Marital status was also a significant factor, as divorced/separated caregivers were associated with lower odds of a positive test (aOR = 0.38, 95% CI: 0.16–0.92, *p* = 0.033) compared to married caregivers, while single caregivers showed a trend of higher odds (aOR = 1.30, CI: 0.66–2.59, p = 0.447).
The different importance profiles for certain variables such as child’s age being highly significant in RF (80.36%) and DT (83.84%) but low in GBM (29.55%), or the source of malaria knowledge being highest in DT (100%) but middle in others, show how different algorithms capture different facets of the underlying risk architecture, with ensemble approaches providing more stable feature importance estimates through their ability to capture complex nonlinear relationships and interactions.