Predictive modelling to differentiate possible bacterial and viral cases of childhood pneumonia in Kilifi, Kenya using protein markers and clinical data

Abstract

To date, accessible diagnostic tools to identify whether a patient's pneumonia is a bacterial, or viral infection, are not accurate or timely enough to prevent preemptive antibiotic administration. Relying on single biomarkers or clinical presentations has been insufficient. We aimed to incorporate a wide range of novel biomarkers and clinical presentations into a multivariable model and to validate its ability to differentiate bacterial from viral pneumonia. A total of 457 children aged 2–59 months, admitted to the Kilifi County Referral Hospital, Kenya, with WHO defined pneumonia were classified using gold standard laboratory diagnostics as possible bacterial (i.e., bacterial/ unknown cases, n  = 229) or possible viral ( n  = 228) infections. Clinical and laboratory data, and levels of CRP, PCT and 6 novel protein biomarkers were used to develop and validate a predictive multivariable Poisson regression model to differentiate pneumonia etiology. The Receiver Operating Characteristic curve was used to assess biomarker performance and validate the model internally. In age-adjusted analyses, chest-wall indrawing, cough, convulsions, crackles, angiotensinogen (AGT), and Serpin Family A Member 1 (SERPINA1) were significantly associated with pneumonia etiology, controlling for age. However, only chest-wall indrawing remained significant in multivariable analyses after controlling for age. The model demonstrated fair, but inadequate, discrimination, with an Area Under the Curve of 0.61. In a separate analysis excluding the unknowns from the bacterial group, chest-wall indrawing, crackles, cough, AGT (>=7.9144MFI), were significant predictors of pneumonia etiology, adjusting for age. This model demonstrated good discrimination, with an Area Under the Curve of 0.78 on the validation dataset. Among children admitted to the hospital with WHO-defined pneumonia, modelling a wide range of biomarkers and clinical presentations still failed to distinguish bacterial from viral pneumonia with enough accuracy to guide clinical management. Further research addressing some of the limitations we encountered in outcome classification, stratifying by disease severity and potentially accounting for variation in biomarker performance by age, may improve predictions.