Stroke: physician interpretation of CT scans was poor.

Clinical bottom line (level 4)

  1. In patients with acute stroke, point-of-care physicians readings of CT scans were not accurate enough to reliably rule out ICH.
  2. In patient computed tomography scans, emergency physicians diagnosed two-thirds correctly with regards to eligibility for thrombolytic therapy.
  3. Neurologists diagnosed four-fifths of scans correctly with regard to eligibility for thrombolytic therapy.
  4. General radiologists diagnosed four-fifths of scans correctly.
Schriger et al: Journal of the American Medical Association 1998; 279: 1293-1297
Expires November 2002

The study

Setting: university teaching hospital, USA

103 patients (aged ?, ?% male) Physicians (38 emergency physicians, 29 neurologists and 36 general radiologists) who had to interpret computed tomography scans of patients with acute stroke. Scans were identified from a library and those exhibiting hemispheric parenchymal haemorrhage or early infarction were used. Scans with calcifications were used as imposters for haemorrhage and those with an old infarction were used as imposters for acute infarction. Normal scans were also identified.

Excluded if
  • scans with other abnormalities (eg. subarachnoid or extracerebral haemorrhage)
  • scans with multiple findings
  • scans that could not be definitively placed in one of the other categories



    • Independent blinded reference standard, applied in some patients from a non-consecutive appropriate spectrum.
    Reference standard:
    • Previously diagnosed (via consensus) as having haemorrhage, acute infarction or normal.
    Diagnostic test: Physicians had to interpret CT scans for evidence of whether or not patients should receive thrombolytic therapy. They had to answer either (1) yes; (2) no, because of haemorrhage; (3) no, because of signs of acute infarction. Physicians were initially given 5 scans (2 difficult haemorrhages, 1 intermediate acute infarction, 1 imposter, 1 normal). If these were all answered correctly, they were put into an advanced track with harder scans, and if they were not answered correctly, physicians were put into slightly easier scans. 15 scans in total were given to each physician.
    • Physicians were told which side of the patients body was affected, and that 20-60% of the scans would have no contraindication to thrombolytic therapy.

    The evidence


    differential diagnosis number of patients prevalence
    (95% CI)
    average correct score (%) of emergency physician 67.0%
    (42.8% to 90.5%)
    average correct score (%) of neurologists 83.0%
    (64.5% to 102%)
    average correct score of general radiologists 83.0%
    (64.5% to 102%)


    diagnostic test number of patients sensitivity for
    haemorrhage
    (95% CI)    		 LR+ LR-
    physicians interpretation of CT scans 82%
    (78% to 85%)
    total

    Comments

    1. Study makes it very uncertain if 'clot-busting' in stroke should be performed with this level of inaccuracy in scan interpretation.

    Citation

    1. Schriger DL, Kalafut M, Starkman S, et al: Cranial computed tomography interpretation in acute stroke: Physician accuracy in determining eligibility for thrombolytic therapy. Journal of the American Medical Association 1998; 279: 1293-1297
    Contributor: Clare Wotton and Bob Phillips, November 1999
    Reviewer:

    Clinical Question.
    Patient Physicians
    Intervention or Exposure interpret CT scan
    Comparison not interpreting correctly
    Outcome patients eligible for thrombolytic therapy