Abstract WP189: Prediction of Apnea Hypopnea Index among Stroke and TIA Patients
INTRODUCTION: Obstructive sleep apnea (OSA) is common among patients with stroke and TIA. Previously validated OSA screening instruments used in the general population are largely based on reported symptoms; when applied to those with cerebrovascular disease, they have shown poor correlation with the presence of OSA diagnosed via polysomnography (PSG). We sought to develop a cerebrovascular disease specific prediction model for OSA that is less reliant on symptomatology. Methods: We used data from a multi-site randomized, controlled strategy trial that included ischemic stroke and TIA patients at two VA hospitals. Data on patient demographics, vital signs, anthropomorphic measurements, past medical history, medications, stroke severity, and sleep questionnaires were obtained. All patients received PSG. Sleep apnea was defined as present if the apnea-hypopnea index (AHI) was ≥5. Because approximately half of subjects had OSA, discrimination was difficult; thus, AHI was used as the outcome for our prediction model. Epworth Sleepiness Scale (ESS)≥10, the Berlin Questionnaire>10, and Sleep Apnea Clinical Score (SACS)≥15 were considered ‘high risk’ for OSA. Bivariate regression models were used to assess the strength of the association between predictors and the outcome (log AHI+1); those statistically significant at the 0.1 level were entered into a multi-variable regression model. Backward elimination was used until all remaining variables were significant at the 0.05 level. Results: Among 194 Veterans, 119 (61.3%) had an AHI≥5. Neck circumference > 16 inches, systolic blood pressure less than 132 mmHg, peripheral vascular disease, and recent hospitalization for congestive heart failure were associated with increase in log AHI; history of chronic obstructive pulmonary disease/asthma was associated with a decrease in log AHI (R2=0.13). ESS≥10 (p=0.86) and Berlin Questionnaire >10 (p=0.80) were not associated with log AHI. The SACS≥15 was associated with AHI (p=0.01); however, it did not remain significant in the final model. Conclusions: A model using blood pressure, neck circumference and past medical history data was predictive of AHI. Further work is required to validate the use of this model in a larger cerebrovascular disease cohort.