Pneumonia in Patients with Acute Stroke: Derivation of a Clinical Prediction Rule

Abstract

Background: Pneumonia is a common and dangerous consequence of acute stroke. Others have derived pneumonia prediction scores to guide management but included information on swallowing function, which may not be routinely available at the time of hospital admission. Methods: We conducted a retrospective cohort study to identify predictors of pneumonia incidence (identified by ICD-9 codes 480-486.99 or 507.xx) among 1,924 patients treated for acute stroke at Harborview Medical Center in Seattle, WA, based upon information available only at admission and without information on swallowing function. Risk estimates were determined after adjustment for age. Multivariate associations were determined using logistic regression, with beta coefficients for the predictor variables providing weights for each item in a prediction score. Results: Pneumonia occurred in 291 (15%). Independent predictors of pneumonia, and their contribution to the pneumonia score were the following: Age over 75 years (2 pts); male (1 pt); National Institutes of Health Stroke Scale score over 10 (2 pts); receipt of mechanical ventilation (4 pts); history of coronary artery disease (1 pt); lack of diabetes (1 pt); and history of chronic obstructive pulmonary disease (1 pt). The resulting 12-point score had an area under the receiver operating characteristic curve of 0.79 (95% confidence interval 0.76 to 0.81). Conclusion: Our pneumonia prediction score achieved fair discriminatory accuracy despite not including information of swallowing function. However, there may be limited clinical utility of a tool to screen for pneumonia in patients with acute stroke that is only as accurate as this one.