Observer variability in ECG interpretation for thrombolysis eligibility: experience and context matter

J Thromb Thrombolysis. 2003 Jun;15(3):131-40. doi: 10.1023/B:THRO.0000011368.55165.97.

Abstract

Background: Despite the known benefit of thrombolysis it remains underutilized among eligible patients with acute myocardial infarction. We sought to determine whether potential errors in ECG interpretation might be a contributing factor and to what extent clinical history, a checklist outlining recognized inclusion criteria and a computerized interpretation would influence reliability and accuracy.

Methods: Seventy-five ECGs were interpreted on 8 separate occasions by 9 clinicians (3 cardiologists, 3 cardiology fellows, 3 medical residents) according to a 2 x 2 x 2 factorial design.

Results: The overall level of agreement among all raters was substantial with a kappa (kappa) of 70.4%. Intra-observer ECG reading reliability was stronger among cardiologists (CC) as compared with cardiology fellows (CF) and medical residents (MR). Similarly, inter-observer reliability was substantial to very good and a gradient was seen with greater reliability among CC, followed by CF, then MR ( P = 0.0013). CC recommended thrombolysis significantly more frequently ( p < 0.001) than either CF or MR. Trainees were biased by the presence of a computerized ECG interpretation resulting in a decision to recommend thrombolysis administration less often.

Conclusion: The reliability of ECG interpretation for deciding to administer thrombolysis was substantial; there was a gradient from lowest to highest commensurate with training and experience. Errors in thrombolysis eligibility are influenced by clinical history and the presence of a computerized ECG interpretation among less experienced clinicians.

MeSH terms

  • Consultants
  • Decision Making
  • Diagnosis, Computer-Assisted
  • Diagnostic Errors
  • Electrocardiography / standards*
  • Humans
  • Medical Staff
  • Observer Variation
  • Patient Selection*
  • Thrombolytic Therapy / statistics & numerical data*