340 - Incorporation of biomarkers into a clinical prediction model for pediatric radiographic pneumonia

Poster Number: 340
Publication Number: 340.204

Sriram Ramgopal, Ann & Robert H. Lurie Children's Hospital of Chicago, Chicago, IL, United States; Lilliam Ambroggio, Children's Hospital Colorado, Aurora, CO, United States; Douglas Lorenz, University of Louisville, Louisville, KY, United States; Samir S. Shah, Journal of Hospital Medicine, Cincinnati, OH, United States; Richard M. Ruddy, University of Cincinnati College of Medicine, Cincinnati, OH, United States; Todd A. Florin, Ann & Robert H. Lurie Children's Hospital of Chicago, Chicago, IL, United States

Background: Predictive models for radiographic pneumonia may facilitate judicious use of chest radiography.

Objective: To evaluate the role of blood biomarkers in predicting radiographic pneumonia in children with signs and symptoms of lower respiratory tract infection (LRTI).

Design/Methods: We performed a single-center prospective study of children 3 months to 18 years evaluated in the emergency department (ED) with signs and symptoms of LRTI and who had a chest radiograph performed for suspected community-acquired pneumonia (CAP); excluding patients with complex medical conditions or recent hospitalization. We evaluated four biomarkers: white blood cell count (WBC), absolute neutrophil count (ANC), C-reactive protein (CRP), and procalcitonin. We evaluated the role of each biomarker in isolation and when added to a previously developed clinical prediction model for radiographic CAP (which included focal decreased breath sounds, age, and fever duration) using logistic regression. Improvement in performance was assessed by comparing concordance (c-)indices and test characteristics between models and using the net reclassification index (NRI).

Results: Of 580 included children, 213 (36.7%) had radiographic pneumonia. CRP was higher among patients with radiographic pneumonia compared to those without; other biomarkers were similar between groups (Table 1; Figure). In multivariable analysis, all biomarkers were associated with radiographic pneumonia, with CRP having the greatest effect size (adjusted odds ratio, 1.79; 95% CI 1.47-2.18) (Table 2). The multivariable model with CRP demonstrated the greatest improvement in c-index (optimism corrected; 0.812) compared to the model with clinical variables alone (0.780). This model also demonstrated improved sensitivity (70.0% vs 57.7%) and similar specificity (85.3% vs 88.3%) compared to the clinical model alone when using a cutpoint (corresponding to a predicted probability of 45.6%) identified by the Euclidean method. The CRP model demonstrated the greatest improvement in case identification (NRI=0.092, Figure). Performance gain with other biomarkers (WBC, ANC and PCT) was limited.Conclusion(s): A model consisting of clinical variables and CRP demonstrated improved performance for the identification of radiographic pneumonia compared to a model with clinical variables alone. CRP may be of value in identifying patients at moderate risk of radiographic CAP to guide management decisions.
Table 1. Summary statistics for included biomarkers, displayed as median, (interquartile range).WBC, white blood cell count; ANC, absolute neutrophil count; CRP, C-reactive protein; PCT, procalcitonin
Table 2. Model performance on the addition of individual biomarkers to the clinical model.WBC, white blood cell; ANC, absolute neutrophil count; CRP, C-reactive protein; PCT, procalcitonin; OR, odds ratio; CI, confidence interval; c-index, calibration index; PPV, positive predictive value; NPV, negative predictive value; PLR, positive likelihood ratio; NLR, negative likelihood ratio
a. Includes variables of age, focal decreased breath sounds, and duration of fever.
b. Determined by Euclidean distance method
c. Defined as the sum of differences in proportions of individuals moving up in probability minus the proportion moving down in probability for those with the outcome, and the proportion of individuals moving down in probability minus the proportion moving up in probability for those without the outcome