805 Towards a more perfect union: interrater reliability between telemedicine and in-person evaluators in sleep disordered breathing

Abstract

Abstract Introduction Telemedicine has rapidly changed the landscape of all of clinical practice, and is now widely employed in sleep medicine. To date, the accuracy of telemedicine in identifying patients at risk for obstructive sleep apnea (OSA) is still unknown. Given differences in technique and fidelity, a perfect correlation between these two types of encounters cannot be assumed. We studied how providers using a telemedicine platform compared to providers using traditional in-person encounters in identifying risk for sleep disordered breathing. Methods 90 participants referred to a comprehensive university sleep program were randomized to this interrater reliability study. Subjects were representative of the gender and ethnic breakdown of the outlying community. The subjects were evaluated by an in-person clinician investigator, then randomized to a second clinician investigator who performed an evaluation online using a common teleconferencing platform. Both types of evaluations included a history and physical exam. The primary outcome was pretest probability for obstructive sleep apnea. Secondary comparators included level of daytime sleepiness, snoring volume, apneas witnessed by a third party, modified Mallampati score, presence/absence of tonsils, degree of overjet bite, and severity of apnea based on home sleep testing. Results Interrater reliability for pretest probability of OSA was reflected in a weighted kappa value of 0.414 (SE 0.090, p=0.002). This suggests moderate agreement between the an in-person and telemedicine evaluator. Kappa values of our secondary outcomes ranged from -0.044 (degree of maxillary overjet) to 0.702 (apneas witnessed by a third party), and were generally higher for historical elements and lower for physical exam findings. Conclusion Evaluation for pretest probability for sleep apnea via telemedicine has a moderate interrater correlation with in-person assessment. A relatively high degree of interrater reliability for historical elements suggests that the accuracy of telemedicine for OSA is tempered by a suboptimal physical exam. Telemedicine evaluations might become more accurate through standardization. For instance, using validated scales for OSA or templated encounter scripts may help with risk-stratification, and ultimately lead to more uniform management. Support (if any) This study was supported by an AASM Foundation Focused Project Award