Objective: Twenty-five percent of people with traumatic brain injury (TBI) have been found to have obstructive sleep apnea (OSA) (Mathias & Alvaro, 2012) – rates that are much higher than in the general population. Sleep disturbances have the potential to negatively impact inpatient rehabilitation, recovery, and outcomes. Therefore, identification of persons at risk for OSA is imperative to address the impact of sleep early after TBI. Prior research has not provided information about the potential impact of TBI severity on the presence or absence of OSA. The purpose of the present study is to examine the relationship between TBI severity and presence of OSA during inpatient rehabilitation.
Design: Cross-sectional analysis of an ongoing, three year multicenter prospective observational cohort study comparing methods to screen for and diagnose OSA.
Setting: Six inpatient rehabilitation centers across the United States.
Main Outcome Measure(s) and Analysis: The primary outcome (OSA) was assessed with Level 1 polysomnography and characterized using the American Academy of Sleep Medicine (AASM) Obstructive Apnea-Hypopnea Index. Participants were classified as either having a diagnosis of OSA or not having a diagnosis of OSA. A binomial logistic regression was performed to ascertain the effects of age, body mass index (BMI), neck circumference, and TBI severity on the likelihood that participants were diagnosed with OSA. 
Results/Conclusions: The sample consisted of 233 participants who were primarily male (83%) with an average age of 43 years (SD=18). The most common causes of injury were motor vehicle collisions (42%) and falls (32%). 67% of participants met AASM criteria for diagnosis of OSA. BMI ranged from 14 to 50 (M = 25.99, SD = 5.15). The average number of days in post-traumatic amnesia was 35.87 (SD = 48.55) and ranged from 0 to 323 days. The logistic regression model was statistically significant, Χ2 46.27, p < .0001. The model explained 41% of the variance (Nagelkerke R2) in OSA and correctly classified 77.6% of the cases. Sensitivity was 85.9%, specificity 59.5%, positive predictive value was 82.3%, and negative predictive value was 65.8%. Of the four predictor variables, only age was statistically significant. Increasing age was associated with an increased likelihood of being diagnosed with OSA (OR = 1.10, p < .0001). Severity of TBI was not found to be a statistically significant predictor of OSA diagnosis (p = .17).
When considering age, BMI, neck circumference, and severity of TBI, the model correctly classified the presence or absence of OSA in 78% of participants. In this sample, age was the only significant predictor of OSA. Contrary to the general population, BMI, and neck circumference did not significantly increase the odds of being diagnosed with OSA.

Risk factors of Obstructive Sleep Apnea among Individuals with Moderate to Severe Traumatic Brain Injury