VECTOR AUTOREGRESSION MODELING OF MALARIA INCIDENCE AND MORTALITY RATES IN MIGORI COUNTY, KENYA

Abstract

Malaria prevalence in poorer countries has been a persistent public health concern, disproportionately affecting vulnerable populations such as children and pregnant women. Despite notable progress in scaling up malaria control interventions in Kenya, malaria incidence rates continue to vary widely across counties, with endemic regions like Migori County experiencing persistent challenges. This study aimed to identify key factors associated with malaria incidence and mortality in Migori County using secondary data from the Kenya National Health Management System. Multiple statistical models, including regression, Vector Autoregression (VAR), and Vector Autoregression with Exogenous Variables (VARX), were applied to examine the temporal dynamics of malaria. While malaria incidence rates declined over time, mortality rates remained relatively stable. Regression results indicated that insecticide-treated net usage and effective treatment significantly influenced both incidence and mortality rates. However, model residuals showed substantial variability and signs of poor fit, highlighting the need for improved model specifications. The VAR model revealed issues of residual autocorrelation, while the VARX model, which incorporated exogenous variables, showed improved but still imperfect performance. Bayesian VAR (BVAR) models provided consistent findings across methodologies but also underscored ongoing challenges in modeling temporal malaria data accurately. Therefore, this study concludes that while current models offer valuable insights, they remain limited in capturing the full complexity of malaria dynamics. It recommends methodological enhancements, such as using advanced techniques like Generalized Method of Moments (GMM) or machine learning, conducting rigorous residual diagnostics, and incorporating environmental, socioeconomic, and behavioral variables. Expanding the dataset across regions and timeframes could also improve the robustness and generalizability of future research aimed at informing more effective malaria control strategies.