Study setting
The study was implemented in six postnatal clinics in the Dowa district of central Malawi in southern Africa (Figure 1). Malawi has a population of about 17,563,749 people (8). In 2017, about 71% of the population was living in extreme poverty according to United Nations indicators (10). Malaria is an endemic disease in Dowa and the surrounding districts, but higher numbers of cases are recorded during and after the rainy season (December to July) due to increased potential breeding environments (9,11). We conducted this study between May and September in 2018 because this period would allow us to capture average malaria cases in the study areas.
Study sample
A multistage descriptive cross-sectional study design was employed to select a representative sample of children aged 2 to 59 months and their mothers in the Dowa district. We randomly selected six out of eight outreach clinics that were part of the Mvera mission hospital. The selected clinics were located in Gogo, Ching’amba, Mkhalanjoka, Kalinyengo, Mvera and Mphande, within approximately five to ten kilometres of the Mvera mission hospital. During the study, the Mvera mission hospital served a population of 27,719 people, of whom 5,240 were mothers with a child under five years old. A total population of 4,527 mothers with children between 2 and 59 months were identified in postnatal registers in the six randomly selected postnatal clinics. Our sample size was determined by a Raosoft sample size calculator (12). A margin of error of 5% with a 95% confidence level and 50% response distribution was set.We used a systematic sampling strategy to select a sample of 523 children and their mothers from the postnatal registers.[1] We randomly picked a name for the first child-mother dyad and subsequently picked every ninth child-mother dyad.
Participant recruitment
We contacted the selected mothers and their children in the six postnatal clinics during the regular monthly health screening for children. The screening program is an initiative of the Malawi government to promote maternal and child health through a framework of a continuum of care for mothers, newborns and children (13). Community health workers who were assigned as research assistants sought informed consent from mothers to take part in the study. All interviews took place in the consultation room at each outreach postnatal clinic using a pre-tested questionnaire. The questionnaire contained questions on household sociodemographic characteristics, household food security, house structure type, use of mosquito nets and maternal exposure to Intimate Partner Violence (IPV). The child’s and the mother’s anthropometric and health status data from their health passports was recorded in the questionnaire after their health screening program was completed (14).
Measures
Outcome variable
The main outcome variable of our study was malaria infection in children 2 to 59 months old. The term child malaria infection was operationalized as the presence of the malaria parasite in children’s red blood cells as recorded in the child’s health passport (15). In all the postnatal clinics, RDTs were used to assess the malaria parasitaemia in children. If the diagnostic test was positive, the child was coded as 1 = malaria infection, and 0 = otherwise.
Explanatory variables
We selected potential covariates of child malaria infection in our model based on current literature in Malawi and other countries in SSA (16–21). These were characteristics of the children, who were our study population, characteristics of the parents that influence childcare practices and characteristics of the household. Our covariate variables were child gender, age and weight at birth. We also included variables capturing the child’s history of other morbidities in the past 30 days as reported by the mother. These included diarrheal episodes and acute respiratory infection (ARI); child deworming in the past 12 months was also considered. Child nutrition status was determined through height-for-age, weight-for-height, and weight-for-age Z-score values. Child stunting, underweight and wasting were categorized as those that were ≤ -2 standard deviations of height‑for-age, weight-for-age and weight-for-height Z-scores (22).
Other independent variables that were considered as risk factors for child malaria include the mother’s age, level of education, pregnancy planning and exposure to Intimate Partner Violence (IPV) perpetrated by their current or recent husband. We assessed cases of IPV using a WHO multi-country study questionnaire on women’s health and life experiences that was validated and used in Malawi (23,24). The questionnaire contains 18 items that make up four sub‑scales measuring different forms of IPV: physical abuse, emotional abuse, controlling behaviour and sexual abuse. Maternal exposure to IPV was operationalized as any mother who reported that they experienced any form of IPV.
The fathers’ characteristics were also considered as risk factors. These include level of education, age and health risk behaviours such as alcohol consumption and smoking. Household malaria predisposing and enabling factors in Malawi such as the use of an Insecticide-treated bed nets (ITN), household poverty, type of dwelling, and presence of animals in the house were included. We asked mothers if the child had an ITN, and whether the child slept under the net the night before the survey. Household poverty was defined based on the international poverty measure of US$1.90 a day (25). The presence of animal kraals/sheds within one to ten metres of the dwelling house was considered a risk factor (26). We also asked the mothers if their houses had been sprayed with insecticides and how many people were sleeping in the house.
Survey enumerators administered the survey on Android tablets using an Open Data Kit (ODK). We used a WHO protocol for conducting research on sensitive topics because some of the questions in our study focused on domestic violence (27,28). Enumerator orientation and questionnaire pre-testing was conducted for five days. A PhD candidate in social work, a clinical officer and an environmental health officer were responsible for training the enumerators. The research team, including the enumerators, had professional training in community health, nutrition and primary health care.
Research Ethics Review
Ethics approval to conduct this study was obtained from the University of Livingstonia research ethics committee in Malawi (protocol number: UNILIA-REC-4/18) and the Research Ethics Board of McGill University in Canada (protocol number: REB File #: 503-0518). Written permission was also sought from the Dowa district commissioner’s office, the Dowa district health office and the Mvera mission hospital management. We obtained oral consent from local health leaders and research participants in the study areas.
Data analysis
The Kolmogorov–Smirnov test was used to test the normality of the distribution of numerical variables. These include age, number of children, number of household members and household food security. We constructed categorical variables from our numerical data because we found that our data was not normally distributed (29). Bivariate logistic regressions were performed to examine significant predictors of child malaria. Significant predictors of child malaria at the bivariate level of (p ˂ 0.05) were included in the final multivariable logistic regression model using the forward enter method.
We tested the multicollinearity of explanatory variables and obtained a variance inflation factor (VIF) of 5,143, which indicated independence among the explanatory variables both at the individual and the cluster level. Consequently, a fixed effects model was used to account for the clustering effect in our analysis. The results of the multivariable analysis have been reported as crude and adjusted odds ratios with a 95% confidence interval (CI). A p value of less than 0.05 was considered statistically significant in our study. The data was analyzed using an IBM Statistical Package of Social Sciences (SPSS) for Windows version 23.0 (IBM Corp., Armonk, NY, USA).
[1] We determined that a sample size of 355 was adequate but increased our sample size because we had resources to do so. The larger sample size increased the accuracy and precision of the estimates.