Bayesian spatial modelling of malaria burden in two contrasted eco-epidemiological facies in Benin (West Africa): call for localized interventions

The study was carried out in the Ouidah–Kpomassè–Tori Bossito (OKT) and Djougou–Copargo–Ouaké (DCO) health districts in Benin, West Africa in 2011, [18]. OKT is located in Southwestern Benin and 50 km from Cotonou (Fig. 1).

DCO is located in northern Benin and is at a 381 km distance from Cotonou (Fig. 1). The surface areas of OKT and DCO are 932 km² and 5,505 km², respectively. The commune of Ouidah, Kpomassè, Tori Bossito, Djougou, Ouaké and Copargo are located at 5 m, 27 m, 42 m, 421 m, 654 m and 396 m altitude respectively. Temperatures range between 22° C and 35° C, with an average of 27° C. In the south, the primary rainy season is from March to July; there is a short dry season from July to September and a short-wet season from mid-September to mid-November. Northern Benin only presents one rainy season (May to September, with most rain in August) and one hot, dry season.

A location map of the two study areas relative to the country was already elsewhere [18]. The OKT and DCO populations were 286,711 and 411,835, respectively, in 2013 [19]. Under-five years old were around 17% of the total population [20]. The characteristics of these areas have been previously described [7, 18]. The data from the 2020 Benin statistical health showed that malaria prevalence was 87% in the Donga department (DCO health zone) and 24% in the Atlantique department (OKT health zone), [21].

Data were extracted from a cross-sectional survey cluster design during the rainy season as described elsewhere in 2016 [18, 22]. A two-stage random sampling technique was applied. The inclusion criterion for villages was a population size of 1000–1800 inhabitants. The target population was children aged 0–60 months, living in the selected villages, whose parents gave their informed consent. Thirty-one and 42 villages were randomly selected in the OKT and DCO health zones, respectively. Each village’s geographic coordinates were recorded using a global positioning system (GPS) provided by Benin’s Institute of National Geography (ING). In Benin's health pyramid, the local subdivision unit where health indicators are calculated before being aggregated at the regional and national level is the health zone. It is also at this level that decisions are taken to improve the health conditions of the population.

A total of 10,367 children aged 0 to 60 months were included. In OKT, 4,348 children were recruited from 31 villages. The median age was 29 (1^stqtle = 14; 3^rdqtle = 45). In DCO, 6,019 children were included from 42 villages. The median age was 29 (1^stqtle = 12; 3^rdqtle = 46). The male /female ratio was 1:1 and 1.1:1.0 in OKT and DCO, respectively.

P. falciparum and P. malariae species were present in both areas. In OKT, among 198 positive thick films, 82.3% were positive with P. falciparum, 16.7% were positive with P. malaria and 0.5% co-infection with P. falciparum + P. malaria species. In DCO, among 331 positive thick films, 96.3% were positive with P. falciparum, 2.7% positive with P. malaria and 0.6% co-infection with P. falciparum + P. malaria species. The mean parasite density of P. falciparum in children was 1113 (CI95%). As P. falciparum was a dominant species, the following analysis focused on this species.

The results of non-spatial and spatial models fitted to data revealed that the prevalence of malaria infection is affected by environmental and demographic exposures within the two regions (OKT and DCO). Exposures such as annual rainfall "bio12_wc30s", moisture index of the moist quarter “mimq_wc30s’’, distance to the coastlines “dst_coastlin”, topography “srtm_topo” and land cover influence significantly the prevalence of malaria infection within the OKT health district since the 95% credibility intervals of their mean estimates do not contain 0 (Table 2). These risk factors except the land cover showed a significant effect on malaria prevalence when considering the Bayesian binomial model without spatial component (Supplementary file 4 and Supplementary file 5). For the prevalence of malaria cases in the OKT region, moisture index of moist quarter “mimq_wc30s’’, moisture index of arid quarter and topography covariates were significant as all of their 95% credibility intervals do not contain 0.

Regarding the DCO region, apart from the local mean prevalence (i.e. the intercept computed at the original scale), the population density "ben_ppp" is the only risk factor that significantly affects the prevalence of malaria infection. None covariate significantly affects the estimated prevalence of malaria cases since the 95% credibility interval of their effects contains 0 (Table 3). Contrarily to the OKT region, only GLM model showed significant effects of covariates on the estimates of malaria cases in DCO region (Supplementary file 3).

Bayesian Generalised linear spatial models implemented using stochastic partial differential equations (SPDE) combined with the INLA approach revealed that the prevalence of malaria infection varies locally within villages throughout the two study regions (Figs. 5 and 6). Malaria infection prevalence is predicted to be mesoendemicity or hyperendemicity (between 40 and 60%) in most locations in the study region, namely the villages within Ouidah and Tori Bossito communes. In comparison, a patch of holoendemicity (> 75%) is predicted to occur in the north-eastern part of Tori Bossito and the western part of Kpomassè (Fig. 5a). Globally, the malaria infection is expected to be low at the south of Kpomassè and Ouidah communes in the OKT health district, but with a higher standard deviation of about 20 to 28% especially in Ouidah (Fig. 5c). Regarding the prediction of the prevalence of malaria cases, moderate prevalence values occurred within the three communes of the OKT health district, with more importance in Tori Bossito than Ouidah and Kpomassè (Fig. 5b). A relatively low standard deviation of the estimates (1.6 to 14.0%) has been observed for the prevalence of clinical cases in the region with the highest values in Ouidah (Fig. 5d).

A contrarily, in the DCO region, a very high level of malaria infection was observed throughout the study area. Some areas located in the central and northern part of Djougou and Copargo communes had an estimated prevalence between 60 and 75% (i.e. hyperendemic areas), and others (extreme north of Djougou) had an estimated prevalence greater than 75% (i.e. holoendemic areas). However, some areas located at Ouaké and the south-eastern part of Djougou in the DCO health district has a mesoendemic transmission of malaria infection (10 to 50%) (Fig. 6a). The predicted prevalence of malaria cases was relatively high (40 to 65%) in some locations of the DCO health district, especially in the extreme north of Djougou (Fig. 6b). Moreover, the standard deviations were relatively low for the estimates of malaria infection (6% < SD < 28%), while they were relatively high for the calculations of malaria cases (1% < SD < 39%) (Fig. 6c-d).

WHO recommended ensuring access to malaria prevention, diagnosis and treatment as part of universal health coverage by determining the most effective mix of interventions according to the local context and needs. This induce data collection or using existing data that reflect a need for subnational level [34]. Recent reforms are being introduced in the health department in Benin in order to reduce the burden of AIDS, Tuberculosis, Malaria, Hepatitis, Sexual Transmitted Infections and epidemic-prone diseases. Their effects are not yet measurable.

This is the first-time spatial analysis of malaria infection and morbidity focused on the OKT and DCO health districts. The two ecological regions do not have the same epidemiological profile. The OKT district was a mesoendemic and/or hyperendemic area with a perennial transmission [7]. Heterogeneity of P. falciparum infection has already been described in the OKT district in a previous study and was not surprising [7, 35]. A contrarily, the prevalence rate of P. falciparum was very high in northern Benin. The DCO district remained a hyperendemic area. The Malaria Atlas Project confirmed a similar pattern in the spatial epidemiology of P. falciparum [36, 37]. In Côte d’Ivoire, a West African country, a higher prevalence rate of P. falciparum infection was also predicted in the north compared to the south [38]. Vectors density, coverage in vector control tools, drugs, immunity, human genetics, social, demographic, and environmental factors could explain the difference in P. falciparum prevalence between northern and southern regions. National malaria programmes should determine the appropriate package of interventions for local level considering transmission intensity, as well as a good understanding of each area’s ecological, epidemiological, and social features. Interventions should be adapted and tailored to specific geographical areas within each country [39]. The lack of qualified human resources in data modelling was a huge obstacle in decision making. Challenges remain in convincing national institutions to actively participate and provide financial support.

A 2010 study conducted at the national level in Benin showed that 34.5% of households in the north versus 28.3% in the south owned at least two LLINs. In opposite, 55.9% of children under five years slept under the LLINs the night before in the north compared to 68.1% in the south [40]. This coverage of vector control tools remains poor and can explain the high prevalence of malaria infection and clinical cases found.

The vector resistance to insecticide was high in all the studied health districts. Anopheles gambiae, the main malaria vector, is highly resistant to the standard insecticides [9, 13]. Data collected during the study showed that use of an LLIN was low (17–57%) and gave only 40 to 50% effectiveness against malaria clinical cases [18].

During the 2011 and subsequent campaigns, community awareness of LLIN use was mainly addressed to mothers, pregnant women and children. The understanding or perception induced by the communication campaigns is that the LLINs is to be used by women of childbearing potential and the children with whom they usually sleep. A large part of children was kept out from intervention by the messages delivered during awareness campaigns. The malaria awareness campaigns should therefore be reviewed in order to improve LLINs use rates among elder children and adult males.

In Benin, malaria control tools (LLINs, ACT, SP etc.) are available at the health facilities. The index of availability of malaria control tracers among health facilities offering this service provided by SARA survey was 82% in Donga department (DCO health zone), compared to 71% in Atlantique department (OKT health zone). However, stock-out of anti-malarial drugs were often noticed in certain health centres. Care-seeking at health centres and utilization of recommended treatment of malaria remains a challenge. One major reason is to be self-medication by herbal tea and other non-recommended drugs such as chloroquine, quinine and paracetamol [41].

Studies suggest a protective effect of maternal breastmilk against P. falciparum [42, 43]. So, in theory, children under six months of age are supposed to be protected against infection compared to the elder children due to EBF [44]. In the current study, in the south OKT district, the infection prevalence was similar among the four age groups. Children under six months of age (supposed to be exclusive to breastfed) were similarly infected than the oldest children. The same observation was seen among clinical cases. In the north, the prevalence of P. falciparum infection and the proportion of clinical patients were higher than in the south. The prevalence rate of infection was lower among children under six months of age than in the oldest. According to knowledge, the prevalence of parasitemia increases sharply beginning at about 20 weeks of age [44].

Nevertheless, children remain extraordinarily resistant to high parasitemia, fever, and severe malaria until about six months of age. This protection is associated with maternal immunoglobulin G (IgG) antibodies specific to Plasmodium antigens, as IgG levels decrease from birth over the first year of life. Another assumption is that the protection of infants may be associated with parasite growth-inhibitory factors such as lactoferrin and secretory IgA found in breast milk and maternal and infant sera [44, 45]. Further, data from the Democratic Republic of the Congo showed that EBF reduced the risk of clinical malaria in children under six months of age [43]. However, the assumption that maternal antibodies against malaria antigens were the basis of this protection was contradicted by at least one study [46]. We did not verify this association in the present study. However, in the DHS data from Benin, children under five years received EBF or breastfeeding accompanied by water and/or other liquids (without milk) for 4.5 months in the DCO area and only 3.0 months in OKT [20]. Thus, breastfeeding seems less important in the south than in the north. After the age of peak of the prevalence rate of infection, the number of clinical attacks of malaria per year dramatically declines. For example, in the DCO district, the prevalence of P. falciparum clinical cases decreased after 23 months, suggesting the beginning of immunological premonition acquisition [47, 48]. Then, from 23 to 60 months, the frequency of clinical cases diminishes. The age of onset of this protection is somewhat earlier with heavier transmission [42, 43], but protection rarely occurs before the age of two years [44]. In southern Benin, the prevalence rate of infection did not vary according to age group, probably due to the moderate level of malaria endemicity [7, 18]. To fully understand the link between EBF and malaria infection and disease burden, more studies should be carried out in Benin.

This study assessed the spatial heterogeneity of malaria risk in two ecological regions in Benin by using 2011 database. It estimated the disease prevalence adjusted for environmental and population covariates and found a significant relationship between temperature seasonality, rainfall, population density, moisture index and distance to the nearest waterways. However, the data were aggregated at village location in place of household. When GPS coordinates are available at the household level, similar models could be implemented to improve the smoothness of malaria risk prediction in the study regions. Considering household or individual-level data can allow the inclusion of other socio-demographic or personal characteristics (e.g. age, sex, access to a mosquito bed net, etc.) that are likely to influence disease transmission but not easy to obtain for aggregated data collected in village or county-level. The climate changes and the variations of temperature and precipitations change significantly with a cycle of 30 years average. However, mayor events like catastrophes and crisis can affect diseases epidemiology. Benin country did not experience this kind of event the last decade. Finally, more complex models could be implemented by assuming a non-stationary spatial process to evaluate its influence on models' outcomes [29].