Validation of an administrative claims-based diagnostic code for pneumonia in a US-based commercially insured COPD population

David M Kern,1 Jill Davis,2 Setareh A Williams,3 Ozgur Tunceli,1 Bingcao Wu,1 Sally Hollis,4 Charlie Strange,5 Frank Trudo2 1HealthCore, Inc., Wilmington, DE, 2AstraZeneca Pharmaceuticals, Wilmington, DE, 3AstraZeneca Pharmaceuticals, Gaithersburg, MD, USA; 4AstraZeneca Pharmaceuticals, Cheshire, UK; 5Department of Medicine, Division of Pulmonary and Critical Care Medicine, Medical University of South Carolina, Charleston, SC, USA Objective: To estimate the accuracy of claims-based pneumonia diagnoses in COPD patients using clinical information in medical records as the reference standard.Methods: Selecting from a repository containing members’ data from 14 regional United States health plans, this validation study identified pneumonia diagnoses within a group of patients initiating treatment for COPD between March 1, 2009 and March 31, 2012. Patients with ≥1 claim for pneumonia (International Classification of Diseases Version 9-CM code 480.xx–486.xx) were identified during the 12 months following treatment initiation. A subset of 800 patients was randomly selected to abstract medical record data (paper based and electronic) for a target sample of 400 patients, to estimate validity within 5% margin of error. Positive predictive value (PPV) was calculated for the claims diagnosis of pneumonia relative to the reference standard, defined as a documented diagnosis in the medical record.Results: A total of 388 records were reviewed; 311 included a documented pneumonia diagnosis, indicating 80.2% (95% confidence interval [CI]: 75.8% to 84.0%) of claims-identified pneumonia diagnoses were validated by the medical charts. Claims-based diagnoses in inpatient or emergency departments (n=185) had greater PPV versus outpatient settings (n=203), 87.6% (95% CI: 81.9%–92.0%) versus 73.4% (95% CI: 66.8%–79.3%), respectively. Claims-diagnoses verified with paper-based charts had similar PPV as the overall study sample, 80.2% (95% CI: 71.1%–87.5%), and higher PPV than those linked to electronic medical records, 73.3% (95% CI: 65.5%–80.2%). Combined paper-based and electronic records had a higher PPV, 87.6% (95% CI: 80.9%–92.6%).Conclusion: Administrative claims data indicating a diagnosis of pneumonia in COPD patients are supported by medical records. The accuracy of a medical record diagnosis of pneumonia remains unknown. With increased use of claims data in medical research, COPD researchers can study pneumonia with confidence that claims data are a valid tool when studying the safety of COPD therapies that could potentially lead to increased pneumonia susceptibility or severity. Keywords: positive predictive value, pneumonia, validation, claims data, medical record review