The analytic database included 85,374 women and 71,864 children living in four conflict-affected states with a total of 297,873 organized violence events and 1,702,818 associated fatalities. Table S1 provides the detailed sociodemographic characteristics of the analyzed sample (see Additional file 2).
Sub-national variation in the quality of PHC services
The average quality index was similar at the national level in the four studied countries ranging from 45.5% in DRC to 53.3% in Cameroon. However, sub-national variation in PHC quality was observed at the first administrative level. Mali had the most considerable sub-national variation, where the quality index ranged from 16.8% in Kidal state to 56.4% in Segou state. In DRC, the overall quality of PHC services ranged from 36.2% in Maniema province to 61.8% in the capital Kinshasa. Similarly, in Nigeria PHC quality index ranged from 30.8% in Taraba state to 63.3% in Benue state. In contrast, the lowest sub-national variation in PHC quality was observed in Cameroon, with an index ranging from 44.6% in the North region to 59.7% in the Douala region. Visually, the relationship between conflict location and the variation in the PHC quality index could not be fully established at the first administrative level (Fig. 1).
The relationship between conflict intensity and each component of the PHC quality index was assessed at the household cluster level. Generally, both categories of conflict intensity had low quality scores in most assessed components; however, the four countries had similar or sometimes better results in the individual quality indicators in clusters surrounded by medium or high-intensity conflict compared with clusters surrounded by no or low-intensity conflict (Fig. 2). This observation was more pronounced in DRC, where five out of the six indicators constituting the PHC quality index had better results in medium or high-intensity conflict than in no or low-intensity conflict.
Disparities in individual quality of care index components:
Disparities in informed choice
Overall, women with no education or in the first economic quintile were less likely to have informed choice than women with secondary or more education or in the fifth economic quintile (see Additional file 2: Tables S2 and S3). For example, Congolese women with no education had 17% less likelihood of informed choice than women with secondary or more education. Similarly, Nigerian women in the lowest economic quintile had 10% less likelihood of informed choice than women in the fifth economic quintile. Meanwhile, point estimates from the concentration index showed that disparities were generally low (concentration index < 0.10) and statistically insignificant when all sub-groups of educational or economic status are considered. Furthermore, adjusting for additional sociodemographic variables (using random intercept models) showed statistically insignificant disparities in the majority of the studied contexts. Statistically significant odds of educational disparity were only observed in Nigeria, where women with secondary or more education had 1.7 times the odds of informed choice than women with no education (p = 0.004). Meanwhile, statistically significant odds of economic disparity were only observed in DRC, where women in the fifth economic quintile had 12.5 times the odds of informed choice compared to women in the first economic quintile (p = 0.026). When considering the intensity of conflict surrounding household clusters, only Congolese and Nigerian neighborhoods had higher educational disparities in clusters surrounded by high or medium intensity conflict compared to those surrounded by no or low-intensity conflict (absolute difference = 25.1% and 17.2% vs. 9.1% and 8.1% for DRC and Nigeria respectively) (see Additional file 3: Table 1 and Additional file 4: Table 2).
Using the concentration index, no statistically significant difference in disparities was found when comparing neighborhoods by conflict intensity in any of the studied contexts (see Additional file 3: Table 1 and Additional file 4: Table 2).
Adjusting for additional sociodemographic variables using random intercept models indicated the statistical significance of educational disparities in Nigeria in both types of neighborhoods. In contrast, no statistically significant odds ratios of economic disparities were observed in any of the studied contexts. Furthermore, none of the four studied contexts had a statistically significant interaction between conflict intensity and either educational or economic disparities in informed choice (see Additional file 3: Table 1 and Additional file 4: Table 2).
Disparities in the quality of ANC
Generally, women with secondary or more education or in the fifth economic quintile had a better quality of ANC than women with no education or in the first economic quintile. The three measures of disparities consistently showed educational and economic inequalities in the majority of the studied contexts (see Additional file 2: Tables S2 and S3). For example, 40% of Congolese women with secondary or more education had quality ANC versus only 18% of women with no education. The concentration index also showed the aggregation of quality ANC among the most educated Congolese women (concentration index = 0.21, p < 0.001). Similarly, adjusting for additional sociodemographic variables showed that Congolese women with secondary education have 60% higher odds of quality ANC than women with no education (p < 0.001).
When household clusters were classified by conflict intensity, DRC, Mali, and Nigeria showed a higher or similar magnitude (using absolute difference) of economic and educational disparities in neighborhoods surrounded by medium or high-intensity conflict versus those surrounded by no or low-intensity conflict (see Additional file 3: Table 1 and Additional file 4: Table 2). For example, the magnitude of economic disparities among Malian women living in medium or high was 26.5% versus 19.4% in no or low-intensity conflict. On the other hand, Cameroonian women living in no or low-intensity conflict neighborhoods had a higher magnitude of economic and educational disparities in quality ANC versus those living in medium or high-intensity conflict neighborhoods (absolute difference = 11.5 vs. 25.8 and 20.0 vs. 30.6 for educational and economic disparities respectively).
When all sub-categories of educational and economic status were considered using the concentration index, the four contexts showed a statistically significant concentration of quality ANC among the most educated and the wealthiest in both types of clusters (see Additional file 3: Table 1 and Additional file 4: Table 2). Point estimates and Z tests showed a statistically significant difference between the two types of neighborhoods in Cameroon and DRC only, where lower educational and economic disparities were observed in high or medium intensity conflict neighborhoods in Cameroon (z = − 2.31, p = 0.021 and z = − 2.63, p = 0.009 respectively), while higher educational and economic disparities were observed in the medium or high-intensity conflict in DRC (z = 2.38, p = 0.017 and z = 3.85, p < 0.001 respectively) in alignment with the results obtained using the absolute difference.
In contrast, adjustment for additional sociodemographic variables in medium or high-intensity conflict (using random intercept model) showed a statistically significant odds ratio of educational disparities only in Nigeria and economic disparity only in DRC. Meanwhile, the odds ratios of educational and economic disparities were statistically significant in low-intensity conflict neighborhoods in the majority of studied contexts. Testing for interaction between conflict intensity and either educational or economic disparities was statistically insignificant in the four studied contexts (see Additional file 3: Table 1 and Additional file 4: Table 2).
Disparities in immunization dropouts
At the national level, children of mothers with secondary or more education or in the fifth economic quintile had lower dropout rates compared to children of mothers with no education or in the first economic quintiles, respectively, in almost all the studied contexts. For example, a considerable magnitude of economic and educational disparities in dropout rates was found in Nigeria for all dropout indicators. Dropout rates were 19.7%, 15.4%, and 17.8% for BCG-measles, DPT1-DPT3, and DPT1-measles, respectively, among Nigerian children whose mothers had secondary or more education versus 38.7%, 37.0%, and 34.4% among those with no education. Similarity dropout rates were 11.4%, 12.4%, and 10.5% among children whose mothers were in the fifth economic quintile versus 39.1%, 36.9%, and 34.8% respectively those in the first economic quintile).
When all sub-groups of educational and economic status were considered using the concentration index, a significant concentration of dropouts among the least educated or the least wealthy in Cameroon, DRC, and Nigeria was also observed. Nigeria was also the country with the largest economic and educational inequalities using the concentration index. Meanwhile, point estimates and tests of significance consistently showed the absence of either educational or economic disparities in all three types of immunization dropouts in Mali, with overall lower dropout rates compared to the other three countries.
Economic and educational disparities in dropout rates were observed in most of the studied contexts even after adjusting for additional covariates as child gender, mother age, mother employment, urban/rural status, and the number of children in the family. Nigeria was the only country consistently showing statistically significant odds ratios of both economic and educational disparities in the three dropout indicators (see Additional file 2: Tables S2 and S3).
When the intensity of conflict surrounding household clusters was considered, mothers of children living in clusters surrounded by the high-intensity conflict had similar (less than 5% difference), or higher magnitude of educational disparities in the four studied contexts and similar or higher magnitude of economic disparities in Cameroon and Mali. Meanwhile, mothers of Congolese and Nigerian children living in clusters surrounded by no or low-intensity conflict had a higher magnitude of economic disparities in dropout than those living in medium or high-intensity conflict clusters (absolute difference = 20.68 versus − 1.91 and 26.49 versus 17.74 for DRC and Nigeria, respectively (see Additional file 3: Table 1 and Additional file 4: Table 2).
When the concentration index was calculated, there was a significant aggregation of immunization dropouts only among Cameroonian and Nigerian children whose mothers were least educated or the least wealthy and living in neighborhoods surrounded by medium or high-intensity conflict. On the other hand, mothers of children living in neighborhoods surrounded by no or low-intensity conflict had statistically significant economic and education disparities, mostly in Cameroon, DRC, and Nigeria. Mali was also the only country consistently showing a lack of educational and economic disparities in the three immunization dropout indicators.
Meanwhile, no statistically significant difference was found between neighborhoods of different conflict intensity in either economic or educational disparities for immunization dropout in almost all the studied contexts (see Additional file 3: Table 1 and Additional file 4: Table 2). The only exception was observed in DRC, where economic disparities in DPT1-DPT3 dropout were statistically significantly higher in no or low intensity conflict clusters (z test = 2.48, p = 0.013).
Meanwhile, adjusting for additional sociodemographic variables using random intercept models revealed more statistically significant and higher odds ratios of educational and economic disparities in no or low-intensity conflict clusters than in medium or high-intensity conflict clusters. However, no statistically significant interaction between conflict intensity and educational or economic disparities existed in Cameroon, Mali, or Nigeria. Only DRC showed a statistically significant interaction between economic disparities and conflict intensity in DPT1-DPT3 dropouts (t = 2.28, p = 0.023), confirming the results obtained by comparing concentration indices (see Additional file 3: Table 1 and Additional file 4: Table 2).
Disparities in the quality of diarrhea case management
Generally, children of mothers with secondary or more education or in the fifth economic quintiles were more likely to have their diarrhea managed according to guidelines than children of mothers with no education or in the first economic quintile, respectively (see Additional file 2: Tables S2 and S3). Using absolute difference, the largest inequality gap was observed in Cameroon, where 47.4% of Cameroonian children whose mothers had secondary or more education had their disease managed according to guidelines versus 20.6% of children whose mothers had no education. Meanwhile, a narrower disparity gap was found in Mali, where children with diarrhea had an absolute difference of only 1.0% and 2.3% for educational and economic disparities, respectively. However, the overall quality score was too low to detect a meaningful difference between sub-groups.
Similarly, point estimates of the concentration index indicated higher economic and educational disparities in Cameroon and lower economic and educational disparities in Mali compared to the other studied contexts. However, adjustment for additional sociodemographic variables using random intercept models indicated the presence of statistically significant economic disparities in Mali, where children of mothers in the fifth economic quintiles had 3.0 times the odds of having their diarrhea managed according to guidelines compared to those whose mothers were in the first economic quintile (p = 0.041).
When the intensity of conflict surrounding household clusters was considered, children living in neighborhoods surrounded by medium or high-intensity conflict had higher levels of economic and educational disparities than children living in no or low-intensity conflict neighborhoods in Cameroon, DRC, and Mali (see Additional file 3: Table 1 and Additional file 4: Table 2). Meanwhile, Nigerian Children faced higher absolute educational disparities in low or no intensity conflict clusters versus medium or high-intensity clusters (absolute difference = 10.4 versus 0.6 respectively).
When all sub-groups of educational and economic status are considered using the concentration index, there was an aggregation of quality management of diarrhea among children whose mothers were more educated or wealthier in Cameroon and Nigeria in both types of clusters (see Additional file 3: Table 1 and Additional file 4: Table 2). Meanwhile, no or very low level of disparity was observed in DRC and Mali using a concentration index. Point estimates generally indicated higher disparities in medium or high-intensity conflict than in no or low-intensity conflict neighborhoods. However, differences were statistically insignificant except for economic disparities in Cameroon (Z = 2.3, p = 0.02).
Adjustment for additional sociodemographic variables also revealed similar or higher odds ratios of economic and educational disparities in medium or high-intensity conflict clusters than in no or low-intensity clusters in most studied contexts (see Additional file 3: Table 1 and Additional file 4: Table 2). A statistically significant interaction between conflict intensity and economic disparities was also observed in Cameroon (t = 3.28, P = 0.001), confirming the results obtained by the concentration index. Meanwhile, Nigeria was the only country where adjustment for additional sociodemographic variables revealed higher educational disparities in no low-intensity conflict clusters than in medium or high-intensity conflict clusters with statistically significant interaction between conflict intensity and educational disparities (t = 0.260, p = 0.002).