Clinical rules for bacterial pneumonia. A systematic review and diagnostic test meta analysis

Abstract

ABSTRACT: Context: Adequate differentiation between viral and bacterial pneumonia in children remains a challenge. Several clinical prediction rules (CPR) to support this differentiation but it is not clear their performance. Objective: to evaluate the diagnostic prediction capacity of existing CPR for the detection of bacterial community acquired pneumonia (CAP) in children. Methods: Systematic review and meta-analysis of diagnostic tests accuracy (DTA-MA). We searched Medline, Embase, LILACs and gray literature up to December 2023. We included clinical trials, observational or diagnostic tests studies that evaluated the diagnostic performance of at least one CPR compared to relevant reference standard. We performed DTA-MA using univariate random effects model and calculated the combined sensitivity (Se), specificity (Sp) and positive/negative likelihood ratios. Results: We included 27 studies (9,442 children). The mean patients age range was 0.5-4.4 years. The studies evaluated 17 CPR. We performed meta-analyses and obtained combined sensitivity and specificity, respectively, for the following CPRs: BPS (Se 89.4%, 95%CI 82.1-93.9%; Sp 76.6%, 95%CI 55.5 to 89.6%, very low and low certainty), and Lynch (Se 90.5%, 95%CI 74.5- 96.9%; Sp 28.4%, 95%CI 12.8 to 51.7%; very low and low certainty). The only CPR that showed good performance with moderate certainty for ruling in bacterial infection was the one by Pervaiz (LR+ 11.05; 95CI 7.70-15.9). Conclusions: With moderate certainty the rule by Pervaiz et al, showed good performance for ruling it the bacterial CAP. No CPR showed good performance for ruling out bacterial CAP with acceptable certainty