Health care providers’ knowledge of clinical protocols for postpartum hemorrhage care in Kenya: a cross-sectional study

Background: Postpartum hemorrhage (PPH) remains the leading cause of maternal death worldwide despite its often-preventable nature. Understanding health care providers’ knowledge of clinical protocols is imperative for improving quality of care and reducing mortality. This is especially pertinent in referral and teaching hospitals that train nursing and medical students and interns in addition to managing emergency and referral cases. Methods: This study aimed to (1) measure health care providers’ knowledge of clinical protocols for risk assessment, prevention, and management of PPH in 3 referral hospitals in Kenya and (2) examine factors associated with providers’ knowledge. We developed a knowledge assessment tool based on past studies and clinical guidelines from the World Health Organization and the Kenyan Ministry of Health. We conducted in-person surveys with health care providers in three high-volume maternity facilities in Nairobi and western Kenya from October 2018-February 2019. We measured gaps in knowledge using a summative index and examined factors associated with knowledge (such as age, gender, qualification, experience, in-service training attendance, and a self-reported measure of peer-closeness) using linear regression. Results: We interviewed 172 providers including consultants, medical officers, clinical officers, nurse-midwives, and students. Overall, knowledge was lowest for prevention-related protocols (an average of 0.71 out of 1.00; 95% CI 0.69–0.73) and highest for assessment-related protocols (0.81; 95% CI 0.79–0.83). Average knowledge scores did not differ significantly between qualified providers and students. Finally, we found that being a qualified nurse, having a specialization, being female, having a bachelor’s degree and self-reported closer relationships with colleagues were statistically significantly associated with higher knowledge scores. Conclusion: We found gaps in knowledge of PPH care clinical protocols in Kenya. There is a clear need for innovations in clinical training to ensure that providers in teaching referral hospitals are prepared to prevent, assess, and manage PPH. It is possible that training interventions focused on learning by doing and teamwork may be beneficial. From 2009 to 2019, Kenya’s maternal mortality rate (MMR) decreased slightly from 309 deaths per 100,000 live births to 280 deaths per 100,000 live births [23]. This decline may be partly attributed to a series of policy changes that made maternity care services more accessible in Kenya during this period. Starting in 2013, the Kenyan government introduced free maternity services in all public facilities, leading to an increase in the rate of deliveries in public health facilities and in the use of postnatal care in public health facilities [24]. Additionally in 2013, health services were devolved from the national government to the county government, which may have contributed to changes such as increased construction of health facilities (particularly levels 2 and 3), increases in the number of specialists, and increased accessibility of skilled delivery services. In 2017, the Kenyan Ministry of Health then launched the “Linda Mama” programme, which was designed to further increase access to delivery services. While several challenges have been reported with these programs – including a lack of support for the costs of referrals, challenges for patients trying to access the services to which they were entitled [25], and persistent socioeconomic disparities in access to care [26]– these changes have been associated with improvements in the continuity of care [26]. However, alongside these improvements, several studies have documented important gaps in the quality of maternity care in Kenya [27–29]. We conducted a cross-sectional study among health care providers in three high-volume referral facilities in Nairobi and western Kenya. Data were collected as part of a larger study on postpartum hemorrhage from October 2018 to February 2019. The research was approved by the Harvard University Institutional Review Board (#IRB00047360) and the Ethics and Research Committee of the Jaramogi Oginga Odinga Teaching and Referral Hospital in Kisumu, Kenya. All providers gave their written informed consent for participation. Kenya’s healthcare system is divided into six levels: 1) Community Health Unit (mostly managed by Community Health Volunteers and Community Health Workers), 2–3) Primary health care facilities (Dispensaries and Health Centres), 4) primary referral facilities/hospitals, 5) secondary referral facilities/hospitals and 6) tertiary referral facilities. The three study facilities serve as level 5 facilities and are both regional referral and training hospitals. These hospitals were purposively selected because they manage high volumes of deliveries (between 17 and 50 per day in 2018) and therefore see large numbers of PPH cases. In these hospitals, medical officers (providers who hold a medical degree) may provide supervision but it is typically nurse-midwives, and nursing students who provide care throughout labor and delivery. Qualified providers were sampled based on their availability and their involvement in another component of the overall PPH study from a roster of approximately 300 qualified health care workers involved in maternity care in the study facilities (a convenience sample). Students were sampled based on their availability when they were present in the facility (no roster was provided for this group). The sample included consultants, medical officers, clinical officers, nurse-midwives, and nurse-midwifery students across all three study hospitals. Consultants are fully trained medical doctor specialists such as obstetrician-gynecologists, surgeons or pediatricians with postgraduate training and medical officers are licensed medical doctors who have completed six years of undergraduate training. Clinical officers are non-physician clinicians who undergo three to four years of training (a diploma degree). Clinical officers receive less training than medical officers, have a more restricted scope of practice and are accredited and licensed. In this setting, however, qualified nurses are the primary caregivers. An open-ended knowledge questionnaire was adapted from the USAID Maternal and Child Health Integrated Program (MCHIP) interview & knowledge test [30] (previously administered in Kenya as part of a large evaluation of the quality of maternity care in 2010–11) [31], the World Bank Kyrgyz Republic results-based financing evaluation health worker knowledge test [32], and a knowledge test used in a recent study of the quality of maternity care in health facilities in Uganda [14]. We extracted knowledge questions from previous tools and developed additional questions in the same open-ended style, with the goal of being comprehensive in our coverage of knowledge domains that are relevant to PPH care. Contextually relevant adaptions were made in accordance with the Kenyan national guidelines for quality obstetrics and perinatal care [33], informed by consultations with clinicians in Kenya and the United States. The final questionnaire included questions on provider characteristics such as training and experience in maternity care, in addition to knowledge of maternal and newborn care. The knowledge component of the interview comprised 20 questions on protocols for delivery of care from admissions through discharge. Participants were asked to freely list the clinical actions that they would take in different scenarios. The questions were read aloud to the participant and their verbal responses were recorded on paper questionnaires. Enumerators were instructed not to prompt health providers on their responses. Interviews were conducted face-to-face in private areas in order to ensure the confidentiality of responses. Interviews lasted approximately an hour. Our analysis focused on knowledge of technical clinical protocols for maternal care. We excluded eight questions related to interpersonal care (for example: “What are the times or situations when a health worker should explain to the woman and/or her companion what is happening?”) and to neonatal care. To measure knowledge in different domains, we classified questions into three categories which were informed by both prior groupings (such as the WHO Standards for improving Quality of Maternal and Newborn Care) and evidence from the literature [5, 15, 34–36]. Additional file 1 documents components of the main studies used. These domains were: risk assessment, prevention, and management. These domains correspond with different phases of care (with risk assessment done before delivery; prevention done around the time of delivery; and management done when emergencies occur). The risk assessment domain included questions about what to check for in a patient’s admission history when admitted, and routine monitoring that should be carried out during labor. The prevention domain included questions on basic equipment that should be prepared before delivery, immediate maternal care after delivery, PPH prevention protocols, and appropriate counseling that should be given prior to discharge such as making patients aware of various danger signs (e.g., difficulty emptying the bladder). Lastly, the management domain included actions that are appropriate for women who present with PPH. All questions from the assessment and the correct responses are shown in Additional file 2. We first described the characteristics of the providers in the study sample, including provider cadres, education level, age, gender, work experience, and participation in in-service training on PPH or Basic Emergency Obstetric and Newborn Care (BEmONC). While we are unable to speak to the specific types of training received at each facility, PPH training typically includes training on the prevention and management of PPH. These tend to be general and are rarely comprehensive. BEmONC training covers the necessary skills for handling obstetric emergencies such as postpartum infection, pre-eclampsia/eclampsia, postpartum hemorrhage, essential newborn care and resuscitation. Second, we analyzed provider responses to each of the included survey questions. We scored the providers’ responses to each question by dividing the total number of correct actions that a provider mentioned by the total number of recommended actions based on clinical guidelines. The possible score for each question ranged from zero to one, with zero indicating that the provider listed none of the recommended actions and one indicating that they listed all of the recommended actions. In this analysis, we included all of the recommended maternity-related clinical actions that providers should have mentioned, even if the actions were not specifically related to PPH. Given the important role that students play in the setting of this study, we compared the average score of each question for qualified health workers (i.e., nurses, clinical officers, medical officers, and consultants) to students. We estimated 95% confidence intervals around these scores using a normal approximation. Third, we measured knowledge in each of the three domains of PPH care: PPH risk assessment, prevention, and management. Knowledge in each domain was defined as the sum of actions a provider mentioned for each domain divided by the total number of recommended actions in each domain. All clinical actions included in this analysis are bolded items in column three of the table in Additional file 2. There is some repetition in potential correct responses across questions about PPH management. For example, providers should have mentioned “administer a treatment uterotonic” in response to the questions about PPH from atonic uterus, PPH from retained placenta, and PPH due to lacerations. Details on how this is incorporated into scores can be found in Additional file 2. Finally, we used Ordinary Least Squares (OLS) linear regression to examine characteristics associated with knowledge of PPH protocols. We ran separate models for each knowledge domain (assessment, prevention, and management). Associated characteristics included provider gender, age, education, specialization, years of experience in maternity care,1 participation in relevant in-service training, and self-reported closeness of relationships with colleagues. The closeness of providers’ relationships with their colleagues was measured using a survey question that asked providers to circle the picture that best represented their relationship with other providers showing four pictures ranging from A-D; option A showed separate circles representing distant working relationships, while option D showed overlapping circles representing “close” working relationships. This question was adapted from the Adapted Inclusion of Others in Self Scale [37] by Ashraf et al. (2016) [38] and is shown in Additional file 3. It was included as a proxy measure to explore the possibility of knowledge spillover from close peers, since maternity care in this setting is generally conducted by teams. All regressions included facility and enumerator fixed effects and used robust standard errors. We defined statistical significance at the α = 0.05 level. All data were analysed using Stata, version 17. We tested several approaches to handling missing covariate data in our regression analysis. In our main models, we used multiple imputation in Stata to impute missing values of covariate data. We describe the details of our main approach and the alternate approaches we tested in Additional file 4. It is also possible that using our measurement method, providers appeared more knowledgeable if they listed more clinical actions, even if they listed unnecessary or harmful actions (since we do not take away points for these additional actions). To assess the sensitivity of our measurement approach to this issue, we evaluated the extent to which this changed knowledge scores. We identified providers who mentioned harmful actions such as asking a patient to walk shortly after PPH identification, conducting a laparotomy, or initiating breastfeeding whilst managing PPH. Harmful practices were informed by the Kenyan Guidelines [36]. We then tested whether the probability of mentioning a harmful action increased as a provider listed more actions. Lastly, we tested the robustness of our findings to logistic regression instead of linear regression.