RESUMO
BACKGROUND: Venous thromboembolism is a serious perioperative complication. We developed a new pre-test probability score for predicting deep vein thrombosis (DVT) before surgery. METHODS: Whole leg ultrasonography was performed on 973 inpatients and outpatients with suspected DVT based on a preoperative D-dimer cut-off value of ≥ 1 µg/ml. We allocated two-thirds (n = 651) of the study participants to a derivation cohort and one-third (n = 322) to a validation cohort. The pre-test probability model was developed from the derivation cohort data. RESULTS: The pre-test probability model for DVT assigned 2 points to D-dimer ≥ 1.5 µg/mL and 1 point each to age ≥ 60 years, female sex, ongoing glucocorticoid therapy, prolonged immobility, and cancer with high risk of DVT. The area under the curve of the pre-test probability score was 0.72 and 0.70 in the derivation and validation cohorts, respectively. The rates of DVT according to pre-test probability scores in the derivation and validation cohorts were 7% and 6% in the low (score = 0-2), 23% and 22% in the intermediate (score = 3-4), and 47% and 50% in the high probability group (score ≥ 5), respectively (p < 0.0001). CONCLUSIONS: The pre-test probability score (Kagoshima-DVT score) was helpful in detecting preoperative DVT in both inpatients and outpatients. We identified low probability group to reduce whole-leg ultrasonography and high probability group to detect more DVT before surgery.