As one of the first studies applying the WHO criteria in a low-resource setting [9–11], we have identified a high occurrence of MNM and death at a university hospital and a regional hospital in Tanzania. Major causes were eclampsia (42%) and postpartum haemorrhage (13%), but there were also nine MNM events and five deaths due to iatrogenic complications, including high spinal anaesthesia, uterine perforations, ureter injuries, magnesium intoxication, and blood transfusion reaction. CS complications accounted for a large proportion of MNM and death, and the risk per 1,000 operations was high.
Strengths and limitations
Since the patient uptake and resources at the two hospitals we investigated are comparable to other hospitals in the region, our results may be transferrable to similar settings. The WHO criteria allowed us to identify women with direct signs and symptoms of organ dysfunction, which more reliably reflects a life-threatening condition than using a diagnosis-based inclusion . The data collection was performed in a structured and consistent manner. Having the same doctor review all medical records prevented different interpretations of the criteria and causes. When assessing cases potentially associated with CS, we considered each one individually, taking into account other factors that might have contributed to the MNM event or death, such as pre-existing conditions and the indication of the operation. By doing so, we hoped to identify adverse outcomes attributed to the CS per se. The manuscript follows the Strengthening the Reporting of Observational Studies (STROBE) recommendations (Table A4, Additional file 2).
Our study has limitation that needs further discussion. The applicability of the WHO criteria in low-resource settings has recently been questioned in two studies from Malawi  and Tanzania . As described above, we could not apply all WHO criteria due to limited resources at the facilities. Since some women had severe complications, yet did not fulfill any clinical criteria, we might have underestimated the MNMR, especially at the regional hospital. At the time we started the study, there were no studies published that applied the new criteria. When we compare our results with recently published work, we can see that other researchers have made a different interpretation of the criterion “uncontrollable fits”, only including women with continuous seizures [11, 26]. This has to be considered when comparing our estimated MNMR with other studies. As we did not screen wards outside the obstetric and gynaecological departments, some pregnant women admitted to other wards potentially could have been missed. We believe, however, that this had little impact on the results, since most pregnant women in these settings were admitted to the obstetric and gynaecological wards, even if they experienced other medical problems such as malaria, HIV, cholecystitis, and postpartum psychosis. Due to practical reasons, we could not follow women after their discharge from hospital. Since some women may have died at home or at another institution within 42 days, we might have underestimated the number of maternal deaths. As autopsies were not performed, the underlying cause of death was based solely on information in the medical record, and must therefore be interpreted with caution.
Occurrence of MNM and death
The occurrence of MNM in our study was considerably higher than reports from middle-income countries [5–7], but agreed with other research from Africa: a study in a rural referral hospital in northern Tanzania found an MNMR of 23.6 per 1,000 live births , and one from a tertiary facility in urban Ghana estimated the MNMR to be 28.6 per 1,000 live births . The high MNMR of 92 per 1,000 live births at the university hospital in our study is probably due to more referred patients (87% compared to between 20.9% to 64.4% in the other studies), but may also be explained by different interpretations of the criteria, as described above. The higher morality index of 12.9%–18% in our and similar studies from Africa [10, 11], compared to 10.4%–11.1% in studies from middle-income countries [6, 7], demonstrates that a larger proportion of critically ill women in low-resource settings die from their complications.
Panorama of MNM and death
The panorama of MNM and death highlights several areas in need of attention. Although eclampsia and postpartum haemorrhage are known as major causes of MNM and death , and efforts are continuously made to reduce their incidence, they accounted for the largest proportion of severe illness in our study. In spite of a 100% antenatal care coverage in Dar es Salaam , hypertension or pre-eclampsia were detected at the antenatal clinic in only 16% of the women who experienced MNM events or died, indicating that the antenatal controls have been insufficient to detect these risk pregnancies. Many women suffered from severe complications such as hysterectomy due to postpartum haemorrhage, even though active management of third stage of labour is promoted at most facilities. The finding that 71% of women with uterine ruptures met their first MNM criterion after arrival undoubtedly raises questions about surveillance during labour. Also, iatrogenic complications involving intravenous infusions and injuries during surgery indicate that there is a need to promote patient safety at these settings.
Risks with CS
Although the risks with CS in middle- and low-resource settings have been described before [16, 17], there is to our knowledge no studies from low-income countries estimating the proportion of MNM and death directly attributed to CS complications or risk per 1,000 operations. Our finding that CS complications accounted for 13% of maternal deaths is coherent with our previous result of an increase in the CS rate accompanied by an increase in the MMR at the university hospital during the last decade . MNM and deaths attributed to CS complications are especially worrying in the light of rapidly rising CS rates in many low-income countries [12, 15, 17, 27]. Reports that CSs are performed on non-medical indications [27, 28] and among low-risk groups  raise concerns about unnecessary morbidity and mortality after CS. Our estimate of the risk of death due to CS complications (between 1.0 and 4.7 per 1,000 operations) compared to a study from the US (0.0087 per 1,000 operations)  illustrates the danger CS might constitute in low-resource settings .
Distribution of resources
The uneven distribution of human and material resources between different health care facilities in the Dar es Salaam region most likely contributed to the high burden of MNM and death found in our study. Although peripheral hospitals receive the majority of patients, their ability to provide adequate maternal care is restricted due to the extensive lack of drugs, sterile packs, postoperative beds, blood for transfusion, electricity, and trained staff. In our study, 87% of MNM events and deaths at university hospital were referred. The referral system in Dar es Salaam is, however, ineffective, and delays might aggravate an already severe medical condition. With a higher standard of care at the peripheral hospitals, many MNM events and deaths might have been prevented.
The resource shortage at peripheral hospitals might also explain the elevated risk of CS complications at the regional hospital compared to the university hospital. As described before, a large proportion of surgery and anaesthesiology at the peripheral hospitals is provided by assistant medical officers instead of medical doctors. Although previous studies have not detected an increased risk of maternal death or other complications if an assistant medical officer, rather than a medical doctor, performs the surgery [13, 23, 29], untrained anaesthesiology staff has been associated with an increased risk of maternal death during or after CS . Insufficient training of anaesthesiology staff might explain some of the MNM events and deaths in our study, such as the cases of high spinal anaesthesia. The occurrence of postpartum haemorrhage and infections after CS, raises concerns about postoperative surveillance and use of prophylactic antibiotics at the peripheral hospitals.
Based on our findings, we suggest measures that might reduce the incidence of MNM and death. As most women with eclampsia fulfilled MNM criteria on arrival, antenatal services need to more effectively detect women at risk for hypertensive disorders and urge them to seek health care early when signs and symptoms arise. Magnesium sulphate should be readily available at health centres and smaller hospitals, since a majority of eclampsia cases were referrals. Measures to decrease the number of severe complications related to postpartum haemorrhage, including emergency hysterectomy, should be undertaken, such as closer monitoring, better availability of utero-tonic drugs, and enforcement of the use of uterine artery ligation and B-lynch suture. Also, surveillance and active management during labour needs to be improved, especially at peripheral hospitals, in order to detect and prevent uterine ruptures. Auditing cases with iatrogenic complications might help to identify risk situations and strengthen patient safety.
We identified an urgent need to decrease the maternal risks associated with CS. As this is most effectively done by avoiding unnecessary CSs, auditing CS indications, introducing a mandatory second opinion for CS decisions, and promoting active management of labour can be appropriate measures . In many hospitals, high CS rates are accompanied by an underuse of instrumental deliveries [15, 28]. Therefore, increasing the use of instrumental deliveries might be another way to avoid unnecessary CSs. Also, safety during and after the procedure needs to be improved, especially at peripheral hospitals. This includes more training of staff in order to avoid anaesthetic and surgical complications, strict use of prophylactic antibiotics to decrease the number of postoperative infection, as well as earlier detection and more effective treatment of postpartum haemorrhage after CS.