A validation and potential modification of the pneumonia severity index in elderly patients with community-acquired pneumonia

OBJECTIVES: To evaluate the discriminatory power of the Pneumonia Severity Index (PSI) in elderly patients with community-acquired pneumonia (CAP) and to improve its performance. DESIGN: Retrospective review of 193 patients from 1999 to 2001 to derive prognostic rules. The rules were prospectively validated in 144 patients from 2002 to 2003. SETTING: Iwata City Hospital, a 400-bed general hospital. PARTICIPANTS: Patients aged 80 and older who had CAP and were admitted to the hospital. MEASUREMENTS: Predictors of 30-day mortality were identified using logistic regression analysis, and several rules were constructed by combining the PSI and the independent predictors. Classification and regression tree analysis was also used for constructing rules. Sensitivity and specificity were calculated. RESULTS: The original PSI, which defines PSI Class Wand V as a high-risk group, did not perform well in discriminating survivors from nonsurvivors (sensitivity 100%, specificity 15%), whereas a modified PSI, which defines only PSI Class V as a high-risk group, performed better (sensitivity 86%, specificity 63%). Three predictors for mortality were identified independent from the modified PSI: performance status (PS) Grade 3 or higher, anorexia, and partial pressure of carbon dioxide of 50 mmHg or greater. By combining the modified PSI and PS, the performance could be further improved (sensitivity 79%, specificity 80%). CONCLUSION: The modified PSI could identify low-risk patients more accurately than the original PSI. In addition, by combining the modified PSI with PS, higher performance was obtained. Such information would aid physicians in clinical decision-making without overestimating the risk for elderly patients with CAP.