Abstract 2695: A Clinical Prediction Rule for Pneumonia after Acute Stroke

Abstract

INTRODUCTION: Dysphagia screens aim to reduce aspiration and pneumonia (PNA) after stroke. The frequency of PNA after stroke is low, especially in those with mild stroke who have not been intubated. We sought to identify a subset of patients with acute stroke who are at such low risk of PNA that bypassing a dysphagia screen would be justified. METHODS: Harborview Medical Center, a tertiary care academic and primary stroke center, maintains a database of all patients admitted with stroke. From this, we identified 1,641 adults admitted to the Neurology service between 2007 and 2010 for acute ischemic or hemorrhagic stroke (HS). We excluded those with subarachnoid hemorrhage, those who had been intubated, except if only for a procedure, and those who lacked admission NIHSS. With PNA as outcome of interest, we sought associations with potential predictors using multiple logistic regression. We then created a score to predict PNA with each item’s weight based upon its magnitude of association with PNA in the multivariable model. RESULTS: For the resulting 1,008 subjects, average age was 64.0 (+/- 18.2), 57.8% were male, 64.2% had ischemic stroke, and average NIHSS was 7.7 (+/- 8.5). PNA was diagnosed in 67 (6.6%) during the hospitalization. Age, stroke type, NIHSS, atrial fibrillation (AF), heart failure (HF), and chronic obstructive pulmonary disease (COPD) had significant univariate association with development of PNA. In multivariable models, all but AF retained significance. Adjusted odds ratios and 95% confidence intervals were: age (per year: 1.02, 1.00 - 1.04), NIHSS (per point: 1.04, 1.02 - 1.06), HS (1.67, 1.00 - 2.80), HF (2.62, 1.37 - 5.04), COPD (2.39, 1.23 - 4.64). The PNA score ranged from 0-10 with points assigned as follows: age ≥ 75 (2 pts), NIHSS 5-11 (1pts), NIHSS ≥ 12 (3 pts), HS (1 pt), HF (2 pts), COPD (2 pts). Only 8 cases of PNA were diagnosed in the 452 patients with a score of ≤1 (frequency 1.8%). The remaining 59 cases occurred in those 556 patients with a score of >1 (frequency 10.6%). This predictive model yielded a receiver operator curve with area under the curve of 0.76. DISCUSSION: We developed a clinical prediction rule to estimate risk of PNA after acute stroke based upon age, stroke type and severity, and presence of HF and COPD. Presence of dysphagia was not one of the factors we examined because we were interested in seeing if we could identify a subgroup of patients with such low risk of PNA that bypassing a swallowing screen would be justified. Validation of the rule in an independent sample is needed before it could be used to influence clinical decisions. Limits of our study are the reliance on an existing database, which lacked potentially important predictors such as stroke location. Also, we do not know whether PNA was due to aspiration and stroke nor whether HF and COPD played a causal role or merely increased likelihood of PNA detection.