State of newborn care in South Sudan’s displacement camps: a descriptive study of facility-based deliveries

A cross-sectional descriptive design was employed using health facility assessments, direct observation of midwife time-use, clinical observation of the labor and delivery period, and exit interviews with recently delivered mothers. The study population consisted of mother-newborn pairs who sought services and birth attendants who provided delivery services between April and June 2016 in five health facilities operated by International Medical Corps (IMC) in displacement camps located in South Sudan: Maban (Gendrassa and Kaya), Juba, and Malakal. Maban hosts refugee camps, whereas Juba and Malakal have internally displaced person (IDP) camps. One hospital and four PHCCs that conduct the majority of deliveries in the camps agreed to participate in the study. The hospital, located in Juba, operated 24/7 with skilled attendants, an in-patient department and comprehensive emergency obstetric care, while the PHCCs have midwives and traditional birth attendants (TBA) conducting deliveries.

Health facilities in the camps were selected based on: (1) provision of delivery care, (2) delivery of at least 20 births per month, and (3) management by IMC. The sample size for these current analyses of baseline data was based on power calculations conducted to ensure that the planned pre-post evaluation would be sufficiently powered to detect changes as a result of the intervention components. Male and female midwives, aged 18 and older, who were working a daytime shift on the date of data collection at one of the study facilities were eligible for time-use observation. A midwife was observed every ten minutes over an eight-hour shift for a four-day period at each facility to reach the necessary sample size. We estimated that midwives spend about 40 min of an eight-hour working day on postnatal care. To detect a 5% difference post-intervention in the proportion of time spent on postnatal care (e.g. 8% to 13%) with an alpha of 0.05, 0.8 power and 5% non-response rate, we required a total sample size of 620 observations at baseline and end line.

Women, aged 15 and older, arriving at a study facility for delivery during the 9-week enrollment period were eligible for recruitment for observation of the labour and delivery period and an exit interview. Similarly, the sample size calculated to see a change in baseline was 185 deliveries based on a two-sample comparison of the primary outcome—proportion of newborns that receive the components of essential newborn care (Table 1). This assumed a 50% baseline prevalence of the primary outcome, a conservative estimate due to a lack of data from similar settings, and a 15% detectable difference with an alpha of 0.05 and power of 0.8. The sample size accounted for a non-response rate of 8% based on a 3% stillbirth rate and 5% referral for complications. The sample was doubled to allow for analyses stratified by facility type: hospital and PHCC. Women delivering a stillbirth were subsequently excluded from analysis of observation data. Recently delivered mothers who participated in a clinical observation and did not have a stillbirth or neonatal death were asked to participate in an exit interview.

Locally-based non-clinicians were hired as research assistants and trained jointly to standardize data collection procedures across the health facilities. The three research teams consisted of three to five research assistants and one site supervisor. A ten-day training was conducted by study co-investigators to orient the teams to newborn clinical practices, research ethics and procedures, and the study instruments. Researchers practiced the interview aloud in both Arabic and Nuer languages to ensure the translation was standard between researchers. To improve reliability, a two-day pilot session was held to standardize observations between researchers, with one day in the classroom observing videos as a group and one day in a facility observing live births with researchers from the same site. Site supervisors received additional training to conduct the health facility assessment.

Following the assessment, we conducted the time-use observations. At each facility, for a four-day period, one midwife was randomly selected each day for time-use observation. Every ten minutes over an eight-hour period, the researcher recorded the main activity in which the midwife was engaged using a structured checklist; activities were categorized into eighteen general tasks for midwives such as postnatal care, intrapartum care, antenatal care (ANC), documentation, supervision, cleaning, and other tasks. Patient census data in the maternity unit was noted on every observation day. Informed consent was attained from each midwife.

For clinical observations, one researcher was assigned per shift to ensure continuous coverage during facility operating hours. Data collectors rotated every eight to twelve-hour shift, depending on the security guidance in each site. Each mother-newborn pair was observed for at least one and no more than 6 h before starting a new observation. The observation was documented using a structured checklist, based on the Maternal and Child Health Integrated Program (MCHIP) Quality of Care Surveys [15]. We observed the following clinical practices: (i) partograph use, (ii) preparation for delivery, (iii) immediate postnatal care, (iv) bag and mask resuscitation for non-breathing babies, and (v) kangaroo mother care (KMC) for low birth weight babies. Additional information on the outcome of the mother and newborn, gravidity, parity and birth weight was extracted from clinical records. Following observation and before the mother was discharged, researchers administered an exit interview using a structured questionnaire (in either Arabic or Nuer, according to the mother’s preference). Information was gathered on the mother’s demographic characteristics and her knowledge of six newborn danger signs without prompting response options. Verbal consent was obtained for both clinical observations and exit interviews from all participating mothers.

Data were entered on Android tablets using Magpi mobile data software (DataDyne Group LLC, Version 3.2.2). Data were monitored daily for quality checks and uploaded to an online server by site supervisors, reviewed for consistency, and exported to Stata (StataCorp LP, Version 13.1) for analysis. Descriptive analyses were performed to explore frequencies, means and interval estimation. Missing and “don’t know” responses were analyzed as missing and excluded from analysis. The activities measured in the time-use observations were classified into three categories: contact time for newborn care (by antenatal, delivery and postnatal care), non-contact productive time (i.e. documentation, meetings, training, supervision and cleaning) and non-productive time (i.e. waiting for patients, break time and socializing). Proportion of the total amount of observed time spent on each activity was estimated and a regression analysis was conducted to estimate associations with facility type. In the clinical observations, we conducted descriptive analyses of up to 65 practices for partograph use, delivery preparation, immediate postnatal care, neonatal resuscitation, and KMC. The primary outcome was defined by the proportion of newborns who received the essential components of immediate postnatal care: (i) thermal care, (ii) infection prevention, (iii) feeding support, and (iv) postnatal monitoring; individual sub-items contributing to these composite components are shown in Table 1. Partograph use was not analyzed among women who were admitted immediately to the labor room because they were assumed to be in the second stage of labour. For the exit interviews, the pre-defined response options for knowledge of newborn danger signs were entered as a binary (yes/no) variable. One point was assigned for each answer that was coded as ‘yes’, giving a maximum score of six. The overall knowledge score was categorized as ‘poor’ (zero to one danger signs), ‘moderate’ (two to three), and ‘adequate’ (four to six). Mother’s age, parity and years living in the camp were measured as a continuous variable and categorized as a binary variable for analysis. A chi-squared statistic was used to examine differences in newborn care practices by type of health worker and facility. Multiple logistic regression, in a stepwise backward manner with a 0.05 cutoff, was used to identify factors associated with newborn practices and maternal knowledge including maternal, newborn and health worker characteristics. In each step, variables that are least significant are removed until all variables in the regression model have a p-value of less than or equal to 0.05. Associations were examined by relative risks (RR), using a modified Poisson approach with a robust variance estimator, and 95% confidence intervals [16].

Among a total of 357 mothers recruited for observation across the hospital and PHCC strata, one refused to participate and 13 were excluded due to stillbirth or referral to another facility prior to delivery, leaving 343 mother-newborn observations for inclusion in the analysis. Less than 2% of the mothers who participated in an observation did not complete an exit interview before being discharged from the health facility.

Among the mother-newborn observations, 44.6% were from the hospital and 55.4% were from the four PHCCs (Table 2). Over two-thirds of observations had more than one health worker assisting with the delivery, although 25.4% were delivered by only TBAs. Maternal age ranged from 15 to 37 years, with a mean of 24.5 years (SD = 4.7). The proportion of mothers who received a visit from a community health worker (CHW) during pregnancy was 60% higher in the PHCCs than in the hospital (80.5% and 20.3%, respectively), while most mothers across all facilities attended at least one ANC visit (98.2%). Covariates with missing data resulted mainly from missing data in participants’ clinical records.

For the time-use observations, a total of 1116 observations were made of thirteen midwives at the five health facilities, with 16.9% of observations at the hospital and 83.1% at the PHCCs. Across the five facilities, 67.6% of the observations were among female midwives. The median age represented among the observations was 28 years (IQR: 26–29), with the majority receiving their midwifery degree in 2015 (63.8%). The observed midwives had an average daily caseload of 9.3 (SD = 8.2) consultations, 1.6 (SD = 1.7) deliveries, and 2.6 (SD = 4.1) postnatal care visits.

Overall, the majority of the five facilities had electricity (3 of the 5) and running water (5 of the 5) during the past month. The two PHCCs that had no electricity reported at least 15 to 30 days without electricity or a generator in the past month. Only two of the five facilities had skilled birth attendants (SBAs) working during all operating hours. The remaining three facilities had skilled attendants primarily in the daytime hours.

Readiness to deliver basic obstetric care was optimal at the hospital; however, none of the critical services (i.e. parenteral administration of oxytocin, assisted vaginal delivery, manual removal of retained products, or neonatal resuscitation) was found to be universally available across all PHCCs. All facilities were found to be lacking the recommended items to provide intrapartum and postnatal care for newborns (Table 3). The hospital, functioning at the referral level, had 20 of the 37 recommended medicines and equipment for newborn care service delivery and the PHCCs had on average 9 out of 25. Newborn-sized bag and mask, a thermometer and tetracycline eye ointment were the most commonly available newborn supplies across facilities. Although all facilities had a neonatal-sized bag and mask, none had a resuscitation table available in their delivery or postnatal area. Similarly, infant weighing scales were available but there were no KMC wraps for low birth weight babies. At the hospital level, the unavailable supplies were newborn clothing, wrap for KMC, neonatal-sized catheter, pediatric infusion set, pediatric pulse oximeter sensor, infant feeding tubes, oral phenobarbitone, amoxicillin (suspension powder), cefotaxime sodium and caffeine.

In a routine workday, the proportion of midwives’ time spent on patient care was 40.2% and the remaining time (59.8%) was spent in non-contact. Of the time spent on non-contact activities, slightly more than half was non-productive (51.9%); productive time (48.1%) was spent on cleaning, documentation, supervision and formal meetings. The main activity observed within patient care was ANC (53.2%) followed by delivery care (31.4%). Of the total observed time, 6.2% of midwives’ time was spent on postnatal care or about 30 min of an eight-hour work day. There was a statistically significant difference in the proportion of time spent on ANC (RR = 1.83, 95% CI: 1.23 to 2.70) and delivery care (RR = 0.53, 95% CI: 0.38 to 0.74) among the midwives observed in the PHCCs, relative to those observed in the hospital.

For the mother-newborn observations, 64.4% of mothers were admitted to the facility before second stage of labour, while others went directly to the labour room upon arrival. More than half of the mothers (57.0%) who arrived before the second stage of labour had a partograph used to monitor their fetal heart rate, with the majority (91.3%) having their fetal heart rate monitored every 30 min. The practice of partograph use was less common for mothers who delivered in a PHCC (25.7%) than a hospital (85.3%); difference: 59.6% (95% CI: 48.1% to 70.2%).

During the delivery period, all observed newborns were dried immediately with a towel and most received essential practices for newborn care (Table 4). Facilities indicated that the routine practice was to sterilize cord-cutting supplies, but we were unable to measure an actual observation of sterility. In the first hour after birth, a significantly higher proportion of mothers placed their newborn skin-to-skin in the hospital (88.9%) as compared to the PHCC (49.7%). Among those who provided skin-to-skin, the median time for skin-to-skin contact was also higher in the hospital (Mdn = 12 mins, IQR: 8–20) than the PHCC (Mdn = 1 min, IQR: 1–10). Nearly all newborns (95.9%) received tetracycline eye ointment for eye care in the first hour after birth, and a higher proportion received tetracycline by discharge. Similarly, a high proportion of mother-newborn pairs initiated breastfeeding, measured by observed latch, within the first hour in the hospital (98.7%) and PHCCs (97.3%).

Essential newborn care was measured as four different components, with each component consisting of the respective immediate postnatal practices (Table 5). Mother-newborn pairs in the PHCCs were significantly more likely to receive infection prevention (RR = 1.28, 95% CI: 1.11 to 1.47) and postnatal monitoring (RR = 3.17, 95% CI: 2.01 to 5.00) practices than mother-newborn pairs in the hospital, without adjusting for potential confounders. This was not found for thermal care, where in the PHCC mother-newborn pairs were 52% less likely to receive thermal care when compared to the hospital (95% CI: 42% to 60%). The likelihood of receiving feeding support was similarly lower for PHCC-born newborns (RR = 0.49, 95% CI: 0.40 to 0.58). Table 5 shows the results of the final multivariate logistic regression model, with the hospital as the reference value. All components of essential newborn care remained strongly associated with facility type even after controlling for potential confounders.

In the PHCCs, about 45.8% of mother-newborn pairs in the study were attended by only a TBA during delivery. There was a statistically significantly higher proportion that received postnatal monitoring (RR 1.59, 95% CI: 1.09 to 2.32) among mother-newborn pairs delivered by a SBA as compared to those delivered by a TBA. However, other essential newborn care components were not statistically different.

Twenty-six (7.6%) of the 343 newborns were considered to be low birth weight (<2500 g) by the health worker(s) using a weighing scale or estimation of weight with no weighing scale. Among those newborns who were considered low birth weight, none were placed immediately in the proper KMC position.¹

A total of 339 recently delivered mothers were interviewed before they were discharged from the health facility: 153 in the hospital and 186 in the PHCC (Table 6). Mothers’ knowledge of danger signs was poor, with about one-fifth of mothers correctly identifying four or more signs (20.4%). Although 44.8% of mothers reported fever as a danger sign for newborns, only 18.0% recalled baby feeling cold to the touch as a potential sign of illness. A low proportion of mothers had knowledge of other newborn danger signs such as reduced activity (27.7%), difficulty breathing (28.9%), and convulsions (11.2%). However, a relatively higher proportion of mothers (41.0%) described not feeding well as a danger sign for newborns. The median number of dangers signs that were recalled was one (IQR: 0–3), while 47.8% of mothers recalled none of the six components.

Mother’s education, having a home visit by a CHW, and number of years living in a camp were associated with mother’s ability to recall newborn danger signs and were shown to be significant (Table 7). Only 16.4% of mothers with no education had adequate knowledge of newborn danger signs; whereas, 32.9% of mothers with at least primary education had adequate knowledge (p = 0.003). Regarding number of years living in a camp, 29.8% of mothers who lived in the camp for one year or less had adequate recall of newborn danger signs compared to 17.1% of mothers who lived in the camp for more than one year (p = 0.034). Additionally, 28.3% of mothers who had a home visit by a CHW had adequate knowledge but only 11.5% of mothers who did not have a home visit were able to adequately recall four to six danger signs (p = 0.001). There was no significant association between the level of maternal knowledge of danger signs and mother’s age (p = 0.525) or mother’s employment (p = 0.364). Nearly equal proportions of recently delivered mothers who were primiparae and multiparae had adequate knowledge of the danger signs (23.1% and 20.2%, respectively).

This study provides a cross-sectional snapshot of the services available in a specific period, and may not be representative of services during an acute phase of the conflict. Additionally, the data are not representative of newborn health services provided by government facilities in South Sudan where bottlenecks that affect care are likely exacerbated. However, findings may represent services seen in other camp-based, NGO facilities that follow the Minimum Initial Service Package (MISP) for Reproductive Health in Crises, in or outside of South Sudan [35]. Limitations to our methods include the potential for the Hawthorne effect by using direct observation of health worker time-use and clinical practices; however, researchers were discrete about the measures being observed and became integrated in the workforce over a nine-week period. Clinical observation methods rely on researchers to use a standard approach to observing practices. Clean or sterile cord cutting was implemented using different procedures across the facilities, and researchers were unable to standardize their observation methods. Furthermore, our sample size was not estimated to measure newborn resuscitation or KMC, which limited our analysis of these practices. Additionally, the exit interviews excluded mothers with a stillbirth or newborn death who may have been less knowledgeable of danger signs.