Performance of various pneumonia severity models for predicting adverse outcomes in elderly inpatients with community-acquired pneumonia

Objective

We aimed to assess the performance of common pneumonia severity scores, such as pneumonia severity index (PSI), CURB-65, CRB-65, A-DROP, and SMART-COP, in predicting adverse outcomes in elderly community-acquired pneumonia cohort and to determine the optimal scoring system for specific outcomes of interest.

Methods

A total of 822 elderly inpatients were included in the retrospective cohort study. Clinical and laboratory results on admission were used to calculate the above scores. The primary outcome was 30-day mortality. Secondary outcomes were in-hospital mortality, need for mechanical ventilation (MV) and ICU admission. Model discrimination was evaluated by the area under receiver operating characteristic curves (AUCs).

Results

The 30-day and in-hospital mortality rates were 6.8% (56/822) and 8.6% (71/822), respectively. One hundred and ninety-eight (24.0%) received MV and 111 (13.5%) were admitted to the ICU. All five scoring systems showed the same trend of increasing rates of each adverse outcome with increasing risk groups (all p < 0.001). PSI had the highest AUC, sensitivity, and negative predictive value (NPV) in predicting 30-day mortality and in-hospital mortality. SMART-COP had the highest AUC for predicting the need for MV and ICU admission, but PSI had the highest sensitivity and NPV for these two outcomes.

Discussion

PSI performed well in identifying elderly patients at risk for 30-day mortality and its high NPV is helpful in excluding patients who are not at risk. Considering their effectiveness and simplicity, SMART-COP and CURB-65 are easier to perform in clinical practice than PSI.