A multilevel analysis of prevalence and risk factors of Malaria in children under five years in the 2018-19 Malaria indicator survey in Uganda

Abstract

Background: Malaria is a life-threatening disease caused by falciparum Plasmodium parasites. In 2019, there were an estimated 229 million malaria cases globally in 87 malaria endemic countries. Children aged under 5 years are the most vulnerable group affected by malaria; in 2019, they accounted for 67% (274 000) of all malaria deaths worldwide. About 95% of malaria deaths globally were in 31countries among which is Uganda. A web of determinants modulated by both individual and environmental determinants influences malaria transmission. This study therefore aimed at; assessing the effect of weighting and cluster variation on estimates of malaria prevalence and risk factors in children under-five years in the 2018-19 malaria indicator survey in Uganda. The information will provide an understanding of the sources of variation that exist at different levels of malaria survey data analysis to improve future study designs and help program and policy interventions. Methods: Data from the Uganda Malaria Indicator Survey of 2018-19 were analysed. Four models were fitted to determine the effect of weighting and/or adjusting for cluster variation on the risk of under-five malaria. To determine sources of variation in the nested data and significant prevalence and risk factors of malaria infection in under-five of Uganda, a multilevel mixed effects logistic regression model was fitted to the data at individual, household, and enumeration area levels. The intraclass correlation coefficient for hierarchical structures was used to determine the existence of cluster variation. Results: Overall, 21.1% of all children tested positive for malaria by rapid diagnostic test and 51.6% were anaemic. Also, most households had at least one bed net (87.4%) but had not sprayed their dwellings within the last 12 months of the survey (86.5%).The model that was weighted and adjusted for, for cluster variation on the risk of malaria infection performed best according to both standard errors and design factor values of model estimates. Adjusting for other covariates, a child’s age was associated with 42% higher odds of malaria (AOR=1.42; 95% CI, 1.33-1.52). Children whose mothers had at least secondary school education had about 47% lower odds of malaria infection (AOR=0.53; 95% CI, 0.30-0.95) compared to uneducated mothers. Children who dwelled in households in the two highest wealth quintiles had lower odds of malaria compared to those from households in the two lowest quintiles (AOR=0.42; 95% CI, 0.27-0.64). Also, an increase in altitude by 1 metre was associated with slightly lower odds of under-five malaria (AOR=0.98; 95% CI, 0.97-0.99). About 77% of the total variance in the malaria positivity was attributable to differences between enumeration areas. The variation in under-five malaria positivity remained statistically significant across all models. The intraclass correlation coefficient at EA was greater than that at other levels across models. Conclusion: In addition to adjusting for clustering and stratification, analysts should prioritize making adjustments for both weighting and cluster variation in outcomes in the contexts of correlated or nested datasets. Enumeration area specific interventions may be more beneficial in reducing the burden of under-five malaria infection.