Study setting
The study was conducted in the health district of Saraya, located in Kedougou region in the south-eastern part of Senegal, at 800 Km from Dakar, the capital. The district shares a border with Mali at the east, Guinea at the south, the region of Tambacounda at the north and the health district of Kedougou at the west. The health district occupies a land area of 6,837 km2. The population is composed by the rural communities of Bembou, Medina Baffe, Sabadola, Khossanto and Missirah Sirimana. The district has 1 health centre, 22 health pots, 28 health huts and 78 villages with a village volunteer for malaria case management (DSDOM, Dispensateur de soins à domicile). The climate is Sudano-Sahelian with a dry season and a rainy season. Malaria is meso endemic and stable in Saraya, with a long transmission season lasting 4 to 6 months from July to December. Transmission intensity remains high with 20 to 100 infectious bites / man / year and high morbidity during the transmission period. The major vectors are Anopheles gambiae, An. arabiensis sl, An. funestus and An. nili [16]. Malaria incidence was 376.6 ‰ in 2017, 487.2 ‰ in 2018 and 379.8 ‰ in 2019 [3].
Study design and participants
A case control study was conducted from November to December 2020 in four health posts (Bembou, Diakhaling, Khossanto and Sambrambougou), selected purposively based on reported malaria morbidity in 2019 and community accessibility during the rainy season. The number of confirmed malaria cases in 2019 was 1,404 in Bembou, 1,485 in Diakhaling, 2,166 in Khossanto and 3,311 in Sambrambougou.
Cases were recruited from participating health posts while the controls were recruited within the same community, in the same village as the case but not in the same house (case and control houses were distant from each other by at least 2 others houses). A case was defined as an adolescent (10-19 years) coming to seek care at one of the participating health posts for uncomplicated malaria episode defined as fever (Temperature>37.5°) or history of fever within the previous 48 hours, with positive malaria RDT according to the National Malaria Control Program (NMCP) guidelines. Control was a person of the same age group, living in the neighborhood of the case, with negative RDT.
Non-inclusion criteria for both the cases and controls included: (i) Less than 10 years old children, (ii) Individual that does not live in the study area, (iii) Subject who received antimalarial treatment during a period of three weeks prior to the study.
Data collection methods
An electronic data collection platform was used. Data were collected using android tablets with the electronic questionnaire developed on REDCap, which is a Research Electronic Data Capture software compatible with android technology [17,18]. Data collected from the tablets were then synced via internet connection to a server hosted at University Cheikh Anta Diop for storage. Data were extracted from the server for cleaning and analysis. Prior to the start of the study nurses and a community health worker in participating health posts were trained on the study procedures including the inform consent and administration of the questionnaire. For each participant who gave informed consent, an electronic standardized questionnaire was administered to collect data on socio-demographic characteristics of the participant and those of the household head, the household assets (water source, type of toilet, ownership of certain items like television, radio, fridge, etc.), ownership and use of LLIN, use of other malaria control measures, individual behaviours during the mosquitos “biting time”(stay outdoor at the evening/night, sleeping outdoor during the night). Home visits were performed for both cases and controls to assess participants living conditions (types of wall, floor and roof), environmental factors such as presence or absence of stagnant water, overgrown vegetation/bushes inside and in the vicinity of the houses of the participants.
Statistical methods
Sample size calculation: For the sample size calculation, we considered the use of long-acting mosquito nets (LLINs) as the main determinant associated with malaria. The proportion of LLIN use among children in the study area is 62% [19]; assuming a risk alpha at 5%, with a ratio of 1 control for 1 case, the study was powered at 80% to detect an odds ratio of 0.6 if 492 individuals were recruited (246 cases versus 246 controls).
Data analysis: Data were extracted from the server and analyzed using STATA software [20]. A composite variable of wealth index was estimated, based on the assets of the households [21]. We have not included housing variables in the wealth index calculation as in the index used in the Demographic and Health Survey (DHS) [19] to avoid correlation between variables. We have assessed separately the impact of housing materials on malaria risk as shown in other studies [21,22]. The index was then categorized in five wealth quintiles (richest, rich, middle, poor, poorest) using Principal Component Analysis method (PCA) [19,23]. Variables characterizing household materials including type of wall, roof and floor were grouped according to the Demographic and Health Survey (DHS) definition [19] as traditional (rudimentary and natural) and modern to create a binary housing variable. Modern houses were then defined as those that have modern wall, modern roof and modern floor; while traditional houses were those that have rudimentary and natural wall, floor and roof [21]. Modern roof materials include cement/beton, wood planks, tile, metal while traditional roof materials include thatch/straw, bamboo. Modern floor materials include cement and ceramic tiles where as traditional floor materials include earth/mud and bamboo. Modern wall materials consisted of cement and wood/planks while traditional wall materials consisted of earth/mud and bamboo/palm.
Percentage was used to assess the frequency of each outcome with a 95% confidence interval. Characteristics of all participants included in the study were tabulated. Proportions were compared using chi square test or Fisher exact test where appropriated (univariate analysis).
Factors associated with malaria were assessed using a logistic regression model. Covariates with p value <0.20 in univariate analysis were introduced in the multivariate model. From the final model, adjusted odds ratios were derived with their 95% confidence interval. Model validity was tested using the Hosmer-Lemeshow goodness of fit test. The performance of the final model was assessed by the area under the curve, and Akaike and Bayesian information criterion. In addition, a test for multicollinearity between variables was done using the variance inflation factor. Significance level of the different tests was 0.05, two sided.
Ethical considerations
The study protocol was approved by the University Cheikh Anta Diop Institutional Review Board (CER/UCAD/AD/MSN /039/2020). Prior to the start of the study, administrative authorization was sought from the regional and district medical authorities in Kedougou and Saraya. In the field participation to the study was strictly voluntary. Prior to any enrollment, written informed consent was obtained from parents or caregivers for adolescents under 18 years old, while 18-19 years old individuals were invited to consent themselves. In addition, a child assent was sought from 15- 17 years old participants. A unique identification code was attributed to each participant. Personal Identified data collected for the household head were de-identified before the data extraction. All analysis were performed using participants identification code to ensure maximum confidentiality. Access to the study data was restricted and information collected only used for the study purpose.