Artificial Intelligence Identified Dyskalemia Using Electrocardiogram (AIDE)

Once the AIDE indicates high risk of dyskalemia, an obvious message by scarlet letter was appeared in the HIS operation interface to corresponding physicians. To avoid the alert fatigue, we selected the cut-off points with expected positive predictive values of ≥40% according to previous data, which was the consensus of enrolled physicians before the trial considering the clinical loading. The physicians received the AIDE alerts as long as they were operating HIS logged in by their account, even if they were caring other patients. Physicians can review the AIDE predictions of patients in the intervention group. Therefore, this was a single-blind study since HIS presented different information for patients in intervention and control groups. The participated physicians understood the likelihood of dyskalemia and cardiac risk for those patients with ECG-dyskalemia, and provided suitable medical care according to patients' conditions.