0785 Validation of a Two-Stage Questionnaire and Actigraphy Screening for Isolated REM Sleep Behavior Disorder in a Multicenter Case-Control Cohort

Introduction Isolated rapid eye movement sleep behavior disorder (iRBD) is a prodromal marker of synucleinopathies. Yet most cases remain undiagnosed due to the poor predictive value of questionnaires and limited access to video-polysomnography (vPSG). We evaluated a 2-stage screening strategy combining a 4-item questionnaire on dream enactment, subjective hyposmia, constipation and orthostatic symptoms (stage 1), followed by home wrist actigraphy (stage 2). Methods Participants aged 40-80 without a diagnosis of neurodegenerative disease were recruited from five Mount Sinai and Stanford cohorts. All iRBD cases were vPSG-confirmed. The full cohort included 396 participants (99 cases, 297 controls; mean age 64 ± 11; 55% male), of which 289 participants completed the questionnaire, 236 completed 2-week wrist actigraphy, and 129 (75 cases, 54 controls) both assessments. The wearable-based algorithm was built on 4 movement features (mean motor activity, activity index, short or long immobile bouts, twitch activity). Questionnaire-based and wearable-based models were trained with nested cross-validation using XGBoost. Results The dream enactment question alone achieved an area under the curve (AUC) of 0.85, which improved to 0.86 with the full 4-item questionnaire. Across the questionnaire dataset (95 cases, 194 controls), dream enactment showed 78% sensitivity and 92% specificity, while the 4-item model achieved 78% sensitivity and 91% specificity. Actigraphy alone achieved an AUC of 0.88, with 82% sensitivity and 84% specificity. At an assumed population prevalence of 1.5%, adjusted positive predictive value was 10% for the 4-item questionnaire and 6% for actigraphy. Among those completing both assessments (75 cases, 54 controls), the 2-stage protocol yielded 68% sensitivity and 100% specificity when preselecting individuals based on the dream-enactment question alone, compared with 73% sensitivity and 100% specificity using the full 4-item questionnaire. Conclusion A 2-stage protocol combining questionnaire and actigraphy demonstrated excellent specificity, good sensitivity, and robust generalization for iRBD across both centers and cohorts. This low-cost, scalable strategy is compatible with widely used wearable devices and warrants validation in community-based populations. Support (if any) Department of Neurology, Icahn School of Medicine at Mount Sinai.