Background: Multiple factors can lead to overuse of diagnostic imaging for venous thromboembolism (VTE). These include fear of missing a potentially fatal diagnosis, low predictive accuracy of clinical gestalt, prediction rules, and risk assessment models for VTE, and increased access to imaging. Diagnostic imaging for deep venous thrombosis (DVT) with doppler ultrasonography, or for pulmonary embolism (PE) with computed tomography angiography (CTA) of the chest or ventilation/perfusion (V/Q) scans, involves exposing the patient to radiation and contrast agents (with risk of nephropathy or allergic reactions), besides causing patient discomfort, and adding to the healthcare costs. We define diagnostic yield as the proportion of diagnostic tests for VTE that report positive results for VTE. The objective of this study is to assess the diagnostic yield of imaging for VTE in a large population-health database, and to determine the patient characteristics associated with low or high diagnostic yield of imaging for VTE.
Methods: This is a retrospective cohort study using claims data from Optum Clinformatics™ Data Mart between October 2015 and June 2019. We included any patient encounter where a diagnostic imaging for VTE was performed. We excluded encounters with a diagnostic imaging for VTE in the prior 3 months, as they could be a follow up scan for a known VTE. We conducted a descriptive analysis to assess the prevalence of VTE and used generalized estimating equations logistic regression model to assess the association between the known predictors or predisposing factors for VTE and finding VTE on diagnostic imaging.
Results: Between October 2015 and June 2019, 1,502,417 doppler ultrasonograms of the extremities, 620,527 CTA of the chest, and 89,736 V/Q scans were performed. Within 3 months of the first test, 23.2% of the patients had ≥1 repeat doppler ultrasonograms, 25.5% had ≥1 repeat CTA of the chest, and 41.6% had ≥1 repeat V/Q scans. Of the tests performed in the emergency room (ER), inpatient, and outpatient setting, DVT was reported in 10.4%, 16.9%, and 6.5% of the tests, respectively, and PE was reported in 6.4%, 18.7%, and 8.8% of the tests, respectively. Of the doppler ultrasonograms performed in the ER, inpatient, and outpatient setting, D-dimer was ordered in 11.9%, 3.9%, and 2.6% of the patients, respectively. Of the diagnostic tests for PE performed in the ER, inpatient, and outpatient setting, D-dimer was ordered in 43.0%, 10.2%, and 11.5% of the patients, respectively. Of the patients who had a doppler ultrasonograms of the extremities, low diagnostic yield was more likely in females, non-white, those with a history of VTE, obesity, heart failure or acute myocardial infarction, and pregnant women; whereas high diagnostic yield was more likely in those with a history of cancer, surgery, central venous access, thrombophilia, respiratory failure, inflammatory bowel disease, and prescription of hormones (Figure 1). Of the patients who underwent imaging to detect PE, low diagnostic yield was more likely in females and those with a history of VTE; whereas high diagnostic yield was more likely in those with a history of cancer, surgery, central venous access, thrombophilia, obesity, heart failure or acute myocardial infarction, varicose veins, and prescription of hormones (Figure 2).
Conclusions: The diagnostic yield of imaging for VTE was low, especially in the ER and outpatient settings. One of the limitations of this study is that the indication for the imaging could have been unrelated to VTE.