Venous thromboembolism diagnosis definition in claims data: implications for research.

Abstract

To examine whether the variability in ICD codes used for venous thromboembolism (VTE) definition among published studies affects VTE rates and associations detected from claims data. We extracted the ICD codes used for VTE definition from three published studies and proposed a new VTE definition based on clinical review of all utilized ICD codes. We compared these four definitions to assess differences in VTE rates and associated variables using a standardized scenario. We used a random 25% sample of the IQVIA PharMetrics® Plus for Academics database and analyzed patients undergoing gastrointestinal cancer surgery. The primary outcome was 90-day post-discharge VTE. The association of preoperative and intraoperative variables with VTE was assessed using bivariate and multivariable main effects logistic regression models. There were substantial differences in the use of ICD codes among the 4 VTE definitions (range 116 to 304 ICD-9/10 codes). Our population included 2,122 eligible patients (median age 59 years, 47% female) and the rate of VTE ranged from 2.3% to 4.4% using the four definitions. Multivariable analysis showed that the associations between VTE and age and type of surgery (esophageal surgery, gastric surgery) varied depending on the VTE definition used while the Elixhauser comorbidity score and liver surgery type were consistently associated with VTE. In this pilot study, differences in the incidence of VTE and associated risk factors were influenced by the choice of ICD9/10 codes used to define VTE. A standardized definition of VTE may improve the reproducibility and rigor of findings based on administrative claims data.