Introduction

Maternal mortality is an important public health indicator, reflecting not only poor health condition of mothers and poor quality of health care services, but also in macro levels, it indicates poor economic, social, cultural, and political status in a community and also poor status of women in their society [1,2,3]. Nowadays, improving maternal health is the priority of most communities and organizations, including the World Health Organization (WHO) [4]. The fifth goal of the eight Millennium Development Goals (MDGs) focused on the reduction of maternal mortality by 75% from 1990 to 2015. Despite intense political attention of the countries toward the MDG5, most of them are still far from achieving their MDG5 targets [3]. The global estimates indicate that 295,000 (UI 279,000 to 340,000) maternal deaths occurred in 2017; 35% lower than 2000 when maternal deaths were estimated at 451,000 (UI 431,000 to 485,000) [5]. Thus, another target for maternal mortality reduction was set by the WHO to reach a global maternal mortality ratio (MMR) below 70 by 2030, which requires reducing global MMR by an average of 7.5% each year from 2015 to 2030. This means that to achieve this goal, maternal mortality must be reduced by more than three times the 2.3% annual rate of reduction observed globally between 1990 and 2015, which would be a difficult task for many countries [6].

Based on the existing evidence, there are significant variations in MMR of countries based on their development and income level [5, 7]. Annually, half a million women die from pregnancy-related deaths, of which 99% happen in developing countries while most of them are preventable [8]. This variation is portrayed by WHO; while MMR has never been passed over 115 cases in Europe, America, and Western Pacific regions since 1990, it has always been over 500 cases in Africa [9].

In order to plan effectively and progress systematically in the reduction of MMR, having a broad knowledge about determinants of maternal mortality is an unavoidable need for policymakers and researchers. These determinants may distantly include socio-economic and cultural factors or intermediately include health status, reproductive status, access to health services, and health care behavior factors, or approximately include biological causes of death [10]. Causes of maternal death, based on the International Classification of Diseases-Maternal Mortality ( ICD-MM), are classified to direct (e.g., pregnancies with abortive outcome) or indirect (e.g., cardiac disease) causes in six groups by WHO [4, 11], with no diversity among them.

The existing review studies on determinants of maternal mortality are generally old, or are non-systematic reviews or only reviewed selected determinants [6, 10, 12]. Recently, only a few studies reviewed the determinants systematically, such as the study of Yakubu et al. [13], but only focused on some micro determinants or other study on limited number of determinants [14, 15]. Given that maternal death is a multidisciplinary phenomenon and different factors are involved, we tried to comprehensively review and overview all determinants, with no limitation, in high quality studies which would be of higher applicability for policymakers and researchers. The aim of this study was to conduct a systematic review on individual and ecological determinants of maternal mortality in the world and to classify them based on the income level of countries. In addition to summarizing distant and intermediate determinants of death in individual or ecological level, we also tried to answer this question that how these determinants might vary among mothers living in countries with various development level.

Methods

Inclusion criteria

All types of studies, except descriptive studies, that aimed to identify determinants of maternal mortality were included. In these studies the relationship of possible determinants with maternal mortality was statistically tested. Studies that only focused on the cause of maternal mortality as an approximate determinant, published in non-English journals, or when their full-texts were unavailable were excluded. We considered WHO definition for maternal mortality as “death of a woman while pregnant or beyond 42 weeks of gestation, irrespective of duration and site of pregnancy, due to any cause related to pregnancy or its management, but not due to accidental causes” [9]. No time limit was considered in searches.

Search strategy

To identify relevant studies, an electronic literature search was first conducted on “PubMed” and “Scopus” databases. Then, some of the most relevant journals were searched specifically (Table 1). Additionally, references of included articles were screened.

Table 1 Sources and search terms and the number of studies identified in each (last searched in 5th August 2021)
Full size table

Except for PubMed database in which “Maternal Mortality” was searched using MeSH term, this term was searched just as a keyword in the rest of search procedure (Table 1). All articles retrieved from electronic databases and other studies were entered into EndNote X8 for further selection.

Study selection

The process of identification and selection of studies is shown in Fig. 1. Initially, 902 and 901 studies were found through electronic searching in databases and journals, respectively. Furthermore, 1855 records were identified based on the bibliography of the included studies. Duplicate studies (n = 98) were removed. Then, screening of titles and abstracts was done by the researchers and studies not relevant were excluded (n = 3308). Then, the assessment of the full text of articles was done for eligibility of 252 studies. Based on the exclusion criteria, 123 studies were excluded due to being only descriptive and not reporting statistical results of relationship between independent variables and maternal mortality (except for review articles), or only focusing on biological causes of maternal deaths, or focusing on the relationship between determinants rather than their relationship with mortality. 16 studies were also excluded as their full-text were not found [16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31] and eight others were not in English [32,33,34,35,36,37,38,39]. In addition, 16 relevant articles were added through searching on Google. Finally, 121 relevant studies obtained from electronic databases, bibliographies, and hand searching were selected based on the inclusion criteria (Table 1 and Fig. 1). No study was excluded after quality assessment. Two reviewers (R.Z. and A.H.) performed the literature searching, checking the inclusion and exclusion criteria and selection of studies and finally data extraction, all phases independently. To reach consensus in the case of variations, in all stages of this review, the reviewers met frequently for discussion, and in the case of lasting disagreement, the third reviewer (M.T.), has been intervened.

Fig. 1
figure 1

PRISMA flow diagram for systematic review

Full size image

Quality assessment

We used the STROBE checklist for original studies and also for ecological studies. It is designed for quality assessment of observational studies especially cohort, case–control, or cross-sectional studies based on 22 items [40]. There were some limitations in quality assessment of the included studies; i.e., some items of this checklist related to participants (item 6), or sample sizes (item 10), or reasons for non-participation (item 13) were not reported in ecological studies that used MMR as a dependent variable. Considering those items as “can’t tell”, the quality scores of those studies decreased. Other items such as clearly defining potential confounders in item 7, reporting potential sources of bias in item 9, way of handling quantitative variables in item 11, and reporting statistical methods used to control confounders or missing data in item 12 of the STROBE checklist were the main limitations of the selected studies, which were not often reported in their methodology. In some included studies, their limitations were not mentioned in the discussion. Each question in this checklist had one score, so the highest quality of the article would obtain score 22.

We used the critical appraisal checklist from Glasgow University [41] for the quality assessment of review articles. As shown in Table 3, except for two studies containing nine review articles, the quality assessment was not performed for the rest of them, and their results were not reported precisely because determinants of maternal mortality were not based on the confidence interval. It is worth noting that the quality of the studies was independently assessed by two researchers in different places. Then, discrepancies were resolved by participation of a third researcher to reduce the risk of bias.

Data extraction

We extracted the required data from the included review articles and original articles using a purposefully designed data extraction form. Year of publication, number of maternal deaths, time span of the study, setting, design, and methods, and maternal mortality determinants were extracted from the included studies as shown in the Tables 2 and 3.

Table 2 Summary of characteristics and findings of original studies (Chronologically ordered)
Full size table
Table 3 Summary of characteristics and findings of review studies
Full size table

Data analysis

The included studies were developed with diverse objectives, used various methods and different statistical techniques, and included participants with different characteristics and were widely distributed among countries. This diversity made formal meta-analysis almost impossible. Some of the included studies were developed with similar objectives, and participants had similar characteristics and used similar methods and measures, but we also classified the determinants of maternal mortality into three main individual, household, and community levels to be higher applicability for policymakers and researchers.

Classification of the included studies based on income levels of countries

We classified each original study based on low-, lower-middle-, upper-middle-, and high-income countries according to World Bank classification. Some points considered in this classification; for example, if a study in a country was in a specific income group for a some years and then its level changed for next years, we chose the income group that were for a longer period. In the study of Simo ˜es et al. [94], Brazil was in the upper-middle-income group for two years (2000–2001) and then changed as a lower-middle-income country in 2002; thus we considered Brazil in the upper-middle-income group. If a study included different countries in two different income groups, we chose the group with more countries. For example, in the study of Graham et al. [131], among 11 different countries, only two countries were in lower-middle group and the rest were in low-income group; so, we considered it as a low-income country. We also chose the last years of income level for the studies that used the same number of years for two different levels of income. For example, in the study of Taguchi et al. [77], Indonesia was lower-middle-income country for two years (1996 and 1997) and was low-income for the other two years (1998 and 1999), so we considered Indonesia as low-income country. Studies with more than two income groups in different countries were considered as global/regional studies, illustrated in the last column of Table 4.

Table 4 Summary of determinants of maternal mortality according to the Table 2
Full size table

Results

The results of this systematic review are structured here in two main sections; first, according to Tables 2 and 3, which are a summary of all included original studies and review articles, respectively, and provide a description of the characteristics of included studies. Then, maternal mortality determinants are reported for review articles and original studies, summarized at individual, household, and community levels by income groups in Table 4.

Description of the included studies

Of 121 included studies, 12 were review articles (Table 3) and the rest were original studies (Table 2) in which 33%, 28%, 27%, 10%, and 2% were ecological, case–control, cross-sectional, longitudinal (including prospective cohort, panel and time-series studies), and cohort studies, respectively. Among the 12 review articles, six of them with a total of 160 studies were conducted systematically and the rest were narrative reviews (Table 3). Most of the literature evaluated the determinants of maternal mortality globally with special focus on low- and lower-middle-income countries. Moreover, the results by countries show that most of the included original studies have been conducted in China, followed by Niagara and Indonesia (details provided in Fig. 2), and most of them were published after 2013 (Fig. 3).

Fig. 2
figure 2

Number of included studies by setting

Full size image
Fig. 3
figure 3

Number of included studies by publication years during 1991–2021

Full size image

Determinants of maternal mortality in the world

In this section, first the results of the review articles then the original study’s findings are reported based on determinant levels and income levels of countries.

Based on our review articles, some macro factors such as economic growth, poverty, inequality, improved water and sanitation, education, accessibility of health care services, etc. (details are shown in Table 3), and some micro factors such as age, parity, type of delivery, early marriage, etc. have led to maternal mortality.

Maternal mortality determinants based on individual, household, and community levels

Determinants at individual level mainly show the health status and reproductive-demographic characteristics of deceased mothers. A summary of the determinants of mothers who died due to pregnancy-related reasons, based on original studies, is described in Table 4. As shown in column three of the table, among demographic factors, maternal age over 34 or below 18 is one of the main factors affecting pregnancy-related deaths, and nearly 70% of 36 examined studies with 7632 aggregate deaths confirmed this association. In addition, higher gravidity and cesarean delivery (as opposed to natural delivery) were recognized as significant threatening factors in 90% and nearly 75% of the studies, respectively. Furthermore, the results of this review show that mothers attended by skilled birth personnel, under prenatal care, as well as educated women are less likely to die from pregnancy-related causes. Moreover, delay in seeking health services and making decisions for delivery in all the eight observational evidences and factors related to mothers’ health (including the history of underlying disease, medical comorbidities, having complications during pregnancy or postpartum period, and hospitalization during pregnancy) were significant contributors according to most studies. However, in a few studies, some variables such as BMI, race, skin color, and caste were not recognized as associated factors. In contrast, each of the factors of birth weight, substance misuse, nationality, and previous pregnancy history were significant factors in one study, which we refused to include in Table 4. At this level, there is no consensus on the effect of age at marriage, age at first birth, birth weight, and postnatal care coverage. Details about other determinants at this level were described in Table 4.

Determinants at household level show that improved water sources and sanitation systems were the main contributors in most studies. In addition, low family income or wealth, and rural residence were associated factors in nearly 87% and 70% of studies, respectively. Also in this level, access to electricity in all the evaluated studies, and higher levels of husband education were recognized as significant factors in 80% of the observational studies (details in Table 4). Other factors mainly related to the condition of living, including the type of flour and toilet, building material of the house, and type of family were the associated factors, which are not shown in Table 4.

At community or social levels, poor human resources in 73% of the examined studies, and poor accessibility or availability of health facilities (including the number of hospitals, primary health care, hospital beds, and equipment, etc.) in 79% of the 19 examined studies were the main contributors that threaten maternal health. Moreover, health expenditure shares (including higher share of out-of-pocket or private expenditures, and lower share of total health of GDP or public expenditures) and higher total fertility rate per woman were other associated factors in 89% and 77% of the included studies, respectively. In this category, there is no consensus on the effect of quality of services, and less use of contraception (details in Table 4). A few of the selected studies also recognized other factors such as life expectancy, need for emergency obstetric care, total calories consumed, and fat residuals as significant factors, which we refused to enter in Table 4. In terms of economic factors that were mainly studied in ecological studies, lower-income countries levels (GNP or GDP) in 89% of the studies, unemployment rate in all studies, and poverty in 57% of the examined studies were significant threatening factors. However, income inequality and foreign investment are not known as significant factors. In contrast, debt rate, agricultural production indices, and private-sector infrastructure were other contributors in a study, which we refused to enter in Table 4. In addition, results show that maternal mortality was also caused by social, political, and cultural factors. Thus, levels of education and governance (corruption index, instability, voice and accountability, etc.) were the main contributors in five ecological studies. Also, gender inequality and human development index both in four original studies were known as main factors in this category (details in Table 4). Also, other factors such as paved roads, annually published scientific papers, percentage of the indigenous population, and education efficiency were significant factors, which we refused to enter in Table 4.

Maternal mortality determinants based on income levels of countries

A summary of our findings is shown in Table 4. As it shows, the attributable risk of maternal death is varied based on income levels and development levels of countries; more than 80% of the examined studies and all four ecological studies found poor income levels of countries (GDP/GNP) and human development index strongly associated with maternal death, respectively.

Among the factors at individual levels, prenatal care, delay in decisions, type of delivery, and hospitalization during pregnancy in most studies (except the type of delivery in lower-middle income group) were significant factors in all different income levels. In six out of seven global studies, skilled birth attendance was a significant contributor in low-income, less developed or developing countries. Also, maternal education as another main factor, was an essential contributor in low and lower-middle-income countries, and more than 80% and 70% of the examined studies have shown a significant relationship, respectively. However, there is no consensus on this factor in upper-middle-income countries. Some factors like history of underlying disease, medical comorbidities, high-risk pregnancy, length of labor, and pregnancy stage were recognized as risk factors in low or lower-middle-income countries (Table 4).

Based on this review, most studies statistically accepted that less access to clean water and improved sanitation negatively affected maternal death in low and lower-middle-income countries, respectively. In addition, in poor countries, improvement in the availability or accessibility of health services and health expenditure shares were essential determinants in the reduction of maternal death. Based on our findings, reduction in the share of out-of-pocket and private health expenditures and increase in the share of total health expenditures as a percentage of GDP and public expenditures were most responsible for pregnancy-related deaths, particularly in low-income countries as well as other income groups. Moreover, the risk of pregnancy-related death was highly significant among the population below poverty line in all existing evidence, and the use of contraception in many studies in low-income countries. However, poverty in upper-middle-income countries was not significantly associated with maternal death. Moreover, gender inequality and gender empowerment measures (autonomy, economic empowerment, and attitudes towards violence) in low and lower-middle-income countries were associated contributors in most studies. However, gender empowerment measures did not affect maternal death in two ecological studies, and there is no consensus on the significance of gender inequality in our global studies. Details of the other variables is shown in Table 4.

Discussion

This study aimed to review and overview maternal mortality determinants in the world and categorize them based on income levels of countries. Our included studies show that most of these studies were conducted ecologically and were case–control studies and 58% of them were carried out in low and lower-middle-income countries.

According to the result of this study, maternal age, gravidity, type of delivery, education of mothers, pregnancy care, skilled birth attendance, and maternal health status are the main factors at individual level. At family level, factors such as access to improved water and sanitation, region of residence, family income or wealth, and other factors related to living conditions were significantly associated with maternal death. However, the systematic review conducted in Iran [156] failed to find the place of living as a risk factor because of having more access to primary health care among rural women in Iran. At community or social level, availability of health services, total fertility rate, health expenditures shares, income level, governance, inequality and education were main contributors.

Our review shows that women aged 18–34 years were less likely to die as compared with age groups of 35–49 or under 18 years. A possible reason for the high risk of maternal death among women aged 35–49 years is due to weakened uterus and anemia, and becoming pregnant is too risky for women older than 35 years and studies showed that the majority of women in this age group were not educated. Studies indicate that most maternal deaths occurred in women with no antenatal care and delay in seeking health services. This may be due to lack of awareness about the seriousness of maternal health. The type of delivery can also affect maternal mortality, as most cases of obstetric hemorrhage and emergency postpartum hysterectomy are associated with CS deliveries [75].

Among original studies, two observational studies in lower-middle income countries [93, 144], and an observational study in the west of Iran [93], as upper middle-income country, demonstrate that skilled birth attendance has no significant relationship with maternal mortality. However, it doesn’t seem that skilled birth attendance has no effects on maternal mortality at this income level as those studies were conducted with a small sample size and limited time span. A longitudinal study in Chile, as a lower middle-income level country, showed this factor as a significant contributor in adjusted models. Some global or regional studies, mainly including less developed or developing countries, accepted this factor [76, 130, 138].

Based on our findings, households’ access to improved water and sanitation especially in low-income countries were negatively associated with maternal death. In line with this, Benova et al. [15] suggested that women in households with poor sanitation were 3.14 times more likely to die compared to women with better sanitation. The concept is that poor access to sanitation and water can provide the conditions for the prevalence of infectious diseases, which can directly affect maternal death. Studies also show that access to improved sanitation is associated with income level of countries [79, 115]; a study shows that clean water and sanitation access improved by 14% due to economic growth in low and middle-income countries from 1990 to 2010 [73]. Based on WHO data [157], access to improved sanitation and water was always over 98% among high-income countries since 1990, while in low-income countries, access to clean water fluctuated around 46% to 66%, and access to sanitation fluctuated around 13% to 28% from 1990 to 2015.

Governance is one of the key factors that affects maternal mortality. Governance can be described as a set of traditions and conventions that determines the practice of authority in a particular country. It comprises not only the processes through which governments are selected, held accountable, monitored, and replaced, but also the capacity of governments to efficiently manage resources and formulate, implement, and enforce appropriate policies and regulations. In addition, governance regulates the level of respect received by the citizens and the state for conventions and laws that govern the economic and social interactions in the community. Through better governance, public spending can effectively and efficiently enhance health and education outcomes.

Some economic and health system-related factors such as the Gross National Income per capita and per capita government expenditure on health showed an inverse correlation with MMR. In contrast, private sector and out-of-pocket health expenditure showed a significant direct correlation with MMR: the more private sectors and out-of-pocket health expenditures in a country, the higher the MMR. Since appropriate government financing can ensure better access to some essential maternal health services, greater levels of health expenditure will be required for developing countries to achieve MDG on maternal mortality. Between 1995 and 2014, the average total health expenditure (%GDP) varied from 5% in low-middle income countries, 48% of which was paid by public expenditure, to nearly 11% in high-income countries, 61% of which was paid by public expenditure. Based on WHO statistics, the average out-of-pocket share (% total health expenditure) varied from nearly 41% in low-middle income countries to nearly 15% in high-income countries in the same period [158].

According to the literature, poverty positively affects maternal mortality as an economic factor, especially in the low-income countries. According to the evidence, poverty is linked to maternal mortality through malnutrition [99]. Malnutrition has been associated with anemia that is one of the main causes of maternal death [159]. Malnutrition may lead to chronic iron deficiency and anemia, which can make women prone to hemorrhage and infections [160]. Furthermore, women who experience malnutrition early in life are usually smaller, which increases the likelihood of obstructed labor [160].

Also, low human development index (HDI) in all evaluated studies was recognized as a significant contributing factor. HDI is one of the powerful indexes, which is a geometric means with three dimensions of life expectancy at birth, education, and GDP per capita. Much evidence shows that MMR reflects the general health status of a country and its development [130].

Moreover, education as a social factor is the other main contributor. As seen in Table 4, among 20 studies that examined the effects of education on maternal mortality at community level, only one did not confirm this. This shows that education is influential in reducing maternal mortality. Educated women are more exposed to having an informed reproductive behavior, and they access reproductive health facilities frequently and timely. When women are educated, they increasingly improve their status, gain autonomy, awareness, responsibility, and control their fertility and reproductive activities such as in the use of contraceptives, prenatal and postnatal care, and health facility delivery [13]. Literature also shows that education in low and lower-middle-income countries is essential, which may have happened due to the low level of social and economic infrastructures in them.

Other factors including households’ income, employment, urbanization, inequalities, gender empowerment, human resource, availability of services were the other main factors particularly in less developed or developing countries (Table 4).

To the best of our knowledge, this is the first systematic review to present a robust summary of individual and ecological determinants of maternal mortality in the world based on income level of the countries. We used a sensitive search strategy in different sources, and finally identified 121 studies (109 original studies and 12 review articles). However, this review has several potential limitations. As the main limitation of this study, we only included studies in English with access to full-text documents. To compensate this, we did a rigorous search on different electronic databases, journals, and paper references, and the abstracts of excluded studies were reviewed to reduce bias. As with any systematic review, we may have missed some studies due to relatively a short list of search terms. Also, some studies did not provide access to the full text and thus they had to be excluded from this review. However, we minimized this by exhausting the search process through key terms, and the search strategy employed multiple academic and grey literature databases. As another limitation, all included studies had analytical observational and review design, so the cause-effect relation between determinants and maternal mortality could not be established. The results of this review can help researchers to understand the main determinants of maternal mortality based on income level of the countries and provide an appropriate space for research. This study can also provide comprehensive views for policymakers to reduce maternal mortality.

Conclusions

This review aims to comprehensively show the determinants of maternal mortality. The results of this study demonstrate the individual-level factors (e.g., age and parity), household-level factors such as region of residence, access to improved water etc., and community-level factors such as socio-economic, cultural, and health care system factors associated with maternal mortality. Based on extracted determinants, providing health services by making them affordable and available, access to quality health care, access to prenatal care, family planning care, and emergency obstetric care, increasing education levels, improving the condition of living and infrastructures, and the use of contraception can play an important role in the reduction of maternal mortality. Further studies are recommended to reach a consensus regarding certain determinants such as age at marriage, postnatal care coverage, etc.