Development of a Bayesian subjective model for predicting the clinical diagnosis of Ebola in the Democratic Republic of the Congo

Abstract

Abstract The symptoms and clinical signs of Ebola virus disease are similar to those of malaria, thus leading to difficulties in terms of making differential diagnoses. Therefore, we developed a subjective model for the clinical diagnosis of Ebola. Excel and SPSS software were used to analyse data. The likelihood ratio, the kappa statistic and various internal evaluation parameters of the model were calculated. These analyses revealed that 4 factors strongly influence the clinical diagnosis of Ebola: haemorrhagic signs, neurological signs, digestive signs and epidemiological links. Among these 4 factors, the combination of haemorrhagic signs and epidemiological links in a patient yields a 60.5% chance of the case being confirmed as Ebola. Therefore, all health care providers in areas with the potential for Ebola must prioritise classifying any patient with these 2 factors as a genuine case of Ebola