This prospective quasi-experimental study was planned to assess the impact of CBA on management of normal and prolonged labor at Kilimanjaro Christian Medical Centre (KCMC), northern Tanzania. The hospital is a referral hospital with 3312 deliveries in 2015, of which 43% were by cesarean section (CS). Five consultants in gynecology and obstetrics and 21 residents constitute the departmental medical staff. The labor and antenatal wards are staffed by 29 midwives.

The partograph was initiated when the women appeared in the antenatal ward and cervix was at least four centimeters dilated, as this is considered as active labour. Normal progression was defined as ≥≥1 cm/h visualized on the partograph as the  Alertline. A delay during the delivery of more than 4 h, was defined as prolonged labour, and the Actionline on the partograph was crossed. The women were referred to the labor ward when the uterine cervix was dilated  7 cm . Midwives with at least one year of midwifery training attended deliveries. A medical intern and a resident were always on call at the labor ward. They were always   supervised by a consultant.

Included were women admitted for delivery at KCMC with a gestational age more than 28 weeks after informed consent. Excluded were women who delivered a baby with a birthweight below 1000 g.  All women and staff members received verbal and written information about the study and informed consent was signed. They   were  assured  that  their  identities  would  be  kept  confidential and participation was voluntary, and that nonparticipation would  not affect the care received.

The study was a full CBA cycle including five steps (Figure 1). The first step was establishing audit criteria (Figure 1) based on international and Tanzanian guidelines ( Mathai et al., 2003) and , Welfare (2015)by two local consultants and two midwives, to reach a consensus. The following criteria were agreed upon regarding partograph: . Prolonged active phase first stage of labor was defined as dilation less than 1 cm/h indicated by the alert line on the partograph. The action line was reached if the progress deviated more than four hours from the alert line. Artificial rupture of membranes should be performed at a crossed alert line .(Brown, et al., 2013).  Augmentation of contractions with oxytocin should be performed at a crossed action line ,(Sorensen,  et al., 2010)and a CS was recommended if these actions did not normalize progress (Appendix 1) .(Yisma,  et al., 2013). Fetal distress were defined as a fetal heart rate <110 or >180.

Demographic data were comparable between the pre- and post-intervention data except the frequency of registered cases of hypertension in pregnancy, which was higher in the post-intervention period (RR=1.55, CI : 1.03 - 2.33), and so was the frequency of  newborns  with a birthweight above 4000 g (RR=2.12, CI : 1.21 - 3.73) (Table 1).

The overall CS rate decreased from 46.1 to 38.4% (RR=0.83, CI: 0.72 - 0.97) (Table 2). The main indications for CS were previous CS, prolonged labor and suspected fetal asphyxia. CS due to prolonged labor met the audit criteria in 28.3% of deliveries pre-intervention and in 40.4% post-intervention (RR=1.43, CI: 0.82 - 2.48) (Table 2). CS due to suspected asphyxia that met the criteria increased from 25.0 to 28.6% (RR=1.00, CI: 0.52 - 1.93) (Table 3).

Before the intervention, artificial rupture of membranes when crossing the alert line increase from 21.6 to 53.5% (RR=2.48, CI: 1.23 - 5.01), while the use of oxytocin when crossing the alert line decreased from 76.2 to 47.3% (RR=0.63, CI: 0.45 - 0.85). Monitoring of fetal heart rate every 5 min in the second stage increased from 0.7 to 4.8% (RR= 6.93, p=0.012) (Table 2).

Attempted VE increased from 1.7 to 3.9% (RR=2.26, CI: 1.02 - 5.01). Regarding the training session and the Objective Structured Assessment of Technical Skills of the staff, mean scores for the partograph test increased from 15.7 to 19.5 (CI: 2.74 - 4.89) (maximum 26 points). In the skills test for VE the mean score increased from 8.39  to  12.06  (CI: 2.25  -   5.10)   (maximum   score   25 points). The mean score for confidence during VE increased from 3.7 to 4.4 (p=0.001, five= maximum score).

The use of CBA in the study demonstrated that CBA can be a usefull tool to map suboptimal care and improve outcome in low-income areas. Although the effectiveness of CBA has been reported elsewhere, it is still mainly conducted as part of research projects, in low-resource countries. (Maaloe et al., 2012a). Thus, more research is needed to explore strategies to disseminate the use of CBA, and assess its long-term effects and feasibility (Maaloe et al., 2012a). The presentation and discussion of baseline results, coupled with short training, significantly reduced the rate of CS, and CS were more often performed with an indication in accordance with the agreed audit criteria. The very high rate of CS prior to the initiation of the study can be problematic, especially as a high proportion of the performed surgeries could not be justified. The World Health Organization (WHO), states that a population-based CS rate above 10% is not associated with a decreased maternal and newborn mortality (Kabore et al., 2016). In low income settings the risk of adverse outcomes to surgery is high, as adequate staff is not available (Cavallaro et al., 2018). One explanation for he observation of a high frequency of CS may be the domino effect, as repeated CS is more likely if women had a previous CS (Vogel et al., 2015). Other documented reasons for the increasing CS rate are fear of labor pain, the impression that CS is safer than vaginal delivery, and concerns of perineal trauma (Betran et al., 2016). In this hospital based study, it was observed that many CS were performed without a sustained indication, which is consistent with other studies conducted in sub-Saharan Africa (Maaloe et al., 2012b; Kabore et al., 2016).

VE is considered a basic emergency obstetric care signal function (Ameh and Weeks, 2009) used sensibly. It can reduce CS rates and be lifesaving to newborn in second  stage   of   labor   (Ghidini   et   al.,   2017).   The proportion of vacuum deliveries in this study before intervention was similar to other observations in sub-Saharan Africa which may even be below 1% of all deliveries (Maaloe et al., 2012b, 2016). However, the training session improved the skills and use of VE significantly and contributed to the reduced CS rate. The use of oxytocin decreased following training, and artificial rupture of membranes was performed more often.

The frequent use of oxytocin without an indication has been reported by others, where 59% were given oxytocin before crossing the alert line (Maaloe et al., 2016). On the contrary, in a study from north-western Tanzania only 9% of the women with prolonged labor received oxytocin because of fear of complications (Maaloe et al., 2012b) that underlines the need to know the baseline management before planning interventions, which is met by the CBA intervention.

The WHO partograph has been questioned for its effectiveness in reducing maternal and neonatal morbi-dity and mortality (Maaloe et al., 2016), but it is still considered a central tool for monitoring progress of labor and decision-making. However, recently WHO has changed their recommendations regarding the 1 cm/h cervical dilatation due to an inaccurate way to identify women at risk of adverse birth outcomes and is therefore not recommended anymore. However, these changes were not published at the start of the study and therefore it could not be included in the inclusion criteria (Library, 2018).

Similar to these findings, other studies have investigated its use and reported substandard recording in the partograph (Maaloe et al., 2012b; Sorensen et al., 2012; Housseine et al., 2016) delayed management  and interventions  for example, CS done without a sustained indication (Maaloe et al., 2012a; Maaloe et al., 2016). In this study, the staff’s skills in filling the partograph, tested through a standardized test improved. However, a systematic review on the effect of training in emergency obstetric care in low-income countries suggested that knowledge and skill changes after a training session, do not necessarily result in changes in clinical performance and outcomes (van Lonkhuijzen et al., 2010).

A   strength   of   the   study   was   that   criteria   were established in cooporation with staff, making them more likely to be achievable in practice. The criteria were clear and unambiguous, minimizing the risk of bias from a subjective assessment. The data collection did not rely on obstetric specialists because the partograph was the basis for most of the data. Another strength was the low risk of selection bias since only two  of the eligible women were excluded in first data collection and one woman in the second data collection.

The limitations of the study may be information bias due to underreporting, mismanagement and not documenting care given. This may have influenced the data. Secondly, it is a risk that staff in research context wants to act up to standard during the data collection period, the so-called Hawthorne effect (Graham, 2009) and the improvements are not longlasting. Thirdly, skills and confidence tests were unblinded, and performed by the investigators, which may have resulted in more positive  assessment following the intervention, even though structured observation lists were used. After the intervention, we observed significantly more frequent birthweight above 4000 g and more women referred from other health facilities. This could have reduced the effect of the intervention as these parameters are correlated to higher rates of CS. Fouthly, whether the improvements may be sustained after the second data collection is unknown. But this may be crucial for such studies, and in the future, it is important to include assessment of sustainability.

The CBA including a short training session may play a positive role in improving labour management. In this study a decrease in  the rates of CS and oxytocin use was demonstrated.  Findings should however, be taken cautiously as other contextual factors and efforts are probably crucial for success and for long term sustainability.

The authors have not declared any conflict of interests.