An AI-enabled ECG system identified serum potassium levels in patients with severe hypokalemia with high diagnostic accuracy, according to a study published in the American Journal of Kidney Diseases.
The tool provided results a mean of 52.5 minutes faster than standard laboratory blood tests, potentially offering a noninvasive method to assist in bedside treatment decisions.
Between September 2019 and August 2024, investigators at three Taiwan-based teaching hospitals examined 191 adults receiving care for severe hypokalemia, characterized as a laboratory potassium measurement of ≤ 2.5 mmol/L, an ECG-derived potassium estimate of < 3.5 mmol/L, and the availability of at least one paired follow-up reading within 24 hours of receiving potassium supplementation. The study found that across 466 paired measurements, AI-ECG estimates showed a strong correlation with laboratory results, with a repeated-measures correlation coefficient of 0.847.
The AI-ECG system performed with an area under the curve (AUC) of 0.920 for identifying laboratory potassium levels of 3.5 mmol/L or lower. Performance varied depending on the underlying cause of the hypokalemia. Patients experiencing acute potassium shifts, such as those caused by thyrotoxic periodic paralysis, showed higher diagnostic accuracy (AUC, 0.952) compared to those with chronic potassium deficits (AUC, 0.870).
The study said that potassium levels increased approximately three times faster in patients with acute shifts than in those with chronic deficits during the first four hours of treatment (0.121 vs 0.039 mmol/L/h). “By enabling personalized, responsive potassium replacement strategies, ECG potassium monitoring may hold the potential to reduce cardiovascular risk associated with severe hypokalemia and its treatment,” the authors concluded.
The research team observed two cases where the AI-ECG system detected severe rebound hyperkalemia before laboratory confirmation. In these instances, the system identified potassium levels of 8.8 and 6.5 mmol/L approximately 42 minutes ahead of the lab results.
“Bloodless AI-ECG monitoring with automated alerts may offer a strategy for the rapid recognition and management of rebound hyperkalemia,” the authors wrote. They emphasized that the technology is intended to serve as an adjunct to standard laboratory testing rather than a replacement for blood draws.
The retrospective design of the study and differences in treatment protocols across the three participating centers present potential limitations. Information regarding inpatient drug protocols was sparse, and selection bias might have been introduced by omitting roughly half of the initially screened patients who did not have follow-up data.
Most participants were treated at a single center, and the study was restricted to patients whose initial AI-ECG estimates indicated hypokalemia. Future prospective studies are required to confirm the accuracy and clinical utility of AI-ECG monitoring in broader patient populations.
The system uses 12-lead ECG analysis to estimate serum potassium levels. While it correlates strongly with laboratory measurements, it is currently positioned as an adjunctive tool to help guide treatment timing.
The study suggests that the underlying pathophysiology affects results. Acute shifts involve rapid intracellular redistribution of potassium, while chronic deficits result from a reduction in total-body potassium, which impacts how the system correlates with lab values.
The main advantage identified by the researchers is the reduction in time required to assess potassium levels. The AI-ECG results were available 52.5 minutes earlier than lab tests, providing clinicians with real-time data to adjust potassium replacement strategies between intermittent blood draws.
Related reading