316 Development of a phenotyping algorithm to identify patients with autoimmune disease in electronic health records for future large scale studies

Autoimmune diseases (AD) are chronic debilitating disorders that collectively affect 5-8% of the population and result in annual direct health care costs exceeding $100 billion in the US. Due to the high prevalence of AD that affect the skin, dermatology bears a substantial proportion of the disease burden created by autoimmunity. While it is well established that having an AD increases the risk of developing other AD, more recent evidence suggests that these patients may also be at increased risk for metabolic and neuropsychiatric disorders. The long-term goal of this project is to leverage the massive amount of data harbored in electronic health records (EHR) of patients with AD to rigorously test for comorbidities not traditionally thought to be immune-mediated, specifically by utilizing resources in the Electronic Medical Records and Genomics (eMERGE) Network. This nationwide consortium of academic medical centers links genetic data to EHR for 105,325 patients. Here we report the development and validation of a phenotyping algorithm to construct an AD cohort from EHR data for future epidemiological and genetic studies of autoimmunity. We identify cases by searching for patients with at least 3 entries of an ICD code for at least one of 522 ICD codes that indicate the presence of 51 different AD. Patients not defined as cases are excluded from controls by the presence of 1 or more of a set of 28,419 ICD codes, or positive results for antibody titer tests. Our phenotyping algorithm was validated by chart reviews conducted for 161 patients by physicians from three clinical domains, which indicated a positive predictive value of 86% for cases and 93% for controls. The rigorous interrogation of large datasets to substantiate neuropsychiatric and metabolic comorbidities could improve the quality of care of AD patients and open new opportunities for gaining mechanistic insight into disease causes and potential therapeutic targets.