Malaria remains a public health concern worldwide and especially in sub-Saharan Africa (SSA), where it is a major killer among children under five years of age [1–4]. This is despite local, national, and international efforts to control the disease. A country or a region in the country may be classified as endemic or holo-endemic or meso-endemic depending on the transmission intensity [5]. Subsequently, control strategies are designed for the whole region or country and are assumed to suit every local setting therein. However, the transmission may be different from one local setting to another within the same region or geographical locality and interventions designed at the regional or national level may fail to meet needs on the ground. It may be assumed that health districts would focus resources for malaria control proportionate to risk of malaria transmission in specific localities if they were aware of which localities were more threatened than others. Knowledge of malaria risk at a local level may prove more informative for formulating control strategies by supporting resource allocation decisions at the health district level.

There is evidence that malaria transmission risk varies even on the smallest scale. Wang et al have demonstrated inter-city variation of malaria prevalence among school children [6]. In Dar es Salaam, malaria prevalence was shown to gradually increase as one moved from the city centre to the periphery. Similar findings were observed in Ouagadougou and Abidjan [6, 7]. These studies were, however, cross-sectional in nature and were limited to urban areas. Some studies have shown that the transmission risk as measured by the entomological inoculation rate (EIR) may differ between neighbouring localities, but few studies have been carried out on the risk in terms of outcomes in the human host. The aim of this study was, therefore, to assess whether there were differences in the risk of clinical malaria (defined as axillary temperature ≥ 37.5°C and parasites density ≥ 5,000 parasites/μl) among children under-five in one town and three villages in north-western Burkina Faso.

There were differences in the risk for clinical malaria between the four sites as shown by the number of episodes per PY and the odds of clinical malaria given the site of residence. The largest incidence rate was observed in Goni, which had about 1.2 times higher incidence than Nouna and Cissé. The odds ratio of clinical malaria for this site was also higher even though the highest odds ratio was found in Koudougou. Such differences were observable despite the relative proximity of the four sites since they are within a 19–44 kilometre radius of each other.

The lower incidence observed in Nouna and Cissé and Kodougou compared to Goni could be explained by differences in the ecological settings of these sites. Goni is located in a plain where there is seasonal rice farming which creates favourable habitats for mosquitoes. Kodougou is a village about 200 metres from a permanent river, which serves as a water source for traditional irrigation. Proximity to irrigated and flooded agriculture practices have been reported to be highly associated with malaria infection [12–14]. The semi-urban setting of Nouna could explain its comparatively lower incidence of clinical malaria. Indeed, urban settings are reported to be less prone to malaria transmission compared to rural settings [15] due to the pollution of water bodies therein. These polluted waters are unsuitable as breeding sites for the malaria vector. The findings of low incidence of clinical malaria from Cissé are surprising. Cissé is located next to a forest and is surrounded by a swamp during the rainy season. These conditions are suitable for an increased risk of malaria transmission. A possible explanation of the low observed risk could be presence of cattle, which serve as an alternative host of the malaria vector. Cattle keepers with their attendant animal enclosures within the villages are the main inhabitants of Cissé. An investigation of the feeding preference of malaria vectors is needed to determine the fraction of vectors feeding on animals.

Regarding the results from the logistic regression, the pattern observed in the crude incidence rate of clinical episodes per 1000 PYs was maintained after accounting for variables that may influence this distribution. Compared to children living in Goni, those living in Kodougou were more likely to have clinical malaria while those from Cissé and Nouna were significantly less likely to have it. Differences in the ecological settings of the sites as discussed above could still explain these differences in risk. Other covariates showed different effects on the risk of contracting clinical malaria.

Older children were better protected than the young ones. The odds were much higher in the age group below 12 months. This was not surprising because children below 6 months have immunity acquired from the mothers, which they progressively loose, as they grow older. Reyburn et al [16] reported a high risk of severe malaria in the age group below 12 months. With increasing age, children develop partial immunity and become less susceptible, although below five years this immunity remains weak.

Studies have suggested that ethnicity has a significant effect on the clinical malaria risk. Studies in this area have shown that the Peuhl (also known as Fulani) are less susceptible to malaria due to their reduced genetic susceptibility compared to other ethnicities [17, 18]. However, our findings did not confirm this. Mossi and Samo have significantly lower risk than the Peuhl. Ethnic differences observed in previous studies may be attributable to differences in local ecological settings since people of the same ethnic group tend to aggregate in the same village. Previous studies did not assess this latter factor. Studies specifically designed to explore this association in this region may be of interest, as understanding genetic susceptibility to clinical malaria could prove useful for vaccine development.

Self-reported usage of a mosquito net was surprisingly not associated with a decrease in clinical malaria risk among children as has been extensively reported in other studies [19–24]. The question was precise enough to get the required information because interviewers asked mothers at each visit whether the child had been sleeping under a mosquito net since the previous visit. Although a distinction was made between normal mosquito nets and insecticide-treated nets (ITN) during data collection, this was not taken into consideration in the analysis. This was because of the very small number of non-impregnated mosquito nets, as an ongoing community trial provided ITN freely.

Among the factors related to the housing conditions, only the presence of farming activities within a 30-metre radius of the household was associated with increased risk of clinical malaria. The agriculture practice in this region is to dig small furrows between the crops to retain rainfall water for better ground infiltration. Such water bodies combined with the shadow provided by the crops create a suitable environment for vector breeding, thereby increasing the risk of clinical malaria.

The rainy season, as expected is associated with an increase of clinical malaria episodes. In the region where the climate is mostly dry, rainfall, occurring during a specific period of the year, is the main source of breeding sites (open water bodies) for the malaria vector. In contrast, during the dry season, the hot and dry weather is unsuitable for vectors, thereby leading to high vector mortality and reduction in infection. These findings are consistent with those from other studies [25, 26]

The implication of these findings is that malaria control strategies should be designed to fit location-specific contexts. The distribution of the factors responsible for malaria transmission may lead to different malaria transmission patterns even within such a relatively small geographical entity. At the level of a health district or even for national malaria control policies, it would be very informative to know the risk difference between locations as small as a village. Scarce resources for malaria control could then be allocated accordingly. While all villages at risk should receive malaria control measures, high-risk villages can be identified and given priority.

Conducting such studies at the village level in every health district is not only time consuming, but it is costly. Additionally there are very few such studies already available for use. It is, therefore, imperative that well functioning health information systems are established and utilized. In practice, this will require health districts to collect routine health information data series (for example over a three-year period) to assess which villages or localities are at the highest risk, accounting for seasonal variation. This risk assessment may then be utilized to decide to which localities malaria control efforts should be focused rather than distributing resources equally or otherwise based on other criteria.

The author(s) declare that they have no competing interests.