Interobserver Velocity Variability Before Reaccreditation

Two opposing errors are in play, and they do not simply average out.

• Angle: the calculated velocity equals the true velocity times cos(actual θ)/cos(assumed θ). If the true angle is 60° but the operator assumes 45°, the reported velocity is about 0.71 times the correct value - a roughly 29% underestimate. At a 15° mismatch the error is in the 30-40% range; it grows steeply beyond 60°. This alone can shift a case across a stenosis category.
• Sample site: technologist A samples at the point of maximal aliasing - which is generally the narrowest point, i.e., the correct location. Technologist B samples 1 cm downstream in the post-stenotic jet, where velocity has already dissipated; this biases low. But post-stenotic jets can also be turbulent and broadened, so a single high number there is not reliable either.
• Net effect: A's site is correct but the angle is at the upper limit of acceptable; B's angle is good but the site is wrong. The measurements are not interchangeable and neither is a valid basis for applying velocity criteria.
• Closer to correct: A, because sampling site is the more fundamental error - a measurement at the wrong location is measuring the wrong thing. But A must still document why 60° was chosen and keep the angle constant across serial exams.

Key point: velocity criteria are only valid when angle and sampling site are standardized.

Immediate

• Re-verify machine velocity calibration with the string phantom - it is 14 months overdue and must be brought current; document the results. A drifting velocity scale would compound the operator errors.
• Check other machine parameters: Doppler gain, wall filter, PRF/scale defaults, and whether any preset was altered.

Corrective

• Revise and enforce the protocol with explicit wording: angle 45-60°, adjusted to vessel course and never exceeding 60°; sample volume at the site of maximal narrowing; record the angle on the image.
• Re-educate both technologists on the physics of angle error and on why the post-stenotic jet is not the measurement site.
• Audit a sample of prior studies to identify reports where a stenosis category may have been misassigned, and notify the interpreting physicians so affected patients can be re-scanned if clinically relevant.

Systemic

• Establish a periodic interobserver variability study with a defined acceptance threshold (for example, coefficient of variation under 10-15% for PSV in the same vessel) and a documented remediation pathway.
• Track the QA loop through to closure with dates and signatures - accrediting bodies expect a corrective-action record, not just an intention.

Design

• Recruit 20-30 consenting patients spanning a range of disease severity, including normal, moderate, and severe stenoses - a sample drawn only from normal carotids will not expose angle and site errors.
• Have each technologist independently scan the same patient on the same day, ideally with the second reader blinded to the first's values.
• Measure PSV at the same defined anatomic landmarks (proximal CCA, bulb, proximal ICA, distal ICA), plus the ICA/CCA ratio.
• Analyze agreement, not just correlation: report the mean difference (bias) and limits of agreement (Bland-Altman), plus the coefficient of variation. Correlation can look excellent while two readers disagree by 30%.

Acceptance

• Bias near zero between operators - a consistent systematic offset means someone is still using a nonstandard angle or site.
• Coefficient of variation under about 10-15% for PSV in the same segment; wider spread means the protocol is still not being followed.
• Category agreement (normal / 50-69% / ≥70% / near-occlusion) should be high, since the clinical decision depends on the category, not the exact number.
• Repeat the audit after any protocol change, equipment change, or new hire, and document the results in the QA file.

Interpret any residual discrepancy against the phantom-verified machine calibration so operator error and equipment error are not conflated.