Factors related to the practice of vacuum-assisted birth: findings from provider interviews in Kigoma, Tanzania

Background: Vacuum-assisted birth is not widely practiced in Tanzania but efforts to re-introduce the procedure suggest some success. Few studies have targeted childbirth attendants to learn how their perceptions of and training experiences with the procedure affect practice. This study explores a largely rural cohort of health providers to determine associations between recent practice of the procedure and training, individual and contextual factors. Methods: A cross-sectional knowledge, attitudes and practice survey of 297 providers was conducted in 2019 at 3 hospitals and 12 health centers that provided comprehensive emergency obstetric care. We used descriptive statistics and binary logistic regression to model the probability of having performed a vacuum extraction in the last 3 months. Results: Providers were roughly split between working in maternity units in hospitals and health centers. They included: medical doctors, assistant medical officers (14%); clinical officers (10%); nurse officers, assistant nurse officers, registered nurses (32%); and enrolled nurses (44%). Eighty percent reported either pre-service, in-service vacuum extraction training or both, but only 31% reported conducting a vacuum-assisted birth in the last 3 months. Based on 11 training and enabling factors, a positive association with recent practice was observed; the single most promising factor was hands-on solo practice during in-service training (66% of providers with this experience had conducted vacuum extraction in the last 3 months). The logistic regression model showed that providers exposed to 7–9 training modalities were 7.8 times more likely to have performed vacuum extraction than those exposed to fewer training opportunities (AOR = 7.78, 95% CI: 4.169–14.524). Providers who worked in administrative councils other than Kigoma Municipality were 2.7 times more likely to have conducted vacuum extraction than their colleagues in Kigoma Municipality (AOR = 2.67, 95% CI: 1.023–6.976). Similarly, providers posted in a health center compared to those in a hospital were twice as likely to have conducted a recent vacuum extraction (AOR = 2.11, 95% CI: 1.153–3.850), and finally, male providers were twice as likely as their female colleagues to have performed this procedure recently (AOR = 1.95, 95% CI: 1.072–3.55). Conclusions: Training and location of posting were associated with recent practice of vacuum extraction. Multiple training modalities appear to predict recent practice but hands-on experience during training may be the most critical component. We recommend a low-dose high frequency strategy to skills building with simulation and e-learning. A gender integrated approach to training may help ensure female trainees are exposed to critical training components. The cross-sectional survey design was set in the region of Kigoma, Tanzania. Kigoma has a population of 2.3 million, a population growth of 2.7% and is divided into eight administrative councils [15]. Most people in the region are subsistence farmers and reside in rural areas. In 2018, Kigoma’s institutional delivery rate was 85% (up from 49% in 2013) and the population-based cesarean delivery rate was 4.5% (up from 2.6% in 2013) [16]. In the last 3 months of 2018, only 36% of the 33 hospitals and health centers in Kigoma provided assisted vaginal delivery [17]. For over a decade, Thamini Uhai, a Tanzanian non-governmental organization, has worked closely with the government and other partners in Kigoma to develop and implement a model program to reduce maternal and neonatal mortality in remote areas. The model was developed in 2005 by the late Dr. Mbaruku (former Kigoma Regional Medical Officer), Thamini Uhai/Vital Strategies (formerly World Lung Foundation) and its partners. It was a response to the challenge of delivering good-quality EmONC to women living far from hospitals by decentralizing comprehensive EmONC services (including obstetric surgery) to the health center level. The program was implemented in 12 health centers in Kigoma. In addition, 3 existing hospitals in the region that already provided comprehensive EmONC were included and given technical support by the program to backstop health centers. Support was provided to the 15 health facilities in this study between 2006 and 2019; the most intensive period of implementation was between 2013 and 2019. To support the delivery of good-quality comprehensive EmONC in health centers, the program constructed/renovated and equipped operating theatres and maternity wards and installed/strengthened electricity and water supply systems. Mid-level health providers, mainly assistant medical officers and nurse-midwives, were trained, mentored and supervised to perform the nine comprehensive EmONC signal functions [18]. In particular, continuous efforts were made to sustain vacuum extraction by ensuring they had functional equipment, providing regular onsite mentoring, conducting periodic skills-based continuing medical education workshops, producing an e-Learning platform for independent learning that included a module on vacuum extraction [19], providing access to senior obstetricians 24/7 via a free phone call and conducting routine clinical audits of caesarean sections and vacuum extractions. This package of clinical support was designed by the program’s implementers to enable providers working in remote parts of the region to provide comprehensive EmONC on their own. With this support, most EmONC signal functions were sustained, such as obstetric surgery and neonatal resuscitation. Yet, performance of assisted vaginal delivery across supported facilities was inconsistent over time [17]. This study aimed to interview maternal health providers about their training experiences with vacuum extraction, their knowledge of and attitudes toward vacuum extraction, and how these related to their recent performance of vacuum delivery. Nearly half of institutional deliveries in the region are managed at the 15 selected facilities [16]. Moreover, the majority of skilled birth attendants expected to perform basic and comprehensive EmONC signal functions are located in these health facilities. Generally, the types and distribution of birth attendants posted at these facilities are similar, with nurses and medical attendants accounting for most of them, with a few assistant medical officers (AMOs) and/or medical doctors (MDs). By default, the hospitals have more staff compared to health centers. The survey was conducted through face-to-face interviews with health providers and managers using a structured questionnaire. The survey questionnaire (originally in English and translated to Swahili) was developed specifically for this study by the investigators, who are all knowledgeable and experienced with the subject (*Additional file 1). To be eligible for inclusion, the provider or manager had to be present during the study visit, report having worked at their current location for at least 3 months and grant informed consent. Providers and managers were interviewed in their work-stations during the last 2 weeks of January 2019. Due to the shortage of health providers in the region, all eligible providers in health centers were invited to participate. In hospitals, only providers working in maternity units were interviewed. Managers included facility supervisors, in-charges and members of the Regional, District or Town Councils’ health management team (R/CHMT). For R/CHMT managers, the study was limited to clinicians. Senior clinicians (AMOs/MDs) and registered nurses/midwives with experience in maternal and newborn health services administered the surveys. They received a 3-day training on the study protocol and survey instruments which included a 1-day pilot test of tools in a nearby health facility. For this study, teams of 3 deployed to the field; data collection often took 2–3 days in hospitals and 1–2 days in health centers. On average, interviews lasted between 30 and 45 min per respondent. Interviews were conducted in private confidential settings within the facility’s premises. Data were entered into Epi info [7] and exported to SPSS version 24 (SPSS, Chicago, IL, USA) for statistical analysis. The primary outcome of interest was recent performance of vacuum delivery, which was self-reported. “Recent” was defined by the immediate 3-month period prior to survey. Three months was chosen to maximize recall accuracy as well as concern that less frequent clinical experience could be problematic for skill set maintenance. The variables selected for exploration related to characteristics of the provider (sex, age, professional cadre, training experience, years since professional qualification, and knowledge and attitudes regarding vacuum extraction). Contextual variables included type of facility and administrative district. All were categorical variables. A composite knowledge score was created for each respondent based on his or her ability to report indications and contra-indications for vacuum extraction. The questions were posed as open-ended but responses were pre-coded and required a respondent to spontaneously list all the indications and contra-indications for vacuum extraction that they could. No prompting beyond “can you think of anything else?” was done. The 10 pre-coded indications were: prolonged 2nd stage of labor, maternal exhaustion, need to shorten 2nd stage of labor for medical reasons, severe anemia, heart diseases/failure, severe pre-eclampsia or eclampsia, suspected or imminent fetal distress in 2nd stage of labor, fetal bradycardia, fetal tachycardia, and thick meconium stained liquor. The 8 pre-coded contra-indications were: breech/face/brow presentation/transverse lie, un-engaged fetal head, gestational age less than 34 weeks, incomplete cervical dilation in nulliparae, cephalo-pelvic disproportion, incomplete cervical dilation in multiparae, moulding grade 3 and HIV-infected pregnant women. Each item mentioned had the value of 1; otherwise the item was given a value of zero. Responses ranged from zero to 18 and were dichotomized into the top quartile and the lower three quartiles. A summary index for training was calculated based on a respondent’s self-reported exposure to 11 different training opportunities. No weighting of the scores was done, thus, each reported opportunity accrued a value of 1; scores could range from 0 to 9 since a person could report only one pre-service option (with or without hands-on experience) and only one in-service option (with or without hands-on experience). This score was presented as a categorical variable: exposure to 6 or fewer training modalities and 7 or more. Since recent performance of vacuum extraction was a dichotomous categorical variable with a yes/no answer, binary logistic regression was the most appropriate method to control for covariates. All independent variables were entered as a block in the regression. All variables were categorical. No multicollinearity among explanatory variables was found. Significance for all statistical testing was set at 95% CI (p < 0.05). Pearson Chi-square tests were used to test for associations between groups for all descriptive analyses. The study received ethical clearance from the National Ethical Review Committee of the National Institute for Medical Research (NIMR) in Tanzania. Approval was also obtained from the Regional and District Medical Officers as well as the respective facility heads. A written informed consent was obtained for all survey participants.