Background
Malaria is an endemic disease and a public health issue in Cameroon. It is a major cause of morbidity and mortality among children less than 5 years. In 2014, the morbidity of malaria was 30% in children and 18% in adults [
1,
2]. Conscious of this situation, the government has considered the fight against malaria to be a national priority and part of the health strategic plan [
3]. Since 2002, the National Malaria Control Programme (NMCP) was created under the coordination of the ministry of public health. The aim was to improve the quality of strategic actions and to raise resources. During the last 10 years, huge investments have been deployed by donors, the international community and the government, to develop strategies and tools for reducing the burden of malaria in the country. According to the national malaria strategic plan of 2014–2018 [
4], the NMCP is implementing interventions to sustain and scale up malaria control. Those interventions include distribution of insecticide-treated nets (ITN) to populations at risk and of sulfadoxine–pyrimethamine to pregnant woman, parasitological confirmation of suspected malaria cases (microscopy or rapid diagnostic test), and treatment of uncomplicated malaria cases by artemisinin-based combination therapy (ACT). Until 2011, the NMCP has distributed ITNs only to vulnerable groups. In 2012, the distribution policy has changed and more than eight million of long-lasting insecticide nets (LLIN) was given to populations at risk [
5,
6]. Before the LLIN mass campaign distribution, two representative surveys were carried out by the National Institute of Statistics: a demographic and health survey (DHS) combined with multiple indicator cluster survey (MICS) and a malaria indicator survey (MIS).
The DHS was the first national malaria survey to collect prevalence data across the country, however for logistic reasons data were collected outside the malaria high transmission season. The NMCP and partners have decided to conduct the MIS during the second and most important rainy season (September–October), when the highest peak of malaria transmission occurs in order to assess the ability of DHS to estimate the malaria burden in the country [
7]. Hence, the objective of this study is to assess the influence of the survey period on the detection of risk pattern by comparing estimates of the malaria parasite risk and the effects of interventions obtained from both surveys. The analysis was carried out using Bayesian geostatistical logistic regression models similar to the ones that have been used for spatial analyses of other DHS and MIS data such as Angola, Senegal, Nigeria, Burkina Faso, Uganda and Sudan [
8‐
13].
Discussion
This study is the first to assess the influence of survey season on the estimates of the geographical distribution of malaria parasite risk and of the effects of interventions, using data collected by DHS and MIS carried out at the same locations and year, but at different malaria transmission seasons. The analysis employed Bayesian geostatistical models because this study was interested in comparing the estimates of the risk pattern across the country rather than at the observed locations.
The DHS collects a large number of indicators on diverse sectors and huge logistics are involved to guarantee the coverage of all clusters, in particular those in rural areas with difficult access. Moreover, the planning of DHS usually avoids the rainy season in Africa because of road’s degradation which challenges the survey implementation. The constraints described above have often an impact on the schedule and duration of DHS. The DHS and MIS surveys in Cameroon provide a unique opportunity to assess the effect of season on malaria survey-based estimates.
Both surveys showed low level of parasitaemia risk (under 5%) in West and Adamawa highlands. These areas are suitable for elimination interventions. Also, both data indicated that, the parasitaemia risk in East region was the highest in the country and above 50%. This high risk level is explained by the important coverage of forest, the predominance of rural areas and the low educational level of the population.
DHS data did not identify a cluster of high malaria parasite risk in the North and Far-North regions as estimated by the MIS. However, evidence from the upsurge of malaria cases that over strain the capacity of the health system during the rainy season and the high malaria mortality risk among children in the northern part of the country does not support the DHS finding [
37,
38]. The non-concomitance between DHS and the malaria seasonal transmission in the north regions may explain the underestimation of malaria parasite risk in that area.
Furthermore, DHS could not capture a malaria cluster in the coastal part which is the estuary of the biggest rivers in the country that pour into the Atlantic Ocean. During the long rainy season that begins in August, some areas are flooded and large ponds of stagnant water are created [
39‐
42]. The high transmission occurs just within the rainy season which is characterized by the increase of mosquito population. The water availability is among the key criteria for mosquito breeding, especially in the North part of Cameroon which is covered by the Sahel and in coastal towns, such as Douala because of poor condition of the pluvial drainage system. The model has identified additional climatic factors in MIS compared to the DHS.
In the Adamawa, North-West, West and Centre regions of the country, MIS estimated lower risk compared to DHS. In the capital, Yaoundé, parasitaemia risk was 6% based on MIS that is half the one obtained by the DHS data. The coverage of household by an ITN among the population of Yaoundé and Douala was 31 and 37%, respectively. Among households with at least one ITN, the percentage of those who use ITN in Yaoundé and Douala was among the highest in the country, i.e. 43 and 52%, respectively. Human behaviour at the beginning of the rainy season changes and people are likely to increase the use of preventive tools, such as ITN, mosquito spraying devices or repellents [
43‐
45].
The altitude and NDVI were identified as important predictors in the cluster level models of both surveys. The presence of forest, EVI and distance to water body were found to be important in modelling the MIS data. As known, the altitude has a negative effect on malaria parasite risk. The effect of distance to water was not linear and households located more than 70 m away from water bodies are at higher risk of malaria compared to those households close to them for a number of reasons including the wind direction and the availability of human hosts [
46]. Rainy season has an influence on vegetation and on human activities, such as farming which exposes people to mosquito bites, and that could be the reason of the positive association between the EVI, NDVI, the presence of forest and the parasitaemia risk [
47,
48].
The analysis of the MIS data showed that the proportion of households with one ITN per two persons was statistically important with a negative effect indicating that the household coverage had an influence on malaria parasite risk among children [
49]. According to the DHS, the ITN coverage indicator with a statistically important and protective effect was the population with access to an ITN. The use of ACT among children under 5 years old with fever in the last 2 weeks before the survey was positively associated to the malaria parasite risk but not statistically important. Similar results regarding ACT have been obtained from the MIS in Uganda and in Burkina-Faso [
11,
12].
The disease risk resembles the pattern of socioeconomic inequalities in the country. In both surveys, the place of residence had an important effect and was negatively associated to malaria parasite risk. The DHS data showed that the effect of only the least poor category of the wealth index was statistically important compared to the most poor baseline category, however the MIS data estimated statistically important effects in all socio-economic categories. The educational level of mothers had a protective effect which was however statistically important only for the DHS. These results suggest that during the high malaria transmission season, the quality of the household environment is more important than the mother’s education. Obviously, children from wealthy households can benefit from additional vectors control tools, such as appropriate malaria treatment, ITNs, sprays products and the sanitized neighbourhood. Wanzirah et al. and Tusting et al. have also shown that high house quality reduces the entry of mosquito vectors and, therefore, lessens the risk of infection [
50,
51].
A gradient of malaria parasite risk was associated to the age and as expected the gender effect was not statistically important. Younger children were at lower risk than older ones, which may be a consequence of the passive immunity given by mothers [
52].
The high residual spatial correlation estimated by the models, especially those that used the MIS data indicates the presence of unmeasured spatially structured factors that influence the geographic distribution of the parasitaemia risk. It is likely that the climatic proxies considered in the model such as day and night LST or NDVI and EVI were not able to capture the entire ground climatic conditions. Similar analyses of other MIS data estimated relative high residual spatial correlation, particularly in recent surveys that climatic factors are confounded from malaria interventions [
10,
12,
30]. The BCI width of the estimated parameters obtained with DHS were tighter than those of MIS, most likely due to the smaller number of survey clusters in the later [
53,
54].
Both, DHS and MIS were used a RDT. RDTs could remain positive for few weeks after a malaria treatment. Therefore our estimates of parasitaemia risk may be slightly overestimated than those based on diagnosis by microscopy [
55‐
57].
DHS and MIS are based on a two-stage cluster sampling design. In the first stage, the number of clusters that are selected at regional level is proportional to the population. This design oversamples clusters in places with high population density and can selects fewer clusters over larger regions with small populations (i.e. East region) where the disease may vary more compared to the urban areas and big cities such as Yaoundé and Douala. Therefore, the DHS/MIS survey design may provide lower precision of the estimates in rural areas.
Since 2011, Cameroon has implemented two mass campaigns of LLINs, introduced preventive treatment of children against malaria in the North region and built two large dams in the East and South regions. There is currently a DHS ongoing in Cameroon and the results of this study will serve as a baseline to assess the changes in malaria risk as a result of disease interventions, climatic effects and environmental modifications [
58,
59].
Authors’ contributions
PV had conceived, designed the study and contributed to the analysis. KC and RW contributed to the design, collect of the DHS and MIS data. SM had analysed the data and drafted the manuscript. PV, SM, KC and RW revised the manuscript and provided the intellectual content. All authors read and approved the final manuscript.