Abstract 2684: Modulators of Normal ECG Intervals Identified in a large Electronic Medical Record
Background: Normal ECG intervals were derived from small studies nearly 50 years ago, mostly in Caucasian males.
Methods: This analysis was conducted in the Vanderbilt Synthetic Derivative, a de-identified image of Electronic Medical Record (EMR), containing data on over 1.7 million subjects for the past 15 years. ECGs were included if they were designated as normal by the reviewing cardiologist, and heart disease, abnormal electrolytes, or interfering medications were all absent. Algorithms were developed using natural language processing and queries of laboratory and billing data to identify normal ECGs. The algorithm was iteratively refined and then tested on a set of 200 records with 96% positive predictive value for identifying patients without exclusions.
Results: The algorithm identified normal ECGs in 30,363 unique subjects, and this analysis used the first recording available. The ethnic makeup was 75% Caucasian, 15% African American, 1% Hispanic, 1% Asian, 1% Other, and 6% Unknown. The range that included 99% of normal PR intervals was 120 –206 msec; QRS 65–108 msec; and QTcB 354 – 448 msec. In the 9,890 ECGs with complete age, height, and BMI data available, we found for each decade of life, PR was 2.8 msec longer and QTcB was 1.1 msec longer; QRS was age-independent. PR and QRS increased 1.0 and 2.4 msec for each decimeter of height, and 0.31 and 0.15 msec for each unit of BMI; QTcB was independent of height and BMI. Women had shorter QRS duration (6.0 msec) and longer QTc (10.7 msec) than men; PR intervals were similar. PR interval was longer by 2.7 msec in African American men and women than Caucasian men. QRS duration was 3.7 msec shorter in African Americans than in other ethnicities.
Conclusion: Analysis of large diverse set of EMR-derived normal ECGs reproduced the known gender/QTcB association, and established new normal ranges. In addition, new covariates were identified, including age (PR, QTcB), height and BMI (PR, QRS), sex (QRS), and ancestry (PR, QRS). Studies of molecular and genomic determinants of the normal and abnormal ECG should take these new findings into account.