Statistics from Altmetric.com
Chalmers et al1 and Loke et al2 present excellent meta-analyses of the value of various tools in predicting mortality from community-acquired pneumonia (CAP). There is a continuing fallacious belief, however, that only patients at high risk of death are at high risk of complications. Of the 47 studies identified by Chalmers and Loke, only 16 made any assessment of the value of these scores in predicting the need for critical care. These are presented in table 1, together with a further five studies previously identified in the field.3 6 11 14 15
Forest plots of sensitivity and specificity are shown in figure 1, allowing the calculation of pooled estimates (with 95% CIs): CRB-65 sensitivity 0.467 (0.428 to 0.506) and specificity 0.825 (0.817 to 0.833); CURB sensitivity 0.484 (0.447 to 0.521) and specificity 0.72 (0.708 to 0.732); CURB-65 sensitivity 0.499 (0.479 to 0.519) and specificity 0.734 (0.728 to 0.74); PSI sensitivity 0.755 (0.743 to 0.767) and specificity 0.486 (0.481 to 0.491).
As concluded in Ewig's editorial,24 none of the existing mortality predictor tools performs adequately in identifying patients who will need high intensity care, and therefore the application of these tools to protocols or guidelines for sites of care should be with caution.