Interrater agreement in the measurement of QRS interval in tricyclic antidepressant overdose: Implications for monitoring and research

Study objective: To assess interobserver agreement when experienced clinicians measure QRS-interval duration in tricyclic antidepressant (TCA) overdose. Methods: We studied the admission ECGs of 231 patients with TCA poisoning. Three of the authors, with experience in the management of TCA poisoning, independently measured QRS intervals manually. Each rater was blinded to patient outcome and the measurements made by the other raters. Our main outcome measure was agreement among raters, particularly as it applies to clinically used QRS cutoff points for the determination of treatment and disposition. Results: Agreement on the measurement of QRS intervals was good (intraclass correlation coefficient, .60; 95% confidence interval [CI], .53 to .66) for transformed data and .82 (95% CI, .78 to .85) for raw data. When assigning patients to categories of QRS interval, the raters agreed on 169 of 231 patients (73%) (weighted kappa=.83; P<.0001). However, the raters did not agree unanimously on whether the QRS interval was less than 100 milliseconds or 100 milliseconds or greater in 45 of 231 of patients (19.5%) (kappa=.69; P<.0001). Conclusion: Reliance on the manually measured QRS interval to determine management in TCA poisoning is not justified because of substantial observer variation in the determination of whether the interval falls below the clinical cutoff point of 100 milliseconds. Agreement is sufficient to make the measurements useful as part of the overall assessment of toxicity.