Diagnostic E-codes for commonly used, narrow therapeutic index medications poorly predict adverse drug events

Objective: We sought to examine the validity of specific hospital discharge codes in identifying drug toxicity precipitating hospitalization, among elderly users of high-risk medications. Study Design and Setting: We conducted a cross-sectional evaluation assessing the diagnostic test characteristics of International Classification of Diseases-9 External-Cause-of-Injury codes (E-codes) compared with a reference standard of medical record review. This study was nested within a prospective cohort of elders using warfarin, digoxin, or phenytoin as identified in the Pharmaceutical Assistance Contract for the Elderly benefit program. Results: We identified 4,803 subjects contributing 11,409 person-years of exposure to at least one of three drug groups. Subjects experienced 8,756 hospitalizations, of which 304 were deemed, by expert review, to be a result of an adverse event of warfarin, digoxin, or phenytoin. The sensitivity, specificity, and positive (PPVs) and negative predictive values for drug-specific E-codes were warfarin-25.5%, 98.3%, 46.6%, and 95.7%; digoxin-84.0%, 99.1%, 56.8%, and 99.8%; and phenytoin-86.7%, 98.7%, 59.1%, and 99.7%. Conclusions: E-codes for digoxin and phenytoin have a high sensitivity, but E-codes for all three medications have poor PPVs, a result that might produce misclassification in studies based solely on discharge coding. Investigators should confirm such rare events via medical record review. (c) 2008 Elsevier Inc. All rights reserved.