Comparison of Diagnostic Sleep Studies in Hospitalized Neurorehabilitation Patients With Moderate to Severe Traumatic Brain Injury

Abstract

BACKGROUND: OSA is prevalent during a time of critical neural repair after traumatic brain injury (TBI). The diagnostic utility of existing sleep studies is needed to inform clinical management during acute recovery from TBI. RESEARCH QUESTION: This study aimed to evaluate the non-inferiority and diagnostic accuracy of a portable level 3 sleep study relative to level 1 polysomnography in hospitalized neurorehabilitation patients with TBI. STUDY DESIGN AND METHODS: This is a prospective clinical trial conducted at six TBI Model System study sites between May 2017 and February 2019. Of 896 admissions, 449 were screened and eligible for the trial, with 345 consented. Additional screening left 263 eligible for and completing simultaneous administration of both level 1 and level 3 sleep studies, with final analyses completed on 214 (median age ¼ 42 years; ED Glasgow Coma Scale ¼ 6; time to polysomnography [PSG] ¼ 52 days). RESULTS: Agreement was moderate to strong (weighted kappa ¼ 0.78, 95% CI, 0.72-0.83) with the misclassification commonly occurring with mild sleep apnea due to underestimation of apnea hypopnea index (AHI). Most of those with moderate to severe sleep apnea were correctly classified (n ¼ 54/72). Non-inferiority was not demonstrated: the minimum tolerable specificity of 0.5 was achieved across all AHI cutoff scores (lower confidence limits [LCL] range, 0.807-0.943), but the minimum tolerable sensitivity of 0.8 was not (LCL range, 0.665-0.764). INTERPRETATION: Although the non-inferiority of level 3 portable diagnostic testing relative to level 1 was not established, strong agreement was seen across sleep apnea indexes. Most of those with moderate to severe sleep apnea were correctly identified; however, there was risk of misclassification with level 3 sleep studies underestimating disease severity for those with moderate to severe AHI and disease presence for those with mild AHI during early TBI neurorehabilitation.

The Impact of Opioid Medications on Sleep Architecture and Nocturnal Respiration During Acute Recovery From Moderate to Severe Traumatic Brain Injury: A TBI Model Systems Study

“…As previously described, the interrater agreement and scoring of sleep and respiratory events were strong to very strong with most the studies rated as interpretable/scoreable (97.2%). 35,36…”

“…Potential participants were consecutive patients enrolled in the TBI Model Systems (TBIMS) lifetime longitudinal study at 6 of the 21 study sites (Tampa, Florida; Seattle, Washington; Dallas, Texas; Columbus, Ohio; Englewood, Colorado; and Elkins Park, Pennsylvania), over a 19-month period (from May 2017 to January 2019), from as part of a separate Patient-Centered Outcomes Research Institute-funded diagnostic comparative effectiveness trial. [33][34][35][36] All participating sites received institutional review board approval for conduct of the study with consent obtained by patients or designated medical proxy using TBIMS procedures. Complete study inclusion/exclusion criteria for the Patient-Centered Outcomes Research Institute trial are described in prior publications and listed in Appendix 1.…”

“…Complete study inclusion/exclusion criteria for the Patient-Centered Outcomes Research Institute trial are described in prior publications and listed in Appendix 1. 35,36 Twenty-five percent (n = 63) of the study sample came from a single Veterans' Administration site.…”

A Flexible Deep Learning Architecture for Temporal Sleep Stage Classification Using Accelerometry and Photoplethysmography

“…Nocturnal PSG and overlapping ACC recordings from 271 participants involved in a study of sleep disordered breathing in patients undergoing rehabilitation from a traumatic brain injury [35] were also included. The study extent and scope is described elsewhere [35].…”

“…Nocturnal PSG and overlapping ACC recordings from 271 participants involved in a study of sleep disordered breathing in patients undergoing rehabilitation from a traumatic brain injury [35] were also included. The study extent and scope is described elsewhere [35]. Of these, 12 were excluded because they were missing either ACC or PPG signals, and further 28 were excluded because they had unstable sample frequency.…”