Article Text

Does armed conflict lead to lower prevalence of maternal health-seeking behaviours: theoretical and empirical research based on 55 683 women in armed conflict settings
  1. Tingkai Zhang1,
  2. Qiwei He1,2,
  3. Sol Richardson1,
  4. Kun Tang1
  1. 1Vanke School of Public Health, Tsinghua University, Beijing, China
  2. 2Institute of International Development Cooperation, Chinese Academy of International Trade and Economic Cooperation, Beijing, China
  1. Correspondence to Dr Kun Tang; tangk{at}


Background Women and children bear a substantial burden of morbidity and mortality due to armed conflict. Life-saving maternal and child health (MCH) services are low-quality in most conflict-affected regions. Previous studies on armed conflict and MCH services have been mostly cross-sectional, and a causal relationship between armed conflict and MCH services utilisation cannot be inferred.

Methods First, we constructed a utility equation for maternal health-seeking behaviour. Next, we extracted MCH data from the Multiple Indicator Cluster Survey led by the UNICEF. Armed conflict data were obtained from the Uppsala Conflict Data Programme; 55 683 women aged 15–49 from Chad, the Central African Republic, the Democratic Republic of Congo (DRC) and the Republic of Iraq were selected as participants. We fitted a difference-in-differences (DID) model, taking before or after the conflict started as an exposure variable to estimate the effects of armed conflict on maternal health-seeking behaviours.

Results According to the results of the DID model, in the regional sample, armed conflict had a positive effect on tetanus vaccination (β=0.055, 95% CI 0.004 to 0.106, p<0.05), and had a negative effect on antenatal care at least eight visits (ANC8+) (β=−0.046, 95% CI −0.078 to −0.015, p<0.01). And, the effects of armed conflict on ANC, ANC4+, institutional delivery and early initiation of breast feeding (EIB) were not statistically significant. As for the country sample, we found that armed conflict had a negative effect on EIB (β=−0.085, 95% CI −0.184 to 0.015, p<0.1) in Chad. In Iraq, armed conflict had positive impacts on ANC (β=0.038, 95% CI −0.001 to 0.078, p<0.1) and tetanus vaccination (β=0.059, 95% CI 0.012 to 0.107, p<0.05), whereas it had a negative effect on ANC8+ (β=−0.039, 95% CI −0.080 to 0.002, p<0.1). No statistically significant associations were discovered in DRC based on the DID model.

Conclusions There might be a mixed effect of armed conflict on maternal health-seeking behaviours. In the absence of humanitarian assistance, armed conflict reduces certain maternal health-seeking behaviours, such as ANC8+. When practical humanitarian health assistance is provided, the damage can be alleviated, and even the prevalence of maternal health-seeking behaviours can be improved, such as tetanus vaccination. Providing humanitarian assistance to conflict-affected regions improved the accessibility of MCH services for women living in those areas. However, the goals of saving lives and alleviating suffering still need to be achieved. In conflict-affected regions, humanitarian assistance on ANC, institutional delivery and breast feeding need strengthening.

  • Maternal health
  • Public Health
  • Child health
  • Health services research

Data availability statement

All data are publicly available. The health data from the Multiple Indicator Cluster Surveys (MICS) can be found publicly (, accessed on 17 October 2022). Armed conflict data used in this study are from UCDP Georeferenced Event Dataset (GED) Global Version 22.1, which are openly available in The Uppsala Conflict Data Program at (accessed on 17 October 2022).

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See:

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


  • In the majority of conflict-affected regions, maternal and child health (MCH) services are of poor quality.

  • Previous studies were mainly cross-sectional studies discussing the associations between armed conflict and maternal health-seeking behaviours. Countries affected by armed conflict have yet to be studied in conjunction, and few studies explored the effects of armed conflict on MCH.


  • This study is the first to construct a theoretical model of utility to explain the effects of armed conflict on maternal health-seeking behaviours and test the model through empirical research.

  • The theoretical model illustrates that armed conflict will negatively affect maternal health-seeking behaviours in the absence of humanitarian health assistance. However, when practical health assistance is provided, armed conflict may lead to an unintentional benefit. Armed conflict is more likely to positively affect maternal health-seeking behaviours in regions with poor health infrastructure, such as sub-Saharan Africa.


  • This study emphasises that armed conflict can positively affect maternal health-seeking behaviours when providing humanitarian health assistance. Policy-makers should strengthen the delivery of lifesaving humanitarian health assistance in conflict-affected regions, especially on antenatal care, institutional delivery and breast feeding.


The number of armed conflicts at regional and national levels worldwide has rapidly increased since 2010.1 There is no fixed definition of ‘armed conflict’; based on interpretations from previous studies, it can be understood in broad terms as including war, organised violence and other international or national conflict events.2 Since 2010, the majority of armed conflicts have occurred in Africa, Asia and the Middle East.3 4 The high number of armed conflicts has resulted in many battle-related deaths (BRDs).1 3 In addition to directly caused deaths, the frequency of armed conflict has also caused severe public health and development crises.5–7

Women and children are disproportionately affected by armed conflict. Approximately 10% of women and 16% of children worldwide were affected by armed conflict in 2017.8 9 In addition to directly contributing to increased mortality rates,8 10 armed conflict often leads to humanitarian crises such as displacement, food shortages, malnutrition and lack of resources for maternal and child health (MCH) services8 11 12; it also results in non-fatal physical injuries and disabilities,13 14 acute and communicable diseases, chronic and non-communicable diseases, mental health problems, and sexual and reproductive health-related issues.8 Furthermore, maternal populations are more vulnerable in conflict settings. According to the WHO, approximately 830 women worldwide die each day from pregnancy or childbirth complications.15 It is estimated that over 99% of maternal deaths occur in developing countries, with over half of these occurring in sub-Saharan Africa and nearly a third in South Asia, with more than half occurring in regions experiencing armed conflict or humanitarian crises.15 16

In this situation, local and international humanitarian organisations have endeavoured to give the conflict-affected populations with substantial humanitarian health assistance, such as prenatal counselling, HIV/AIDS, delivery assistance, child nutrition supplementation, vaccination and other services.9 17 The provision of humanitarian health assistance, however, is limited by the fact that the humanitarian organisations generally determine the scale, setting and mode of delivery.18 As a result, accessing the services is challenging, and they are not delivered in a systematic or comprehensive approach in conflict settings.9 18 19

For women and newborns, especially in conflict settings, timely and sufficient maternal health services, such as antenatal care (ANC), institutional delivery and postnatal care, are essential.20–22 Maternal health-seeking behaviour (MHSB) is defined as the utilisation of MCH services by women during pregnancy to ensure the well-being of both the mother and the fetus throughout pregnancy, delivery and the postpartum period.23 24 Previous studies usually used MHSB to measure MCH services utilisation,25–28 for they are one of the direct pathways that contribute to the associations between socioeconomic status and health outcomes.29 Studies in Nepal, Afghanistan, Colombia, Burundi, Pakistan and Nigeria have all found significant negative associations between armed conflict and MCH services utilisation, such as ANC, institutional delivery and caesarean sections.7 30–36 However, previous studies have also found that some indicators of MCH services utilisation are positively correlated with armed conflict. Skilled assistance at delivery has been found to be higher among conflict-affected women in Uganda.37 Similar findings were found in Somalia and South Sudan.38 39

Previous studies on this topic have been primarily cross-sectional and focused on the correlation between armed conflict and MHSB. Most are empirical studies based on a single country or conflict event and lack interpretable theoretical models. Using a theoretical model of utility equation and an empirical study in multiple countries’ panel datasets, this study aims at answering the core question: What effects do armed conflict have on MHSB?


Theoretical exploration

Before the empirical study, we constructed an economic utility equation to theoretically project the effect of armed conflict on MHSB (online supplement 1). We found that as the severity of armed conflict rises, the amount of time individual women spend on MHSB decreases, which may lead to a low prevalence of MHSB in conflict regions. And in most areas with low levels of basic health facilities and medical services, MHSB should increase as humanitarian health assistance increases. Based on the theoretical investigation, in conflict-affected undeveloped regions, humanitarian health assistance may be able to lessen or even reverse the devastation of MHSB caused by armed conflict. Therefore, we devised an empirical study to estimate the actual effects of armed conflict on MHSB in selected countries.

Supplemental material


We used maternal health data from the UNICEF Multiple Indicator Cluster Survey (MICS). MICS has completed six rounds of surveys (MICS1–MICS6) worldwide since 1993 for low-income and middle-income countries. Stratified random sampling was applied for data collection. Indicators for demographic and socioeconomic variables, epidemics, child nutrition indicators, sexual and reproductive health, water and sanitation, and mental health can be obtained from this database.40

Our study employed MICS4 (2009–2013) and MICS6 (2017–present) data. The measurement and availability of variables for the two rounds of MICS data were generally consistent. We initially collected and matched the complete data sets of 12 African and Middle Eastern countries in MICS4 and MICS6, including Algeria, Central African Republic (CAR), Chad, Democratic Republic of Congo (DRC), Gambia, Ghana, Guinea-Bissau, Iraq, Madagascar, Sierra Leone, the State of Palestine and Tunisia. We then merged data on armed conflict. Countries included in the study were selected according to the following criteria: (1) MCH indicators in the MICS4 and MICS6 databases were matched; (2) no or few armed conflicts broke out in the country during the year of data collection, while severe outbreaks of armed conflict occurred during the period between the two rounds of data collection and (3) armed conflicts data could be obtained from Uppsala Conflict Data Program (UCDP).3 4 Four countries: Chad (MICS4: 2010, MICS6: 2019), CAR (MICS4: 2010, MICS6: 2018–2019), DRC (MICS4: 2010, MICS6: 2017–2018) and Iraq (MICS4: 2011, MICS6: 2018) were finally involved in this study based on the country selection criteria. Figure 1 shows the standardised BRDs, which reflects the severity of armed conflict, by year in these four countries during the intersurvey period of MICS4 and MICS6. More details about the occurrence of armed conflicts in each country are in online supplement 2.

Figure 1

Interannual variation in standardised battle-related deaths (BRDs) in four countries during intersurvey period of MICS4 and MICS6. CAR, Central African Republic; DRC, Democratic Republic of Congo; MICS, Multiple Indicator Cluster Survey.

Data on MHSB and other variables of interest were retrieved for 100 275 women in these four countries. Women (1) who were aged 15–49 years old; (2) who had given birth within 2 years prior to MICS data collection, were included in the study (figure 2). A total of 55 683 women from Chad (14 058), CAR (8020), DRC (13 363) and Iraq (20 242) were finally included in the study, of which 29 497 samples were from MICS4, and 26 186 samples were from MICS6.

Figure 2

Flow diagram of the data cleaning process. MICS, Multiple Indicator Cluster Survey; UCDP, Uppsala Conflict Data Program.

Patient and public involvement

Patients or the public were not involved in the design, or conduct, or reporting, or dissemination plans of our research.

Conflict data

Armed conflict events data were applied from UCDP Georeferenced Event Dataset Global Version 22.1. The primary sources of UCDP conflict records include global news reports, global monitoring, translation of local news and secondary sources.3 The basic record of armed conflict is an event, which is defined as an organised actor using armed forces against another organised actor or civilians and directly causing at least one person to die at a specific time and place.3 4 Figure 3 shows the geographical distribution of armed conflict events recorded in the UCDP database in the period between MICS4 and MICS6 data collection for the four selected countries.

Figure 3

Armed conflict events distribution of Chad, CAR, DRC and Iraq, MICS4–MICS6. CAR, Central African Republic; DRC, Democratic Republic of Congo; MICS, Multiple Indicator Cluster Survey.

Measures and variables

Armed conflict

This study applied the number of BRDs from UCDP to measure the severity of the armed conflict. BRD refers to the number of deaths directly caused by an armed conflict event. To measure the severity of local armed conflict in each provincial administrative area (province in DRC; region in Chad and CAR; governorate and district in Iraq), the BRD of all armed conflict events that occurred within the provincial administrative areas during the period between MICS4 and MICS6 data collection were added together. The provincial administrative areas used in this study and the actual provincial administrative divisions in the four countries were displayed in online supplement 3-5. Then the total BRD was divided by the number of years (N) and population for standardisation to obtain the standardised BRD per 100 000 people per year:

Embedded Image

There is no one-size-fits-all definition of conflict exposure in conflict epidemiology, and different authors use different definitions according to the context, research design and objectives.41 Previous studies used quintile, trisection or qualitative assessment as criteria to estimate the severity of different groups’ conflict exposure.30 34 41 42 However, as no standard threshold exists,41 we used the median of standardised BRD for all provincial administrative areas as the division of conflict-affected regions (standardised BRD≥0.8) and non-conflict regions (standardised BRD<0.8) in this study.

Maternal health-seeking behaviours

This study investigated six MHSB as the indicators of MCH services utilisation: at least one ANC, at least four ANC visits (ANC4+), at least eight ANC visits (ANC8+), tetanus vaccination, institutional delivery and early initiation of breast feeding (EIB). All six MHSB variables are binary (yes/no). It is worth noting that the newly published WHO ANC guidelines increased the recommended ANC for pregnant women throughout the pregnancy cycle from four visits to eight.43 Considering that the countries included in the study were all low-income and middle-income countries with poor MCH services utilisation, ANC, ANC4+ and ANC8+ were all dependent variables. Tetanus vaccination, institutional delivery and EIB are all defined according to WHO recommendation.44–47

Other covariates

Gross domestic product per capita, nominal (constant 2015 US$), population density (people per km2 of land area), and current health expenditure per capita (US$) were included as national development indicators. Data were adapted from the World Bank48 opening data. National development indicators of all four countries were indicators of the year when MICS were collected or the last year of data collection if the data collection period spanned more than 1 year.

The demographic and socioeconomic characteristics of women and households were also used as covariates in this study. Maternal characteristics included ages (15–19/20–34/35–49), residence (rural/urban), household economic status (quintile wealth index) and maternal education (preschool and lower/primary school/above). Partnership status was used as a covariate; partnered respondents were those married or in long-term cohabitation. The total number of livebirths (1/2/3/4 and 5/above 5) and previous death of a child (yes/no) were also included.

Statistical analysis

We performed descriptive analyses of the distribution of national development indicators, demographics and socioeconomic characteristics, partnership status, reproductive history and conflict-affected regions in the regional sample and all four countries, respectively. All descriptive statistics were weighted. For the regional sample population, the weight used for descriptive statistical analysis was generated by adjusting the population of each country. For country analysis, the original weights from the MICS database were used.

The effects of armed conflict on the utilisation of MCH services were estimated by comparing changes in the natural logarithm of MHSB prevalence for conflict-affected regions to those of non-conflict regions at the time of MICS4 and MICS6 data collection. That is, the estimates were based on a difference-in-difference (DID) model given by:

Embedded Image

In the equation above, n represents different types of MHSB like ANC and EIB. i represents the provincial administrative division, and t represents time (the time of MICS4 or MICS6 data collection). Embedded Image represents the natural logarithm of the prevalence of MHSB in provincial administrative area i and time t, which is the dependent variable in this model. Embedded Image represents a dummy variable that varies with provincial administrative, taking a value of 1 if the provincial administrative area is a conflict-affected region and 0 otherwise. Embedded Image represents a dummy variable that varies with time, taking the value of 0 for the time of MICS4 data collection and 1 for the time of MICS6 data collection. Embedded Image represents a vector of demographic, socioeconomic and other covariates at the country and provincial levels. Embedded Image is a fixed effect of time (before or after the armed conflict outbreak), while Embedded Image is a fixed effect of country and/or provincial district. Embedded Image is the residual term. After controlling for fixed effects of time and fixed effects of national and/or provincial administrative regions and other covariates, the estimate of β can be considered as the effect of armed conflict on MHSB.

In addition to the basic DID model, the propensity score matching (PSM) method was used together with DID to apply a PSM-DID model to meet the premise of the parallel trend assumption. We used all the demographic and socioeconomic characteristics of women and households as matching indicators. In this study, DID and PSM-DID were applied to estimate the effects of armed conflict on six MHSB. Regressions were applied across the sample overall and individual country samples. The country fixed effects, provincial administrative area fixed effects, time fixed effects, national development indicators, maternal demographic and socioeconomic indicators, marital and family status, and childbearing experience were included as covariates in the regression models. Regression at the national level was not applied to CAR, as all provincial administrative regions in CAR are conflict-affected regions.

We also created a categorical conflict variable (low/moderate/high) using the standardised BRD trisection as a threshold. Three DID regression models use two sets of binary variables (low/moderate and low/high) and a three-category variable (low/moderate/high) as exposure variables, respectively, as a robustness check for the prior DID models.

The SEs at the household level were clustered to account for serial correlation (based on the cluster formed by MICS). Multiple imputations were used for missing values to enhance the validity of this study. Notably, MICS6 Chad reported errors in the data collection stage that resulted in a sizeable number of missing values for ANC, prenatal counselling and HIV testing.49 Therefore, in the Chad country analysis, ANC, ANC4+ and ANC8+were not included. All data analyses in this study were performed using Stata/SE V.16.0.50


Descriptive statistics

Armed conflict

Figure 4 presents the standardised BRD distribution of the provincial districts in the four countries. It can be seen that the regions with more severe armed conflict are concentrated in the Lac and BET regions of Chad; the Kasai region in the south and the northeast region of DRC; the western, northern and central regions of Iraq; and the entire territory of CAR. Figure 5 shows the distribution of conflict-affected regions and non-conflict regions in the four countries based on standardised BRD in each of the provincial districts. The identified conflict-affected regions include Sila, BET and Lac in Chad; Bandundu, Bas-Congo, Orientale, Sud-Kivu, Kasai Occidental and Nord-Kivu in DRC; Kerbala, Wassit, Dahok, Erbil, Babylon, Baghdad, Diyala, Kirkuk, Salah al-Din, Ninewa, Thi-Qar and Anbar in Iraq and the entire territory of CAR.

Figure 4

Standardised battle-related deaths in four countries, MICS4–MICS6. CAR, Central African Republic; DRC, Democratic Republic of Congo; MICS, Multiple Indicators Cluster Survey.

Figure 5

Distribution of conflict-affected regions and non-conflict regions in four countries, MICS4–MICS6. CAR, Central African Republic; DRC, Democratic Republic of Congo; MICS, Multiple Indicators Cluster Survey.

Characteristics of participants

Table 1 illustrates the characteristics of the regional sample. Of the 55 683 women aged 15–49 years in the four countries surveyed in MICS4 and MICS6 who had given birth in the last 2 years, 9.91% were adolescents aged 15–19 years, 69.63% were 20–34 years old and 20.46% were 35–49 years old. 23.66% of women received no education or only preschool education, and only 40.03% received lower secondary education or above. More than half of the women lived in rural areas (56.81%), and more than 40% lived in a household with the second poorest (21.65%) and poorest (22.82%) family economic status. Most women in the four countries experienced multiple livebirths: only 18.44% of women had one child and more than half had three or more children. 18.80% of the women experienced previous death of a child. Despite ANC (74.67%) and institutional delivery (74.99%), the prevalence of ANC4+ (43.68%), ANC8+ (6.63%), tetanus vaccination (36.25%) and EIB (39.81%) were all below 50%.

Table 1

Characteristics of women aged 15–49 with children under 5 of the regional sample, MICS4–MICS6

Table 1 also shows the characteristics of the two rounds of MICS cross-sectional data overall and stratified by conflict. Between MICS4 and MICS6, the prevalence of maternal health-seeking activities decreased in the regional sample and non-conflict regions, except for institutional delivery. However, in the conflict-affected regions, the prevalence of ANC, tetanus vaccination, institutional delivery and EIB all increased. The MHSB and all demographic and socioeconomic factors, except for the age group of women living in conflict-affected regions, show significant differences between MICS4 and MICS6 according to the χ2 test. The supplement displays the characteristics of women in four counties (online supplement 6) and the MHSB of women stratified by conflict-affected regions (online supplement 7) and grouped by provincial administrative areas (online supplement 8). online supplement 9-12 show maps of the changes in the prevalence of MHSB between the two rounds of MICS data collection of the four countries at the provincial administrative level.

Statistical analysis

Table 2 presents the regression results for the regional sample. The DID and PSM-DID model results were generally consistent. Using the DID model results as an example: armed conflict had a positive effect on tetanus vaccination (β=0.055, 95% CI 0.004 to 0.106, p<0.05), and had a negative effect on ANC8+ (β=−0.046, 95% CI −0.078 to −0.015, p<0.01). That is, compared with women in non-conflict regions, armed conflict significantly improved the tetanus vaccination of women living in conflict-affected regions by 5.5%, but devastated ANC8+ by 4.6%.

Table 2

Results of the DID regression and for its modified model of Chad, DRC and Iraq

Table 2 also shows the regression results of Chad, DRC and Iraq. PSM-DID results were generally consistent with the DID results. The DID results showed that armed conflict negatively affected EIB in Chad (β=−0.085, 95% CI−0.184 to 0.015, p<0.1). In DRC, no statistically significant associations were found on MHSB. In Iraq, the effects of armed conflict on ANC (β=0.038, 95% CI −0.001 to 0.078, p<0.1) and tetanus vaccination (β=0.059, 95% CI 0.012 to 0.107, p<0.05) were positive, and the effect on ANC8+ (β=−0.039, 95% CI −0.080 to 0.002, p<0.1) was negative. In addition, the PSM-DID results showed a significant positive effect of armed conflict on institutional delivery (β=0.036, 95% CI −0.001 to 0.074, p<0.1).

Results from the robustness assessment were generally consistent with those from the previous DID models (online supplement 13). For instance, in the regional sample, women were 7.3% more likely to receive the tetanus vaccination in regions with high conflict severity than in regions with low conflict severity. And in a country-by-country examination, this beneficial effect maintained. This demonstrates that armed conflict has an actual effect on tetanus vaccination, and other MHSB, notwithstanding categorical differences in armed conflict severity. It is also worth noting that, moderate conflict has adverse associations with multiple MHSB. However, the adverse effect of high conflict on MHSB is usually not as great as that of moderate conflict, and even has a considerable positive effect.


The most important finding of the empirical research is that armed conflict had a positive effect on tetanus vaccination in the regional sample but an adverse effect on ANC8+. In Iraq, the effects of armed conflict on tetanus vaccination and ANC8+ were the same as the regional sample, and armed conflict has a positive effect on ANC. In Chad, the armed conflict harmed EIB. In addition, the results of DID and PSM-DID models demonstrated that armed conflict has no statistical effect on MHSB in DRC.

First, armed conflict was shown to have positive effects on tetanus vaccination, and ANC in Iraq. This finding is consistent with previous studies in multiple countries with armed conflicts. Armed conflict positively associated with specific MHSB in Somalia, Uganda, South Sudan, Syria, etc.37–39 51 Previous studies have implied that although armed conflict may result in a decline in health services, it may also attract humanitarian assistance, which increases the healthcare accessibility and utilisation for women in conflict settings.9 52 Before 2016, in the northeast region of Nigeria, reproductive health, maternal, newborn and child health conditions were poor, and related health services coverage was low. However, after the arrival of international humanitarian assistance due to the conflicts that broke out in 2016, some MCH indicators improved even as armed conflict continued.33 Another study on vaccination in conflict settings found that from 2016 to 2019, the coverage of DPT-3 immunisation in Cameroon fell by 42% as a result of the armed conflict. The crisis is also linked to a serious decline in monitoring indicators and diseases that can be prevented by vaccination. However, after providing routine humanitarian immunisation services, the performance of DPT-3 coverage in 2020 has improved compared with that in 2019 even with the burden of armed conflict and COVID-19.53

Thus, the positive effects of armed conflict on the utilisation of tetanus vaccination, as well as ANC in Iraq, found in this study may be related to the delivery of humanitarian medical and health assistance. Humanitarian assistance data and reports confirm this hypothesis. In Iraq, the ongoing conflict has adversely affected health systems and significantly reduced the quality of medical service available to the Iraqi people.54 However, a profusion of diversified humanitarian funds and assistance emerged as a response to the armed conflict. During the civil war in Iraq, the Iraq Humanitarian Fund (IHF) was established in June 2015, and was responsible for supporting to conflict-affected people.55 Reports showed that more funding and service assistance was provided to the western and northern regions, which were experiencing more severe armed conflict, than to other regions of the country.56 A similar allocation of humanitarian funds was reported in the first quarter of 2017. Although there was still a gap between those supported by IHF and the 8.7 million conflict-affected population reported in the Humanitarian Response Plan,57 58 utilisation of ANC and ANC4+ increased as a result of the investment of humanitarian funds, despite the ongoing conflict. Another study on immunisation in Iraq indicated that children living in high-conflict areas are more likely to receive immunisations against tuberculosis and measles than those living in low-conflict areas. This study suggested that the most likely reason for this finding was the broad presence of international aid organisations in conflict-affected areas, and WHO indeed provided initiatives and programmes to enhance vaccination in these regions.59 This demonstrates, to a certain extent, the effectiveness of humanitarian emergency responses.

Notably, previous studies have also suggested that the higher fertility rates and MHSB prevalence in conflict-affected areas are strategies for coping with armed conflict.37 As extended families provide greater social and economic security, women may consciously attempt to maintain or expand their family size by reducing contraception and adopting behaviours more conducive to fetal health.35 37

Another finding of this study is that armed conflict harmed ANC8+in the regional sample and Iraq. Previous studies have found armed conflict has a negative effect on ANC and other MCH services, due to the collapse of medical and health systems caused by armed conflict.60–62 However, in Iraq, we found that armed conflict has a positive effect on ANC at least one time, but significantly harms ANC8+. We believe that the positive results of armed conflict on ANC were highly related to the provision of humanitarian assistance. Médecins Sans Frontières (Doctors Without Borders) began implementing on prenatal care projects for Syrian refugees and Iraqi displaced persons in 2014.63 But researchers evaluated that even though around 50%–80% women could access to at least one ANC under humanitarian settings, the coverage of ANC4+ remained suboptimal.63 64 Another study discovered that the average number of ANC visits of Syrian refugees in Jordan and Lebanon was about 4–5 times, and the first ANC visit occurred about 3–4 months after conception.64 We believe that humanitarian assistance may be useful in delivering fundamental MCH services, including at least one ANC and tetanus immunisation. Women in conflict-affected regions may find it challenging to undergo ANC8+ throughout pregnancy due to security issues, the collapse of the medical system, and the unsustainability of health assistance. Therefore, armed conflict positively associated with ANC in Iraq but harmed ANC8+, which can be explained by the limitations of humanitarian health assistance in sustainability.

We also found that armed conflict harmed EIB in Chad. This may be due to the lack of breastfeeding-related humanitarian assistance in Chad. A wide range of humanitarian assistance, including food supplies, medical care, temporary camps, clean water and sanitation facilities, was provided through joint efforts made by the international community in Chad.58 However, previous studies found that the EIB prevalence in Chad has been low, and EIB has not been a priority for humanitarian assistance in Chad.65 66 As a result, breastfeeding advocacy and support were scarce and could not reverse the negative impact of armed conflict on EIB in Chad.

The statistically insignificant effects of armed conflict based on DID model results in DRC may be related to the 2016 crisis in the Kasai region, which was triggered by tensions between customary chiefs and the regional government.67 Essential health was poor in DRC, with great disparities in MCH services across different provinces and regions. Until 2016, the persistently unstable northeast and southeast regions received the majority of the country’s humanitarian funding, with over US$80 million invested in healthcare alone.68 In contrast, before the outbreak of armed conflict in 2016, the Kasai region in the south received less humanitarian funding,68 with around 74% of the population living below the poverty line.69 Since the start of the conflict crisis, the lack of humanitarian assistance and corresponding delay of external humanitarian assistance delivery have resulted in a sizeable affected population in the Kasai region.69 70 In 2017, it was estimated that 42% of households were food insecure and 23 500 people crossed the border as refugees.70 With less than 2 years between the outbreak of the Kasai crisis in 2016 and the MICS6 data collection, the health system in 2018 was likely still impacted by the armed conflict. Moreover, the benefits of humanitarian assistance may not have been present at that time or the effectiveness was delayed. Another possible explanation might be the inequality in the distribution of humanitarian health assistance across the country. A previous study focused on eastern DRC found that MCH indicators in these conflict-affected provinces are higher than those in other parts of the country.41 71 72 Researchers also emphasised the significance of humanitarian assistance, which provided these regions with funding for more MCH services.41 73 MCH services are frequently provided without user fees and fully sponsored by non-governmental groups in some conflict-affected areas.41 74 These methods have improved the accessibility of MCH services. The eastern DRC has benefited from humanitarian health assistance for longer than the Kasai region, which may effectively mitigate the adverse effects of armed conflict. It may also help to explain why the armed conflict effect on MHSB in the DRC is not statistically significant at the national level.

In addition, no statistically significant associations were found between armed conflicts with ANC, ANC4+, institutional delivery and EIB in the pooled analysis. There are also many non-significant results in country analyses. On the one hand, this may be because small-scale armed conflicts did not severely impact these MHSB. On the other hand, the effects of armed conflicts and humanitarian assistance on MHSB may be counteracted. More detailed researches are needed in the future.

Previous studies and the discussion here demonstrate how armed conflict both positively and negatively impacts women’s utilisation of MCH services. The emergence of armed conflict has devastating effects on health systems. In the absence of relevant humanitarian assistance, armed conflict may significantly harm the utilisation of health services for pregnant women (eg, ANC8+ in the regional sample and EIB in Chad). However, ongoing armed conflict can also attract humanitarian aid. Sustained investment in humanitarian health assistance in regions with severe armed conflict may help minimise the adverse effects of armed conflict on MHSB and, in some cases, even result in conflict-affected women having more and better access to MCH services than those living in non-conflict regions. In such cases, armed conflict may have a positive effect on MHSB (eg, tetanus vaccination in the regional sample and ANC in Iraq). In conclusion, the effect of armed conflict on MHSB may be mixed. Armed conflict normally has negative impacts on MHSB, but damage can be mitigated and even the prevalence of MHSB can be improved by provision of effective humanitarian health assistance. This mixed effect is also demonstrated by results that compared with high level of conflict, moderate conflict normally has a greater detrimental effect on MHSB (online supplement 13). Regions with high level of conflict usually received greater humanitarian health assistance, which mitigates the harmful consequences of armed conflict.

The strengths of our study include a combination of two high-quality databases, rich covariate data and proposing a simple mathematical model to analyse the effects of armed conflict on MHSB. And the mathematical model was interpretable and consistent with the results of the empirical study.

This study has several limitations. First, as the data on MHSB used in this study were derived from the participants’ self-reports in the MICS database rather than on healthcare records or health insurance files, it is possible that recall bias affected the data collection process. Second, the division of conflict-affected regions and non-conflict regions used in the study was based on a standardised provincial-level BRD. If accurate latitudinal and longitudinal positioning of the sample household addresses were available, an alternative measurement method for identifying conflict-affected regions based on 50 km proximity to armed conflicts could be applied.75 Third, due to the requirements of the DID model, the independent variable used in this study was a dichotomous variable rather than a continuous variable. Even though we used a three-category variable for robustness check, further research is needed to optimise the study with other models. Moreover, we use provincial administrative areas as the units of conflict exposure. This measure may ignore the differences in conflict exposure between different regions within a provincial administrative area. Because one part of a provincial administrative area may be unaffected by conflict while it is concentrated in a particular region.


Armed conflict leads to the destruction of national health systems, deterioration of living environments and decline of health services. As a result, armed conflict has a negative impact on the utilisation of MCH services. However, in conflict-affected regions, where humanitarian actors have mobilised funds and facilities to provide essential humanitarian assistance for those severely affected by armed conflict, MCH service utilisation in conflict-affected regions was found to have significant improvement. Nonetheless, local and international humanitarian assistance goals of saving lives and promoting health remain far from being achieved. First, women living in conflict-affected regions have higher levels of postconflict utilisation of MCH services compared with women in non-conflict regions yet still show a decline in the prevalence of MHSB between two rounds of MICS. Second, armed conflicts do not show statistically significant associations with ANC, institutional delivery and EIB in the regional sample. This suggests that assistance on these dimensions needs to be strengthened. Furthermore, women living in the least developed non-conflict countries and regions of the world need to be addressed, as humanitarian assistance is primarily concentrated in conflict-affected areas, especially regions with high level of conflict severity. To achieve Sustainable Development Goals, policy-makers and humanitarian organisations must pay greater attention to women affected by armed conflict and increase the deployment and operational capacity of humanitarian assistance to conflict-affected regions while also providing more support to people living in underdeveloped regions with low and moderate conflict severity. Especially the promotion of humanitarian assistance on ANC, institutional delivery and breastfeeding services needs to be sustained.

Data availability statement

All data are publicly available. The health data from the Multiple Indicator Cluster Surveys (MICS) can be found publicly (, accessed on 17 October 2022). Armed conflict data used in this study are from UCDP Georeferenced Event Dataset (GED) Global Version 22.1, which are openly available in The Uppsala Conflict Data Program at (accessed on 17 October 2022).

Ethics statements

Patient consent for publication

Ethics approval

This study involves human participants and the health data used in this study were from the Multiple Indicator Cluster Surveys (MICS) led by UNICEF, and can be found publicly (, accessed on 17 October 2022). The survey was held and approved by the Global MICS Team of UNICEF. Participants gave informed consent to participate in the study before taking part.


We appreciate Ms. Rie Takesue’s information about the impacts and mechanisms of armed conflict on maternal health-seeking behaviours in conflict-affected areas. Ms. Takesue works for UNICEF office in the Democratic Republic of the Congo.


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.


  • TZ and QH are joint first authors.

  • Handling editor Seye Abimbola

  • TZ and QH contributed equally.

  • Contributors TZ, QH and KT designed the study. TZ performed all review activities. TZ and QH constructed the theoretical model. TZ and QH extracted the data and performed the statistical analyses. TZ and QH accessed and verified the data. QH contributed to the modification of the empirical research model. TZ produced the descriptive analysis. TZ, QH, SR and KT contributed to the writing and review of subsequent versions of the manuscript. TZ and QH contributed equally to this article. KT served as the guarantor for this work, assumed full responsibility for the overall content, the conduct of the study, access to the data, and the decision to publish. All authors reviewed and approved the final version before submission.

  • Funding This study was funded in full by the National Natural Science Foundation of China, grant number 72074130.

  • Map disclaimer The inclusion of any map (including the depiction of any boundaries therein), or of any geographic or locational reference, does not imply the expression of any opinion whatsoever on the part of BMJ concerning the legal status of any country, territory, jurisdiction or area or of its authorities. Any such expression remains solely that of the relevant source and is not endorsed by BMJ. Maps are provided without any warranty of any kind, either express or implied.

  • Competing interests None declared.

  • Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.