The interobserver reliability of pretest probability assessment in patients with suspected pulmonary embolism

Marc A. Rodger, Elana Maser, Ian Stiell, Heather E.A. Howley, Philip S. Wells

Research output: Contribution to journalArticlepeer-review

73 Scopus citations

Abstract

Introduction: Pretest probability assessment and objective testing are combined to appropriately manage patients with suspected pulmonary embolism (PE). However, the interobserver reliability of pretest probability assessment has not been investigated. We sought to determine (for patients with suspected PE) the interobserver reliability of pretest probability assessment (by overall impression (gestalt) versus an explicit clinical model). Materials and methods: A prospective cohort study was conducted at an urban university hospital. For patients referred for ventilation and perfusion (V/Q) scanning for suspected PE, structured assessments (11 history and 4 physical examination parameters) were performed by a referring physician and a designated thrombosis physician. The referring and thrombosis physicians also assigned a pretest probability for PE (low, moderate, or high) by gestalt. An explicit seven-point clinical model for suspected PE was later applied to each structured assessment to determine the pretest probability. Assessments were performed independently and prior to diagnostic test results. Interobserver reliability (two rater unweighted Kappa (κ) statistic) was determined for each parameter on the structured assessment and the pretest probability assessments (gestalt vs. explicit clinical model). Results: One hundred and ten patients with suspected PE received duplicate assessments. Historical features demonstrated substantial to almost perfect interobserver reliability (κ=0.60-0.95). For the physical findings, only heart rate had substantial interobserver reliability (κ=0.60). Pretest probability assessment was not reliable when using physician's gestalt (κ=0.33), but produced substantial interobserver reliability using the explicit clinical model (κ=0.62). Conclusions: Given the inadequate interobserver reliability of pretest probability assessment by overall impression (or gestalt), physicians should use explicit clinical models in the diagnostic management of patients with suspected pulmonary embolism.

Original languageEnglish
Pages (from-to)101-107
Number of pages7
JournalThrombosis Research
Volume116
Issue number2
DOIs
StatePublished - 2005
Externally publishedYes

Keywords

  • Interobserver reliability
  • Physical examination
  • Pretest probability
  • Pulmonary embolism

Fingerprint

Dive into the research topics of 'The interobserver reliability of pretest probability assessment in patients with suspected pulmonary embolism'. Together they form a unique fingerprint.

Cite this