Overuse of the d-dimer to screen for possible pulmonary embolism (PE) can have negative consequences. This study derives and tests clinical criteria to justify not ordering a d-dimer. The test threshold was estimated at 1.8% using the method of Pauker and Kassirer. The PE rule-out criteria were derived from logistic regression analysis with stepwise backward elimination of 21 variables collected on 3148 emergency department patients evaluated for PE at 10 US hospitals. Eight variables were included in a block rule: Age < 50 years, pulse < 100 bpm, SaO(2) > 94%, no unilateral leg swelling, no hemoptysis, no recent trauma or surgery, no prior PE or DVT, no hormone use. The rule was then prospectively tested in a low-risk group (1427 patients from two hospitals initially tested for PE with a d-dimer) and a very low-risk group (convenience sample of 382 patients with chief complaint of dyspnea, PE not suspected). The prevalence of PE was 8% (95% confidence interval: 7-9%) in the low-risk group and 2% (1-4%) in the very low-risk group on longitudinal follow-up. Application of the rule in the low-risk and very low-risk populations yielded sensitivities of 96% and 100% and specificities of 27% and 15%, respectively. The prevalence of PE in those who met the rule criteria was 1.4% (0.5-3.0%) and 0% (0-6.2%), respectively. The derived eight-factor block rule reduced the pretest probability below the test threshold for d-dimer in two validation populations, but the rule's utility was limited by low specificity.