Administrative codes inaccurately identify recurrent venous thromboembolism: The CVRN VTE study

Background: Studies using administrative data commonly rely on diagnosis codes to identify venous thromboembolism (VTE) events. Our objective was to assess the validity of using International Classification of Disease, 9th Revision (ICD-9) codes in identifying recurrent VTE. Materials and methods: Among 5497 adults with confirmed incident VTE from four healthcare delivery systems in the Cardiovascular Research Network (CVRN), we identified all subsequent inpatient, emergency department (ED), and ambulatory clinical encounters associated with an ICD-9 code for VTE (combined with relevant radiology procedure codes for inpatient/ED VTE codes in the secondary discharge position or outpatient codes) during the follow-up period. Medical records were reviewed using standardized diagnostic criteria to assess for the presence of new, recurrent VTE. The positive predictive value (PPV) of codes was calculated as the number of valid events divided by total encounters. Results: We identified 2397 encounters that were considered potential recurrent VTE by ICD-9 codes. However, only 31.1% (95%CI: 29.3-33.0%) of encounters were verified by reviewers as true recurrent VTE. Hospital or ED encounters with VTE codes in the primary position were more likely to represent valid recurrent VTE (PPV 61.3%, 95%CI: 56.7-66.3%) than codes in secondary positions (PPV 35.4%, 95%CI: 31.9-39.3%), or outpatient codes (PPV 20.3%, 95%CI: 18.3-22.5%). PPV was low for all VTE types (29.9% for pulmonary embolism, 38.3% for lower and 37.7% for upper extremity deep venous thrombosis, and 14.1% for other VTE). Conclusions: ICD-9 codes do not accurately identify new VTE events in patients with a prior history of VTE.