0711 REM Sleep Behavior Disorder Diagnostic Code Accuracy and Implications in the Outpatient Setting

Abstract Introduction Isolated REM sleep behavior disorder (iRBD) carries increased risk for neurodegenerative parkinsonian disorders or dementia (NPD), but is difficult to accurately screen for in the community. Healthcare data offer the opportunity to identify large numbers of iRBD cases among outpatients. We aimed to determine the positive predictive value (PPV) of an RBD international classification of disorders (ICD) code for actual iRBD based on manual review of the electronic health record (EHR), examine risk of NPD diagnosis, and explore whether a statistical model developed using selected EHR data can identify individuals with RBD ICD code that have high probability for actual iRBD. Methods In this retrospective cohort study, an EHR search at a single healthcare system was conducted to identify outpatient cases who received ICD9 or ICD10 RBD code in 2011-2021. The EHR for each case was manually reviewed. Cases with secondary causes of RBD were excluded. Remaining cases were classified as no iRBD or actual iRBD. Incident cases of NPD were identified. PPV of presence of RBD ICD code for actual iRBD was calculated. Cumulative incidence of NPD with death as a competing event was compared in those with vs without iRBD. Least absolute shrinkage and selection operator (LASSO) regression was used to build a prediction model for iRBD and the model was validated in an independent dataset. Results Among 1,130 cases with RBD ICD code, 499 had secondary causes of RBD. For the remaining 628 cases, determination based on EHR review was that 168 (26.8%) did not have iRBD; PPV of RBD ICD code was 73.25%. Compared to the no iRBD group, the iRBD group had a higher risk of NPD (sub-distribution hazard ratio=10.4 (95% CI 2.5-43.1)). The LASSO prediction model for iRBD had an Area Under the Receiver Operating Characteristic Curve of 0.844 (95%CI 0.806-0.880). Conclusion PPV of an RBD ICD code is moderate. In the real-world setting, patients with iRBD had a high risk of diagnosis of NPD. Results indicate feasibility of using statistical models developed using EHR data to accurately predict iRBD. Support (if any) This work was partially funded by the University of Pittsburgh Dean’s Faculty Advancement Award (Chahine PI).