Abstract 17174: Revisiting the QT Correction Formulas: A Generalized Additive Models for Location, Scale, and Shape Approach

Abstract

Introduction: Various mathematical equations have been proposed to correct QT interval for heart rate (QTc). However, with most formulas, QTc remains dependent on heart rate (HR) especially at low and high HR values. Hypothesis: A spline correction function would perform better than standard mathematical formulas by allowing the data to determine the form of the relationship. Methods: A series of regression models using the generalized additive models for location, scale, and shape framework was applied to 10,000 completely normal electrocardiogram data from the National Health and Nutrition Examination Surveys II and III. Evaluation of the model’s performance was performed using the R 2 coefficient of determination and the root mean squared of the errors between the predicted and observed QTc. The new regression models were compared to the Bazett’s and Fredericia’s formulas, which are the 2 most widely used formulas in clinical practice. Results: When boxplots of the QTc for each formula are plotted, grouped by HR in intervals of 5 beats/minute, QTc determined by the penalized spline regression was almost heart rate independent for both male and female as the slope of the regression line was almost zero (-0.003 for female and -0.001 for male) (Figure 1 A—B). By contrast, QTc by Bazett had a positive correlation (regression slope of 0.86 for female and 0.89 for male) while Fredericia’s had a negative correlation (regression slope of -0.14 for female and -0.13 for male) with HR (Figure 1C—F). In all 3 formulas, there was no significant difference between male and female. Conclusions: A new QTc formula was developed, which is almost independent of HR thereby providing a more accurate estimate of QTc for clinical management. Automatic QTc calculation and percentiles estimation could easily be incorporated in an online calculator or app for easy integration in everyday clinical practice.