Predicting pneumonia mortality using CURB-65, PSI, and patient characteristics in patients presenting to the emergency department of a comprehensive cancer center

The prognostic accuracy of the CURB-65 criteria and pneumonia severity index (PSI) in immunocompromised cancer patients with pneumonia is unknown. We sought to determine whether CURB-65 and PSI predict 28-day mortality in cancer patients with pneumonia, and identify other factors that predispose cancer patients with pneumonia to a high mortality risk. We assessed sensitivities, specificities, predictive values, and areas under the receiver operating curve area under the curve (AUC) of the CURB-65 and PSI in predicting the 28-day mortality of cancer patients presenting to our institution's emergency department with pneumonia. We used the DeLong and Clarke–Pearson approach to determine whether the addition of other risk factors improved the scales' performances. The overall and pneumonia-related 28-day mortality rates were 20.2% (n = 44) and 17.4% (n = 38), respectively. In predicting 28-day mortality, the CURB-65 score had sensitivity of 45% and specificity of 81%, and the PSI score had sensitivity of 82% and specificity of 34%. The CURB-65 and PSI discriminated poorly between fatal and nonfatal pneumonia cases (AUCs, 0.664 and 0.658, respectively; 95% confidence interval [CI], 0.57–0.75 for each). The addition of radiation therapy (RT) within 4 weeks and stem cell transplant (SCT) significantly improved the AUCs of the CURB-65 (0.75; 95% CI, 0.67–0.83) and PSI (0.73; 95% CI, 0.65-0.82). Inadequate performances of CURB-65 and PSI demonstrate that a tool for predicting pneumonia-related mortality in cancer patients and other immunocompromised populations is needed. Pneumonia patients who have undergone recent RT or (SCT) are at a high risk of dying from pneumonia and require special consideration when assessing pneumonia-related mortality risk.