Applicability of the WHO maternal near miss tool in sub-Saharan Africa: a systematic review

A total of 710 citations were identified through our initial search. After removal of duplicates and screening of titles and abstracts, 82 articles were retained for full text review, of which 67 were excluded. Main reasons for exclusion were that the studies did not contain data on MNM (18), did not report any relevant data (16), did not apply the WHO MNM criteria (12) or were duplicate publications from the same database (8) (Fig. 1).

Methodological quality of the remaining 15 studies is shown in Table 2. Matching in design and appropriate handling of missing data were not applied in all studies. Overall, four studies were rated as strong [29‐32], three as moderately-strong [33‐35] and eight as moderately-weak [11, 36‐42].

All studies were cross sectional in design, although sometimes reported as being prospective or retrospective cohort studies. The median MNM ratio was 24.2 per 1000 live births and ranged from 4.4 in a population-based study from South Africa to 198 per 1000 live births in a private rural referral hospital in Nigeria. For each maternal death, 6.2 MNM cases were reported ranging from 1.3 in Zanzibar to 15.4 in Rwanda (IQR 2.6–6.8). Mortality index ranged from 6% in Rwanda to 43% in Zanzibar with a median of 14 (IQR 12.9–27.7). The maternal mortality ratio ranged from 71 in South Africa to 2875 per 100,000 live births in Rwanda (Table 3).

Eight studies discussed challenges related to using the WHO MNM criteria [11, 29, 34, 36‐40]. A thorough discussion and adaptation was done in one study (Haydom near miss criteria) [37], and another study utilized these adapted criteria [34]. The Haydom criteria adapted the WHO MNM tool to a local hospital in Tanzania, Haydom Hospital. These criteria comprised of all the WHO clinical-based (n = 11), two out of eight laboratory-based criteria, and three out of six management-based criteria of the 2009 WHO MNM criteria. Additional criteria (admission to intensive care unit, eclampsia, sepsis/severe systemic infection, and uterine rupture), which were not part of the 2009 WHO criteria, were added and the threshold for the number of units of blood transfusion was lowered from five or more units of blood to one or more [9]. A study by Kalisa et al. reported another adaptation applied in Rwanda (the Ruhengeri Hospital criteria). In this adaptation they included all the WHO clinical criteria (n = 11), four out of eight laboratory-based criteria, and five out of six management-based criteria from the 2009 WHO MNM criteria. Additionally, admission to an intensive care unit, eclampsia, sepsis/severe systemic infection, and uterine rupture [11] were included. The remaining studies reported limitations with the use of some management-based (dialysis for acute renal failure, use of continuous vaso-active drugs) and a majority of the laboratory- based criteria (measuring pH, lactate, bilirubin, creatinine, arterial blood gas PaO2/FiO2) [29, 36, 38‐40] (Table 4). Seven studies did not describe limitations with regard to the use of the WHO MNM tool [30‐33, 35, 41, 42]. In general, suggested changes in near miss inclusion criteria included lowering the threshold of units of blood given for transfusion from five or more units to one or more [34, 37], four or more [32] or five or more units ordered but not transfused due to shortage [40]. Criteria that were suggested to be included were additional clinical criteria (eclampsia, uterine rupture, and sepsis or severe systemic infections) [11, 34, 37, 38]; and including admission to an intensive care unit as additional management-based criterion [11, 32, 34, 37, 38]. One study compared the WHO criteria with the Sequential Organ Failure Assessment (SOFA) score [43] and reported better utility of the WHO criteria in obstetrics [32]. SOFA is used to quantify organ dysfunction and predict prognosis of severely ill patients [44, 45]. Utility of SOFA score in women with MNM or admitted to intensive care unit was previously validated [43, 46, 47]. Details of reported limitations and suggested adaptations are summarized in Table 4.