2022 Volume 75 Issue 2 Pages 23-34
[Purpose] Devices capable of performing quick blood tests in outpatient clinics have become popular. The purpose of this study is to construct an “estimated discriminant of microorganism” that can infer the causative microorganism from the obtained blood test results and to help prescribing antibacterial agents.
[Method] The discriminant was constructed based on the hematological and CRP results, age, and the results of causative pathogens in pediatric pneumonia (n=682) under 6 years old obtained in a large-scale multicenter study. For each test item, AUC, sensitivity, specificity, and threshold (cutoff value) were calculated from the respective ROC curves. Based on these results, an estimated discriminant of microorganism” was constructed using three items: WBC, CRP, and age by multivariate analysis. We also verified the validity of the discriminant.
[Results] The constructed discriminant was Y1=3.45-4×WBC (cell/μL)+0.847×CRP (mg/dL)+0.610×age-6.849. Estimates of the causative pathogens, bacterial or viral, were expressed as percentage (%).The sensitivity and specificity of discriminant was 88.5% and 88.6% respectively. The false positive and false negative rates were 11.4% and 11.5%. In many false positive cases, either WBC or CRP was high. Half of mycoplasmal pneumonia was determined to be viral by the discriminant. The error in validation was around 11%.
[Conclusion] The discriminant is easy to understand because the probabilities of bacterial and viral infections are displayed in%. The false positive rate was 11.4% and the false negative rate was 11.5%, but in many false positive cases, either WBC or CRP was high, and many of the false negative cases included mycoplasma infection, so antibacterial. It will contribute to the decision when prescribing the drug and the explanation to the parents of the patient.
