Embedded Files
Birmingham, and Duke University) with special expertise in large data set analyses to execute RISE data requests. These centers will work with de-identified data extracts; any identified data, including text from clinical notes, will remain within the clinical data warehouse and will be analyzed centrally. Analysis of initial data in RISE. We analyzed data in RISE from October 1, 2014 through September 30, 2015, including the characteristics of participating practices and patients. We examined both sociodemographic characteristics of patients (age, sex, race/ethnicity, insurance status and US region), as well as clinical characteristics (smoking status, blood pressure, billing and problem list diagnoses as defined by the International Classification of Diseases, Ninth Revision [ICD-9] codes, and medications). All sociodemographic characteristics and clinical data were examined from the most recent clinical encounter. When querying diagnoses, we generated a list of ICD-9 codes for common rheumatic diseases and calculated the frequency of these diagnoses at the last available clinical encounter. We did not require these diagnoses to be mutually exclusive; in other words, patients with 2 diagnoses are represented twice. Because a key initial focus of the registry is rheumatoid arthritis (RA), we examined clinical data for the subset of patients with this diagnosis in further detail to demonstrate the type of data in RISE. First, we examined active medications at the end of the most recent RA clinical encounter, including all nonbiologic DMARDs (methotrexate, hydroxychloroquine, leflunomide, sulfasalazine, azathioprine, minocycline, cyclosporine, cyclophosphamide, penicillamine, and gold), as well as biologic and targeted synthetic DMARDs (etanercept, adalimumab, infliximab, golimumab, certolizumab, abatacept, anakinra, tocilizumab, rituximab, and tofacitinib). To create these categories, we built algorithms that collated instances where a single drug (e.g., methotrexate) appeared more than once as an active medication. This meant removing duplicate prescriptions and also collapsing instances where 2 formulations of the same drug were listed. Similarly, we built algorithms to create mutually exclusive categories for the biologic DMARDs. For example, we manually reviewed all records in which patients were listed as taking 2 tumor necrosis factor–inhibiting biologic drugs (1.1% of records). We contacted a sample of practices to understand the workflows leading to these within-class duplicates. Reasons identified included incomplete medication reconciliation or clinicians routinely ordering more than 1 drug at time of initiation while waiting for insurance authorization. In these instances, we either looked into Figure 1. The Rheumatology Informatics System for Effectiveness (RISE) registry data flow for practice improvement and research. The RISE registry uses a lightweight connector to establish a connection between a practice’s electronic health record (EHR) server and the centralized registry servers. Local practice data are iteratively mapped by the RISE team to ensure accuracy, and verified extracts enter the RISE clinical data warehouse. Data are then aggregated and analyzed for federal quality reporting, research, and practice-based improvement. Authorized users in a practice can access the RISE data dashboard through a web-based registry portal. 1868 Yazdany et al future encounters to see which drug persisted and selected only the persistent drug, or randomly sampled one of the drugs. Among individuals in RISE, we also examined performance on selected quality measures, including those recently endorsed by the National Quality Forum (NQF), such as assessment of disease activity (NQF
Report abuse