Maternal mortality estimation methodologies: a scoping review and evaluation of suitability for use in humanitarian settings

Abstract

Background

Around the world, a maternal death occurs approximately every two minutes—most of these deaths are preventable. The maternal mortality ratio is a key indicator for the Sustainable Development Goals, yet we have no reliable way to estimate maternal deaths in refugee or internally displaced persons (IDP) camps and settlements. The goal of this study was to understand the methodologies most suited for adaptation for use to estimate the proportion of maternal mortality due to abortion complications in these settings.

Methods

We conducted a scoping review of methodologies to estimate maternal mortality and evaluated them using a predetermined set of criteria. We evaluated nine original methodologies using eleven categories related to implementation in refugee or IDP camps and settlements: data sources, definitions, sample size, timing of point estimate relative to data collection, bias, human resources, time needed for implementation, data collection training, statistical training, digitalization, and cost. Each category could be assigned zero to four points, for a total score of 44 points. After evaluating each original methodology, we reviewed the original publication's citations or searched for other implementations through October 2022. We revised the original scores and developed a rank-order list of the methodologies according to their suitability for implementation in refugee and IDP camps.

Results

We identified 124 publications that estimated maternal mortality. The Maternal Deaths from Informants/Maternal Death Follow on Review (MADE-IN/MADE-FOR) (33.5), hospital- or facility-based (33.5), and community informant-based (32.5) methodologies ranked highest due to low costs, short time interval needed for implementation, small sample sizes and close timing of point estimate relative to data collection, easy digitalization, and the need for no statistical training.

Discussion

Similar to the lack of a “perfect” methodology to estimate maternal mortality in stable settings, there are compromises to consider when applying these methodologies to humanitarian settings. The most promising methodologies are adaptable to practical constraints in refugee and IDP camps and settlements. New methodologies that adapt and strengthen the MADE-IN/MADE-FOR, hospital- or facility-based, and community informant-based methodology show promise and must be further developed.

Plain language summary

Around the world, a maternal death occurs approximately every two minutes—most of these deaths are preventable. The goal of this study was to understand the methodologies most suited for use to estimate maternal deaths due to abortion complications in refugee and internally displaced persons (IDP) camps and settlements. We conducted a scoping review of methodologies to estimate maternal mortality and evaluated them using a predetermined set of criteria. We used eleven evaluation categories: data sources, definitions, sample size, timing of point estimate relative to data collection, bias, human resources, time needed for implementation, data collection training, statistical training, digitalization, and cost. Each category could be assigned zero to four points, for a total score of 44 points. We identified 124 publications that estimated maternal mortality. The Maternal Deaths from Informants/Maternal Death Follow on Review (MADE-IN/MADE-FOR) (33.5), hospital- or facility-based (33.5), and community informant-based (32.5) methodologies ranked highest due to low costs, short time interval needed for implementation, small sample sizes and close timing of point estimate relative to data collection, easy digitalization, and the need for no statistical training. Similar to the lack of a "perfect" methodology to estimate maternal mortality in stable settings, there are compromises to consider when applying these methodologies to humanitarian settings. The most promising methodologies are adaptable to practical constraints in refugee and IDP camps. New methodologies that adapt and strengthen the MADE-IN/MADE-FOR, hospital- or facility-based, and community informant-based methodology must be developed.

Background

In 2020, a maternal death occurred approximately every two minutes—for a maternal mortality ratio (MMR) of approximately 223 deaths per 100,000 live births globally [1]. These are deaths to women of reproductive age (aged 15–49 years old—WRA), who are pregnant, postpartum, post-abortion, or delivering, and whose death was caused by or aggravated by the pregnancy [2]. Approximately 75% of these deaths are caused by severe bleeding, infection, high blood pressure, other complications from delivery, and unsafe abortion [1]. Most of these deaths are preventable.

The 1999 annual UNFPA report states that “post-abortion complications account for some 25–50 per cent of maternal deaths in refugee situations” [3]. Globally, 7.9% of maternal deaths are attributable to unsafe abortion, which implies that the UNFPA estimate for abortion-related maternal mortality in refugee situations represents an increased risk of about three to six times [4]. Since the report's publication, this 25–50% estimate has been widely cited in peer-reviewed journal publications [5, 6], despite a lack of information in the report, including the methodology, geography, and time-period under review [3]. Although some studies have since attempted to estimate or understand maternal mortality in forcibly displaced populations [7,8,9,10,11,12,13], only one included abortion as a cause of death, though the investigators did not report any abortion deaths [8].

Decreased access to contraception and increased sexual violence put forcibly displaced individuals at a higher risk of unintended pregnancy. One in four of the over 89 million forcibly displaced individuals globally is at risk of unintended pregnancy [14]. Health facilities in humanitarian settings often experience stock outs of contraceptive methods or may not have the requisite healthcare providers or equipment to provide high-quality reproductive health services [15]. This reduced access to contraceptives may increase the number of unintended pregnancies in these populations [16]. Forcibly displaced individuals are more likely to experience sexual violence and exploitation than host communities or individuals in other settings [17]: in complex humanitarian emergencies, the reported prevalence of sexual-based violence is 21.4% [18], compared to about 10% globally [19].

Once forcibly displaced individuals become pregnant, they may face barriers to safe abortion care if they wish to end their pregnancy. A scoping review of barriers and facilitators to safe abortion care in humanitarian crises found that the legal environment, context, stigma, economic factors, and service delivery constraints prevent access to services [20]. While there is little research exploring forcibly displaced individuals' access to comprehensive abortion care, it is reasonable to expect that they may face issues similar or greater than the barriers that general populations face at the policy, community, health facility, interpersonal, and individual level [21,22,23,24]. The AMoCO study in Nigeria and the Central African Republic documented high levels of abortion-related morbidity and mortality in hospitals in fragile settings, demonstrating the challenges individuals in humanitarian crises face in accessing safe services [25].

Maternal mortality, and especially abortion complication-related mortality, is challenging to estimate even in stable populations [26,27,28]. While many methodologies have been created and tested in stable populations, they suffer from significant limitations. Direct methodologies measure mortality using vital registration systems or facility-based data and indirect methodologies rely on other techniques to elicit information about maternal deaths [28]. Methodologies that use routinely collected information may not find reliable data if there are large gaps in reporting or inaccurate reporting. Central to many of these challenges is that maternal death is relatively rare [2]. This means that many indirect methodologies require large samples to be considered reliable [28]. Methods that require surveying next of kin may suffer from response bias if there are no eligible respondents or selected households are unavailable for participation; when next of kin are available, social desirability bias may affect the cause of death reported. Even when social desirability bias is not an issue, respondents need to know about the decedent's death, their pregnancy status at the time of their death, and their cause of death, for enumerators to properly classify the death. Estimating abortion-related deaths is especially challenging given its taboo status in much of the world. Despite the constraints present in accurately estimating maternal mortality, the maternal mortality ratio is a key indicator for the Sustainable Development Goal #3.1, which calls for a reduction in global maternal mortality to less than 70 deaths per 100,000 live births by 2030 [29].

While past reviews have investigated methodologies to estimate maternal mortality in low- and middle-income countries [26,27,28] and humanitarian emergencies [30], the purpose of this paper is to (1) catalogue and review the existing methodologies for estimating maternal mortality, (2) understand their potential applicability to forcibly displaced populations, specifically refugee and internally displaced persons (IDP) in camps and settlements (hereafter referred to as “camps”), and (3) make a recommendation for future estimates of maternal mortality, and specifically maternal mortality due to abortion complications, in these populations. This information is imperative for monitoring and evaluating sexual and reproductive health programs in refugee and IDP settings.

Methods

In brief, we conducted a scoping review of methodologies to estimate maternal mortality and evaluated them using a predetermined set of criteria. We created an a priori list of nine methodologies to estimate maternal mortality based on our experience in these settings and the literature on surveillance implementation in humanitarian settings [26,27,28]. Our strategy was to identify items for each of the methodologies currently in use. There are only a finite number of methodologies used to estimate maternal mortality, to date, so in essence we conducted a systematic search of each methodology, its adaptations, and subsequent implementation. We evaluated each of the original methodologies using eleven categories related to implementation in refugee or IDP camps. Each category could be assigned zero to four points, for a total score of 44 points. After evaluating each original methodology, we reviewed the original publication’s citations, and if there were not at least ten additional applications of the methodology, searched for other implementations through October 2022. We revised the original scores accordingly and developed a rank-order list of the methodologies according to their suitability for implementation in refugee and IDP camps. Figure 1 presents a process diagram of the different steps we took to evaluate the existing methodologies to estimate maternal mortality.

Fig. 1

Process diagram of steps to review and evaluate methodologies to estimate maternal mortality

Assemble a priori list of methodologies

We created an a priori list of methodologies based on our experience in the humanitarian sector that included: census- or survey-based, hospital- or facility-based methodology, Maternal Deaths from Informants/Maternal Death Follow on Review (MADE-IN/MADE-FOR), motherhood, neighborhood, prospective informant methodologies, community informant-based, Reproductive Age Mortality Survey (RAMOS), and direct sisterhood. The direct methodologies that we included in this review are motherhood, hospital- or facility-based, census- or survey-based, and direct sisterhood. The indirect methodologies included in this review are MADE-IN/MADE-FOR, neighborhood, indirect sisterhood, RAMOS, and community informant-based.

Locate original implementation of the methodology

We located the original field implementation of each methodology by searching PubMed with the methodology name and reviewing article references to find the first instance when a methodology was cited; we continued the process until we found the article or another published item that explicitly stated it was presenting a novel methodology. If the original authors later issued corrections or changes to the methodology, we included these new publications as though they were part of the original methodology and revised our evaluation accordingly. Hospital- or facility- based, census- or survey-based, and community informant-based methodologies were considered not to have original implementations captured by specific research publications. We did not constrain original methodology searches by year or publication status, but excluded items that were not published in the English language and that were not instances of the methodology’s implementation.

Evaluation of the original methodologies

Given the constraints that are present in refugee and IDP camps, we developed eleven criteria upon which to evaluate each methodology relevant to implementation in this specific environment for these unique populations, and as they relate to the specific task of identifying abortion-related deaths. We developed these criteria by reviewing the literature and through our own experience in humanitarian settings and estimating mortality due to abortion complications. Several of these criteria (1, 3, 4, 7, 8, 9, 10) deal explicitly with resource constraints that are common in humanitarian settings and the importance of training to assist enumerators with collecting data on sensitive topics, like abortion [31]. Others are concerned with the reliability of estimates (2, 5, 6, 9, 11) given refugee and IDP camps' open populations, the likelihood of incomplete data sets, and the added difficulty of estimating mortality due to abortion [32]. The criteria are listed below in Table 1.

Table 1 Criteria for evaluating methodologies to estimate maternal mortality in refugee and IDP camps in low resource settings

We evaluated the methodologies by assigning between zero and four points—zero to indicate the lowest possible score and four to indicate the highest possible score—with regard to their suitability to implementation in refugee and IDP camps to estimate proportion of maternal deaths due to abortion complications in relation to other methodologies. The total possible number of points, or highest score, was 44, and categories were not weighted. For the first two methodologies under review, NP and BEO independently assigned scores, compared results, and recalibrated ratings. For all other methodologies, BEO evaluated the methodologies, NP confirmed scoring, and DF, RH, and SM reviewed scores for inconsistencies.

Scoping review of additional implementations

To locate additional implementations of a given methodology, we first reviewed the citations for each original implementation listed in PubMed, Google Scholar, and Research Gate. If the original implementation did not have at least ten citations, we conducted additional searches on the same databases for the time period through October 2022, and evaluated as many implementations as were available. We selected a minimum of ten citations before conducting additional searches to try to capture a broad application of the methodology. The terms for the methodologies where we conducted additional searches are available in Table 2. Since hospital- or facility-based, census- or survey-based, and community informant-based methodologies were not considered to have original implementations, all data for these methodologies were collected through this search methodology. We did not constrain results by type of publication. Any publication that used the methodology in question was retained. We also included items located through references reviews of items located in each search.

Table 2 Search terms used on PubMed to locate additional implementations of each methodology

Evaluation of additional implementations

For this step in the assessment process, we evaluated each new implementation of a given approach according to the eleven categories and considered whether specific adaptations in each implementation, if any, strengthened or weakened the original methodology vis-à-vis its suitability to estimate the number of maternal deaths and proportion of maternal deaths due to abortion complications in refugee and IDP camps. If subsequent implementations strengthened the original methodology, we adjusted the score up for the relevant evaluation category; if they weakened the original methodology, we did not adjust the original score.

Rank-order list creation

After completing the evaluation of original methodologies and subsequent implementations, we consolidated the results for each methodology, incorporating strengthening elements from implementations into the original methodology. We weighed each evaluation category evenly in the final assessment. We calculated the total number of points per methodology by summing the value assigned to each category and ranked the methodologies accordingly.

Reporting

We report the results of the scoping review according to the Preferred Reporting Items for Systematic reviews and Meta-Analyses extension for Scoping Reviews (PRISMA-ScR) Checklist [39]. The methodologies are grouped into three sets in the Results: those with the highest scores, those with average scores, and those with the lowest scores.

Results

We reviewed and evaluated nine methodologies to estimate maternal mortality: motherhood, hospital- or facility-based, census- or survey-based, direct sisterhood, MADE-IN/MADE-FOR, neighborhood, indirect sisterhood, RAMOS, and community informant-based. Of the methodologies with original implementations, all but one were implemented first in Africa and/or Asia—the first implementation of the direct sisterhood method also included data from Bolivia in addition to Sudan and Egypt. Table 3 lists the original methodologies, whether the methodology produces a direct or indirect estimate, the publication title, year of publication, authors, geography of the original implementation, and final score. Table 4 presents the mean, median, and range of scores for each criterion.

Table 3 Original methodologies evaluated in this scoping review, listed by total score, from highest to lowest ranking

Table 4 Mean, median, and score range for each criterion included in our evaluation

We conducted additional searches for the MADE-IN/MADE-FOR, motherhood, and neighborhood methodologies. For MADE-IN/MADE-FOR, we found two items through citation reviews and no additional items from searches. For the motherhood methodology, we did not find any additional implementations from citations or searches. For the neighborhood methodology, we found one relevant citation and no additional items through searches. For the methodologies considered not to have original implementations (census- or survey-based, community informant-based, and hospital- or facility-based), we reviewed six, seven, and sixteen items from searches, respectively. We retrieved 34 articles and twenty additional articles from citation reviews for the indirect sisterhood and RAMOS methodologies. We identified no items from citations and fourteen items from searches for the direct sisterhood methodology. Figure 2 shows a summarized flow diagram for all methodologies and Fig. 3 for each methodology individually. A complete list of the publications considered as implementations of each original methodology is available in Additional file 1.

Fig. 2

Flow diagram with summarized sources of additional implementations of all methodologies under review

Fig. 3

Flow diagram with sources of additional implementations of each methodology under review

Highest scores

The methodologies that received the highest scores after review of both the original implementation and subsequent implementations were MADE-IN/MADE-FOR (33.5), hospital- or facility-based (33.5), and community informant-based (32.5). The evaluation forms, with scores, for each of these three methodologies are available in Additional files 2–4. These methodologies all scored highly for sample size (4, 4, 4 points respectively for MADE-IN/MADE-FOR, hospital- or facility-based, and community informant-based), timing of point estimate relative to data collection (3, 3, 4), time needed for implementation (2.5, 3, 3), statistical training (3, 4, 4), digitalization (4, 4, 4), and cost (3, 4, 3). These methodologies have small sample sizes and can be quickly digitized using data collection forms which require no statistical training to implement. Due to the way these data are collected, the cost to implement the methodologies is relatively low compared to other methodologies, and because estimates can be generated after one or two meetings (MADE-IN/MADE-FOR), retrospective review (hospital- or facility-based), or after retrospective data collection or ongoing prospective data collection (community informant-based), the time needed for implementation is insubstantial and the timing of point estimate relative to data collections is very close to the time of data collection.

The highest scoring methodologies differed in several other evaluation categories. While MADE-IN/MADE-FOR scored 3 points for data sources; hospital- or facility-based and community informant-based methodologies scored 1 and 2 points, respectively. This is because MADE-IN/MADE-FOR requires at least two data sources (reports from healthcare workers and verification through verbal autopsy), whereas the hospital- or facility-based methodology requires only facility records, though maternal death review committees can be involved, and the community informant-based methodology can use a myriad of sources to identify deaths within a population. Definitions scores likewise differed, with the methodologies receiving 3, 2, and 2 points, respectively. While MADE-IN/MADE-FOR provides a clear and standardized definition of pregnancy-related or maternal deaths, hospital- or facility-based and community informant-based methodologies provided definitions that were not standard and/or not followed in the methodologies' actual implementation. In terms of bias, the hospital- or facility-based methodology received one point because deaths may not be reported to a facility and then would not be included in counts, causing selection bias, whereas community informant-based and MADE-IN/MADE-FOR methodologies received 2.5 points due to their use of multiple sources with ongoing reporting to minimize recall bias; while these two latter methodologies include additional data sources for cause of death, they assume that recall among next of kin is accurate. While the hospital- or facility-based methodology received four points for data collection training, due to the straightforward nature of collecting data from maternal death records, MADE-IN/MADE-FOR received 3.5 points, and the community informant-based methodology received 2 points. Similarly, the hospital- or facility-based methodology received 3.5 points for human resources, due to the small amount of data collection, management, and analysis required, whereas the community-informant based methodology and MADE-IN/MADE-FOR received two points apiece due to more intensive staff needs.

Average scores

The methodologies that received the average scores after review of both the original implementation and subsequent implementations were motherhood (29.5), neighborhood (29), and indirect sisterhood (26). The evaluation forms for each of these three methodologies, with scores, are available in Additional files 5–7. These methodologies all scored highly for data collection training (4, 3, 3 points respectively for motherhood, neighborhood, and indirect sisterhood), statistical training (4, 4, 3), digitalization (4, 4, 4), and sample size (3, 3, 2). While the original neighborhood methodology cites a half-day training requirement [42] and the original indirect sisterhood methodology makes mention of a one-hour training plus a half-day of supervision [48], the motherhood methodology does not require training [41]—thus, data collection training for these methodologies are simple to implement. Likewise, the calculation for maternal deaths in each of these three methodologies is clear and straightforward, requiring no additional statistical training. Due to the ease with which data are programmed into pre-existing digital tools, these methodologies are easily digitized for data collection. Neighborhood and motherhood methodologies received three points for sample size due to the lack of a prerequisite number of deaths for the methodologies to be implemented, whereas indirect sisterhood has a larger sample size requirement of 3000–6000 in most instances, but documentation provides clear guidance on calculation.

The motherhood, neighborhood, and indirect sisterhood methodologies similarly scored two points or fewer for data sources (2, 2, 2 points respectively), timing of point estimate relative to data collection (1, 1, 1), and bias (2, 2, 2). Each of the three methodologies utilizes at least two data sources—for neighborhood, information is elicited through discussions with community members and confirmed via verbal autopsies, within the motherhood methodology, pregnant individuals share information that is verified with vaccination records and verbal autopsies with next of kin, and in indirect motherhood, census records lead to representative surveys of households in a given geography. All three methodologies received one point for the timing of point estimate relative to data collection due to the significant difference between implementation and when the estimate is valid: for the motherhood methodology, 2 years and 3 months before the study period, for the neighborhood method, 2 years and 42 days, and for indirect sisterhood, 5 years plus the amount of time spent in implementation. Lastly, each methodology received two points for bias due to the risk of selection bias, non-response bias, and recall bias.

The motherhood, neighborhood, and indirect sisterhood methodologies were similar in their middling scores in the human resources (3, 2.5, and 2 points respectively), time needed for implementation (2.5, 2.5, 2), and cost categories (3, 3, 2). For human resources, the fewest number of individuals are needed for the motherhood methodology, where facilitators lead group discussions and occasionally collect verbal autopsies, as opposed to the neighborhood method, where enumerators gather information on deaths in public places and then routinely collect verbal autopsies, and the indirect sisterhood methodology, wherein households are randomly sampled for participation in the survey without knowing if/when maternal or other deaths will be reported. In terms of time needed for implementation, both the neighborhood and motherhood methodologies report the amount of time for the original studies, 42 days and 30 days, respectively, and then estimates for other studies, whereas the indirect sisterhood method suggests that for a shorter implementation time, a large cadre of staff are needed. While cost is not reported, it can be extrapolated from other information that indirect sisterhood would be the most expensive of these three methodologies, though neighborhood and motherhood methodologies also require considerable inputs for data collection and processing.

The methodologies differed most in scoring for definitions (1, 3, 3 points respectively for neighborhood, motherhood, and indirect sisterhood). While the indirect sisterhood, at least in later implementations, and neighborhood methodologies provide clear definitions of maternal death in line with the International Classification of Diseases (ICD)-10 and standard WHO maternal death classifications, the motherhood methodological paper does not describe the deaths that they are looking, other than to say that they are maternal—it is unclear whether participants know how to identify these deaths too.

Lowest scores

The methodologies that received the lowest scores after review of both the original implementation and subsequent implementations were RAMOS (23.5), census- or survey-based (22.5), and direct sisterhood (22.5). The evaluation forms for each of these three methodologies, with scores, are available in Additional files 8–10.

There are only two categories in which all three of these methodologies scored highly: definitions (4, 3, 3 respectively for RAMOS, direct sisterhood, and census- or survey-based) and digitalization (4, 4, 4). All surveys received high scores for definitions due to the consistent use of WHO standard maternal death definitions in line with ICD-10; each specific implementation of RAMOS methodology stated the exact deaths they were targeting. For digitalization, the methodologies all received four points because the tools are already digitized through the Demographic and Health Surveys and online verbal autopsy tools, which can be easily adapted by other implementers.

The methodologies all received two or fewer points for sample size (1, 1, 1 points respectively for RAMOS, direct sisterhood, and census- or survey-based methodologies), human resources (1, 1, 1), time needed for implementation (2, 1, 1), cost (2, 1, 1), data collection training (2, 2, 2), timing of point estimate relative to data collection (1, 2, 2), and bias (2, 2, 2). Each of these methodologies require large sample sizes with multiple thousands of respondents to surveys, which is why each received only one point in this category. These large sample sizes require extensive human resources, including field supervisors and teams of enumerators. With large sample sizes and despite sizeable study teams, the time needed for implementation can be prolonged, especially in wide-ranging geographies. Though for the Demographic and Health Surveys (DHS), large sample sizes are related to the nationally-representative nature of the surveys, the direct sisterhood methodology still requires large numbers of respondents for precise estimates. These factors all lead to increased cost for these methodologies compared to other methodologies for estimating maternal mortality. Due to the vast time range (5 + years) for which respondents are requested to provide information about maternal deaths, each methodology received low scores for timing of point estimate relative to data collection, with RAMOS receiving two points due to the flexibility of the methodology to adjust for shorter time periods. Similar to all other methodologies, RAMOS, direct sisterhood, and census- or survey-based methodologies may be affected by significant levels of recall, non-response, and selection bias unless they are well-designed even in ideal scenarios.

In terms of data sources, the RAMOS methodology received three points, whereas direct sisterhood and census- or survey-based methodologies received two points. The RAMOS methodology calls for the triangulation of data from several sources, while census- or survey-based and direct sisterhood methodologies rely only on accurate census data and interviews with members of randomly selected households from a sampling frame. Due to the extremely simple and direct calculation of maternal mortality from census- or survey-based and direct sisterhood methodologies, they each received 3.5 points for statistical training, whereas the RAMOS methodology received only one point due to the lack of explanation for how the calculation is derived from the collected data in the original or any subsequent implementations of the methodology.

Discussion

In this scoping review, we synthesize nine methodologies to estimate maternal mortality and abortion-related maternal deaths and determine their suitability for use in refugee and IDP camps in resource-scare settings. We do this by using eleven categories that account for practical implementation constraints in these environments and specifically for estimates of abortion-related mortality. The results of the review point to the tradeoffs inherent in existing methodologies—tensions between methodologies that are simple to implement and less accurate and those that are more technically complex but produce more reliable estimates. Even in stable settings, choosing the right methodology to estimate maternal mortality can be difficult, and these difficulties are amplified in humanitarian emergencies, including forcibly displaced persons camps. Resource scarcity, competing priorities, and incomplete data sets can make data collection especially challenging. Stigma associated with abortion may influence enumerator and provider recorded and informant given information about cause of death. While other reviews have evaluated these methodologies more broadly [26,27,28, 49], and despite the challenges of calculating maternal mortality estimates in these settings, we completed this review because believe that these figures are necessary to accurately understand and respond to the needs of refugee women and girls.

These tradeoffs are readily apparent in the top scoring methodologies in this scoping review. The MADE-IN/MADE-FOR methodology relies on more data sources than hospital- or facility-based methodology or the community informant-based methodology, which causes MADE-IN/MADE-FOR costs to be higher and require more human resources. On the other hand, MADE-IN/MADE-FOR and the community informant-based methodology are less likely to introduce bias to estimates than the hospital- or facility-based methodology. In these methodologies there are clear differences in the comprehensiveness of the approach and the time intensity to accomplish a more thorough estimate. Humanitarian practitioners will need to take these limitations into account when designing studies to estimate maternal mortality.

Though there are risks and benefits to the use of any of these methodologies in any setting, there is an opportunity to test new methodologies that combine the best qualities of the existing ones. For example, novel methodologies could investigate what biases may be introduced when using data sources that tap into the expertise of community members, such as community leaders, health workers, or pregnant mothers, and more readily available data, such as death records available in health facilities. While this may introduce new biases, it may also prevent bias resulting from incomplete data sets in humanitarian emergency settings by capturing deaths that may or may not be reported to authorities. Methodologies with clear definitions of maternal death and cause of death that use straightforward data collection training that are easily digitized due to their lack of advanced statistics may be more practical in non-stable settings where there are competing priorities.

Any adaptations or new methodologies should be tested and validated against stable-setting “gold standard” methodologies, i.e., RAMOS or census- or survey-based methodologies, in refugee and IDP camps. We also believe that any study using human subjects should undertake the most caution to ensure confidentiality and data privacy, especially when dealing with sensitive topics like maternal death and deaths due to abortion complications. Though we certainly can’t solve all issues related to the trickiness of estimating maternal mortality in these special circumstances, there are adaptations and opportunities worth exploring to estimate this significant indicator in refugee and IDP camps.

Limitations

This study has several limitations. We conducted a scoping review of the literature, and not a systematic review, so we may have missed some methodologies. Similarly, we limited our review to published literature and items that were published in English, which means that grey literature and items in other languages were excluded. In almost all cases, we extrapolated cost based on other information provided in manuscripts, because many studies did not directly report expenses in publications—however, the results remain the same whether or not cost is included in the final score. Another limitation is that although this is an evaluation and ranking for refugee and IDP camps, we did not actually test each methodology in these environments. Instead, we reviewed the available evidence for each methodology, and using our expertise in humanitarian emergencies, considered and assessed using pre-determined categories, the application of each methodology in these populations. Ideally, we would implement all of these methodologies in a site with known mortality and determine their accuracy compared to the truth. This scoping review is subjective—we, the authors, determined the categories as well as their relative ranking, based on our expertise in humanitarian settings. Despite these limitations, the scoping review reveals insights into potential feasibility and applicability of existing methodologies to measure maternal mortality in refugee and IDP settings.

Conclusions

To adequately address the specific sexual and reproductive health needs of pregnant and postpartum forcibly displaced persons, we must be able to accurately estimate the burden of maternal mortality in the population, and specifically, maternal mortality due to abortion complications. Three methodologies currently used to estimate maternal mortality and mortality related to abortion complications show the most promise for suitability to use in refugee and IDP settings. Thus, it is imperative to develop new methodologies that adapt and strengthen the MADE-IN/MADE-FOR, hospital- or facility-based, and community informant-based methodologies for research in forcibly displaced settings.