The Paul Coverdell National Acute Stroke Registry provides feedback to states on adherence to guidelines of care to improve quality of care for hospitalized patients with stroke and transient ischemic attack in hospitals across 6 states (Georgia, Massachusetts, Michigan, Minnesota, North Carolina, and Ohio).6
Hospital participation is voluntary. Trained abstractors from participating hospitals collect detailed information on stroke and transient ischemic attack admissions concurrent with or soon after patient care using standard data definitions provided by the Centers for Disease Control and Prevention.6,7
The main outcome for this study, pneumonia, was defined as nosocomial, hospital-acquired pneumonia occurring during hospitalization, at least ≥48 hours after admission, documented by a physician, and requiring antibiotic treatment. The main explanatory variable, swallow screening, and results of the screening had to be documented by a healthcare professional in the medical record before oral intake of food, fluid, or medications. If there was no documentation as to the performance of a swallow screen being done or the results of the screen were not documented, then abstractors were to assume it was not done or results were missing, respectively. If the swallow screen was contraindicated, then the contraindication must be documented in the medical record. Patients in whom a swallow screen was contraindicated were not included in this analysis. A screening test was not necessarily a formal evaluation of swallowing by a speech and language pathologist but had to be a standardized method of swallowing assessment accepted by the institution. Acceptable methods of dysphagia screening included the following: bedside swallow assessment, simple water swallow test, Burke water swallow test, simple standardized bedside swallowing assessment, barium swallow, video fluoroscopy, double-contrast esophagoscopy, radionucleotide studies, manometry, endoscopy, and formal evaluation by a speech–language pathologist. Documentation of the National Institutes of Health Stroke Scale (NIHSS), testing of cranial nerves, or checking the gag reflex were not considered to be swallow screens.
Included in our analyses were patients with ischemic and hemorrhagic stroke aged ≥18 years discharged between March 1 and December 31, 2009. Excluded were patients who remained nothing by mouth throughout the hospital stay and also patients who had a documented reason for not undergoing DS (eg, intubation). We also excluded 6452 patients whose length of stay was ≤2 days reasoning that it usually takes at least 3 days for the development and detection of pneumonia as well as 5 patients with length of stay >120 days.
Patient characteristics and outcomes were compared according to patient DS status (passed screening, failed screening and unscreened). The χ2 was used for categorical variables and the Wilcoxon-Mann-Whitney rank test for continuous variables.
Logistic regression was used to assess the association between the outcome variable of interest, pneumonia during hospitalization, and DS status after adjusting for age, sex, race, and presence of clinical features (aphasia, weakness, altered level of consciousness). Stroke severity indices, the NIHSS for patients with ischemic stroke and Glasgow Coma Scale (GCS) for patients with hemorrhagic stroke, were missing in a large fraction of patients. Hence, clinical features, collinear with the severity indices, were used as surrogates of severity in the models. The adjusted OR were obtained along with 95% CIs. All statistical analyses were performed at the Centers for Disease Control and Prevention using SAS 9.2 (Cary, NC).