FLEXIBLE PARAMETRIC MODELING OF MALARIA SURVIVAL: APPLICATION TO BURKINA FASO

Objectives:Severe malaria remains a major cause of childhood mortality in sub-Saharan Africa. Identifying prognostic factors associated with death among hospitalized children is therefore an important public health priority. This study aimed to identify the parametric model that best describes hospital survival among children with severe malaria and to determinethe clinical factors associated with death.Methods:We analyzed a cohort of 444 children aged 1–59 months who were hospitalized for confirmed malaria at Dori Regional Hospital, Burkina Faso. Five parametric survival models were fitted and compared: Exponential, Weibull, three-parameter Weibull, Beta-Weibull, and modified Beta-Weibull models. Model performance was assessed using the Akaike information criterion (AIC) and Bayesian information criterion (BIC). A simulation study was conducted to assess estimator stability. Clinical factors associated with death were investigated using the selected model under an accelerated failure time (AFT) formulation.Results:The three-parameter Weibull model provided the best fit (AIC = 595.11; BIC = 607.40). Its shape parameter was k=0.912, indicating a decreasing hazard. The estimated threshold parameter was approximately one day. The Beta-Weibull and modified Beta-Weibull models showed identifiability problems. In the multivariable analysis, respiratory distress (TR =0.209; p<0.001) and shock (TR = 0.147; p<0.001) were the factors most strongly associated with shorter survival times. These findings correspond to reductions of 79% and 85% in survival time, respectively.Conclusions:Flexible parametric modeling, particularly the three-parameter Weibull model, provided a suitable framework for characterizing hospital mortality among children with severe malaria. Respiratory distress and shock were major warning signs and may warrant early intensive management.