Repeated pneumonia severity index measurement after admission increases its predictive value for mortality in severe community-acquired pneumonia

J Formos Med Assoc. 2009 Mar;108(3):219-23. doi: 10.1016/S0929-6646(09)60055-3.

Abstract

Background/purpose: Severe community-acquired pneumonia (CAP) is associated with high hospital mortality, and accurate assessment of patients is important for supporting clinical decision making. The Pneumonia Severity Index (PSI) is a good tool for predicting disease severity, especially in the low-risk group of patients with CAP. We investigated whether the change in PSI measurement after admission could identify patients at high risk of mortality from CAP.

Methods: We prospectively studied 250 inpatients with CAP. PSI was measured at admission and 72 hours later at a tertiary referral medical center from May 2005 to February 2006. The initial and repeated PSI results were compared. Hospital mortality was used as the outcome measure.

Results: Initial PSI in high-risk patients (PSI class > IV) had a low specificity (37%), and a low positive predictive value (PPV) (17%). Increased repeated PSI score, as compared with initial score, was associated with an increased mortality rate (from 7.8% to 33.3% in class IV, and 25.3% to 53.3% in class V; p < 0.0001), and improved the predictive value, with 94% specificity and a PPV of 46% for mortality in high-risk patients.

Conclusion: Increased PSI score, 72 hours after admission, for patients with CAP improved the predictive value of PSI score and more accurately identified patients with a high risk of mortality.

Publication types

  • Comparative Study

MeSH terms

  • Community-Acquired Infections / diagnosis
  • Community-Acquired Infections / mortality*
  • Female
  • Follow-Up Studies
  • Hospital Mortality / trends
  • Humans
  • Male
  • Middle Aged
  • Patient Admission*
  • Pneumonia / diagnosis
  • Pneumonia / mortality*
  • Prognosis
  • Prospective Studies
  • Severity of Illness Index*
  • Survival Rate / trends
  • Taiwan / epidemiology
  • Time Factors