CCERW: a new rule to predict treatment failure in patients with community-acquired pneumonia in middle aged and elderly people

Abstract

To develop and validate a clinical rule to predict treatment failure in middle-aged and elderly patients suffering from community-acquired pneumonia (CAP) in China, and to compare it with other prognostic rules. Data of 58 variables affiliated to 6 aspects, including demographics, underlaying diseases, previous status, complications, symptoms, signs and laboratory examination results from the CAP patients aged≥45 years admitted to the respiratory departments in three university affiliated hospitals between December 17, 2006 and December 22, 2008 were enrolled prospectively and then validated in two groups to create a derivation cohort with 75% of the patients for rule development and an internal validation cohort with the other 25% for internal test. An external validation cohort was formed at the same time with patients admitted to the other university affiliated hospital for external test. The single outcome was treatment failure at the time of 14 days after admitted or at discharge from hospital. Univariate analysis, multivariate analysis and receiver operating characteristics (ROC) curve were used for the rule establishment, assessment and comparison among the pneumonia severity index (PSI), CURB65 [confusion, blood urea nitrogen>6.8 mmol/L, respiratory rate (RR)≥30 breaths per minute, systolic blood pressure (SBP)<90 mm Hg (1 mm Hg=0.133 kPa) or diastolic blood pressure (DBP)≤60 mm Hg, age≥65 years] and CRB65 (confusion, RR≥30 breaths per minute, SBP<90 mm Hg or DBP ≤60 mm Hg, age≥65 years). The data of a total of 539 patients in three hospitals were enrolled for analysis. Of those, 400 and 139 patients were randomly allocated into the derivation cohort or internal validation cohort respectively. Meanwhile, 159 patients were enrolled in the external validation cohort. Analyzing the data from 400 patients in the derivation cohort, 33 variables of 6 aspects had significant difference between cure and treatment failure outcome in the univariate analysis. Then, in the multivariate analyses, five independent predictive factors showed significant difference, including confusion (C), creatinine<60 μmol/L, electrolyte disturbances (E), respiratory failure (R), white blood cell (WBC)>7.5×10(9)/L. A clinical prediction rule CCERW based on these variables showed that the treatment failure outcome increased directly with increasing scores: 5.5%-9.1%, 12.8%-20.0% and 31.0%-40.5% for scores of 0-1, 2 and 3-6, respectively. ROC curve analysis yielded an area under the curve (AUC) for CCERW of 0.709 [95% confidence intervals (95%CI) 0.638-0.780], 0.725 (95%CI 0.613-0.838) and 0.686 (95%CI 0.590-0.782) in the derivation, internal and external validation cohorts respectively; and in the same manner, of 0.710 (95%CI 0.659-0.761) for total 698 patients, which was better than PSI, CURB65 and CRB65, at 0.667 (95%CI 0.614-0.719), 0.648 (95%CI 0.592-0.705), and 0.584 (95%CI 0.530-0.638), respectively. CCERW can help physicians to distinguish high and low risk leading to treatment failure in middle-aged and elder patients with CAP, and has better predictable capability than PSI, CURB65 and CRB65. We prudent recommend the simple rule can be used in the middle-aged and elder patients with CAP of Han race people in China.