Prediction scores do not correlate with clinically adjudicated categories of pulmonary embolism in critically ill patients

Christina Katsios, Marco Donadini, Maureen Meade, Sangeeta Mehta, Richard Hall, John Granton, Jim Kutsiogiannis, Peter Dodek, Diane Heels-Ansdell, Lauralyn McIntyre, Nicholas Vlahakis, John Muscedere, Jan Friedrich, Robert Fowler, Yoanna Skrobik, Martin Albert, Michael Cox, James Klinger, Joseph Nates, Andrew BerstenChristopher Doig, Nicole Zytaruk, Mark Crowther, Deborah Cook

    Research output: Contribution to journalArticlepeer-review

    6 Citations (Scopus)


    BACKGROUND: Prediction scores for pretest probability of pulmonary embolism (PE) validated in outpatient settings are occasionally used in the intensive care unit (ICU). OBJECTIVE: To evaluate the correlation of Geneva and Wells scores with adjudicated categories of PE in ICU patients. METHODS: In a randomized trial of thromboprophylaxis, patients with suspected PE were adjudicated as possible, probable or definite PE. Data were then retrospectively abstracted for the Geneva Diagnostic PE score, Wells, Modified Wells and Simplified Wells Diagnostic scores. The chance-corrected agreement between adjudicated categories and each score was calculated. ANOVA was used to compare values across the three adjudicated PE categories. RESULTS: Among 70 patients with suspected PE, agreement was poor between adjudicated categories and Geneva pretest probabilities (kappa=0.01 [95% CI -0.0643 to 0.0941]) or Wells pretest probabilities (kappa=-0.03 [95% CI -0.1462 to 0.0914]). Among four possible, 16 probable and 50 definite PEs, there were no significant differences in Geneva scores (possible = 4.0, probable = 4.7, definite = 4.5; P=0.90), Wells scores (possible = 2.8, probable = 4.9, definite = 4.1; P=0.37), Modified Wells (possible = 2.0, probable = 3.4, definite = 2.9; P=0.34) or Simplified Wells (possible = 1.8, probable = 2.8, definite = 2.4; P=0.30). CONCLUSIONS: Pretest probability scores developed outside the ICU do not correlate with adjudicated PE categories in critically ill patients. Research is needed to develop prediction scores for this population.

    Original languageEnglish
    Pages (from-to)36-42
    Number of pages7
    JournalCanadian Respiratory Journal
    Issue number1
    Publication statusPublished - 2014


    • Intensive care unit
    • Prediction models
    • Pulmonary embolism


    Dive into the research topics of 'Prediction scores do not correlate with clinically adjudicated categories of pulmonary embolism in critically ill patients'. Together they form a unique fingerprint.

    Cite this