Study context
Togo is a country in West Africa, with an population of 7,886 million inhabitants for a density of 152 inhabitants/km2 in 2021 (13). It is bordered to the north by Burkina Faso, to the south by the Gulf of Guinea, to the east by Benin and to the west by Ghana. It has six health regions and forty-three health districts in total (Figure 1).
The health system in Togo is organized as a three-level pyramidal structure (3) (figure 2). The first level is composed of the central administration and the different departments and programs where the guidelines and national policies are developed. The regional (or intermediate) level includes six health regions which provide coordination and technical support to the third level health districts. The peripheral level is represented by the health district which is the most decentralized operational entity comprising of 43 health districts and 944 peripheral health units.
Study Type And Sample Design
This was a secondary data analysis project using cross-sectional data from the Togo Malaria Indicator Survey (TMIS) 2017. These data provide information on malaria prevalence among children under five and pregnancy women in the country(14). The TMIS carried out a two-stage sampling method to select the sample. Using information from the last general population census in Togo in 2010 (15), each region was subdivided into Enumeration Areas (EAs). Each area of study was separated into urban and rural parts to form the sampling strata. At the first level, 171 EAs were drawn with a probability proportional to the size of which 60 were urban and 111 rural. To have an update in the selected enumeration areas, a household enumeration was carried out, which provided a complete list of occupied households in each EA selected. At the second level, 30 households were drawn systematically in each EAs. All women aged 15-49 usually living in the selected households or present the night before the interview with or without children under five were eligible to be interviewed. Selected households with neither a woman aged 15-49 nor a child under five were only included for the household questionnaires. In total, the sample consisted of 171 EAs, 5,130 households (1,800 in urban areas and 3,330 in rural areas), 4,895 women from 15 to 49 years old (1,684 in urban areas and 3,211 in rural areas), and 3,271 children under five (2,441 in rural areas and 830 in urban areas)(16).
Tools and data collection
Three questionnaires were used in the context of the 2017 TMIS: a household questionnaire, a women questionnaire, and a biomarker questionnaire. The household questionnaire recorded all household members and visitors who slept there the night before the interviewer visited the household, water source, types of toilets, habitat characteristics, possession of durable goods, and use of mosquito nets were collected. The women questionnaire was used to collect information on socio-demographic characteristics, knowledge of malaria and prevention measures, births over the last 5 years, prevalence and treatment of fever in children under five. Finger prick blood samples for malaria testing were taken from all children aged 6-59 months in the selected households, for whom the parents or responsible adults had previously given their informed consent. Screening for malaria was done with a rapid diagnostic test (RDT), namely the SD Bioline Malaria Ag Pf/Pan with a sensitivity of 94.0% and specificity of 91.4%. Children who tested positive for malaria, or who had other signs of severe malaria or other serious illnesses, were referred to the nearest health facility for "advice and action" following the national health policy in Togo. Blood collection on slides was carried out to confirm the infection status of all children using a microscope. After drying and fixing the blood smears, the prepared slides were stored in special boxes containing cold accumulators and humidity controllers. Blood samples on slides, accompanied by the transmission sheets, were regularly collected in the field and transported to the parasitology laboratory of the “Institut National d’Hygiene (INH)” to be registered, checked, and analysed. After being stained with Giemsa, the slides were examined for the presence of the parasite. Each slide was analysed independently and blindly by two different biologist technicians. In case of discrepancies between the results of the two technicians, the slide was re-examined by a senior biologist technician (16).
Study Variables
The dependent variable in this study is the infection status of the child (positive/negative) on the microscopic examination of the malaria parasitaemia. The main independent variable was the region. Other independent variables included environment related factors such as altitude, household density, main water source, type of toilet facility, main material of floors, main material of walls and main material of roof. Human related factors included household wealth quintiles, age, gender, possession and use of a mosquito bed net, mother's education level, knowing mosquitoes as vectors of malaria, being exposed to malaria prevention messages, as well as ethnicity and religion. We also took into account if the child had fever in last two weeks before the study. The household wealth quintile variable was constructed using principal component analysis on household’s assets and amenities(17).
Data analysis
For each of the factors recorded, we assessed associations with malaria infection in univariate logistic regression using Generalized Linear Models (GLM), calculating odds ratios. Any association that was found to be statistically significant at a level of p≤0.10 in univariate analysis was assessed as a potential confounder in the association between region of residence and malaria infection. We then included in a multivariable model, region of residence and all the potential confounders as well as any other factors that were significantly associated in univariate analysis (p ≤ 0.10). We then removed one at a time, starting from the one with the highest p-value, each of the secondary exposures. If this resulted in a change of more than 10% in the odds ratio of region or if the likelihood ratio test comparing the complex model with the simple model was significant, the secondary exposure was retained. We then tried to remove the next secondary exposure, using the same criteria. This process was continued until all remaining secondary exposures were either important as confounders or their removal would result in a significantly less precise model(18). With the variables retained (p≤0.05 or confounders), we checked for interactions between our main variable of interest (region of residence) and each of the secondary exposures. For this purpose, categorical variables with more than two levels were recoded to binary. Dose response curves were generated using the predictor effect plots (19) to show the trend between the malaria prevalence and certain associated risk factors. Data were analysed using the software R version 4.0.4.
[1] Sources of drinking water were recoded as improved (protected well, tube well or borehole, public tap/standpipe, protected spring) and unimproved (unprotected spring, unprotected well, rainwater, surface water, river/dam/stream/irrigation channel)
[2] Type of toilet facility were recoded as improved (pit toilet latrine, ventilated improved pit latrine, pit latrine with slab, flush toilet, flush to piped sewer system) and unimproved (no facility/bush/field, pit latrine without slab/open pit, bucket toilet)
[3] Main materials of floors were recoded as 'improved materials' (ceramic tiles, cement, vinyl asphalt strips, parquet and polished wood) and 'unimproved materials' (rudiments, wooden planks, earth, sand, palms, bamboo and others)
[4] Main materials of wall were recoded as 'improved materials' (cement, stone with lime, cement blocks and bricks) or 'unimproved materials' (cane/palm/trunks, earth, bamboo or stone with mud, uncovered adobe, and others rudimentary materials)
[5] Main materials of roofs were categorised into "improved materials" (roofing shingles and cement) and "unimproved materials" (thatch, rudimentary, rustic matting, palm/bamboo, wooden planks, cardboard, and no roof)