Validity of Pneumonia Severity Assessment Scores in Low- and Middle-Income Countries: A Systematic Review and Meta-Analysis

Abstract

Background: Pneumonia treatment decisions are often guided by the use of severity assessment scores, such as the well-known and validated pneumonia severity index and CURB-65. Several scoring systems have been developed, but the evidence of their utilisation in low- and middle-income countries remains limited.Objective: In this review, we sought to systematically investigate and provide a comprehensive synthesis of the evidence around the validity and predictive value of the existing pneumonia severity assessment scores in adult patients diagnosed with community-acquired pneumonia in low- and middle-income countries.Methods: Medline (Ovid), Embase (Ovid), Cochrane Central Register of Controlled Trials, Scopus, and Web of Science were searched for eligible articles up to May 2020. Following the screening and eligibility assessment, relevant data were extracted, and 2x2 contingency tables were constructed according to the definition of severe pneumonia and the patients' reported endpoints. The association between high severity scores and the studied outcome was tested. Using the bivariate model, the performance of severity scores at their usual cut- off in predicting the reported outcome was examined. Heterogeneity was assessed using the I2 index.Key Findings: The search yielded 12,897 articles; 11 were included following the screening. Of these, only six studies with sufficient data were included in the final analysis that involved examining CURB-65 and CRB-65 severity scores. In patients with pneumonia, both scores at a threshold ≥3 were related to an increased mortality risk, with pooled relative risks of 8.58 (3.48-21.18) and 4.83 (2.52-9.28) for CURB-65 and CRB-65, respectively. The predictive performance of CURB-65 and CRB-65 at the usual cut-off, respectively, were as follows: the pooled sensitivity, 0.69 (0.25-0.94) and 0.04 (0.00-0.40); the pooled specificity, 0.89 (0.72- 0.96) and 0.99 (0.95-1.00); and the area under the summary receiver operator characteristic curves, 0.90 (0.87-0.92) and 0.86 (0.83-0.89). Conclusion: CURB-65 and CRB-65 at a cut-off ≥3 are strongly associated with mortality and appear to be valid scores for mortality prediction. Although CURB-65 exhibited higher sensitivity and overall accuracy, it is likely that CRB-65 is also an appropriate assessment score to employ where urea levels are unavailable.