
Researchers at the Mayo Foundation for Medical Education and Research, Rochester, Minnesota, used an all-patient cohort to train a convolutional neural network to produce an artificial intelligence (AI) algorithm that can detect hyperkalemia from the surface electrocardiogram (ECG). The research team, led by John J. Dillon, recently conducted a validation study to examine the network’s performance among patients in the emergency department (ED) and in the intensive care unit (ICU).
Results were reported during a poster session at the American Society of Nephrology Kidney Week 2023. The poster was titled Noninvasive Artificial Intelligence (AI)-Enhanced Electrocardiographic Detection of Hyperkalemia in the Emergency Department (ED) and ICU.
All adult patients who had a standard 12-lead supine ECG and a blood potassium value within 4 hours of the ECG who presented to the ED at all Mayo Clinic sites between February and August 2021 (ED cohort) and patients admitted to the ICU at Mayo Clinic St. Mary’s Hospital, Rochester, Minnesota, between August 2017 and February 2018 (ICU cohort) were eligible for the validation study.
The network analyzed leads I and II of the 12-lead ECG to calculate the probability of hyperkalemia. Hyperkalemia was defined as potassium >6 mEq/L. The two cohorts were analyzed separately. The researchers performed exploratory subgroup analyses for patients with estimated glomerular filtration rate (eGFR) <45 mL/min/1.73 m2 and those with eGFR <30 mL/min/1.73 m2.
The ED cohort included 40,128 patients and the ICU cohort included 2636 patients. The prevalence of hyperkalemia was 0.9% in the ED cohort and 3.3% in the ICU cohort. The AI-ECG area under the curves (AUCs) were 0.88 in both cohorts, with sensitivities and specificities ≥80%. Negative-predictive values (NPVs) were >99% in both cohorts.
While positive AI-ECGs quadrupled the probability of hyperkalemia, positive predictive values (PPVs) were relatively low: 3.5% in the ED cohort and 14% in the ICU cohort. The low values were due, in part, to low prevalences of hyperkalemia. Patients in the low eGFR subgroups had higher prevalence of hyperkalemia and higher PPVs.
In conclusion, the authors said, “The AI-ECG demonstrated excellent discrimination with AUCs of 0.88 in both cohorts. It was highly effective at ruling out hyperkalemia with NPVs >99% in both cohorts, but with much lower PPVs, suggesting that it is most useful as a screening test to exclude hyperkalemia. One method by which PPVs can be increased is by limiting testing to high-risk populations, such as those with reduced eGFR.”
Source: Dillon JJ, Liu K, Dugan J, et al. Noninvasive artificial intelligence (AI0-enhnced electrocardiographic detection of hyperkalemia in the emergency department (ED) and ICU. TH-PO012. Abstract of a poster presented at the American Society of Nephrology Kidney week 2023; November 2, 2023; Philadelphia, Pennsylvania.