Background: The aim of this meta-analysis was to determine if severity assessment tools can be used to guide decisions regarding intensive care unit (ICU) admission of patients with community-acquired pneumonia.
Methods: A search of PUBMED and EMBASE (1980-2009) was conducted to identify studies reporting pneumonia severity scores and prediction of ICU admission. Two reviewers independently collected data and assessed study quality. Performance characteristics were pooled using a random-effects model.
Results: Sufficient data were collected to perform a meta-analysis on five current scoring systems: the Pneumonia Severity Index (PSI), the CURB65 score, the CRB65 score, the American Thoracic Society (ATS) 2001 criteria and the Infectious Disease Society of America/ATS (IDSA/ATS) 2007 criteria. The analysis was limited due to large variations in the ICU admission criteria, ICU admission rates and patient characteristics between different studies and different healthcare systems. In the pooled analysis, PSI, CURB65 and CRB65 performed similarly in terms of sensitivity and specificity across a range of cut-offs. Patients in CURB65 group 0 were at lowest risk of ICU admission (negative likelihood ratio 0.14; 95% confidence interval 0.06-0.34) while the ATS 2001 criteria had the highest positive likelihood ratio (7.05; 95% confidence interval 4.39-11.3).
Conclusion: Large variations exist in the use of ICU resources between different studies and different healthcare systems. Scoring systems designed to predict 30-day mortality perform less well when ICU admission is taken into account. Further studies of dedicated ICU admission scores are required.