Background: The diagnosis of pulmonary embolism (PE) may be a difficult task. The diagnostic performance of imaging techniques is limited and pre-test probability of PE, estimated from basic clinical data, must be taken into consideration to interpret their results. The aim of this study was to evaluate the accuracy and agreement of clinicians in estimating PE probability.
Patients and methods: We reviewed the charts of 116 patients admitted to hospital for suspected PE. Basic clinical data (symptoms and signs, arterial blood gases, chest X-ray and EKG) were extracted and given to five clinicians, who were asked to estimate the probability of PE. We determined the inter-clinician agreement and compared their estimates with the final diagnoses.
Results: Among patients with a final diagnosis of PE, clinical estimations of PE probability were: high in 63%, intermediate in 21%, and low in 16%. The accuracy of estimates varied between 67 and 80%. Actual PE prevalence was 81% among cases estimated as having high probability and 42% in those considered as low probability. The global inter-clinician agreement rates ranged from 56 to 72%, whereas the average kappa coefficient was 0.44.
Conclusion: Basic clinical data seem to be more useful to predict PE than to exclude it. The accuracy and agreement between estimates from different clinicians are only moderate.