Identifying age-common and age-specific factors of Plasmodium infection in Nigerian children under five: Application of a cluster-aware multistage selection framework to the 2018 Nigeria Demographic and Health Survey

Despite developmental heterogeneity, children under five are often analyzed as a single group, obscuring age-related differences in Plasmodium infection. Using the 2018 Nigeria Demographic and Health Survey data, we analyzed 2,914 younger children (6-23.9 months) and 5,553 older children (24-59 months) to identify age-common and age-specific factors. We applied a Cluster-aware Multistage Selection (CMS) framework integrating penalized regression, interaction testing, and bidirectional selection while accounting for sampling weights, clustering, and stratification. Plasmodium prevalence was 37.3% in younger children and 49.4% in older children. Each 10-percentage-point increase in community-level livestock and agricultural land ownership was associated with 2.5% (PR = 1.025; 95% CI: 1.011-1.039) and 4.6% (PR = 1.046; 95% CI: 1.029-1.062) higher prevalence, respectively. Socioeconomic disadvantage was associated with higher prevalence. Children in the lowest wealth quintile had 80.7% higher prevalence than those in the richest quintile (PR = 1.807; 95% CI: 1.467-2.224), and children whose mothers had no formal education had 55.8% higher prevalence than those with higher education (PR = 1.558; 95% CI: 1.225-1.980). Severe maternal anemia was associated with 22.5% higher prevalence (PR = 1.225; 95% CI: 1.012-1.483). Stunting and household insecticide-treated net (ITN) ownership were associated with 8.0% (PR = 1.080; 95% CI: 1.024-1.139) and 10.4% (PR = 1.104; 95% CI: 1.017-1.198) higher prevalence, respectively. Conversely, sleeping under an ITN (PR = 0.916; 95% CI: 0.858-0.977), breastfeeding (PR = 0.845; 95% CI: 0.764-0.934), maternal internet use (PR = 0.638; 95% CI: 0.475-0.857), and overweight-for-height (PR = 0.761; 95% CI: 0.631-0.917) were associated with lower prevalence. Age-specific associations were observed only in younger children, including paternal lack of education (PR = 1.301; 95% CI: 1.025-1.652) and rural residence (PR = 1.462; 95% CI: 1.266-1.688). These findings support age-tailored Plasmodium prevention and highlight the utility of CMS for complex population data.