Practical utility of clinical prediction rules for suspected acute pulmonary embolism in a large academic institution

Introduction: In an attempt to standardize clinician's approach to the determination of pretest probability (PTP) in pulmonary embolism (PE), two simplified scoring models have recently been proposed. We sought to determine the utility of these algorithms in patients with suspected PE in a large, tertiary, academic medical center. Methods: We performed a retrospective analysis of 295 inpatients and outpatients from our institution who were evaluated for suspected PE. Pretest probability (PTP) was calculated using two previously formulated scoring systems by Wells et al. (Canadian score) and Wicki et al. (Geneva score). Our primary endpoint was the prevalence of PE within each strata of PTP. Results: The prevalence of pulmonary embolism in our cohort was 30%. The prevalence of PE in the low, intermediate and high PTP groups using the Canadian score was 15.3% (95% CI 9.5-23.7%), 34.8% (95% CI 27.9-42.4%), and 47.2% (95% CI 32.0-63.0), respectively. When compared with the low PTP group, the odds ratios of the likelihood of PE was 2.95 (95% CI 1.56-5.59) in the intermediate PTP group and 4.95 (95% CI 2.11-11.64) in the high PTP. The Wicki analysis was divided into "Geneva pure" and "Geneva presumed", where the fractional inspired oxygen concentration was known and presumed to have been sampled on room air, respectively. Neither of the Geneva scores showed statistical significance in the prevalence of PE among the PTP groups. Conclusions: The Wells' clinical prediction score is easily applied and meaningfully risk stratifies patients with suspected PE. In our population, the Geneva score was less useful.