Study design and settings

We conducted a cross-sectional study at MNH in 2012. MNH is the highest referral hospital in Tanzania, and is a teaching hospital for Muhimbili University of Health and Allied Sciences (MUHAS). The hospital is situated in Dar es Salaam, which, according to the Tanzania Demographic Health Survey (TDHS 2010), has a population of about 4.5 million and an annual population growth rate of 4.3%. As one of the four consultant hospitals in Tanzania, MNH serves as a referral centre for the city of Dar es Salaam and the neighbouring coastal region. The annual number of deliveries in the hospital in 2012 was between about 8000 and 9000, of which 70% were self-referrals and hence the majority were also low-risk deliveries. Furthermore, out of 15 to 40 daily deliveries, 30 to 60% delivered by CS [32]. The peak delivery period was March to April and the lowest delivery period was December to January. According to the birth registry, during 2012–2015, the MMR of women who delivered at MNH increased from 96 to 146/100,000 live births, and the fresh stillbirth ratio and early neonatal deaths decreased from 32/1000 to 26/1000 and from 27/1000 to 16/1000, respectively.

The obstetric wards were attended by 35 obstetricians, 25 obstetrics and gynecology residents, 2 registrars and 25 nurse midwives. Daily routines included a daily grand ward round for 24-h call reports, and major ward rounds in the labour room, obstetric ICU and admitting antenatal/postnatal ward. The outpatients’ clinic and elective obstetrics and gynecological surgeries were included among daily hospital routines. Doctors on call work for 24 h in a team comprised of one intern doctor, two residents (one in the labour ward and another in the obstetric theater), and one specialist and one consultant obstetrician who were mostly called for a second opinion or in case of emergency. The length of working hours for nurses was 8 h on the morning, 6 h on the evening and 11 h on the night shift. The nurse-parturient ratio averaged at 1:3–5 on the morning and evening shifts, but 1:6–8 on the night shifts. After a normal vaginal delivery, the mothers and babies were observed in hospital for 6–10 h. During this time, babies are vaccinated with BCG and polio vaccinations before being discharged. Babies delivered by CS or those who were sick were admitted to the neonatal ward, which also admits sick babies from other lower referral hospitals.

Study population and sampling

All parturients who delivered at MNH during the study period were identified from the midwifery registry and assessed for eligibility for recruitment. The exclusion criteria included patients that had a high risk of, or preexisting conditions that could lead to, poor maternal and/or perinatal outcomes: history of peripartum complication such as abruption placenta, placenta previa and eclampsia; and neonatal complications, including congenital anomalies and intrauterine fetal deaths confirmed antenatally by ultrasound or macerated stillbirths. All parturients who delivered <28 weeks of gestation, those who delivered twins, and those delivered by elective CS were also excluded.

Data collection

Data were collected for 5 months (June to December) from the partogram and case notes using a checklist and questionnaire. The variables of interest included: socio-demographic characteristics, such as age, marital status and level of education; obstetric characteristics, including parity, gestation age, source of admission, number of gestations, nature of labour and mode of delivery; and perinatal outcomes, including 5th minute Apgar scores, birth trauma, intrapartum stillbirth and early neonatal death. The data were collected by one of the researchers from the midwifery registry, however, in case of incomplete or ambiguous information from the registry, mothers were interviewed for clarification. Neonates who were admitted to the neonatal unit were followed up for seven days to document the early neonatal outcome. Alternatively, neonates who were not admitted after delivery were considered to have good outcomes, hence they were not followed up. Verification of the information recorded by questionnaire and checklist was completed daily, and hence, no case was excluded from the analysis.

Definition of terms

Adverse perinatal outcomes were defined as; Apgar score < 7 at 5 min, fresh stillbirth, early neonatal death, and physical birth trauma (including skull fracture, clavicle fracture, facial nerve palsy and brachial plexus injuries). The perinatal period commenced from 28 completed weeks of gestation or when the neonatal birth weight was above 500 g.

Early neonatal death was defined as death of a newborn within 7 days post-delivery. A stillbirth was defined as the delivery of a dead newborn within the perinatal period, and fresh stillbirth was the birth of a dead fetus without postmortem changes based on physical appearance, including blistered or peeled-off skin. A perinatal death was a neonatal death during the perinatal period and hence included stillbirths and early neonatal deaths. Primipara referred to a woman experiencing her first delivery, while Multipara was defined as a woman having her second or more delivery. A work shift meant a period of working hours within a 24-h cycle that was divided into a morning shift from 7 am to 2 pm (8 h), an evening shift from 2 pm to 7 pm (6 h), and a night shift from 7 pm to 7 am (12 h).

Data analysis

Data entry and cleaning was performed using Epi Info™ version 6, and then data were transferred to SPSS version 21.0 (SPSS, Armonk, NY: IBM Corporation) for statistical analyses. Data cleaning included removing typographic errors and duplicated information and amending incomplete or incongruent information using case notes and ward report logs and by interviewing the mother. Statistically significant differences in associations between maternal characteristics, mode of delivery and perinatal outcome to time of birth were assessed using chi-squared χ² test for categorical variables, while Student’s t-test was used to assess continuous variables at a significance level of p = 0.05. Multiple logistic regression analysis was used to measure independent associations between selected predictors of adverse outcomes in relation to night shifts.

Main findings

The night shift was significantly associated with adverse perinatal outcomes, including low Apgar score, early neonatal death, and fresh stillbirth when compared with the morning and evening shifts. Our findings were consistent with some previous findings [10, 12, 13]; although refuted by others [17, 18] from settings with access to standardized care. The Dar es Salaam public hospitals birth registries, including that of MNH, confirmed that most of the daily deliveries occurred during the night shift compared to the morning and evening shifts. [32–35]. The most common groups of patients who were admitted to the delivery room during the night shift included patients from antenatal wards, referrals from other hospitals, and a significant number were abnormal breech deliveries. These groups of patients presumably possessed an independent risk of fresh stillbirth that was associated with maternal and fetal characteristics or obstetric risk factors that either led to their admission from the antenatal care or caused by delays in care as they travelled through the referral system. Therefore, the demonstrated risk of poor perinatal outcomes during off-hours in this study could have been associated with timing of delivery rather than the quality of care provided during off-hours. Alternatively, the exclusion of parturients with a known history of pregnancy complications to eliminate the confounding effect of antenatal complication might have reduced the resulting proportions of adverse perinatal outcome, either among antenatal admissions or referred patients.

In our study, teenagers were not associated with adverse perinatal outcomes and the majority delivered in the morning shift. However, teenage pregnancy has been found to have higher perinatal risks, including anaemia [36], placental problems [37], preterm delivery [36–38] and low birth weight [15]. Nevertheless, although not part of this study, teenagers require special attention due to the above-mentioned risks and others, including adherence to antenatal care and late booking [39], substance abuse [40], and possibly financial instability and dependence compared to adult and/or married pregnant women [41, 42]. The characteristics of our study participants included a majority of multiparas who delivered at term, and from whom one-third had secondary or college education, hence a protective effect on the exposure to obstetric risks.

Consistent with previous studies [42, 43], high CS rate was associated with very low use of LCVE, contrary to the WHO recommendation of appropriate use of emergency obstetric and newborn care (EmONC) services. Further, despite evidence of CS being associated with maternal near miss and death [44], CS was not independently associated with an increased risk of perinatal outcome, and was evenly distributed across the daily shifts. The LCVE rate was less than 2%, consistent with previous studies [42, 43]; and, therefore, in a way confirmed the tendency of obstetricians to avoid difficult obstetric procedures [44]. On the other hand, the low LCVE rate implied an opportunity to improve EmONC.

Interpretation of the results

Increased prevalence of perinatal morbidity and mortality during off-hours could generally be associated with pre-existing problems in low-resource settings, including delays in reaching and receiving care among referral cases [45, 46]. A high proportion of referred patients in labour during the night hours might have overwhelmed the available resources allocated to the perhaps underestimated number of anticipated night deliveries. This underestimation could either be due to the failure of hospital management to appreciate the daily trend of deliveries or in an attempt to evenly distribute the limited resources; which, in this case, was counterproductive.

The majority of antenatal ward admissions who gave birth during the night shift were presumably pregnant women with poor medical and obstetric histories, as an indication for admission, and hence, they experienced an added recurrent risk of poor perinatal outcome compared to other admissions. On the other hand, some of the admissions might have included low-risk self-referrals, as MNH also received some private patients, although the majority of these came from home. As previously demonstrated [47–50], referrals from other hospitals experienced first-, second- or even third-levels of delay in receiving care at the referring point. The transfer process through the primary and secondary referral facilities could be commonly occurring in the morning and afternoon before the pregnant women eventually reach MNH during the night shift. One can reasonably argue that an increased prevalence of poor perinatal outcome at the highest referral point should be anticipated during the night shift rather than the morning or evening shifts.

The review of the birth registry [32] demonstrated that not all reasons for referral to MNH were due to obstetric complications. The majority of referral indications were non-obstetric, including a lack of sutures, a non-functional theater bed or anaesthetic machine, consumables being out of stock, among others [51, 52]. This highlights that an overwhelmingly increasing workload of deliveries is placed at the highest referral point, without a corresponding increase of resources. The relatively compromised quality of care in terms of human and physical resources during the night compared to the daytime could place increased demand of resources and intensity of work, and hence elevate the level of poor outcomes during off-hours. For example, on-call doctors commence work at 8 am and receive 15 to 40 deliveries per day, among which 30% to 60% are CS, which can reasonably cause severe exhaustion to the care providers in the night shift; and therefore, reduce the quality of the process and outcome of care [53, 25].

We found high rates of CS, which agreed with a previous report [43], which addresses an increasing rate of CS in low-risk groups, and exposed questionable CS indications based on the absence of concomitantly improved neonatal outcomes. Despite a lack of association between CS and poor perinatal outcome in our study, assigning unnecessary CS interventions in resource-limited settings, such as those in Tanzania, indirectly deprives others who are in need of the intervention; hence adding to the risk of poor perinatal outcome. Importantly, a maternal near-miss study on CS at Dar es Salaam public hospitals found CS to be associated with life-threatening events [54], and other studies found delays in the timeline for intervention after taking the decision to perform CS [51, 52], the result of which is that the negative effect might be higher during the night shift when there were more deliveries than during morning or evening shifts, which have less deliveries [32].

Supported by another author [12], our study setting has unlimited access to senior doctors and midwives during the morning shifts compared to off-hours shifts. The benefits of the presence of senior staff during the morning shifts include better workforce and clinical and organizational decision making compared to other shifts, when seniors may be off work. However, the benefits of professional seniority might not necessarily mean better skills in obstetrics and midwifery. Nevertheless, the positive impact of the presence of senior professionals depends on the adequacy of resources and the participation of senior doctors in the daily routine. During this study, senior and experienced nurse midwives were mostly administrators and hence performed less clinical work. Similarly, senior doctors were usually available during the grand rounds, past the 24-h admission report and occasionally during the service morning ward rounds. Senior doctors who were on call attended emergency cases, when called to do so during their 24-h on-call duty. Similar to previous findings [25], the absence of senior doctors during off-hours compromised decision making, hence causing delays or precipitating suboptimal interventions that could have led to poor delivery outcomes. The strengthening of leadership skills during working hours and the off-hours shift, as well as the teamwork skills during the on-call emergency shift, has been found to improve practice [51, 52]. The presence of senior staff and the proportion of working staff per shift positively influences perinatal outcome with some variation from hospital to hospital [53, 25].

High prevalence of perinatal outcome (i.e., 89/1000 deliveries for Neonatal distress, 42/1000 for ENND, and 64/1000 for fresh stillbirth), in a relatively low-risk group of parturients in our study, highlights the overall deficiencies in EmONC [55]. However, the observed independent association of adverse perinatal outcome such as fresh stillbirths with referrals from other hospitals, antenatal admissions and ABD during the night shift stressed a need for special considerations for 24/7 standardization of EmONC within the existing health care system [17].

Study limitations

Our study had the advantage of having a large sample size, hence it yielded high power. However, as a cross-sectional design we failed to make causal inferences in the highlighted predictors of poor perinatal outcome. A possible bias included an overlapping time of case management from one shift to another, which might have over-expressed adverse perinatal outcome in a particular shift. Caution should be taken in the generalization of hospital-based findings to the general population due to discrepancies in the proportions of the number and mode of deliveries, and in the availability of equipment, supplies and skilled human resources. Importantly, because of the low prevalence of some variables, such as LCVE and ABD, a bias in the comparability of events could be present.