Diagnostic accuracy of Gram staining when predicting staphylococcal hospital-acquired pneumonia and ventilator-associated pneumonia: a systematic review and meta-analysis

Ranzani, Otavio; Motos, Ana; Chiurazzi, Chiara; Ceccato, Adrian; Rinaudo, Mariano; Bassi, Gianluigi Li; Ferrer, Miquel; Torres, Antoni

doi:https://dx.doi.org/10.1016/j.cmi.2020.08.015

Diagnostic accuracy of Gram staining when predicting staphylococcal hospital-acquired pneumonia and ventilator-associated pneumonia: a systematic review and meta-analysis

Citation

Ranzani OT, Motos A, Chiurazzi C, Ceccato A, Rinaudo M, Bassi GL et al. Diagnostic accuracy of Gram staining when predicting staphylococcal hospital-acquired pneumonia and ventilator-associated pneumonia: a systematic review and meta-analysis. Clin Microbiol Infect. 2020 Nov;26(11):1456-1463. DOI: 10.1016/j.cmi.2020.08.015

Abstract

Background: There is no clear guidance on empirical antibiotic coverage against Staphylococcus aureus for hospital-acquired pneumonia (HAP) and ventilator-associated pneumonia (VAP). Objective: To evaluate whether the presence of clusters of Gram-positive cocci in Gram staining of respiratory samples predicts S. aureus as HAP/VAP pathogen. Methods: Data sources were MEDLINE, PubMed, Embase, Scielo, CINAHL and Scopus, from inception to 15/07/2017 (update on 31/10/2019), and original data from a single-centre database (PROSPERO: CRD42017072138). We included studies reporting the diagnostic accuracy of a Gram-staining evaluation suggestive of Staphylococcus compared with a positive culture for S. aureus in any type of lower respiratory tract sample. Participants were adult patients with HAP/VAP. The index test was morphological evaluation of Gram staining of respiratory samples. We followed PRISMA guidelines and assessed risk of bias and applicability with the QUADAS-2 tool. We conducted a meta-analysis using a bivariate random effects model. Results: We selected five studies that included only VAP and data from a single-centre database including VAP and HAP. We pooled six studies for VAP and analysed 1665 respiratory samples. Pooled sensitivity was 68% (95%CI 49-83 and specificity 95% (95%CI 86-98). The pooled positive likelihood ratio was 12.7 (95%CI 5.1-31.6), negative likelihood ratio 0.34 (95%CI 0.20-0.57), diagnostic odds ratio 38 (95%CI 13-106) and area under the summary receiver operating curve (SROC) 0.91 (95%CI 0.88-0.93). There was great heterogeneity between sensitivity and specificity. In scenarios in which the prevalence of S. aureus was between 5% and 20%, the positive and negative predictive values were 62% (95%CI 47-77) and 95% (95%CI 82-100), respectively. Conclusions: Detection of Gram-positive cocci in clusters in respiratory samples of patients with VAP has the potential to guide risk assessments of S. aureus for more personalized antibiotic coverage. Randomized clinical trials with patient-centred outcomes are needed for strong clinical recommendations.