Cross-sectional study evaluating the effectiveness of the Mozambique–Canada maternal health project abstraction tool for maternal near miss identification in Inhambane province, Mozambique

Abstract

Objectives

The objectives of this study are to determine whether the additional clinical criteria of the Mozambique maternal near miss abstraction tool enhance the effectiveness of the original WHO abstraction tool in identifying maternal near miss cases and also evaluate the impact of sociodemographic factors on maternal near miss identification.

Design

Cross-sectional study.

Setting

Two secondary referral hospitals in Inhambane province, Mozambique from 2021 to 2022.

Participants

From August 2021 to February 2022, 2057 women presenting at two hospitals in Inhambane Province, Mozambique, were consecutively enrolled. Eligible participants included women admitted during pregnancy, labour, delivery, or up to 42 days post partum. Selection criteria focused on women experiencing obstetric complications, while those without complications or with incomplete medical records were excluded.

Primary and secondary outcome measures

The primary outcome was identifying maternal near miss cases using the original WHO Disease criterion and the additional clinical criteria from the Mozambique-Canada Maternal Health Project abstraction tool. Secondary outcomes included the association between sociodemographic factors and maternal near miss identification. All outcomes were measured as planned in the study protocol.

Results

The new Mozambique-Canada abstraction tool identified more maternal near miss cases (28.2% for expanded disease and 21.1% for comorbidities) compared with the original WHO tool (16.2%). Hypertension and anaemia from the newer criteria were strongly associated with the original WHO Disease criterion (p<0.001), with kappa values of 0.58 (95% CI 0.53 to 0.63) and 0.21 (95% CI 0.16 to 0.26), respectively. Distance to health facilities was significantly associated, with women living over 8 km away having higher odds (OR=2.47, 95% CI 1.92 to 3.18, p<0.001). Type of hospital also influenced identification, with lower odds at Vilankulo Rural Hospital for Expanded Disease criterion (OR=0.70, 95% CI 0.57 to 0.87, p=0.001), but higher odds for comorbidities criterion (OR=3.13, 95% CI 2.40 to 4.08, p<0.001). Finally, older age was associated with higher odds of identification under the comorbidities criterion, particularly for women aged 30–39 (OR=3.06, 95% CI 2.15 to 4.36) as well as those 40 years or older (OR=4.73, 95% CI 2.43 to 9.20, p<0.001).

Conclusions

The Mozambique-Canada Maternal Health Project tool enhances maternal near miss identification over the original WHO tool by incorporating expanded clinical criteria, particularly for conditions like hypertension and anaemia. Sociodemographic factors, including healthcare access, hospital type and maternal age, significantly impact near miss detection. These findings support integrating the expanded criteria into the WHO tool for improved identification of maternal near misses in Mozambique and similar low-resource settings. Future research should examine the tool’s effectiveness across varied healthcare contexts and populations.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• Rigorous adaptation of the WHO maternal near miss tool to suit local clinical contexts in Mozambique.

• Consecutive sampling ensured the comprehensive inclusion of eligible participants.

• Homogeneity of the study population may limit generalisability to diverse groups.

• The absence of an external control group could affect the robustness of the comparisons.

Introduction

Maternal mortality remains a critical global health issue. Despite United Nations’ targets in the Millennium Development Goals and Sustainable Development Goals to reduce maternal deaths, the rate remains around 151 per 100 000 live births, especially in Africa, Asia and Latin America, where clinical and sociocultural factors elevate the risk.¹,³ Although maternal mortality is high, maternal near misses (MNMs)—where a woman nearly dies from complications during pregnancy or childbirth—occur more frequently.^{4 5} MNM is valuable for assessing and enhancing maternal care quality and influenced the development of an abstraction tool by the WHO that aids clinicians and researchers in identifying these cases.^{6 7} The WHO MNM abstraction tool includes three criteria: disease based, intervention based and organ dysfunction. The disease-based criterion flags five initial clinical indicators as ‘potentially life-threatening,’ marking the start of a pathological progression towards death.⁷ The intervention criterion uses four hospital-based parameters to predict and reduce death risks, while the organ-dysfunction criterion identifies seven critical organ system failures representing the final life-threatening stage (online supplemental figure S1).^{7 8} The organ-dysfunction criterion, deemed most effective by the WHO, is both sensitive in detecting severe cases and specific enough to reduce unnecessary clinical loads, making it the preferred choice in MNM predictions.^{5 9 10}

However, subsequent studies reveal significant challenges in relying solely on the organ-dysfunction criterion in low-income and middle-income countries due to limited diagnostic equipment and skilled clinicians.^{7 10} Research indicates that the disease-based criterion is more effective for identifying MNMs in these settings, as pregnant women often present with multiple clinical indicators, reflecting delays in accessing obstetric care.^{3 9 11} Overwhelming evidence show that haemorrhage, anaemia and hypertension are the primary clinical indicator of MNM in low-income regions,¹⁰,¹³ thus, prompting revisions to the WHO tool to enhance its utility in resource-limited settings.

The sub-Saharan Africa (SSA) MNM Abstraction tool is a key adaptation of the WHO tool and has served as the basis for country-specific versions, such as the Nigeria tool.¹¹ The Mozambique-Canada Maternal Health (MCMH) Project further tailored the Nigerian version to the resource availability in Mozambique (online supplemental figure S2). The MCMH abstraction tool retains all indicators from the intervention and organ dysfunction. Crucially distinct from the original, it expands the disease-based markers with new “Expanded-Disease” and “Comorbidities” criteria. The expanded-disease criterion details each marker from the original WHO clinical category, while the comorbidities criterion includes non-obstetric conditions like HIV/AIDS and malaria. Unique to this tool, it also integrates sociodemographics affecting maternal outcomes, providing a more comprehensive clinical and demographic profile.¹⁴,²⁵ The tool was used in collecting data in the MNM 1.0 study in Inhambane, Mozambique.

The current study therefore aimed to evaluate the effectiveness of the MCMH tool in enhancing MNM identification compared with the original WHO disease criterion. Additionally, it sought to examine the influence of sociodemographic, health systems and geographic factors on MNM identification in Mozambique’s Inhambane province. The findings provide valuable insights into the tool’s potential to improve maternal care outcomes in resource-limited settings.

Materials and methods

Study setting, design and population

Eligible participants were women who presented consecutively at two hospitals, Vilankulo Rural Hospital (HRV), a rural referral hospital with a large rural catchment area of approximately 46 543 inhabitants, and Inhambane Provincial Hospital (HPI), the central tertiary hospital in the Inhambane province, from 16 August 2021 to 18 February 2022, during pregnancy, labour, delivery, or up to 42 days post partum or termination of pregnancy (including abortion and ectopic pregnancy). A total of 2057 medical records were retrieved, with 1255 participants from HRV and 802 from HPI.

Inhambane province is one of Mozambique’s 11 provinces, located on the southeastern coast of Africa with a population of about 1.4 million as of the 2017 census.²⁶ Both hospitals are secondary referral facilities, but HPI is a referral centre for the entire province, including the Vilankulo district and other secondary hospitals. Participants were recruited by well-trained health personnel who approached women in the maternity units in both hospitals. Interested women gave consent to access their records and to participate in the study. The trained personnel also collected data through participant interviews and medical records, using the MCMH abstraction tool, and weekly meetings were held to ensure data quality.

In this cross-sectional study, patients were not involved in the design or conduct of the research. The nature of our study, being cross-sectional, involved obtaining information from participants at a single point in time, without establishing prior or follow-up contact with these individuals. Therefore, there was no active engagement of patients in the planning, execution, analysis or dissemination of the study findings. However, the study benefitted significantly from collaborative efforts with the Inhambane Provincial Health Directorate (DPSI, acronym in Portuguese, Mozambique’s lingua franca), the Ministry of Health and key physicians from our recruiting sites. Their invaluable contributions were integral to the planning, methodology and dissemination of results. Specifically, the DPSI provided essential guidance and support in aligning the study objectives with national health priorities. Physicians from both recruiting sites played crucial roles in refining the study design, ensuring methodological rigour and interpreting findings within the local healthcare context. Their collaborative efforts have enhanced the study’s relevance, validity and potential impact on health policy and practice in Mozambique.

Data collection tool

The MCMH abstraction tool collected the clinical and sociodemographic information of participants (MCMH tool included in the online supplemental material, figure S2). The information was divided into the following sections, respectively: (1) hospitalisation information, (2) sociodemographic data, (3) vitals and pregnancy history, (4) organ-dysfunction markers, (5) intervention indicators, (6) disease-based markers, (7) expanded-disease markers, (8) comorbidities indicators, and (9) fetal or neonatal outcome in women with delivery.

An MNM case could fulfil definitions from any or all five criteria (sections (4) through (8)). Thus, fulfilment of each criterion was counted as an MNM identified by that criterion. For example, an HIV-positive woman with severe bleeding and neurological issues admitted to the intensive care unit (ICU) met disease-based (6), intervention (5), organ dysfunction (4) and comorbidities (8) criteria. Biomedical data in sections (4) through (8) were abstracted exclusively from medical records. Four data collectors and a supervisor, based at each hospital, received training before data collection for each research site. Data quality was ensured through frequent checks for uniformity, completeness and accuracy. Following data collection, two additional personnel further cleaned the dataset before analysis.

Data analysis

χ² tests of independence were conducted to examine the relationship between each additional clinical criterion and the original WHO disease criterion. Kappa statistics were used to assess the agreement between indicators of the expanded disease criterion and the original disease criterion. Agreement levels were interpreted based on predefined thresholds.²⁷ To explore hospital-specific differences, the data were stratified, and both χ² tests of independence and kappa statistics were re-run on health facility-specific data.

To determine the association between specific sociodemographic factors and MNM outcome, multivariable logistic regressions were performed. Initially, bivariate analyses were conducted to assess the relationship between each independent variable and the outcome. Significant variables (p<0.05) were included in multivariate analyses. Interaction terms were tested using the 2-log-likelihood method, and confounding effects were evaluated. Based on the literature review, sociodemographic factors with potential confounding were systematically included in the model. Confounding effects were assessed by comparing adjusted ORs with crude estimates, with a change exceeding 10% indicating significant confounding. This approach allowed us to reduce bias from confounders and enhance the accuracy of our adjusted effect estimates about MNM outcomes. Goodness-of-fit was assessed using the Hosmer-Lemeshow²⁸ estimates, with a χ² value and p>0.05 indicating a good model fit. ORs, 95% CIs and p values were reported for each model. Facility-based models were obtained through data stratification and following the same steps. Only statistically significant models per hospital were reported.

Patient and public involvement

Patients were not directly involved in the research; however, the governing body of the Mozambique Health Directorate and selected health professionals were engaged from the planning stage. They provided input on defining the research questions, participant selection criteria and data collection processes, ensuring that the research priorities aligned with local healthcare needs. Health professionals also helped assess the feasibility and burden of data collection methods to minimise participant inconvenience. To maintain transparency and collaboration, periodic reports, webinars and workshops were used to update stakeholders on the study’s progress and findings. The dissemination plan involves sharing study results through these same channels, tailored to decision-makers feedback on content and timing, ensuring that findings are communicated effectively to both health professionals and the broader health system.

Results

Study population characteristics

The distribution of population characteristics is presented in table 1. Approximately half (49.8%) of the women were 24 years or younger and nearly one-third of all study participants were either married or lived maritally with their partners. Most women (68%) resided close to the study health facilities, within 8 km. Less than half of the population (39%) completed secondary school education and even less than one-tenth completed a bachelor’s or graduate degree. The majority of these women were unemployed (86.7%) and practised Christianity (93.7%) (table 1).

Table 1. Characteristics of study participants, maternal near miss study, Inhambane Province, Mozambique (N=2057).

Variable	Frequency (%)
Hospital admitted (n=2055)
Inhambane Provincial Hospital	800 (38.9)
Vilankulo Rural Hospital	1255 (61.1)
Distance to health facility (n=2054)
Within 8 km to hospital	1397 (68.0)
More than 8 km	657 (32.0)
Educational level (n=2051)
None	246 (12.0)
Primary school 1	255 (12.4)
Primary school 2	545 (26.6)
Secondary school	633 (30.9)
Post secondary	285 (13.9)
Bachelors	77 (3.7)
Graduate	10 (0.5)
Profession (n=2053)
Unemployed	1735 (86.7)
Unqualified employment	161 (8.2)
Semi-qualified employment	92 (4.6)
Professional	11 (0.5)
Religion (n=2045)
Islam	60 (3.1)
Christianity	1867 (93.7)
Traditional	63 (3.1)
Other	2 (0.1)
Marital status (n=2052)
Single	518 (25.9)
Married	96 (5.0)
Live maritally	1374 (68.6)
Divorced	9 (0.4)
Widow	2 (0.1)
Age categories (n=2057)
≤19	437 (21.8)
20–24	558 (27.8)
25–29	433 (21.6)
30–34	354 (17.9)
35–39	165 (8.2)
≥40	55 (2.7)

MNMs identified by different criteria

The expanded disease (28.2%) and comorbidities criteria (21.1%) identified the highest MNM cases while the organ-dysfunction criterion yielded the least (2.7%) (online supplemental table S3 and figure S4). Each criterion comprises prominent markers that helped identify MNMs for that category (online supplemental table S5). Hypertension was the leading contributor to cases in the original WHO disease criterion (66.6%) while infection represented the least (1.5%). Blood transfusion contributed to the most MNM events under the intervention criterion (71.3%); no patient underwent interventional radiology. Neurological dysfunction accounted for the most cases under the organ-dysfunction criterion (58.2%), while liver dysfunction showed the least contribution (1.8%). Hypertension remained the highest contributor to the expanded disease criterion (36.6%) while infection remained the least (1.6%). MNMs under the comorbidities criterion were mainly attributed to HIV/AIDS (42.9%), anaemia (6.5%) and malaria (1.9%). Medical conditions like kidney, heart, liver and lung diseases, as well as cancer, were not identified in any study participants.

The data were stratified by hospitals (online supplemental table S6). Hypertension, admission to ICU, neurological dysfunction and HIV/AIDS remained the predominant markers of MNMs in both sites. Notably, MNMs associated with hypertensive disorder (16.2% for WHO disease criterion, and 20.6% for expanded disease criterion) and with ICU admission (2.9%) remained highest among patients in the provincial hospital, while neurological dysfunction (2.1%) and HIV/AIDs (24.2%) were identified as the most prominent causes associated with MNMs in HRV.

Indicators of the expanded disease criteria that are associated with the WHO disease criterion

As presented in table 2, all variables in the expanded disease criterion were statistically associated with the original WHO disease criterion (p<0.001). Within this category, hypertension had the strongest association with the original disease category (χ²= 678.5 (d.f. = 1, p<0.001)) while obstructed labour had the weakest association with the original disease criterion (χ²= 24.0 (d.f. = 1, p<0.001)). Variables in the comorbidities criterion were also statistically related to the original WHO disease criterion (p<0.05) except HIV/AIDS (p>0.05). Among the statistically significant, anaemia was most strongly associated with the original WHO disease criterion (χ²= 200.1 (d.f. = 1, p<0.001)) while malaria showed the weakest association with this original WHO disease criterion (χ²= 4.6 (d.f. = 1, p<0.005)).

Table 2. Overall results from χ² test of independence and kappa statistic between the WHO disease criterion versus the Mozambique-Canada Maternal Health (MCMH) expanded disease and comorbidities criteria.

Association variable	χ²	d.f.	P value	Kappa value	CI
					Lower	Upper
MCMH expanded disease indicators with WHO disease criterion:
Bleeding and original disease criterion	46.7	1	<0.001	0.11	0.06	0.16
Infection and original disease criterion	35.6	1	<0.001	0.05	0.02	0.08
Hypertension and original disease criterion	678.5	1	<0.001	0.58*	0.53	0.62
Obstructed labour and original disease criterion	24.0	1	<0.001	0.10	0.06	0.15
MCMH comorbidities indicators with WHO disease criterion:
HIV/AIDS and original disease criterion	0.7	1	0.409†	0.02	−0.027	0.06
Malaria and original disease criterion	4.7	1	0.032	0.02	−0.00	0.04
Anaemia and original disease criterion	200.1	1	<0.001	0.21*	0.16	0.26
Embolic disease and original disease criterion	10.4	1	0.001	0.01	−0.00	0.02

MCMH, Maternal Near Miss Abstraction Tool.

^*kKappa values greater than 0.2 fair or higher reliability.

^†P value greater than 0.05 suggesting a lack of association between variables.

The kappa estimates, which evaluate the corroboration between the original WHO disease criterion and MCMH expanded-disease indicators, revealed an overall ‘weak’ agreement between the two categories (table 2). Only hypertension, one specific condition within MCMH expanded disease criterion, showed a ‘moderate’ degree of agreement with the original WHO disease criterion (κ=0.58, 95% CI 0.53,0.63). Generally, the kappa estimates for the MCMH comorbidities criterion and WHO disease criterion were lower than those for the MCMH expanded disease criterion and WHO disease criterion. Most comorbidities showed no agreement with the original WHO disease criterion, with embolic disease having the lowest level of agreement (κ=0.01, 95% CI 0.00, 0.02). Anaemia was the only factor to have a ‘fair’ agreement with the original WHO disease criterion (κ=0.21, 95% CI 0.16, 0.26).

Similar tests were conducted on hospital-stratified data to identify hospital-specific characteristics. In provincial hospital-specific data, hypertension remained strongly associated (χ²= 336.7, d.f. = 1, p<0.001) and showed an improved agreement, from ‘moderate’ to ‘substantial’, with the original WHO disease criterion (κ=0.65, 95% CI 0.58, 0.72). For the comorbidity variables, HIV/AIDS continued to lack an association with the original WHO disease criterion (p>0.05) (online supplemental table 7). In HRV data, hypertension retained the strongest association with the original WHO disease criterion (χ²= 317.1, d.f. = 1, p<0.001). Hypertension had a ‘moderate’ agreement with the original WHO disease criterion (κ=0.50, 95% CI 0.43 to 0.57), and HIV/AIDS still showed no association with the WHO disease criterion (p>0.05) (online supplemental table 8).

Sociodemographic factors that are associated with identifying MNMs defined by each clinical criterion

Table 3 displays results of the multivariable analysis on the relationship between sociodemographic factors and WHO abstraction tool-defined MNMs. The first multivariable analysis examined the relationship between sociodemographic factors and the original WHO disease criterion-defined MNMs. The Hosmer-Lemeshow test indicates a good fit (χ²= 4.8, d.f. = 8, p=0.780 > 0.05) with the model successfully predicting observed data points (84%). None of the interaction terms tested were statistically significant and no confounding effects were observed. Only distance to the health facility was statistically significant in this model (online supplemental figure S20), indicating that individuals living over 8 km from the study hospitals had over twice the odds of being identified as MNMs compared with those within 8 km (OR=2.47, 95% CI 1.92, 3.18).

Table 3. Multivariable analysis on the association between maternal characteristics and WHO disease criterion for identifying maternal near misses.

Variables	Adjusted OR	P value
Hospital admitted
Inhambane Provincial Hospital	1 (Ref)
Vilankulo Rural Hospital	0.84 (0.65 to 1.09)	0.186
Distance to health facility
Within 8 km	1 (Ref)
More than 8 km	2.47 (1.92 to 3.18)	<0.001
Educational level
None	1 (Ref)
Primary school	0.88 (0.59 to 1.33)	0.554
Secondary school or post secondary	1.09 (0.72 to 1.65)	0.684
Bachelors or graduate	1.89 (1.02 to 3.52)	0.043
Age categories
≤ 19	1 (Ref)
20–29	0.90 (0.66 to 1.23)	0.511
30–39	1.11 (0.77 to 1.59]	0.589
≥40	1.50 (0.73 to 3.09)	0.272
Marital status
Single	1 (Ref)
Married or live maritally	1.20 (0.89 to 1.61)	0.229
Divorced or widowed	0.47 (0.06 to 3.78)	0.474

All boldened p- values indicate statistically significant values (p<0.05).

Hosmer-Lemeshow Ggoodness of fit: (χ2=4.790, [d.f. = 8, p-value=0.780> 0.05]).

The second multivariable analysis explored the association between sociodemographic predictors and MCMH expanded disease-defined MNMs (online supplemental table S17). Hosmer-Lemeshow test showed that the final multivariable model is well fitted (χ²= 4.9, d.f. = 7, p=0.668 > 0.05) and has a 71.9% accurate prediction of the observed events. No interaction was determined through statistical testing nor was confounding effects observed. Distance from the health facility remained highly associated with MNM identification, with individuals living over 8 km to the hospital having over twice the odds compared with those within 8 km (OR=2.26, 95% CI 1.83 to 2.78). Additionally, the type of hospital was a significant factor, with 29% lower likelihood of identifying an MNM case among the rural hospital cases compared with provincial hospital cases (OR=0.70, 95% CI 0.57 to 0.87) (table 4, online supplemental figure S20).

Table 4. Multivariable analysis on the association between maternal characteristics and the Mozambique-Canada Maternal Health Project expanded disease criterion maternal near misses.

Variables	Adjusted OR	P value
Hospital admitted
Inhambane Provincial Hospital	1 (Ref)
Vilankulo District Hospital	0.70 (0.57 to 0.87)	0.001
Place of residence
Within 8 km	1 (Ref)
More than 8 km	2.26 (1.83 to 2.78)	<0.001
Educational level
None	1 (Ref)
Primary school	1.00 (0.71 to 1.37)	0.946
Secondary school or post secondary	1.19 (0.87 to 1.64)	0.283
Bachelors or graduate	1.47 (0.85 to 2.52)	0.167

All boldened p- values indicate statistically significant values (p<0.05).

Hosmer-Lemeshow Ggoodness of fit: (χ2=4.937, [d.f. = 7, p value=0.668> 0.05)].

Table 5 shows the results from the third logistic regression model between sociodemographic factors and MCMH comorbidities-defined MNMs. Results from the Hosmer-Lemeshow test demonstrated that the final multivariable model is well fitted (χ²= 5.8, d.f. = 8, p=0.674 > 0.05) with a 79.7% accurate prediction of observed events. No interactions between covariates were observed. However, both maternal educational level and type of hospital produced confounding effects (|adjusted - crude ORs|>10%). In the adjusted model, individuals living over 8 km were about 57% more likely to be identified as MNMs than those within that distance (OR=1.58, 95% CI 1.24, 2.01). Unlike previous MNM groups (WHO disease criterion defined and MCMH expanded disease-defined MNMs), admission to the rural hospital, rather than the provincial hospital, increased the odds of identifying comorbidities-defined MNMs (OR=3.13, 95% CI 2.40 to 4.10) (table 5, online supplemental figure S20).

Table 5. Multivariable analysis on the association between maternal characteristics and the Mozambique-Canada Maternal Health Project comorbidities criterion maternal near misses.

Variables	Adjusted OR	P value
Hospital admitted
Inhambane Provincial Hospital	1 (Ref)
Vilankulo Rural Hospital	3.13 (2.40 to 4.08)	<0.001
Distance to health facility
Within 8 km	1 (Ref)
More than 8 km	1.58 (1.24 to 2.01)	<0.001
Educational level
None	1 (Ref)
Primary school	0.70 (0.49 to 0.98)	0.040
Secondary school or post secondary	0.67 (0.47 to 0.96)	0.028
Bachelors or graduate	0.45 (0.22 to 0.91)	0.027
Age categories
≤ 19	1 (Ref)
20–29	1.73 (1.25 to 2.40)	<0.001
30–39	3.06 (2.15 to 4.36]	<0.001
≥40	4.73 (2.43 to 9.20)	<0.001

All boldened p- values indicate statistical significance of variable (p<0.05).

Hosmer-Lemeshow Ggoodness of fit: (χ2=5.757, [d.f. = 8, p-value=0.674 > 0.05]).

Discussion

This study assessed if the additional clinical criteria, expanded disease and comorbidities, as measured by the MNM abstraction tool developed by the Mozambique-Canada Maternal Health Project improved the scope of the original WHO disease-criterion in identifying potential MNMs. It also examined what sociodemographic factors influenced the identification of MNMs in Inhambane, Mozambique.

It is well argued that MNM cases represent the severe end of obstetrical complications; it has been reported that for every maternal death, there may be up to 12 MNM cases.²⁹ Therefore, broadly identifying potential MNM cases, especially in low-resource settings, has value clinically. This enables healthcare providers to be vigilant and to work towards preventing severe obstetrical risks, ideally, or failing that, to care for those with risk for severe obstetric outcomes and arrest the progression to mortality. In other words, timely and complete identification of potential MNM cases and treating them is expected to reduce maternal deaths; with salutary results not only for the mother and family, but also for the nation.

The study findings revealed that the expanded disease criterion identified more MNMs than the original WHO disease criterion, with hypertensive disorders as the main contributor to this enhanced identification of MNMs. Both the expanded disease and original WHO disease criteria target similar populations, with the former identifying more MNMs. There was less overlap between comorbidities indicators and the original WHO disease criterion, except for anaemia. The organ-dysfunction criterion was the most conservative for identifying MNMs in the current study. These results are generally consistent with previous literature.^{12 30 31} A clinical implication of these findings is that women who present with chronic hypertensive disorders or anaemia are at further risk for severe obstetric outcomes and will need rapid and targeted care to avoid negative outcomes, including death.

Similar to the study findings, the SSA MNM abstraction literature demonstrates that hypertensive disorder is one of the biggest clinical markers of MNMs. Research shows that hypertensive disorders account for about 20%–53% of all MNM cases in SSA.³²,³⁵ Other studies also include dystocia and anaemia as the top causes of maternal morbidity.^{412 36},³⁸ Although haemorrhage is also reported as a prominent marker for MNMs in SSA, this was not observed in our study. Most studies report postpartum haemorrhage as the highest cause of MNMs in SSA ranging from 20% to 57% of all cases.^{1213 33},^{35 39} However, in the current study, haemorrhage contributed only to 3.3% of all cases, which may be partly explained by an effective hospital management protocol for haemorrhage in the hospitals participating in this study, such as the rapid intervention on transfusion of blood or its products.⁴³ This management protocol is further evidenced by the fact that about 72% of all MNMs identified by the intervention criterion received a blood transfusion. Alternatively, the low haemorrhage cases could be attributed to a renewed focus on continuous clinical training on managing obstetric haemorrhage, in part, as a result of the training provided by the MCMH project within the Inhambane province.

Although HIV/AIDS was prevalent, it consistently lacked an association with the original WHO disease criterion in both hospitals. Other comorbidity markers, like malaria and embolic disease (thrombo/amniotic/air embolism), showed weak associations with the original WHO disease criterion as well. Researchers assert that comorbid or pre-existing non-obstetric indicators only account for a small subset of all MNMs. For instance, a study showed that only 2.5% of cases were associated with HIV/AIDS and 4.1% with malaria.⁴⁴ Oladapo et al noted that these comorbid diseases contributed only marginally to the overall MNM cases in their study.³² However, they also observed that these diseases disproportionately contributed more to maternal deaths. Thus, in their study, while only 6.8% of all MNMs were attributed to comorbid disease, about 19.6% of maternal deaths were associated with these underlying non-obstetric markers.³² Overall, the results indicate that the comorbidities criterion, except for anaemia, does not help improve the capacity of the original WHO disease criterion in identifying MNMs in both hospitals.

Overall, the results from the MCMH tool highlight the need to expand the WHO criteria to capture more MNMs. A recent study conducted in Turkiye at a tertiary referral hospital identified MNM cases based on WHO management-based criteria and expanded inclusion to patients with severe pre-eclamptic and Hemolysis, Elevated Liver enzyme levels, and Low Platelet levels (HELLP) syndrome, which represent high-mortality conditions in maternal health.⁴⁵ Similar to our findings, hypertensive disorders emerged as a major contributor to MNMs. The inclusion of severe pre-eclampsia and HELLP in the Turkiye study highlights the importance of customising MNM criteria to address specific high-risk conditions prevalent in regional settings. Although our study focused mainly on the disease criterion while theirs focused both on the clinical-based and management-based criteria, their findings support our study’s rationale for adapting MNM criteria in Mozambique, where local morbidity and health system contexts may necessitate tailored MNM definitions to improve case identification and care strategies.

To further understand the profile of a potential MNM, it is necessary to understand the structural factors underpinning the condition. Distance from the hospital was consistently associated with MNMs, regardless of the clinical definition. Thaddeus and Maine stipulated, in their three-delay framework, that distance from a health facility was an essential determinant of the second type of delay that increases the risk of MNM and/or death.⁴⁶ In many rural areas within SSA, the paucity of public transportation, high transport costs and/or poor road infrastructure exacerbate delays in reaching these facilities during obstetric complications.^{30 37 47 48} The unavailability of suitable transportation forces some women to walk the distance during an obstetric complication.⁴⁷,⁴⁹ A study showed that walking for more than 1 hour to a health facility was associated with about four times higher odds of MNMs.⁴⁸ Another study revealed that delays caused by the lack of vehicles increased the odds of MNMs by eight times.⁵⁰ Furthermore, Hadush and colleagues observed that delays in reaching a health facility contributed to about 40% of the maternal morbidities in their study.⁵¹ Our results reveal similar trends; it suggests that significant delays potentially occur for women who live greater than 8 km from a health facility due to transport-related issues which subsequently increases their odds of MNMs.

The type of hospital, essentially a provincial hospital drawing referrals from the entire province or a district hospital with a large rural catchment area, was also significantly associated with the identification of MNMs by using the additional clinical categories. As compared with the provincial hospital, it was less likely for women to be identified as MNMs in the district hospital when using the expanded disease criterion. One potential reason is that, as a provincial referral hospital, the HPI receives women in more critical clinical conditions than those seen at the HRV. Conversely, the odds of identifying MNMs using the comorbidities criterion were higher in the rural hospital than in the provincial hospital because the HRV receives more pregnant women with non-pregnancy-related comorbid conditions, such as HIV/AIDS and malaria. It is interesting to note that the type of hospital did not influence the identification of MNMs under the original WHO disease criterion. Again, this suggests that the WHO abstraction tool is more conservative in MNM identification.¹³ Therefore, incorporating this expanded disease criterion possibly expands the range of structural factors to consider, such as the type of hospital, when creating a potential MNM profile.

As expected, older age at delivery increased the odds of identifying MNMs in the comorbidities criterion.⁴⁷⁵²,⁵⁵ It is well established that the risk of maternal morbidities is higher at two periods in a woman’s reproductive life cycle.^{26 53} The first is during adolescence (10−19 years) and the second is at the end of a woman’s reproductive period (35 years and above).^{54 56} One study shows that for women 10−15 years, their risk of maternal mortality is about fivefold higher than women between 20 and 24 years.⁵⁶ Another study presented that women older than 35 years were 74% more likely to develop an MNM than women between 25 and 34 years.⁵² Women above 35 years especially run a higher risk of developing comorbid disease such as hypertension, heart and thyroid disease, and diabetes that complicate their pregnancies and make them more susceptible to MNMs.⁵² This may also explain why the odds of identifying MNMs in women over 35 years was almost fivefold within the comorbidities criterion in our study as well. Although this sociodemographic factor did not show a significant association with MNMs identified by the other clinical criteria, it is important to realise that a similar trend was seen with the odds of identifying MNMs defined by the original WHO disease criterion. In summary, age may be an important factor to consider when building the MNM profile especially when relying on the comorbidities definition within secondary hospitals across Inhambane.

Our study did not find any consistent associations between maternal education, marital status, profession or religion and MNMs defined by all clinical criteria. This is potentially because of the homogeneity of the study population in our sample. A majority of women had not completed formal schooling, were married, were unemployed and declared their religion as Christian. Different groups of women with varying demographics should therefore be included in future studies.

This study has several strengths. To the best of our knowledge, it is one of the first studies that attempts to evaluate an adapted MNM evaluation tool, that is, MCMH abstraction tool, and to refine the WHO clinical definition of MNMs based on local Mozambican clinical contexts. It also sheds more light on specific sociodemographic factors that help influence MNM identification in Inhambane. Moreover, the design of the MCMH abstraction tool—to include both the additional (expanded disease criterion) and original WHO disease criterion—helps to better compare data and improve internal validity. This sharply contrasts with other studies where the WHO tool is completely separated and is different in design from the adapted versions, potentially causing inconsistencies in the data collection and results comparison process.^{32 33 38 52} Again, this study worked directly with local clinicians who provided relevant clinical and cultural information to help contextualise the investigation. Another major strength was our considerable study sample which was comparable to the sample sizes present in the literature.

Nevertheless, some limitations must be acknowledged. For instance, we observed some data quality issues, such as participants receiving the same unique identifiers on separate hospital admissions. However, these errors were corrected during the data-cleaning phase, therefore minimal to or no effect on results. Also, the rigour of the study could have been further strengthened by introducing external control. Specifically, a clinician could have performed their clinical diagnoses of potential MNMs, independent of any abstraction criterion, as another source of comparison with both the WHO disease and additional clinical criteria. Another limitation is the potential for selection bias due to the inclusion of only two hospitals, which may not fully capture the diversity of maternal health experiences across the broader reproductive age population in Inhambane province. While we included a provincial referral hospital and a rural district hospital with a large catchment area to reflect varying healthcare access levels, women who do not seek hospital care or who receive only community-based maternal care, and who experience MNM for whatever the reason, may not be well represented in our sample, hence potentially skewing the study findings. Despite these limitations, our study unveiled important findings concerning MNM identification in rural Mozambique settings. We hope that these findings can be applied to the current maternal health practices to improve obstetric care, especially within health facilities across the province of Inhambane and beyond.

To further advance the identification of MNMs using the MCMH model, future research is needed. For example, more investigation is required to determine if the MCMH abstraction tool produces similar results within different levels of care, such as primary-level and tertiary-level facilities. Although some sociodemographic factors were not statistically associated with MNMs in our study, further research should be done to test these associations within more diverse study populations. Furthermore, other sociodemographic factors based on the available literature that were not originally captured should also be tested. Finally, qualitative studies should be conducted to provide more insight that complement the findings of this present study.

Conclusion

This study demonstrated that using expanded disease criteria identified more potential MNM cases than using the WHO abstraction tool, especially the disease group as a criterion. Incorporating markers from an expanded disease criterion to identify MNMs strongly corroborated the identification of MNMs using the original WHO disease criterion. Conversely, using markers like HIV/AIDS within the comorbidities criterion to identify MNMs consistently lacked corroboration with the original WHO disease criterion, indicating less overlap between their populations. Therefore, relying on non-pregnancy-related comorbidities, that is, HIV/AIDS, to profile women at greater risk for MNMs should be done cautiously. Geographical and sociodemographic factors such as longer distance to a health facility and advanced age increased the odds of identifying MNMs.

In summary, the study supports using the expanded disease criterion alongside the original WHO disease criterion to identify a wider range of MNMs. It also emphasises assessing specific georgraphical, health access, and sociodemographic factors, such as distance to the birth facility, type of health facility, and age, to guide the identification of these cases.

The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Data availability statement

Data are available upon reasonable request. All data relevant to the study are included in the article or uploaded as supplementary information.