Development of a score model to predict long‐term prognosis after community‐onset pneumonia in older patients.

Background and Objective: The identification of factors associated with long‐term prognosis after community‐onset pneumonia in elderly patients should be considered when initiating advance care planning (ACP). We aimed to identify these factors and develop a prediction score model. Methods: Patients aged 65 years and older, who were hospitalized for pneumonia at nine collaborating institutions, were included. The prognosis of patients 180 days after the completion of antimicrobial treatment for pneumonia was prospectively collected. Results: The total number of analysable cases was 399, excluding 7 outliers and 42 cases with missing data or unknown prognosis. These cases were randomly divided in an 8:2 ratio for score development and testing. The median age was 82 years, and there were 68 (17%) deaths. A multivariate analysis showed that significant factors were performance status (PS) ≥2 (Odds ratio [OR], 11.78), hypoalbuminemia ≤2.5 g/dL (OR, 5.28) and dementia (OR, 3.15), while age and detection of antimicrobial‐resistant bacteria were not associated with prognosis. A scoring model was then developed with PS ≥2, Alb ≤2.5, and dementia providing scores of 2, 1 and 1 each, respectively, for a total of 4. The area under the curve was 0.8504, and the sensitivity and specificity were 94.6% and 61.7% at the cutoff of 2, respectively. In the test cases, the sensitivity and specificity were 91.7% and 63.1%, respectively, at a cutoff value of 2. Conclusion: Patients meeting this score should be considered near the end of life, and the initiation of ACP practices should be considered. We aimed to identify factors influencing the long‐term prognosis of elderly patients with pneumonia and to develop a prediction score model for the effective initiation of advance care planning. Performance status (PS) ≥2, hypoalbuminemia ≤2.5 g/dL and dementia were significant factors. The scoring model demonstrated 94.6% sensitivity and 61.7% specificity. [ABSTRACT FROM AUTHOR]