0447 Assessing The Performance Of Billing Codes To Identify Patients With Central Sleep Apnea From The Electronic Medical Record

Abstract

Introduction: Research on central sleep apnea (CSA) has often relied on International Classification of Disease (ICD) billing codes. However, the accuracy of this approach has not been evaluated. We aimed to assess the validity of an ICD code approach for identifying patients with CSA. Method(s): We evaluated the validity of CSA ICD codes in a sleep and cardiovascular disease cohort of 5,181 patients aged >=18 years. Patients were selected into this cohort if they underwent an in-lab diagnostic sleep study in one of six University of Pittsburgh Medical Center sleep laboratories between 01/2004 and 12/2017 and either right heart catheterization or echocardiography. ICD diagnosis of CSA was based on ICD-9 codes 327.21, 327.27 and 786.04 or ICD- 10 codes R06.3, G47.31, and G47.37. Medical records including all sleep study reports as well as inpatient and outpatient clinical notes were extracted and reviewed by two physicians in all 101 patients with a CSA ICD code and a random sample of 135 patients without a CSA ICD code in order to verify CSA status. Result(s): Among the 101 patients with a CSA ICD billing code, a CSA diagnosis was confirmed in 62 patients for a positive predictive value of 61.4%. Among the 135 patients without a CSA ICD billing code, a CSA diagnosis was excluded in 131 patients for a negative predictive value of 97.0%. After accounting for sampling probabilities, the sensitivity and specificity of CSA based on presence of a CSA ICD billing code were 29.2% and 99.2% respectively. Conclusion(s): In a clinical cohort selected for the presence of cardiovascular disease, the use of ICD billing codes to identify CSA has low positive predictive value which could reduce power to detect the impact of CSA on health outcomes.