Accurate Interpretation of Electrocardiograms: Validation and Identification of Challenges

Celia Wing-Sze Lai, Douglas Lee

Abstract


Objective: This study was primarily designed to validate electrocardiogram (ECG) interpretations conducted by nurse abstractors and computer software against the interpretations of an expert electrocardiographer.

Method: ECGs collected must have arrived at ICES no later than June 12, 2009 (n=429). The ECGs were evaluated independently by 3 different subject groups: cardiologist, nurse abstractors, and computer interpretative software. Interpretative fields required included rhythm, rate, PR interval, QRS duration, type of bundle branch block and type of ischemic changes. The ECG interpretations of the cardiologist served as the comparison standard. A library of synonyms equated computer output with corresponding ECG features of interest to facilitate analysis. Kappa analysis and crude agreement were computed for each interpretative field comparing cardiologist’s responses against those of nurse abstractors and computer software. Separate sub-group analyses for selective ECG characteristics (paced rhythm, presence of bundle branch block, and number of lead groups with significant Q waves) were conducted.

Main outcome measures: The validity criteria of this study was determined to be ? > 0.7 or crude agreement >70%. It was found that all nurse abstractors’ interpretations met the validity criteria. However, nurse abstractor’s ability to detect atrial fibrillation/ flutter was significantly lower when the rhythm was paced (a difference of 24% in crude agreement between the responses of the cardiologist and nurse abstractors).

Conclusions: Validity of nurse abstractors’ interpretation was established. ECG training course for nurse abstractors can be adapted to focus on their weak areas to further improve ECG interpretations.

Keywords


electrocardiogram; nurse abstractor; cardiologist; computer; ECG validation

Full Text:

PDF


© University of Toronto Journal of Undergraduate Life Sciences.