Abstract T P384: Development of a Screening Tool for Obstructive Sleep Apnea in the Stroke Population
Background: Obstructive Sleep Apnea (OSA) occurs in >50% patients with stroke/ TIA; treatment can improve outcomes. Current OSA screens do not perform well in stroke patients. The objective of this study is to develop an OSA screening tool for use in stroke patients.
Methods: The cohort consisted of patients seen in a cerebrovascular clinic who completed the STOP survey Jan 2011 - Dec 2012 and underwent polysomnography (PSG) for OSA within 1 year. The STOP tool, a commonly used OSA screen, served as the base model to enhance. Six prediction models were created. The dependent variable was presence of OSA (defined by AHI ≥ 10) identified on PSG. The predictor variables, in addition to STOP items, were based upon lit review and electronically extracted from the EHR, except for neck circumference which was recorded at the time of the PSG. Receiver operating characteristic (ROC) curves were created and sensitivity and specificity of various cutoff scores were calculated for each model. Validation was done using bootstrap methods with 200 iterations.
Results: There were 208 patients in cohort with mean age 55.4 yrs, 49.0% were female. Sleep apnea was diagnosed in 61.0%. ROC curves for the models are displayed in the Figure. Models with the highest c-statistics included BMI, age and gender, plus the STOP items (STOP-BAG). Addition of neck circumference and other variables did not significantly improve the models.
The STOP-BAG2 model, using continuous variables, had a sensitivity of 0.94 (95% 0.89-0.98) and specificity 0.60 (95% CI 0.49 - 0.71) at a cutpoint= 0.395. The STOP-BAG model, which used dichotomous versions of the variables, had a sensitivity of 0.91 (95% CI 0.85-0.96) and specificity of 0.48 (95% CI 0.37 - 0.60) at cutpoint= 3.
Conclusion: The STOP-BAG, consisting of 7 variables available in most EHRs, can be used to identify cerebrovascular patients at increased risk for OSA. A version that uses continuous variables (STOP-BAG2), can be used if automated score calculation can be done.
Author Disclosures: I. Katzan: None. N. Thompson: None. K. Uchino: None. N. Foldvary: None.
- © 2015 by American Heart Association, Inc.