Introduction and objectives: We aimed to develop a new prediction model of mortality in cancer patients with pneumonia and to compare its performance with CURB-65 and the Pneumonia Severity Index (PSI).
Methods: Active cancer patients who were diagnosed with pneumonia at the Emergency Department (ED) from 7/1/2014 to 12/31/2014 were consecutively included. Clinical data were collected through a medical chart review. The primary outcome was the 28-day mortality, and clinical factors were analyzed using logistic regression analysis.
Results: Among a total of 218 analyzed patients with a median age of 64.0 years (IQR, 56.8-71.0) and a male proportion of 72%, 42 (19.3%) died within 28 days of ED admission. By multivariate logistic regression analysis, an ECOG performance status (PS) 3 (OR: 8.54, 95% CI: 3.42-21.33) or 4 (OR: 13.17, 95% CI: 3.19-54.32), SpO2 <90% (OR: 3.06, 95% CI: 1.17-8.00), and elevated lactic acid levels (OR: 1.42, 95% CI: 1.12-1.81) were significantly associated with mortality. With these three variables, a new prediction model with total scores ranged from 0 to 6 was generated. The area under the curve of the new prediction model was 0.840, compared with 0.673 and 0.586 for CURB-65 and PSI, respectively.
Conclusion: In cancer patients with pneumonia, a poor ECOG PS, SpO2 <90%, and lactic acid elevation are independent predictors of mortality. The new prediction model, comprising three predictors, performs better in predicting mortality in cancer patients than CURB-65 or PSI.
Keywords: cancer; mortality; pneumonia; prediction.
© 2016 John Wiley & Sons Ltd.