0496 Multi-Site Validation of a Simple EHR-based Algorithm for Obstructive Sleep Apnea. (27th April 2018)

• Record Type: Journal Article
• Authors: Keenan, B T; Kirchner, H; Veatch, O J; Borthwick, K M; Davenport, V A; Gendy, M; Pack, F; Sirikulvadhana, L; Malow, B A; Morgenthaler, T I; Zee, P C; Robishaw, J D; Derose, S F; Pack, A I
• Is Part Of: Sleep. Volume 41(2018)Supplement 1
• Page Start: A186
• Page End: A187
• Publication Date: 2018-04-27
• Subjects: Sleep -- Physiological aspects -- Periodicals; Sleep disorders -- Periodicals; Sommeil -- Aspect physiologique -- Périodiques; Sommeil, Troubles du -- Périodiques; Sleep disorders; Sleep -- Physiological aspects; Sleep -- physiological aspects; Sleep Wake Disorders; Psychophysiology; Electronic journals; Periodicals; 616.8498
• DOI: 10.1093/sleep/zsy061.495
• Languages: English
• ISSNs: 0161-8105

Abstract

Abstract: Introduction: A simple algorithm to accurately distinguish cases and controls using the electronic health record (EHR) is crucial for large-scale analyses leveraging available clinical biorepositories. Combining this algorithm with de-identified genomic data linked to EHRs offers the opportunity to leverage existing resources to decipher heterogeneity in complex sleep disorders, like obstructive sleep apnea (OSA). We examined the performance of such an algorithm to define OSA across 5 health centers in the United States. Methods: Cases were defined as individuals with at least 2 instances of ICD-9 and/or ICD-10 diagnostic codes reflective of OSA in their medical record (on separate dates) and controls as individuals with zero OSA-related diagnostic codes. Using chart reviews on at least 100 cases and 100 controls at each site, diagnostic performance [sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV)] was evaluated at Geisinger, University of Pennsylvania, Kaiser Permanente Southern California, Northwestern University and Vanderbilt University Medical Center. Results: Chart reviews were performed in 1, 087 randomly selected individuals (571 EHR-cases, 516 EHR-controls) across sites; the sample was middle-aged (57.3 ± 13.8 years), obese (33.4 ± 11.1 kg/m 2 ) and a majority female (54.5%) and Caucasian (85.4%). The algorithm showed excellent performance among all patients, with estimates (95% CI) of sensitivity [96.1 (94.2, 97.6)], specificity [95.6 (93.4, 97.2)], PPV [96.0 (94.0, 97.4)], and NPV [95.7 (93.6, 97.3)] all >95%. The algorithm performed similarly well within each site, with all estimates of diagnostic performance ≥90%, apart from specificity [86.4 (78.5, 92.2)] and PPV [87.5 (80.2, 92.8)] at one site. A modified algorithm of ≥3 codes resulted in estimates of ~95% for all characteristics at this site, but excluded the 18.3% of cases with only 2 codes. On average, EHR-cases were similar in age to controls (p=0.459), but 6.7 kg/m 2 more obese (p<0.0001) and more likely to be male (39.0% vs. 51.3%, p<0.0001). Conclusion: A simple EHR-based algorithm for OSA showed excellent predictive characteristics in a multi-site sample from the US. We are currently leveraging this algorithm for large-scale genetic association studies of OSA. Support (If Any): NIH Grant R01 HL134015.