, namely neonatal resuscitation and postpartum hemorrhage. The outcomes included quality of neonatal resuscitation management, successful neonatal resuscitation, quality of postpartum hemorrhage management, and stabilization (i.e., controlling or ending) of postpartum bleeding. A detailed description of each primary and secondary outcome, including specific index components, is provided in Table A in S1 Appendix.
Power
Based on prior work by members of our study team, we assumed that the mean quality of routine maternity care in the control group would be 58%, with a standard deviation of 13%. With an expected 100 fully observed childbirths per cluster, 20 clusters per study arm, intra-cluster correlation of 0.35, significance level of 0.05, and power of 80%, we calculated a minimum detectable effect size of 6.88 percentage points.
Statistical analysis
We first descriptively assessed baseline characteristics of facilities, providers, and patients whose births were observed in the intervention and control facilities. We then examined, at endline, provider attendance at in-service trainings within the preceding six months. To assess the impact of the MENTORS program, we estimated the intention-to-treat effect on provider-level and individual childbirth observation-level outcomes at endline using linear models. Outcomes were regressed on a binary variable indicating whether the facility was in the intervention group, the baseline facility birth volume tertile (stratification variable), and an indicator for whether the endline data were collected in wave 1 or 2. We did not collect and therefore do not adjust for measures of each outcome at baseline. Accordingly, our findings should be interpreted as endline differences rather than change-from-baseline effects. For childbirth observation-level outcomes, models were estimated on the sample of fully observed childbirths. Due to the facility-level randomization, heteroskedasticity robust standard errors were used and clustered at the facility level. For select exploratory outcomes, we generated figures to visualize qualitative trends. Analyses and data visualization were done in Stata version 17 and R version 4.2.1.
Sensitivity analysis
We conducted three sensitivity analyses. First, to improve precision and account for potential facility- and county-level differences, we estimated adjusted models that included controls for facility and county. Second, although primary and secondary outcomes were organized within pre-specified domains to account for multiple comparisons and reduce the risk of false positives, we additionally assessed robustness to multiple hypothesis testing by applying Benjamini-Hochberg adjustments to the P-values for the pre-registered outcomes. Third, during baseline data collection, we identified three facilities in the intervention arm that were concurrently participating in other research studies aimed at improving care for specific obstetric emergencies, and we repeated analyses excluding these facilities.
Results
Among the 40 study facilities, five were Level 3 health centers, 32 were Level 4 hospitals, and three were Level 5 county referral hospitals. A total of 360 providers were recruited at baseline (~60% of all maternity care providers in study facilities), and 357 were eligible and consented to be interviewed. Meanwhile, 376 providers were recruited at endline, all of whom consented to be interviewed; 242 (68%) of these providers were also interviewed at baseline.
Facility and provider characteristics were well-balanced between the intervention and control groups at baseline (Table 1). On average, facilities had 15 providers working in the maternity department and 23 beds for maternity care. Nearly all facilities (98%) offered around-the-clock obstetric and neonatal care. On average, facilities were equipped with 76% of the essential equipment for maternity care and 90% of essential drug categories. 68% of facilities had a functional blood bank. The number of providers interviewed at each facility ranged from 5-18. Most providers (78%) were female, with an average age of 25. Providers on average had worked at the study facility for 4.3 years, with 5.4 years of experience working in maternity care. 10% of providers had at least a bachelor’s degree.
A total of 4,377 admissions were observed for normal labor and childbirth (Fig 1), with 983 observations being incomplete (most commonly because the observation began after the first exam, the patient was referred for cesarean section or transfer upon arrival, or the enumerator’s shift ended before the observation was complete). Patient characteristics from incomplete childbirth observations are reported in Table B in S1 Appendix. Among the remaining fully observed births, patient characteristics were similar between the intervention and control groups (Table 2). For additional clinical context, Table C in S1 Appendix reports rates of maternal and neonatal complications and adverse outcomes observed during childbirth; these outcomes may be influenced by care received during labor and delivery and are therefore presented separately from the childbirth sample characteristics.
The average age of mothers giving birth in control facilities was 25.7 compared to 25.4 among intervention facilities. Pregnant women also attended a similar number of antenatal care visits, with an average of 3.7 visits for women observed in control facilities and 3.6 in intervention facilities. The average documented birthweight for observed births was 3,078.9 grams in control facilities and 3,026.1 grams in intervention facilities.
Receipt of training varied substantially across study facilities. The number of CME sessions and simulation drills conducted over the study period ranged from 5 to 19 and 8–17, respectively. At endline, provider interviews indicated that a higher proportion of providers in intervention facilities attended in-service trainings, across all eight types assessed, compared to control facilities (Fig 2). These included trainings conducted by Jacaranda Health and in-service trainings offered as part of providers’ routine workflow. For routine care in-service trainings, 59% of providers in intervention facilities compared to 29% of providers in control facilities reported attending a training in the last 6 months.
Timeliness of care outcomes were not analyzed due to substantial measurement error in recorded time stamps, with 25% of observations having missing or implausible values.
Provider knowledge
At baseline, the average provider knowledge index was 0.52 (out of 1) in intervention facilities, and 0.49 in control facilities, indicating that providers correctly responded to half of questions on a knowledge assessment of maternal and newborn health topics. By endline, provider knowledge scores improved to 0.58 in the intervention group and 0.57 in the control group, with no statistically significant differences (95% CI: -0.038, 0.051; P = 0.772) between groups (Table 3). Fig 3A details provider scores across each of the 10 knowledge topics. Providers scored lowest on labor monitoring, with only 28% in intervention facilities and 29% in control facilities correctly specifying monitoring frequency for vital signs and vaginal bleeding during the fourth stage of labor. Scores were highest on pre-eclampsia diagnosis, with 94% of providers in the intervention group and 92% in the control group correctly diagnosing pre-eclampsia.
Quality of routine and emergency obstetric and newborn care
Based on the 20-item quality of care index by Tripathi et al., providers in intervention facilities correctly performed 58.6% of the assessed items, compared to 57.0% in control facilities, with an observed intra-cluster correlation of 0.36 (Fig 3B). Essential actions were least likely to be completed during initial client assessment and examination. Providers in intervention facilities asked mothers about their HIV status in only 18.8% (11.7% in control facilities) of births, inquired about headaches or blurred vision in 13.1% (11.7%) of births, and washed hands or used alcohol rubs in just 14.5% (21.2%) of births. The highest rates of adherence were in the second and third stages of labor where providers in intervention facilities correctly performed 91.0% of essential items (93.0% in control facilities). In the immediate newborn and postpartum period, palpation of the uterus 15 minutes after birth occurred in only 12.6% of births in intervention facilities and 14.8% in control facilities, while checking vital signs at this interval occurred in just 14.2% and 8.8%, respectively. At endline, there were no significant differences in the overall quality of routine maternity care between intervention and control groups (95% CI: -0.029, 0.065; P = 0.446; Table 3).
Quality of care in neonatal resuscitation simulation
Providers in intervention facilities scored higher on the neonatal resuscitation simulation, correctly performing 79.1% of simulation steps compared to 70.0% of simulation steps in control facilities (95% CI: 0.000, 0.188; P = 0.050; Table 3). In exploratory analyses, we found providers in the intervention group were 17.2 percentage points (95% CI: 8.1, 26.3; P < 0.001; Table 3) more likely to achieve a perfect score on the simulation compared to a control group rate of 3.9%.
Secondary outcomes
Analysis of pre-registered secondary outcomes showed no significant improvement in provider motivation and satisfaction, management of maternal and neonatal complications (in aggregate), quality of infection prevention, quality of partograph completion, and patient treatment during labor and childbirth (Table 3). In exploratory analyses of standardized childbirth observations, we found no statistically significant improvements in neonatal resuscitation quality and successful resuscitation, as well as postpartum hemorrhage management. However, the estimated effects for neonatal resuscitation quality (95% CI: -0.001, 0.083; P = 0.056) and successful resuscitation (95% CI: -0.009, 0.013; P = 0.088) were positive but did not meet the pre-specified 5% significance threshold.
Sensitivity analysis
In adjusted analyses, the improvement in the primary simulation-based neonatal resuscitation outcome was only marginally significant (Table D in S1 Appendix); after applying Benjamini-Hochberg adjustments for multiple comparisons and excluding the three facilities exposed to other training interventions, the improvement was no longer statistically significant (Tables E and F in S1 Appendix). In contrast, improvements in the proportion of providers achieving a perfect score on the resuscitation simulation remained significant across sensitivity analyses (Tables D and F in S1 Appendix).
Discussion
We conducted a cluster RCT to evaluate the impact of an onsite nurse mentorship and simulation training intervention on provider knowledge, quality of childbirth care, and maternal and neonatal health outcomes in Kenya. The bundled mentorship intervention included monthly CME lectures and simulation drills focused on basic and emergency obstetric and newborn care over a period of nine months. While the intervention increased the likelihood that providers had received training, we did not observe improvements in provider knowledge or care metrics such as quality of routine maternity care, partograph completion, infection prevention, and management of neonatal and maternal complications in aggregate. However, providers in intervention facilities demonstrated higher-quality neonatal resuscitation care in simulation-based assessments, with suggestive evidence of concordant improvements during standardized childbirth observations. Although these specific findings were not robust across all sensitivity analyses, improvements in the proportion of providers achieving a perfect simulation score were robust at the 5% level. These findings are particularly relevant given that birth asphyxia is among the most common obstetric emergencies in Kenya and a leading cause of neonatal mortality. In contrast, we did not observe improvements in postpartum hemorrhage management, which, unlike neonatal resuscitation, was not separately evaluated through a standardized simulation assessment at endline.
These findings contribute to a mixed literature on the effectiveness of provider training interventions in low-resource settings. While several studies have reported positive impacts of training programs on provider knowledge and performance, many rely on pre-post designs without control groups, limiting their ability to attribute observed changes to the intervention itself [10,30]. Among randomized trials, results have been inconsistent. One randomized evaluation in Mexico found improvements in selected clinical practices following simulation-based training, though these gains were limited to targeted behaviors rather than broader measures of care quality [31]. A large cluster RCT in Kenya and Uganda reported reductions in neonatal mortality, specifically among preterm and low-birthweight newborns, when simulation-based training was embedded within a broader quality improvement package [32]. In contrast, an RCT conducted in Tanzania by Larson et al. (2020) found limited impacts of in-service training combined with mentorship and infrastructure support on quality improvement [33]. A separate trial of a coaching-based implementation of the WHO Safe Childbirth Checklist improved adherence to essential practices but did not translate into reductions in maternal or perinatal mortality [19].
The limited impact of the MENTORS intervention may reflect aspects of how the training was delivered rather than, or in addition to, its overall frequency. While facilities conducted regular CME lectures and simulation drills over the nine-month period, these sessions covered a broad range of topics, which may have limited opportunities for repeated practice and reinforcement of specific skills. Prior studies demonstrating positive impacts from mentorship interventions often involved a low-dose high-frequency training model with providers receiving several consecutive days of training supplemented by refresher sessions [17,34,35]. Additionally, although a higher proportion of providers in intervention facilities reported receiving in-service training on routine childbirth care compared to controls (59% vs. 29%), a substantial proportion of providers in the intervention arm still had not received recent training, potentially diluting any impact of the intervention.
Implementation challenges likely further constrained impact. During the study period, at least six mentor turnovers were reported across participating facilities, although the true number may have been higher. Providers may have also crossed over between intervention and control facilities during the study period, reducing differences between the study groups. Facilities also experienced staffing shortages that disrupted scheduling and training. It is possible that challenges such as limited staffing, competing responsibilities, and lack of protected time made it difficult for providers to attend training sessions, suggesting a potential mismatch between the intervention design and day-to-day provider workflows. Additionally, variability in baseline facility infrastructure and resource ability may have limited providers’ ability to implement new skills and/or maintain improvements in care quality [36]. These factors likely reduced both the fidelity and the intensity of implementation. At the same time, improvements in provider knowledge and care practices observed in both intervention and control facilities across the study period raise the possibility that external or concurrent quality improvement efforts we were unaware of may have influenced outcomes across both arms.
Finally, the limited impact of the intervention on our primary outcomes may reflect variation in implementation of the MENTORS curriculum across facilities. Although the program addressed both routine and emergency obstetric and newborn care, in-facility mentors tailored training to local needs, and routine care may not have been emphasized uniformly across sites. Nevertheless, we selected routine maternity care quality as a primary outcome because these practices affect a large share of patients, offering greater statistical power to detect meaningful changes. By contrast, obstetric emergencies, while clinically critical, occur relatively infrequently, limiting power to detect differences in such outcomes. Moreover, improving the quality of routine care may offer broader opportunities to prevent complications and avertable maternal and neonatal deaths.
Still, the intervention improved performance on the broadly administered neonatal resuscitation simulation, the sole primary outcome focused on emergency care, with exploratory analyses suggesting concordant improvements in childbirth observations. More widespread incorporation of simulations or drills targeting routine care practices – as well as secondary outcomes like partograph completion, infection prevention, and respectful maternity care – may likewise have produced stronger effects on those outcomes. This aligns with earlier work showing that frequent, focused drills can lead to sustained improvements in emergency preparedness and neonatal outcomes [10,17].
Key strengths of this study include its randomized design and the broad geographic coverage of study facilities, which included 40 facilities across eight counties in Kenya. We assessed the impact of the MENTORS intervention across multiple levels, including provider knowledge and skills, changes in clinical practice, and improvements in maternal and neonatal health outcomes. A particularly notable strength was the large volume of childbirth observations collected by trained enumerators at each of these study facilities using standardized protocols.
Nevertheless, there are several limitations to our study. To the extent that women gave birth in facilities in which they sought antenatal care, births observed in intervention facilities were likely of women who were also exposed to the PROMPTS digital health platform. While births were balanced on measured patient characteristics, PROMPTS may have introduced unmeasured differences in maternal knowledge, preparedness, or engagement with care that could have influenced observed delivery care quality. Our findings are also limited in generalizability, as our study was restricted to mid- to high-volume facilities (50–400 births per month); thus, the generalizability of our findings to lower-volume or more remote facilities remains uncertain. We also acknowledge that unmeasured post-randomization differences between facilities that were not directly related to the intervention cannot be fully characterized and may have contributed to outcome differences. In addition, the study was not sufficiently powered to detect effects on rarer emergency obstetric outcomes, such as postpartum hemorrhage and eclampsia, which were areas emphasized in the program’s content. While Jacaranda Health tracked mentorship activities held at the facility level, data on individual provider attendance were not collected. Thus, we could not link training participation to the providers observed at endline and are unable to comment on the extent to which individual providers were directly exposed to the intervention. Additionally, the types of trainings delivered varied across facilities, as facility mentors selected trainings based on their assessed need, which further introduced variation in exposure to the intervention. While the improvement in the primary simulation-based neonatal resuscitation outcome was not robust across all sensitivity analyses, the direction and magnitude of the effect – as well as concordant exploratory results in childbirth observations – suggest a meaningful difference between intervention and control facilities. Nevertheless, because providers in intervention facilities had greater exposure to simulation-based training, improved performance may partly reflect increased familiarity with the simulation format and assessment environment, in addition to gains in clinical skill. Finally, while our primary measure of childbirth quality was based on direct childbirth observations, which allowed us to directly capture provider actions rather than rely on self-reports or facility records, they may have been subject to the Hawthorne effect, wherein providers altered their behavior due to being observed. We consider this risk to be minimal, as overall quality remained low across both groups and provider scores were consistent between earlier and later observation days.
To conclude, despite modest increases in training exposure and the quality of neonatal resuscitation care, we did not identify improvements across other broad measures of provider knowledge or routine and emergency childbirth care quality. Our results suggest that insufficient intervention intensity, limited workflow integration, and/or lack of ongoing reinforcement may prevent mentorship programs from achieving their intended impact. Future research should explore how mentorship interventions can be optimized for greater effectiveness, whether through increased frequency and duration, integration with existing supervision systems, more focus on simulation-based training, and/or better alignment with provider incentives and facility capacity. Further work is needed to understand the specific mechanisms through which mentorship affects provider behavior and to identify which domains of care are most amenable to change. In addition, future research should examine the role of implementation fidelity, including mentor quality and the consistency and quality of training sessions, in shaping program effectiveness.
Supporting information
S1 Appendix. Fig A. Map of study facilities.
Fig B. MENTORS training-of-trainers program structure. Fig C. Timeline of study activities. Table A. Index components for primary and secondary outcomes. Table B. Patient characteristics among incomplete childbirth observations. Table C. Frequency of childbirth complications among completed childbirth observations. Table D. Adjusted intervention impact on primary and secondary outcomes. Table E. Intervention impact on pre-registered outcomes with Benjamini–Hochberg adjusted P-values. Table F. Intervention impact on outcomes excluding three facilities with ongoing similar interventions.
https://doi.org/10.1371/journal.pgph.0006073.s001
(DOCX)
Acknowledgments
We are deeply grateful to all study participants who generously shared their time and experiences with us. We also acknowledge the dedication of our study enumerators, whose careful and consistent data collection made this evaluation possible. We are grateful to the entire team at Jacaranda Health for their collaboration and support. In particular, we thank Jay Patel and Henry Njogu for their meticulous curation and provision of administrative data that informed our analyses, as well as Valerie Scott (Harvard T.H. Chan School of Public Health) for their assistance in preparing the study data.
Disclaimer: CK and SR are Co-Executive Directors; AW is Director of Research, Evaluation and Design; SL is Global Medical Director; and SA is a Research Manager at Jacaranda Health. Jacaranda Health had no other role in study design, data collection and analysis, decision to publish, or preparation of the study manuscript.
References
- 1.
World Health Organization. Trends in maternal mortality 2000 to 2023: estimates by WHO, UNICEF, UNFPA, World Bank Group and UNDESA/Population Division. 2025. https://data.unicef.org/resources/trends-in-maternal-mortality-2000-to-2023/
- 2. Austin A, Langer A, Salam RA, Lassi ZS, Das JK, Bhutta ZA. Approaches to improve the quality of maternal and newborn health care: an overview of the evidence. Reprod Health. 2014;11(2):S1.
- 3.
UNICEF. UNICEF Data. https://data.unicef.org/topic/child-survival/neonatal-mortality/. 2024. Accessed 2024 August 21.
- 4. Doctor HV, Nkhana-Salimu S, Abdulsalam-Anibilowo M. Health facility delivery in sub-Saharan Africa: successes, challenges, and implications for the 2030 development agenda. BMC Public Health. 2018;18(1):765. pmid:29921275
- 5.
Chari AV, Okeke EN. Can institutional deliveries reduce newborn mortality? Evidence from Rwanda. In: SSRN Scholarly Paper, 2014. https://doi.org/10.2139/ssrn.2800311
- 6. Shajarizadeh A, Grépin KA. The impact of institutional delivery on neonatal and maternal health outcomes: evidence from a road upgrade programme in India. BMJ Glob Health. 2022 Jul 6;7(7).
- 7. Miller S, Abalos E, Chamillard M, Ciapponi A, Colaci D, Comandé D, et al. Beyond too little, too late and too much, too soon: a pathway towards evidence-based, respectful maternity care worldwide. The Lancet. 2016;388(10056):2176–92.
- 8. Lusambili AM, Naanyu V, Wade TJ, Mossman L, Mantel M, Pell R, et al. Deliver on your own: disrespectful maternity care in rural Kenya. PLoS One. 2020;15(1):e0214836. pmid:31910210
- 9. Sharma J, Leslie HH, Kundu F, Kruk ME. Poor quality for poor women? inequities in the quality of antenatal and delivery care in Kenya. PLoS One. 2017;12(1):e0171236. pmid:28141840
- 10. Ameh CA, Mdegela M, White S, van den Broek N. The effectiveness of training in emergency obstetric care: a systematic literature review. Health Policy Plan. 2019;34(4):257–70. pmid:31056670
- 11. Rowe AK, Rowe SY, Peters DH, Holloway KA, Chalker J, Ross-Degnan D. Effectiveness of strategies to improve health-care provider practices in low-income and middle-income countries: a systematic review. Lancet Glob Health. 2018;6(11):e1163–75. pmid:30309799
- 12. Broek N van den, Ameh C, Madaj B, Makin J, White S, Hemming K, et al. Effects of emergency obstetric care training on maternal and perinatal outcomes: a stepped wedge cluster randomised trial in South Africa. BMJ Glob Health. 2019;4(6). pmid:10.1136/bmjgh-2019-001670
- 13. Carlo WA, Goudar SS, Jehan I, Chomba E, Tshefu A, Garces A, et al. Newborn-care training and perinatal mortality in developing countries. N Engl J Med. 2010;362(7):614–23.
- 14.
Jhpiego. Low Dose, High Frequency: A Learning Approach to Improve Health Workforce Competence, Confidence, and Performance. 2016. https://hms.jhpiego.org/wp-content/uploads/2016/08/LDHF_briefer.pdf
- 15. Manzi A, Hirschhorn LR, Sherr K, Chirwa C, Baynes C, Awoonor-Williams JK, et al. Mentorship and coaching to support strengthening healthcare systems: lessons learned across the five Population Health Implementation and Training partnership projects in sub-Saharan Africa. BMC Health Serv Res. 2017;17(Suppl 3):831. pmid:29297323
- 16. Feyissa GT, Balabanova D, Woldie M. How Effective are mentoring programs for improving health worker competence and institutional performance in Africa? a systematic review of quantitative evidence. J Multidiscip Healthc. 2019;12:989–1005. pmid:31824166
- 17. Ugwa E, Kabue M, Otolorin E, Yenokyan G, Oniyire A, Orji B, et al. Simulation-based low-dose, high-frequency plus mobile mentoring versus traditional group-based trainings among health workers on day of birth care in Nigeria; a cluster randomized controlled trial. BMC Health Serv Res. 2020;20(1):586. pmid:32590979
- 18. Ndwiga C, Abuya T, Okondo C, Akinyi S, Wickramanayake A, Warren CE. Effect of mentorship and a mHealth application in updating provider skills and knowledge in maternal and newborn care in two informal settlements of Nairobi. BMC Womens Health. 2023;23(1):580. pmid:37940919
- 19. Semrau KEA, Hirschhorn LR, Marx Delaney M, Singh VP, Saurastri R, Sharma N, et al. BetterBirth Trial Group. Outcomes of a Coaching-Based WHO Safe Childbirth Checklist Program in India. N Engl J Med. 2017;377(24):2313–24. pmid:29236628
- 20. van Tetering AAC, de Vries EL, Ntuyo P, van den Heuvel ER, Fransen AF, van der Hout-van der Jagt MB. Mono-professional simulation-based obstetric training in a low-resource setting: stepped-wedge cluster randomized trial. JMIR Med Educ. 2025. pmid:40344206
- 21. Blair AH, Openshaw M, Mphande I, Jana O, Malirakwenda R, Muller A, et al. Assessing combined longitudinal mentorship and skills training on select maternal and neonatal outcomes in rural and urban health facilities in Malawi. J Transcult Nurs. 2022;33(6):704–14. pmid:36062416
- 22. Vatsa R, Chang W, Akinyi S, Little S, Gakii C, Mungai J. Impact evaluation of a digital health platform empowering Kenyan women across the pregnancy-postpartum care continuum: a cluster randomized controlled trial. PLOS Medicine. 2025;22(2):e1004527.
- 23.
World Health Organization, United Nations Population Fund, Mailman School of Public Health. Averting Maternal Death and Disability, United Nations Children’s Fund (UNICEF). Monitoring emergency obstetric care: a handbook. Surveill Soins Obstétricaux Urgence Man Util. 2009;152.
- 24. Reisman J, Martineau N, Kairuki A, Mponzi V, Meda AR, Isangula KG, et al. Validation of a novel tool for assessing newborn resuscitation skills among birth attendants trained by the Helping Babies Breathe program. Int J Gynaecol Obstet. 2015;131(2):196–200. pmid:26283225
- 25. Clarke-Deelder E, Opondo K, Achieng E, Garg L, Han D, Henry J, et al. Quality of care for postpartum hemorrhage: a direct observation study in referral hospitals in Kenya. PLOS Glob Public Health. 2023;3(3):e0001670. pmid:36963063
- 26. Tripathi V, Stanton C, Strobino D, Bartlett L. Development and validation of an index to measure the quality of facility-based labor and delivery care processes in Sub-Saharan Africa. PLoS One. 2015;10(6):e0129491. pmid:26107655
- 27.
American Academy of Pediatrics. Helping Babies Breathe 2nd Edition. https://www.aap.org/en/aap-global/helping-babies-survive/our-programs/helping-babies-breathe/helping-babies-breathe-2nd-edition/. 2016. Accessed 2026 May 1.
- 28. Mbindyo PM, Blaauw D, Gilson L, English M. Developing a tool to measure health worker motivation in district hospitals in Kenya. Hum Resour Health. 2009;7:40. pmid:19457237
- 29. Berger BO, Strobino DM, Mehrtash H, Bohren MA, Adu-Bonsaffoh K, Leslie HH, et al. Development of measures for assessing mistreatment of women during facility-based childbirth based on labour observations. BMJ Global Health. 2022;5(Suppl 2):e004080. pmid:34362791
- 30. Kruk ME, Vail D, Austin-Evelyn K, Atuyambe L, Greeson D, Grépin KA, et al. Evaluation of a maternal health program in Uganda and Zambia finds mixed results on quality of care and satisfaction. Health Aff (Millwood). 2016;35(3):510–9. pmid:26953307
- 31. Fritz J, Walker DM, Cohen S, Angeles G, Lamadrid-Figueroa H. Can a simulation-based training program impact the use of evidence based routine practices at birth? Results of a hospital-based cluster randomized trial in Mexico. PLoS One. 2017;12(3):e0172623. pmid:28319122
- 32. Walker D, Otieno P, Butrick E, Namazzi G, Achola K, Merai R, et al. Effect of a quality improvement package for intrapartum and immediate newborn care on fresh stillbirth and neonatal mortality among preterm and low-birthweight babies in Kenya and Uganda: a cluster-randomised facility-based trial. Lancet Glob Health. 2020;8(8):e1061–70. pmid:32710862
- 33. Larson E, Mbaruku GM, Cohen J, Kruk ME. Did a quality improvement intervention improve quality of maternal health care? Implementation evaluation from a cluster-randomized controlled study. International Journal for Quality in Health Care. 2019;32(1):54–63.
- 34. Rouse M, Mphande I, Mwale N, Kapira S, Sayenda M, Mvula MG, et al. Exploring the long-term impact of a nurse-midwife mentorship intervention in Neno district, Malawi: a secondary data analysis of maternal and neonatal complications. BMC Glob Public Health. 2024;2(1):74. pmid:39681971
- 35. Gomez PP, Nelson AR, Asiedu A, Addo E, Agbodza D, Allen C, et al. Accelerating newborn survival in Ghana through a low-dose, high-frequency health worker training approach: a cluster randomized trial. BMC Pregnancy Childbirth. 2018;18(1):72. pmid:29566659
- 36. Larson E, Mbaruku GM, Cohen J, Kruk ME. Did a quality improvement intervention improve quality of maternal health care? Implementation evaluation from a cluster-randomized controlled study. Int J Qual Health Care. 2020;32(1):54–63. pmid:31829427