JMIR Publications

ABSTRACT

Background:

Cardiac dysrhythmia is an extremely common disease among people today. While severe arrhythmias often cause a series of complications including congestive heart failure, fainting or syncope, stroke, and sudden death.

Objective:

The aim of this study was to predict incident arrhythmia prospectively within the next one year to provide early warning of impending arrhythmia.

Methods:

Retrospective (1,033,856 subjects registered between October 1, 2016 and October 1, 2017) and prospective (1,040,767 subjects registered between October 1, 2017 and October 1, 2018) cohorts were constructed from electronic health records integrated in the state of Maine. An ensemble learning workflow was built through multiple machine learning algorithms. Differentiated features including acute and chronic diseases, procedures, health status, laboratory tests, prescriptions, clinical utilization indicators, and social-economic determinants were compiled for incident arrhythmia assessment. The predictive model was retrospectively trained and calibrated using an isotonic regression method, and prospectively validated.

Results:

The cardiac dysrhythmia case finding algorithm (the areas under the receiver operating characteristic curve ROC AUC is: retrospective 0.854; prospective 0.819) divided the validation population into five risk subgroups: 53.348%, 44.832%, 1.757%, 0.060% and 0.003% cases in the very low-risk, the low-risk, the medium-risk, the high-risk, and the very high-risk subgroups. 51.85% patients in the very high-risk subgroup were confirmed with a new incident cardiac dysrhythmia within the next one year.

Conclusions:

With the promise to predict future one-year incident cardiac dysrhythmias in a general population, we believe that our case finding algorithm can serve as early warning system to allow statewide population-level screening and surveillance to improve cardiac dysrhythmia care.