An AliveCor algorithm has identified people with high potassium levels from electrocardiogram (ECG) data. Limited prospective clinical testing suggests pairing the algorithm with AliveCor’s smartphone and Apple Watch-based ECGs could enable the remote, noninvasive monitoring of potassium levels.
Hyperkalemia, the medical term for high potassium levels, is typically asymptomatic and identified by doctors through blood tests ordered to diagnose or monitor other conditions. Given hyperkalemia can severely affect cardiac function, AliveCor thinks there is value in identifying the condition sooner so people can take preventive actions, such as dietary changes.
Building on research correlating hyperkalemia to ECG findings, AliveCor worked with the Mayo Clinic to develop an algorithm capable of linking the dots. This entailed training and testing an algorithm on a dataset covering more than 700,000 patients, 2 million ECGs and 4 million serum potassium values.
Having been trained on two-thirds of the dataset, AliveCor turned the algorithm on the other third to assess its performance. The algorithm chalked up sensitivity and specificity figures of 85% and 72%, resulting in an area under the curve (AUC) of 0.87. AUC scores above 0.80 are typically seen as good, although exactly how good depends on the disease and the available alternatives.
AliveCor followed up by running a small prospective study of the algorithm. That entailed asking 10 patients to use AliveCor’s smartphone-based ECG device to generate four-hour recordings during two dialysis sessions and to undergo concurrent blood testing. The AUC for that group was 0.91.
Presenting the data at the American College of Cardiology’s 67th Annual Scientific Session, AliveCor and Mayo researchers stated the algorithm “might facilitate remote monitoring and enhanced drug titration in patients with renal and cardiovascular disease.” In time, AliveCor may incorporate the capability into its FDA-cleared Apple Watch ECG accessory, KardiaBand.