Predictive evaluation of pediatric patients based on their typhoid fever status using linear discriminant model

Abstract

Epidemiologic studies have established a relationship between pediatric patients and typhoid fever infection. This study was carried out to ascertain if specific hematological measurements of the pediatric patients discriminate between their positive and negative status to typhoid infection and to produce a rule for classifying other pediatric patients. Discriminant analysis was applied to predict the probability of a specific categorical outcome based on several explanatory variables (predictors). This study analyzed the differentiation between two hundred pediatric patients attending Landmark University Medical Centre based on their typhoid fever status. The hematological parameters considered were Packed Cell Volume, White Blood Cell count; Neutrophil, Erythrocyte level, Hemoglobin and Platelet count, Assay of samples were performed using standard procedures. Fisher’s Linear Discriminant Method was used for classification of variables in this study. With the use of the Fisher’s Linear Discrimination method for classification of the obtained data, a minimum value of −0.0067 was obtained implying that any new pediatric patient with a discriminant score above −0.0067 would be diagnosed to be typhoid negative; otherwise, they would be classified as typhoid positive pediatric patients. The efficiency of this method of classification was tested using two approaches; Retribution estimate approach and leaving-one out approach which showed a prevalence rate of typhoid positive patients at 75.8% and 74.7% respectively. This data analysis hypotheses that typhoid fever is highly endemic amongst our study subjects. A point-of-care diagnosis with a strong positive predictive value, which improves pediatric enteric fever diagnosis, is strongly advocated.