Setting

The Zanzibar archipelago, a semiautonomous part of Tanzania, struggles with poverty and a resource constraint health system. Half of the 1.3 million Zanzibarians live below the poverty line [15]. In 2011, the maternal mortality ratio was reported at 287 deaths per 100 000 live births, of which the majority occurs during or shortly after childbirth [16]. Though little is known about perinatal mortality, estimates from 2010 suggest a rate of 50 perinatal deaths per 1000 total births, as opposed to 36 per 1000 in mainland Tanzania [17].

The governmental Mnazi Mmoja Hospital is the only tertiary care facility on the archipelago, and at the time of data collection, the only hospital on the biggest island providing comprehensive obstetric and neonatal care around the clock. In 2014, 13 291 women delivered in the hospital, corresponding with an average of 36 deliveries daily. Of these, 16 % were caesarean sections (CSs), and 44 maternal deaths were counted. According to the official registers, 41 babies were stillborn in 2014 per 1000 total births, and 17 neonatal deaths occurred up to discharge per 1000 live births.

The hospital plays a leading role in clinical training of future Tanzanian health care providers. Yearly, approximately 60 intern doctors do their initial clinical rotations at the department, and more than 200 Tanzanian nursing and medical students are trained.

Intrapartum care is located in two rooms only. In the labour room’s 18 beds, women are assisted during the first stage of labour. Postpartum women needing extra surveillance stay in this room as well. When women reach the second stage of labour, they walk 15 m to the delivery room with only three beds. There is one, occasionally two, theatres available for obstetric surgery. For more than a decade, it has been the aim to apply the World Health Organization’s (WHOs) composite partograph for all women in labour, which is a graphic monitoring sheet including foetal heart rate (FHR), labour progress, and maternal vital signs during latent and active phase of labour [18]. A treatment and observation sheet for severe hypertensive disorders has been available since June 2014 and includes evidence-based guidelines on anticonvulsant and antihypertensive treatment [19].

After uncomplicated delivery, the woman and baby are usually observed in the labour room for two hours before referral to the postnatal ward for another four hours where no routine observations are done. Babies with birth asphyxia, with birthweights <1500 gram (g), or delivered by CS are admitted to the Neonatal Intensive Care Unit, which is adjacent to the delivery room.

At day time, there are on average three nurse-midwives and two registrar doctors on duty in the labour and delivery rooms as well as a fluctuating number of intern doctors. During evenings and nights, there are two or three nurse-midwives on duty for all obstetric patients and one registrar doctor assisted by two interns are in charge of the entire department, which also includes an average of 30 gynaecological patients. A second doctor is on call from home. At the department, there are two specialist obstetricians, of which one is member of the study team (TM).

Study population

All stillbirths defined as late foetal deaths ≥1000 g [20] were identified between 1st October 2014 and 31st January 2015 in the admission, delivery, and theatre registers, and their case files were searched for. In addition, to give the best possible estimate of the actual hospital-based stillbirth rate, non-registered stillbirths were sought by systematically going through all case files from the four months. Due to a breakdown in the hospital’s storage system of case files prior to October 2015, four months was the maximum feasible period for the study. The control group was identified in the same registers and from the same time period. The first control included in October 2014 was selected by throwing dices and afterwards every tenth delivery was identified, resulting in a case control ratio of approximately 1:4. These tasks were conducted by three research assistants (NH, RSK, and AGM), who had weekly meetings with TM or NM to assure quality of files retrieval.

Afterwards, all case files were checked individually by NM to determine whether the inclusion criteria were met for in-depth criterion-based audit (Fig. 1).

All singleton stillbirths were included for the audit and divided into three groups: birthweight 1000–1999 g (very preterm stillbirths) [21]; birthweight ≥2000 g without positive FHR on admission (pre-hospital stillbirths), and birthweight ≥2000 g with positive FHR on admission (intra-hospital stillbirths). Hence, pre-hospital stillbirths included not only cases in which the stillbirth diagnosis was made on admission, but also cases where there was no documentation of FHR (neither present nor absent) during the admission. While deliveries from multiple gestations were included in the overall estimation of the stillbirth rate, they were not included in the further audit process as their case files were often too ambiguous; e.g. frequently only one FHR was registered throughout labour. The 2000 g cut-off was decided as this reflects gestational age of 32–34 weeks, where lung maturity no longer plays a major role in survival, and the newborn is less dependent on dexamethasone treatment prior to delivery as well as advanced intensive neonatal care [22]. Data on the very preterm stillbirths are only presented in Additional file 1.

Inclusion criteria for controls were singletons with birthweight ≥2000 g and Apgar score ≥7; hence a group with immediate good neonatal outcome. It varied whether one, five or 10 min (min.) Apgar scores were recorded. It was therefore decided to use the latest. The case-control study was unblinded.

Criteria of realistic best quality of care

Criteria reflecting locally best possible labour management with the available resources were formulated and agreed upon. This developmental process was conducted by a participatory approach including both local skilled birth attendants and seven external specialists in midwifery and obstetrics. It was decided that the standards should include routine labour care as well as management of frequent complications of labour, related to FHR and foetal distress, labour progression by dilatation and descent, and maternal vital signs with specific focus on severe hypertensive disorders and fever. Selected criteria were adapted from the Active management of labour package, modified by WHO, and supplemented with other evidence-based guidelines [23–31]. Frequency of routine assessments was reduced to reflect local reality. For example, FHR assessments every 30 min. for all women in active labour were kept as optimal practice, but assessments within intervals of <90 min. were applied as an acceptable audit criterium.

In addition, information was collected on background and admission characteristics as well as outcome parameters. Finally, if information was available, time from last FHR assessment till delivery or diagnosis of intrauterine foetal death was calculated, as well as the admission to delivery interval.

Data extraction and analysis

Data was extracted into a structured entry form based on the pre-selected audit criteria, using EPI INFO 7 software (Centres for Disease Control and Prevention, Atlanta, GA, USA). Differences were analysed by logistic regression in SAS Enterprise Guide 6.1 (SAS Institute, Inc., Cary, NC). P-values <0.05 were considered statistically significant.

Background characteristics

Of multiparous women experiencing stillbirth, 30/83 (36 %) had also previously experienced the loss of one or more children, compared to 30/134 (22 %) controls (OR 1.96, 95 % CI 1.07–3.59). A history of previous CS occurred in 19/83 (23 %) compared to 18/134 (13 %) controls (OR 1.91, 95 % CI 0.94–3.90; Table 1).

There were three maternal deaths in the study population. They were all associated with stillbirth and suffered from severe delays in intrahospital surveillance and management (Additional file 2).

Admission and partograph use

Of women reaching active labour and admitted before second stage, significantly more in the pre-hospital stillbirth group did not have a partograph filled in, when compared to both intra-hospital stillbirths (OR 9.78, 95 % CI 2.56–37.42) and controls (OR 3.39, 95 % CI 1.52–7.56; Fig. 2). In all groups, 237/276 (86 %) women with a partograph applied had the first cervical dilatation appropriately plotted on the alert line (Table 3).

Foetal heart rate (FHR)

Labour progress

In 27/69 (39 %) women experiencing intra-hospital stillbirth, ≥5 h elapsed between any two vaginal examinations during active labour, compared to 40/207 (19 %) controls (OR 2.68, 95 % CI 1.48–4.86; Fig. 2). This resulted in delays in diagnosing poor labour progress, which was a common complication among intra-hospital stillbirths when compared to controls (Table 5). After crossing the alert line, in 18/33 (55 %) and 9/51 (18 %), respectively, ≥3 h elapsed before next vaginal examination (OR 5.60, 95 % CI 2.07–15.13). After crossing the action line, in 2/16 (13 %) intra-hospital stillbirths and 9/21 (43 %) controls, membranes were still intact, and in an additional 3/16 (19 %) and 5/21 (24 %), there was no information regarding membranes. Moreover, severe delays in surveillance were found after crossing the action line.

Oxytocin for labour augmentation was administered in 26/72 (36 %) of intra-hospital stillbirths, compared to 58/249 (23 %) controls (OR 1.86, 95 % CI 1.06–3.27). However, in 8/26 (31 %) of those, there was no indication for augmentation, and in an additional 7/26 (27 %) the infusion was started between the alert and action line with the membranes still intact. Likewise, 34/58 (59 %) controls had the infusion started before crossing the alert line (Fig. 3). In none of the cases, information on oxytocin titration and maintenance dose was documented, and FHR and contractions were never assessed half hourly after starting the infusion. Preceeding 13/35 (37 %) of CSs resulting in stillbirth, oxytocin was started for labour augmentation. This was also the case in four of six uterine ruptures, of which three had a history of one previous CS.

Maternal vital signs

Prevalence of severe hypertensive disorders was significantly higher among women experiencing stillbirth (27/139, 19 %) than in controls (10/249, 4 %; OR 5.76, 95 % CI 2.70–12.31). An additional 33/130 (25 %) and 81/235 (35 %) of women reaching active phase of labour, had no recordings of blood pressure during active labour (OR 0.65, 95 % CI 0.40–1.04; Fig. 2). Overall, 13/27 (48 %) of all stillbirth cases with severe hypertension had significant proteinuria (≥2+ on urine dipstick). However, urine analysis was not recorded in an additional 6/27 (22 %) cases, and information about clinical symptoms were too sparse to analyse for signs of organ failure. In 18 (67 %) of all 27 stillbirth cases with severe hypertension, there were no recordings of relevant antihypertensive treatment. Of the severe pre-eclampsia cases, 4/13 (31 %) had no documentation of having received magnesium sulphate.

In 43/139 (31 %) stillbirths and 104/249 (42 %) controls, there was no maternal temperature recording from admission till delivery (OR 0.63, 95 % CI 0.40–0.97; Fig. 2). Intrapartum fever or infection were rare diagnoses with five stillbirths related to infection and none among controls.

Causes of stillbirths

As suggested in other studies, the high number of intra-hospital stillbirths appeared primarily to be a sensitive indicator of substandard quality of care [3, 32, 33]. For instance, primigravid women suffered an increased risk of intra-hospital stillbirths, which may be associated with their admission earlier in labour, and longer and more complicated labour duration [34]. In addition, increased vulnerability and need for quality care among all women experiencing intra-hospital stillbirth is emphasized by the association with severe hypertension and prior loss of child.

While staff often referred to late referral of women as a major cause to adverse labour outcome, 85 % of women experiencing intra-hospital stillbirth directly sought care at the study site, and 58 % were admitted before active labour.

As found in other East African studies [8, 11, 35], not only was correct management often delayed, unnecessary and even harmful management was sometimes initiated. The presence of under- as well as over-treatment became particularly apparent when reviewing cases with poor progress of labour, where CSs often were found unnecessary. Yet, as only 4 % of foetuses had a documented positive FHR when deciding to perform CS, it is doubtful how big an impact e.g. possible delays in the decision-to-delivery interval for CS had on causing stillbirths; the unnecessary CSs may rather have caused avoidable maternal risks. In contrast, the under-use of vacuum extraction may be seen as an indicator of poor FHR monitoring, leading to undisclosed foetal distress in the second stage of labour.

Likewise, there appeared to be a dangerous over-use of oxytocin for labour augmentation. This is similar to studies from Bangladesh and Pakistan, where misuse of oxytocin was associated with stillbirth and birth asphyxia [36, 37]. Notably, the Pakistani study draws attention to the danger of insufficiently trained healthcare workers administrating this highly potent drug [36].

Similar to intrapartum care, antenatal visits did not appear consistent with effective antenatal surveillance and treatment, and as in other studies this appeared to be a central determinant of both pre- and intra-hospital stillbirths [33]. For instance, while nearly all women attended antenatal care at least once, a severe hypertension prevalence of 19 % among stillbirths suggests missed opportunities [38]. Furthermore, the fact that less than half of the study population had attended four antenatal care visits indicate a lost chance for continuity in care [39].

Maternal risks

Substandard quality of risk assessment on admission as well as poor intrapartum surveillance and decision-making were associated with profound maternal risks and appeared to be major determinants of the death of three women (Additional file 2). Women with foetal death on admission were the most neglected. While they were in particular high intrapartum risk due to the often underlying morbidity and further at increased postpartum risk of e.g. obstetric fistula, labour progress and vital signs were often undocumented throughout active labour [40, 41].

CS is generally not indicated when there is foetal death [42]. However, a high proportion of CSs were done on doubtful indications, and many were related to insufficient management of prolonged labour. Thus, 26 % CSs among stillbirths is unacceptably high; in particular as the vast majority had either foetal death diagnosed or did not have FHR recorded prior to surgery. Except for an even higher rate found at three hospitals in Mozambique [11], this is markedly higher than other studies from low- and middle-income countries [33]. A high proportion of CSs were done in the second stage of labour without an attempt of operative or destructive vaginal delivery. While short- and long-term maternal risks of suboptimally treated prolonged labour and unnecessary CSs are widely established [34, 43, 44], lack of transparency as to when to perform CS is found in other African studies too [11, 45, 46].

Six women suffered from uterine rupture. When considering the low level of surveillance in 14 % of stillbirth cases with one or more previous CSs, and the misuse of oxytocin, many more appeared at risk of rupture. Furthermore, while foetal bradycardia is an early sign of impending rupture [47], substandard FHR assessments made it less useful in timely detection.

Of the 19 % with severe hypertensive disorders experiencing stillbirth, more than half had severe pre-eclampsia or eclampsia. A Nigerian study found a similar prevalence and comparable insufficient antenatal and intrapartum surveillance and treatment of these dangerous conditions [48]. Recent data from well-resource settings emphasize suboptimally treated severe hypertension as an important contributor to maternal mortality [38]; in our study, 67 % of women with severe hypertension did not receive any antihypertensive treatment.

Clinical implications

At this East African referral hospital, facility births were frequently not accompanied by skilled intrapartum attendance. While widespread insufficiency in quality of routine and emergency labour care may partly be caused by massive structural constraints, suboptimal care often resulted in more risk associated, complicated, and resource draining interventions. Some of the revealed deficiencies may be addressed even without high costs in manpower and other resources, and the main risks and determinants are crucial in effectively designing low cost interventions (Table 6).

For many years, effectiveness of using the WHO partograph has been questioned [49, 50]. Yet, when analysing quality of intrapartum care at low resource facilities, partograph use for timely surveillance and decision-making appears central in ending preventable complications [11, 14, 18, 35]. In 86 % of all cases where the partograph was applied, first cervical dilatation in active phase of labour was plotted correctly on the alert line, and knowledge on accurate recording did not appear to be a major challenge. However, similar to other studies, WHO’s recommendations for frequency of recordings were not followed and also did not seem achievable with the resources available [35, 51, 52]. In the present study, even though the majority of intrapartum decision-making did not seem influenced by partograph use or evidence-based guidelines, it would be premature to conclude ineffectiveness of the WHO partograph. However, for the partograph to assist in surveillance and management, it must be coupled to a locally achievable and relevant labour management protocol. Although often not prioritised in evaluations of partograph use, this has previously proven effective [14, 18]. For instance, when considering the low resources at the study site, it seems unrealistic to assure close monitoring and titration of oxytocin augmentation if more than a few women are treated simultaneously [53].

This study identified 25 % underreporting of stillbirths in the official hospital registers, and even though a systematical surge was conducted through all piles of case files, a considerable number of files remained missing (Fig. 1). Initially, it was the intention also to include early neonatal deaths in the study. However, data collection revealed frequent default record keeping between the obstetric and neonatal units as well as substantial underreporting of very early neonatal deaths in all registers, which resulted in reluctance to include them. Furthermore, missing documentation in medical records – or “blanks” – was a frequent finding, which is likely to have affected patient care and labour outcomes. Incomplete health information systems are notoriously linked with poor health outcomes [3, 54]. It is warranted that the underlying factors for these “blanks” in medical recording are evaluated, and that quality of documentation and record keeping as well as use of the data are improved.

Strengths and limitations

The present pragmatic study was found suitable as a structured, simple, and low-cost method to identify central challenges in intrapartum care at this real-world setting with limited information available. Classification in pre- and intra-hospital stillbirths was a useful, more achievable, and simple alternative to ‘fresh’ versus ‘macerated’ stillbirths, which, as in a study from Ghana, was found unreliable [55]. Moreover, intra-hospital stillbirths may be seen as an even stronger indicator of intra-hospital quality of care than ‘fresh’ stillbirths. However, in 30 % of pre-hospital stillbirths there was no FHR documentation on or after admission, which may potentially hide an even higher proportion of intra-hospital stillbirths.

Selected audit criteria were unambiguously applicable. Yet, though intensive efforts were made for adapting international evidence-based guidelines to reach local reality, some criteria, such as <90 min. between FHR recordings, might be too optimistic as a sensitive audit standard for detecting quality improvements at this setting.

A central limitation to the study is that a criterion-based audit does not allow exploration of underlying determinants of substandard care, such as structural needs for supplies, space, and knowledge/skills among staff. Another limitation is the varying quality of data, which might bias results; there may be a tendency of staff to forget reporting given care or to under-report mismanagement. Participant observations during the study period also identified the issues presented in the current paper, and qualitative analysis opened up to a complex tangle of both structural and process related underlying challenges influencing health providers’ ability to deliver acceptable quality of care.