In the nonpregnant general population, deep venous thrombosis (DVT) can be excluded at the initial presentation not only by clinical assessment and review of risk factors, but also by the clinician's assessment of probability using subjective criteria or prediction rules. Because of the absence of a validated prediction rule for diagnosing DVT during pregnancy, diagnosis in pregnant patients depends largely on the use of compression ultrasonography, the diagnostic reference standard that establishes the presence or absence of DVT. The aim of this cross-sectional study was to investigate the predictive accuracy of clinicians' subjective assessment of the probability of DVT in women with suspected DVT and to identify useful objective prediction variables. The participants were 194 unselected pregnant women with suspected first DVT. Potential study patients were identified by clinicians during the initial examination when the clinical assessment was consistent with DVT, and led to consultation with thrombosis experts who obtained informed consent for participation in the study. Before compression ultrasonography testing, the experts examined each study patient and categorized their pretest probability of DVT based on subjective criteria as low, moderate, or high. The sensitivity, specificity, negative predictive value, and likelihood ratios of the clinicians' subjective pretest probability were calculated. Patients defined as DVT positive had abnormal diagnostic compression ultrasonography results at the initial presentation or during serial testing, or had symptomatic DVT or pulmonary embolism on follow-up. Patients with negative compression ultrasonography at presentation and no venous thromboembolism on follow-up were defined as DVT negative. DVT was diagnosed in 17 women (8.8%). The prevalence of DVT in patients with low pretest probability (two-third of patients) was 1.5% (2/131), with a 95% confidence interval of 0.4% to 5.4% (P < 0.001). The prevalence among those with moderate or high pretest probability was considerably higher, 24.6% (15/61), with a 95% confidence interval of 15.5% to 36.7% (P < 0.001). Multivariate analysis identified 3 independent variables that were highly predictive of DVT in pregnant women: left leg symptoms, difference in calf circumference of 2 cm or more, and presentation during the first trimester of pregnancy. DVT was unlikely in the absence of all 3 variables, and its probability increased substantially if 1 or more was present. These findings demonstrate that probability of DVT in pregnant women with suspected DVT may be assessed with use of 3 easily determined objective variables. The data show that DVT is very unlikely with a very low pretest probability.