Abstract TP397: Validation of a Pneumonia Score in Acute Stroke
INTRODUCTION: Accurate prediction of pneumonia (PNA) risk after stroke would 1) allow clinicians to target interventions to patients at highest risk, and 2) help researchers to determine the efficacy of those interventions. We previously derived a PNA risk score based on data available at the time of admission, purposefully leaving out information on swallowing function, which may not be available at time of admission. Items for the 11-point score and point values were: age > 75 (2), male (1), National Institutes of Health Stroke Scale score > 10 (2), mechanical ventilation (4), coronary artery disease (1), chronic obstructive pulmonary disease (1). In a retrospective single-hospital cohort of 1,924 patients with acute ischemic stroke and intracranial hemorrhage (ICH), we used medical records and discharge diagnosis codes to derive the score: C-statistic = .79 (95% CI, .76 - .81). In the current study, we tested whether the score could accurately predict PNA in two other cohorts.
METHODS: The one cohort (n=398 with acute ischemic stroke or ICH) was obtained retrospectively and presented to various hospitals within the same city during a time period prior to the derivation cohort. Data on predictor variables and the outcome of PNA were obtained from medical records and discharge diagnosis codes. The other cohort (n=89 with acute ischemic stroke) was a subset of the derivation cohort. Data were collected prospectively, and the diagnosis of PNA was ascertained using rigorous criteria that included clinical, radiographic, and culture data.
RESULTS: Within the retrospective cohort, PNA was diagnosed in 46 (12%), and the score achieved a C-statistic of .71 (95% CI, .66 -.75). Within the prospective cohort, pneumonia was diagnosed in 9 (10%), and the C-statistic for the score was .88 (95% CI, .79 -.94).
CONCLUSION: The predictive value of the PNA score was validated in two additional cohorts, one with data collected retrospectively and the other, prospectively. The score performed best within the prospective cohort, but sample size was relatively small and the 89 patients were a subset of the derivation cohort. Further refinement and validation of the score is planned.
- © 2012 by American Heart Association, Inc.