ObjiectiveTo obtain reliable evidence of diagnosis and treatment through evaluating the validity of pneumonia severity index (PSI), CURB-65 and acute physiology and chronic health evaluationⅡ(APACHEⅡ) scores in predicting risk stratification, severity evaluation and prognosis in elderly community-acquired pneumonia (CAP) patients.MethodsClinical and demographic data were collected and retrospectively analyzed in 125 in-hospital patients with CAP admitted in Shanghai Dahua Hospital from January 2012 to April 2015. The severity of pneumonia was calculated with PSI, CURB-65 and APACHEⅡgroups during 1 to 3 days after admission. Mortality and intensive care unit (ICU) admission rates were evaluated among patients in each scores and was categorized into three classes, namely mild, moderate and severe groups during 1 to 3 days after admission. Mortality and ICU admission rates were evaluated among patients in each severity level. Through evaluating the sensitivity, specificity, the predicting values and the area under receiver operating characteristic (ROC) curve (AUC) among PSI, CURB-65 and APACHEⅡ, the validity and consistency of these three scoring systems were assessed.ResultsUsing PSI, CURB-65 and APACHEⅡ scoring systems, the patients were categorized into mild severity (48.8%, 64.0% and 52.8%, respectively), moderate severity (37.6%, 23.2% and 32.0%, respectively) and severe severity (13.6%, 12.8% and 15.2%, respectively). In PSI, CURB-65 and APACHEⅡ systems, the mortality in high risk groups was 41.3%, 62.5% and 47.4%, respectively; The ICU-admission rate in high risk groups was 88.3%, 100.0% and 94.7%, respectively. The sensitivity of PSI, CURB-65 and APACHEⅡ was 50.0%, 71.4% and 64.3% in predicting mortality, and was 46.8%, 50.0% and 59.3% in predicting ICU-admission, respectively. PSI, CURB-65 and APACHEⅡ showed similar specificity (approximately 90%) in predicting mortality and ICU admission. ROC was conducted to evaluate the sensitivity of PSI, APACHEⅡ and CURB-65 in predicting mortality and ICU admission. The AUC had no significant difference among these three scoring systems. The AUC of PSI, CURB-65 and APACHEⅡwas 0.893, 0.871, 0.880, respectively for predicting mortality, and was 0.949, 0.837, 0.949, respectively for predicting ICU admission. There was no significant difference among these three scoring in predicting mortality and ICU admission (all P>0.05).ConclusionsPSI, CURB-65 and APACHEⅡ performed similarly and achieved high predictive values in elderly patients with CAP. The three scoring systems are consistent in predicting mortality risk in elderly CAP patients. The CURB-65 is more sensitive in predicting the risk of death, and more early in identifing patients with high risk of death. The APACHEⅡ is more sensitive in predicting the risk of ICU admission, and has good value in identifying severe patients and choosing the right treatment sites.