Background: Deep vein thrombosis (DVT) is a difficult to diagnostic disease. The aim of this study was to determine the utility and accuracy of a risk stratification questionnaire and a diagnostic strategy, which were applied to patients with suspected DVT on lower extremities in an emergency department.
Patients and method: A prospective cohort study was performed in 569 outpatients with clinical suspected DVT during 14 months. The applied questionnaire stratified patients into three pre-test probability categories. Items included signs, symptoms, risk factors and potential alternative diagnosis, which were based on a modified Wells clinical model. DVT was diagnosed by the combined use of clinical model, compression ultrasonography (CUS) and follow-up CUS one week later in those moderate-high risk patients with an initial normal test. These patients were followed over three months for the development of venous thromboembolic complications.
Results: Two hundred three (35.7%) patients were classified as having a low, 186 (32.7%) moderate and 180 (31.6%) high clinical probability. Overall, DVT was diagnosed in 153 patients (26%; CI95%, 23.2-30.7%): 144 (96%) at the initial CUS, 6 (3.5%) at the second testing and 3 over the 3-month follow-up period. 22 patients had a low pretest probability (11%; CI95%, 7-16%), 43 (23%; CI95%, 17-30%) moderate, and 88 (49%; CI95%, 41-56%) high pretest probability. The difference in the prevalence of DVP among risk categories was significant (p < 0.00001). When the high and moderate groups were joined, the model had a 86% sensitivity, a 90% negative predictive value and a 43% specificity for diagnosis of DVT.
Conclusions: The clinical model used in this study is accurate and feasible, though it is not enough to take clinical decisions. The diagnostic strategic used is effective but not efficient.