Surveillance bias and deep vein thrombosis in the National Trauma Data Bank: The more we look, the more we find

Background: Deep vein thrombosis (DVT) has been identified as a marker of quality of care by various governmental and consumer groups. However, the lack of standardized DVT screening systems across trauma centers may introduce surveillance bias in the rates of DVT reported. We hypothesize that trauma centers with higher rates of duplex ultrasound detect more DVTs and subsequently report higher DVT rates to the National Trauma Data Bank. Methods: We queried the National Trauma Data Bank version 6.1 and calculated ultrasound rates and DVT rates per trauma center. We excluded hospitals that did not report performing any ultra-sounds or any complications. Simple and multiple linear regressions were used to describe the association between ultrasound and DVT rates among hospitals. Results: One hundred forty-seven hospitals (16%) met the inclusion criteria, accounting for 578,252 patients (39% of the total patients in the dataset). When dividing hospitals into quartiles by duplex ultrasound rate, the DVT rate in the highest quartile was 7-fold higher than the average combined DVT rate in the first three quartiles (1.52% vs. 0.22%; p < 0.001). Multivariable analysis suggested that hospitals with an ultrasound rate <= 2% had a 1.07% increase in reported DVT rate for every 1% increase in ultrasound rate (95% confidence interval 1.05-1.09; P < 0.001). Conclusions: More aggressive screening procedures may be associated with higher DVT rates. Trauma centers that screen more and report higher DVT rates may be falsely labeled as having decreased quality of care. Using DVT rate alone as an independent quality measure should be reevaluated because of the potential for surveillance bias.