Context: Experienced clinicians' gestalt is useful in estimating the pretest probability for pulmonary embolism and is complementary to diagnostic testing, such as lung scanning. However, it is unclear whether recently developed clinical prediction rules, using explicit features of clinical examination, are comparable with clinicians' gestalt. If so, clinical prediction rules would be powerful tools because they could be used by less-experienced health care professionals to simplify the diagnosis of pulmonary embolism. Recent studies have shown that the combination of a low pretest probability (using a clinical prediction rule) and a normal result of a D-dimer test reliably excludes pulmonary embolism without the need for further testing.

Objective: To evaluate and demonstrate the accuracy of pretest probability assessment for pulmonary embolism using clinical gestalt vs clinical prediction rules.

Data sources: The MEDLINE database was searched for relevant articles published between 1966 and March 2003. Bibliographies of pertinent articles also were scanned for suitable articles.

Study selection: To be included in the analysis, studies were required to have consecutive, unselected patients enrolled; participating physicians in the studies, blinded to the results of diagnostic testing, had to estimate pretest probability of pulmonary embolism; and validated diagnostic methods had to be used to confirm or exclude pulmonary embolism.

Data extraction: Three reviewers independently scanned titles and abstracts for inclusion of studies. An initial MEDLINE search identified 1709 studies, of which 16 involving 8306 patients were included in the final analysis.

Data synthesis: A clinical gestalt strategy was used in 7 studies, and in the low, moderate, and high pretest categories, the rates of pulmonary embolism ranged from 8% to 19%, 26% to 47%, and 46% to 91%, respectively. Clinical prediction rules were used in 10 studies, and 3% to 28%, 16% to 46%, and 38% to 98% in the low, moderate, and high pretest probability groups, respectively, had pulmonary embolism.

Conclusions: The clinical gestalt of experienced clinicians and the clinical prediction rules used by physicians of varying experience have shown similar accuracy in discriminating among patients who have a low, moderate, or high pretest probability of pulmonary embolism. We advocate the use of a clinical prediction rule because it has shown to be accurate and can be used by less-experienced clinicians.