Assessing latent factors in Malaria health behaviors and access to preventive care using frailty models

Abstract

Malaria remains a major public health challenge in sub-Saharan Africa, particularly among children under five. Despite progress in prevention, disparities in adoption persist due to socioeconomic, geographic, and individual factors. This study examines the influence of unobserved household and cluster-level factors on malaria prevention behaviors and access to care in Kenya using inverse Gaussian frailty models. Secondary data from the 2020 Kenya Demographic and Health Survey (KDHS) is analyzed to assess the impact of malaria-endemic zones, digital connectivity, and child-specific factors on preventive care adoption. A semi-parametric inverse Gaussian frailty model quantifies latent variability at household and cluster levels, providing deeper insights into malaria prevention and treatment behaviors. Results show that frailty models significantly outperform the Cox model (ΔAIC > 18,000), with gamma frailty performing best in high variability settings (θ > 0.2, ΔAIC = 7.68). Mobile phone access strongly predicts ITN adoption (HR ≈ 1.21, p < 0.001), while internet use and social media have no effect (p > 0.70). Children in endemic areas adopt ITNs earlier (Estimate = 0.233, p < 0.001), with moderate frailty variance (θ = 0.006). Younger children exhibit 3.4× higher preventive care uptake (β = −3.368, p < 0.001), with high frailty variance (θ = 0.5), indicating substantial unobserved heterogeneity. Digital connectivity shows minimal cluster effects (θ = 0.005), suggesting uniform access. These findings highlight the need for targeted mobile health strategies, age-specific prevention programs, and frailty-adjusted methodologies to enhance malaria control interventions.