Bacterial meningitis (BM) is still considered an event of interest in public health as it is related to high lethality rates and multiple neurological sequelae (1). This entity, among other infectious diseases, represents a relevant proportion of the burden of disease in low-income countries and is among the top causes of years lived with disability and premature death in both children and adults (2–4). This condition also has higher morbidity rates in vulnerable populations and at extreme ages (5). Likewise, BM represents a threat due to the potential severity of the clinical course, its tendency to generate outbreaks and epidemics in specific geographic regions (6), and the costs associated with its treatment [1.6 billion USD in 2006 (7), or up to 34% of the gross domestic product per capita (8)]. It is the sixth most lethal infection in the world and the third in newborns (9).

According to the GIDEON (Global Infectious Disease and Epidemiology Network), in 1997, there were 1,200,000 bacterial meningitis cases worldwide (41.66% in Africa) with an 11.35% lethality (10, 11). Over the last decades, high and upper-middle-income countries have relatively succeeded in reducing BM frequency by implementing standardized epidemiological surveillance systems and mass vaccination campaigns against the most frequent etiologic agents (3, 12). On the contrary, a region located in Sub-Saharan Africa, comprising more than twenty countries, is known as the African Meningitis Belt (AMB) because of the ongoing recurrence of BM outbreaks and epidemics. The Global Burden of Disease (GBD) estimates that almost 4 out of 5 cases in Africa are registered in this region. For 2016, this region accounted for >45% of cases worldwide (13).

Bacterial meningitis (mainly caused by N. meningitidis, H. influenzae type B, or S. pneumoniae), is a communicable disease associated with different climatological and social characteristics (14). In North America and Europe, BM cases peak during the winter (15–17), while in the AMB there is an evident seasonality as most patients demonstrate symptoms during the dry season (around the 6^th epidemiological week) (18, 19). Sub-Saharan Africa has high temperatures due to its relative distance to the Equator and geographical proximity to the desert, which also influences the air concentration of particulate matter (mineral dust) (20).

Studies addressing socioepidemiological features were mainly published between 1970 and 2000 and carried out in Europe (16, 21, 22), Oceania (23, 24), and Africa (25–27). In high-income countries, researchers like Haneberg et al. (28) identified distinct exposures including psychologically stressful events. During the meningococcus outbreak in Wales (2000), Fitzpatrick et al. evaluated the association of BM with psychological stressors including the death of a family member, moving home, traveling, and receiving bad news; after the multivariate adjustment, the latter remained associated with a meningococcus carrier state (22). In English preschoolers, BM incidence increased with recent exposure to dust, household overcrowding, and parental involvement in marital conflicts or legal issues (16). These results are consistent with associations found by Jones et al. regarding social deprivation, unemployment, and lack of house ownership (21).

Environmental exposures, for example, passive smoking (16, 23), alcohol abuse or dependence (29, 30), pacifier sharing (23), breastfeeding (24), and vitamin A supplementation (31) have also been proposed to be linked to BM.

Complementarily, Greenwood et al. analyzed a BM epidemic in Nigeria (1977) and suggested a possible association between incidence and socioeconomic status because few cases occurred in inhabitants with high-income (25). They also investigated variables such as social class, occupation, literacy, and nutritional state, among others, in the BM outbreak at The Gambia (1982) and found that the density of cattle was higher in regions that were not affected by the disease (26). This index has been considered a surrogate feature in the context of poverty (32). Subsequent publications have described the importance of social determinants of health on BM incidence but mainly from the individual scope. In 2001, Hodgson et al. applied a case-control design to analyze risk factors for meningococcal meningitis in survivors of the epidemic in the north of Ghana (1997); significant associations were found with exposure to smoke produced by firewood stoves and sharing the bedroom with a case (27).

Although different studies have already demonstrated that socioepidemiological conditions are associated with BM in individuals from the AMB, one could question why the cumulative incidence importantly diverges between two Sub-Saharan countries that share land borders. In other words, what are the unique conditions that the AMB populations face that make them more vulnerable to BM in contrast to the rest of Africa?

The main objective of this research was (1) to identify country-level statistics that contribute to explaining the differences in BM cumulative incidence between AMB and non-AMB African countries. For this purpose, we also (a) described the incidence of BM (attributed to N. meningitidis, H. influenzae, and S. pneumoniae) in the AMB compared to that in non-AMB African and non-African countries and (b) studied the epidemic curve per AMB subregion and the relationship with climate variables. As a secondary objective, we aimed (2) to determine if the same features were associated with the number of BM incident cases per country worldwide. We hypothesized that climate variables and country-level statistics related to poverty (i.e., unmet basic needs) were the main socioepidemiological macro-determinants of BM in both Africa and worldwide.

This is a completely-ecological study with a multiple-group design that was conducted as a secondary analysis of aggregate, environmental, and global measures (BM incident cases in 2016 and country-level statistics) obtained mainly from the records of the GBD, the MenAfriNet Consortium, the World Bank and the World Health Organization (WHO).

According to Morgenstern, ecological studies “focus on the comparison of groups, rather than individuals” by analyzing aggregate (i.e., proportion of smokers), environmental (i.e., air-pollution levels), and global measures (i.e., population density) (33). Levels of analysis can be classified as completely-ecological (all variables are aggregate measures), partially-ecological (some joint distributions are known), and multilevel (both individual and aggregate measures of interest are available). There are three different designs relevant to ecological studies: multiple-group (“comparison of disease rates among regions during the same period”), time-trend (“comparison of disease rates over time in one geographically defined population”), and mixed study (a combination of the previous two) (33). A detailed description of ecological studies has been published elsewhere (34–36).

The main three advantages of ecological studies are the study of exposures of interest from the collective scope (social and environmental dynamics), the quick obtention of results at a low cost allowing preliminary decision-making in public health, and the generation of hypotheses that must be confirmed or refuted with further research (37).

With this research, we analyzed socioepidemiological macro-determinants that possibly explain some of the differences between the AMB and non-AMB African countries. Our results support previous findings from studies at the individual and subnational levels, as similar variables were significant from the ecological scope. Population research conducted in specific AMB countries has shown 2.1 times higher incidence in areas with high population density, 34.8 times in those with very high levels of absolute poverty, and 41.6 times where the inhabitants exhibit low literacy rates (57). It has also been proposed that the use of closed kitchens with wood ovens, unhealthy housing, and low household income increase the likelihood of BM (27).

In middle-income countries from America, some other studies also recognize the relevance of socioepidemiological conditions on BM incidence. Molina et al. unveiled a possible association with air pollution sources in Cuba and reported higher incidences in municipalities with elevated population density and high percentages of infants and/or elders. These associations varied among cases attributed to N. meningitidis, H. influenzae, and S. pneumoniae (58). In Brazil, Fonseca et al. determined a relationship between BM incidence and poor living conditions (low income, housing in poor areas, literacy, access to drinking water, and household occupancy) (59). Except for the population density, the rest of the micro-determinants found in these studies were also significant at the country level whether in the bivariate post hoc or the multivariate analyses pursued here.

The population at the extremes of life has a greater susceptibility to developing BM due to the characteristics of their immune system. As synthesized by Simon et al., in newborns, the monocyte line is immature (60), which leads to impaired phagocytosis, inadequate response of natural-killer cells, poor cytokines and chemokines secretion, and limited concentrations of interferons. Moreover, the levels of complement factors are lower than in adults, and there is a high tolerance to self-antigens (61). In older adults, there is immune senescence characterized by absolute lymphopenia, decreased phagocytic and cytotoxic ability of neutrophils, and abnormal cytokine production, among others (62), etc.

In this sample, the main variables that make the AMB a distinct region and partially explained BM incidence worldwide were the temperature, gross national income per capita, household occupancy, and malaria incidence.

The annual median temperature in the AMB (27.56 °C, IQR 25.67–29.11) is not only higher than that in the other continents but also in the rest of Africa (~4 °C). Sub-Saharan African territories with high rates of BM demonstrate high temperatures and this seems related to their geographical location (intertropical convergence zone) (20). Likewise, it has suffered the impact of climate change due to the El Niño phenomenon (southern oscillation, ENSO), which favors the appearance of droughts (63); it is expected that the number of months in summer with temperatures above five standard deviations for the year 2100 would be 75% and up to 90% in west Africa (64).

Both temperature and rainfall are fundamental to understanding the dynamics of meningococcus since it has been demonstrated it can survive outside the human body, for up to 8 days, in glass, plastic, or cloth under favorable conditions of 30°C (65) and 22% relative humidity (66). The retrieved data show that temperature has a poor correlation with the volume of rainfall (ρ = 0.14, p = 0.052), while there is a clear relation with relative humidity (ρ = −0.29, p = 0.013). The association of this variable with the incidence of BM is also greater in comparison to that of rainfall (IRR 0.95 95% CI 0.93–0.96, p = 0.000 vs. 0.99, 95% CI 0.99–9.99, p = 0.000). This suggests that a fraction of the risk is attributed to low humidity beyond what is expected due to low volumes of rainfall during the dry season (Figure 4).

At the meteorological level, the beginning of an epidemic in the AMB coincides with a greater velocity of the dry winds (Harmattan) coming from the north and dragging mineral dust (a fraction of particulate matter 2.5) from the Sahara (6^th±2 epidemiological weeks, R² = 0.85) (18). This source of pollution is significantly concentrated in the AMB (40.04 μg/m³, IQR 32.68–50.05), where it is 1.5 times that of the rest of Africa. In Nigeria, the most populous AMB country, PM 2.5 was associated with 50,900 deaths (35,700–73,200) in 2015, while worldwide 675,000 (492,000–889,000) deaths attributed to lower respiratory tract infections were associated with air contaminants (67).

It is believed that detrimental changes in temperature, precipitation, and relative humidity favor colonization by meningococcus (68) and pneumococcus (69) after epithelial injury to the oral and nasopharyngeal mucosa, facilitating the translocation of the pathogen to the brain through the olfactory nerve. Mueller et al. synthesize the impact of the environment by proposing a model in which dusty weather increases bacterial invasiveness at the community level by up to 100 times (hyperendemicity) after the rainy season. Subsequently, an epidemic cofactor such as viral respiratory infection facilitates the carrier state and increases the incidence, generating localized epidemics during the dry season (70).

The relationship of BM incidence with the gross income per capita is an indicator of poverty and social deprivation as has been informed by multiple authors (21, 25, 57, 59). Intra-household and social overcrowding are also well-known determining factors that increase the risk of infections such as BM (27), tuberculosis, pneumonia, typhus, gastroenteritis, scabies, etc (71). In Auckland, Baker et al. found that two additional adolescents or adults in a 6-room house double the risk of children contracting meningococcal meningitis (23).

The incidence of malaria in west Africa increases during the dry season, similar to BM (72). Furthermore, it seems that higher temperatures increase the incidence of malaria since they facilitate the growth of the parasite and the survival of the infected mosquitoes (Anopheles sp.) (73). In the multivariate regressions, incident cases of BM were correlated with those of malaria, pointing to co-infection, as previously observed in Burkina Faso, where 11.8% of meningitis or bacteremia patients were infected as well by Plasmodium sp. (74). Both cases of cerebral malaria (p = 0.031) and BM (p = 0.014) have been associated with a deficiency of zinc levels compared to healthy controls from Ghana (75). We considered and collected the prevalence of zinc deficiency by country as an additional variable; in the univariate regression, it was associated with both the incidence of BM (IRR 1.05 CI 95% 1.04–1.05, p = 0.000) and malaria (IRR 1.04 CI 95% 1.03–1.04, p = 0.000). Unfortunately, this design is not adequate to define the impact of zinc deficiency as a common pathophysiological mechanism in these entities.

Given the low proportion of lumbar punctures, it could be possible that BM cases without a CSF microbiological study were actually caused by enteric pathogens whose route of transmission (fecal-oral) is related to the use/access to sanitation services (76) (an unmet basic need at the AMB). Particularly, group B streptococcus (5) is transmitted through this route and is considered the most common cause of BM in infants aged < 3 months (77). Even more, Berkman et al. carried out a study in Turkey, where 37% of specimens were positive for Enterobacter sp. or Klebsiella sp. while S. pneumoniae was found in only 6.5% (78), and in a series from Ghana, E. coli (3.4%), Salmonella sp. (3.4%), and Pseudomonas sp. (2.5%) were the most frequent bacteria after S. pneumoniae (77.7%) (79). Enterovirus can cause up to 80% of viral meningitis cases (54.6% of infectious meningitis) (80). From the populational perspective, improved sanitation and drinking water access are variables inversely associated with BM as a cause of death [see the GBD study (13)]. This has been also suggested in the past by Coulehan et al. who studied BM in Navajo Indians (81).

In this study, it was especially striking that immunization coverages against pneumococcus [effectiveness >90% (82)] were removed from the regression models since these are evident determinants of the individual risk for suffering BM. Unfortunately, we could not retrieve the coverage for the meningococcal vaccine of most high-income countries (83). Even though access to vaccination in Africa has increased by 38 times and the implementation of more than 20 mass immunization campaigns (268 million doses) has reduced the incidence of meningococcal serogroup A to a tenth, BM continues to be a critical public health problem (84). Due to the latent risk, the WHO and the Global Alliance for Vaccines and Immunization (GAVI) have determined that, as with cholera and yellow fever, it is necessary to increase the reserves of vaccines against meningococcus and ensure transportation logistics (85).

Regarding the study's limitations, it is not possible to make accurate inferences on the individual level due to the assumptions of the design (ecological fallacy). It was also complex to control for all potential confounders because important information (i.e., vaccination rates against N. meningitidis) were not available for every country. Also, the temporality of variables for every country might have been a source of bias (33, 36). Data was compiled from multiple sources, implying a variable information quality, and there was an important proportion of missing data as a minority of variables were available for every observation.

According to the MenAfriNet-WHO bulletins, lumbar puncture was performed for less than half of the AMB cases and it was not possible to reach a higher level of certainty about the etiological agent culprit of the clinical manifestations. This represents an information bias that depends solely on the available information. Similarly, the incident cases estimated by the GBD are calculated through meta-regression from different types of primary sources and it is also not possible to guarantee a microbiological confirmation. In the same sense, the calculation of BM incidence did not consider cases attributed to Listeria monocytogenes because this specific information is not recorded by the GBD; as this bacteria is responsible for an important number of cases in elders, future research will need to address this matter.

On the other hand, using multiple international databases allowed us to build a comprehensive panorama with diverse variables surrounding the BM phenomenon. By combining an evidenced-based rationale with a strict stepwise statistical analysis of multicollinearity, we attempted to reduce the effect of confounders [Simpson's paradox (86)]. The main strength of this study lies in addressing the BM public health challenge from a worldwide and continental ecological perspective by providing insights regarding a disease that has been mostly studied under traditional observational designs, which are susceptible to the individualistic fallacy. From this scope, the study confirmed most of the previous individual and national results and hopefully will inspire the development of novel hypotheses as some relevant determinants usually do not vary enough to be found as significant within populations, but they do when explored between populations (87). The results are also important for policy-making and reaffirm the relevance of social and structural determinants of health as its impact is clear on diseases such as BM that are “deadly, expensive and preventable” (38).

Future research should confirm or reject these findings by incorporating additional variables in multilevel designs that contribute to understanding better the link between epidemiological and pathophysiological mechanisms of BM.

Consistent with the literature, these ecological results reinforce the relationship between a high cumulative incidence of BM in the AMB and socioepidemiological macro-determinants. Countries located within this region of Sub-Saharan Africa suffer from social inequity, poverty, unmet basic needs, low access to vaccines, and higher rates of other infections. From the environmental point of view, the AMB is characterized by high temperatures, low humidity, and increased levels of mineral dust contamination, which favor epidemic waves of the disease during the dry season, mainly in the western subregion. Modifiable socioepidemiological macro-determinants associated with BM cumulative incidence should be targeted by optimal public health policies.

The original contributions presented in the study are included in the article/Supplementary material, further inquiries can be directed to the corresponding author.

GP-M: study conception and design, data collection, and analysis and interpretation of results. GP-M, NL-L, MG, EM-H, JR-S, and JQ-B: draft manuscript preparation. All authors reviewed the results and approved the final version of the manuscript.