Validation of a claims-based diagnostic code for Stevens-Johnson syndrome in a commercially insured population

PURPOSE: To validate the administrative claims identification of a diagnosis of Stevens-Johnson syndrome (SJS) using medical records as the "gold standard" in a large, commercially insured US population. METHODS: Patients with >1 medical claim with the International Classification of Diseases, Ninth Revision, Clinical Modification code 695.1x between 1 July 2000 and 31 May 2007 were queried in the HealthCore Integrated Research Database(SM) , which contains administrative claims data for 14 commercial health insurance plans. Trained nurses and pharmacists abstracted pertinent information from the identified patients' medical records, which were then reviewed by two independent dermatologists to identify criteria to determine SJS diagnosis. Positive predictive values (PPVs) based on the claims and chart data were computed for all the cases. RESULTS: Medical charts for 200 claims-identified cases, with the International Classification of Diseases, Ninth Revision, Clinical Modification code 695.1x, were abstracted and reviewed by the dermatologists. A total of five cases (PPV = 2.50%, 95%CI = 0.8%-5.7%) were determined to be SJS with clinical certainty. PPVs varied with data stratification: PPV for inpatient claims only (PPV = 2.00%, 95%CI = 0.24%-7.04%), inpatient claims with 695.1x in first diagnosis field (PPV = 4.11%, 95%CI = 0.86%-11.54%), and final decisions of either clinical certainty or probable cases of SJS (PPV = 6.00%, 95%CI = 3.14%-10.25%). CONCLUSION: These findings demonstrate the difficulties associated with identifying rare disorders, which lack specific diagnostic criteria, within administrative claims databases. They underscore the challenges of using claims data to monitor ill-defined clinical conditions as well as the need to validate claims-identified cases with information from other sources, such as medical charts. Copyright © 2012 John Wiley & Sons, Ltd.