Development and Validation of a Deep-Learning Model to Screen for Hyperkalemia From the Electrocardiogram.

Importance For patients with chronic kidney disease (CKD), hyperkalemia is common, associated with fatal arrhythmias, and often asymptomatic, while guideline-directed monitoring of serum potassium is underused. A deep-learning model that enables noninvasive hyperkalemia screening from the electrocardiogram (ECG) may improve detection of this life-threatening condition. Objective To evaluate the performance of a deep-learning model in detection of hyperkalemia from the ECG in patients with CKD. Design, Setting, and Participants A deep convolutional neural network (DNN) was trained using 1 576 581 ECGs from 449 380 patients seen at Mayo Clinic, Rochester, Minnesota, from 1994 to 2017. The DNN was trained using 2 (leads I and II) or 4 (leads I, II, V3, and V5) ECG leads to detect serum potassium levels of 5.5 mEq/L or less (to convert to millimoles per liter, multiply by 1) and was validated using retrospective data from the Mayo Clinic in Minnesota, Florida, and Arizona. The validation included 61 965 patients with stage 3 or greater CKD. Each patient had a serum potassium count drawn within 4 hours after their ECG was recorded. Data were analyzed between April 12, 2018, and June 25, 2018. Exposures Use of a deep-learning model. Main Outcomes and Measures Area under the receiver operating characteristic curve (AUC) and sensitivity and specificity, with serum potassium level as the reference standard. The model was evaluated at 2 operating points, 1 for equal specificity and sensitivity and another for high (90%) sensitivity. Results Of the total 1 638 546 ECGs, 908 000 (55%) were from men. The prevalence of hyperkalemia in the 3 validation data sets ranged from 2.6% (n = 1282 of 50 099; Minnesota) to 4.8% (n = 287 of 6011; Florida). Using ECG leads I and II, the AUC of the deep-learning model was 0.883 (95% CI, 0.873-0.893) for Minnesota, 0.860 (95% CI, 0.837-0.883) for Florida, and 0.853 (95% CI, 0.830-0.877) for Arizona. Using a 90% sensitivity operating point, the sensitivity was 90.2% (95% CI, 88.4%-91.7%) and specificity was 63.2% (95% CI, 62.7%-63.6%) for Minnesota; the sensitivity was 91.3% (95% CI, 87.4%-94.3%) and specificity was 54.7% (95% CI, 53.4%-56.0%) for Florida; and the sensitivity was 88.9% (95% CI, 84.5%-92.4%) and specificity was 55.0% (95% CI, 53.7%-56.3%) for Arizona. Conclusions and Relevance In this study, using only 2 ECG leads, a deep-learning model detected hyperkalemia in patients with renal disease with an AUC of 0.853 to 0.883. The application of artificial intelligence to the ECG may enable screening for hyperkalemia. Prospective studies are warranted.

[1]  C. Slovis,et al.  The ability of physicians to predict hyperkalemia from the ECG. , 1991, Annals of emergency medicine.

[2]  Michael J Ackerman,et al.  Noninvasive blood potassium measurement using signal-processed, single-lead ecg acquired from a handheld smartphone. , 2017, Journal of electrocardiology.

[3]  E. Finkelstein,et al.  Development and Validation of a Deep Learning System for Diabetic Retinopathy and Related Eye Diseases Using Retinal Images From Multiethnic Populations With Diabetes , 2017, JAMA.

[4]  K. Alexander,et al.  Use of aldosterone antagonists at discharge after myocardial infarction: results from the National Cardiovascular Data Registry Acute Coronary Treatment and Intervention Outcomes Network (ACTION) Registry-Get with the Guidelines (GWTG). , 2013, American heart journal.

[5]  Deborah Silver,et al.  Feature Visualization , 1994, Scientific Visualization.

[6]  J. Wetmore,et al.  Serum Potassium Levels and Mortality in Hemodialysis Patients: A Retrospective Cohort Study , 2016, American Journal of Nephrology.

[7]  E. Peterson,et al.  Consistency of Laboratory Monitoring During Initiation of Mineralocorticoid Receptor Antagonist Therapy in Patients With Heart Failure. , 2015, JAMA.

[8]  C. D'Orsi,et al.  Diagnostic Performance of Digital Versus Film Mammography for Breast-Cancer Screening , 2005, The New England journal of medicine.

[9]  G. Bakris,et al.  Patiromer in patients with kidney disease and hyperkalemia receiving RAAS inhibitors. , 2015, The New England journal of medicine.

[10]  Xavier Robin,et al.  pROC: an open-source package for R and S+ to analyze and compare ROC curves , 2011, BMC Bioinformatics.

[11]  Michael J. Ackerman,et al.  Novel Bloodless Potassium Determination Using a Signal‐Processed Single‐Lead ECG , 2016, Journal of the American Heart Association.

[12]  D. Ransohoff,et al.  Multitarget stool DNA testing for colorectal-cancer screening. , 2014, The New England journal of medicine.

[13]  Bruce Kupelnick,et al.  K/DOQI Clinical Practice Guidelines on Hypertension and Antihypertensive Agents in Chronic Kidney Disease , 2004 .

[14]  Lis Neubeck,et al.  iPhone ECG application for community screening to detect silent atrial fibrillation: a novel technology to prevent stroke. , 2013, International journal of cardiology.

[15]  K. Bailey,et al.  Errors of Classification With Potassium Blood Testing: The Variability and Repeatability of Critical Clinical Tests , 2018, Mayo Clinic proceedings.

[16]  S. Aslam,et al.  Electrocardiography is unreliable in detecting potentially lethal hyperkalaemia in haemodialysis patients. , 2002, Nephrology, dialysis, transplantation : official publication of the European Dialysis and Transplant Association - European Renal Association.

[17]  Americanjournalofkidneydisease K/DOQI clinical practice guidelines on hypertension and antihypertensive agents in chronic kidney disease. , 2004, American journal of kidney diseases : the official journal of the National Kidney Foundation.

[18]  Dennis A. Tighe,et al.  Computer-assisted image processing 12 lead ECG model to diagnose hyperkalemia. , 2017, Journal of electrocardiology.

[19]  P. Lavin,et al.  Sodium zirconium cyclosilicate in hyperkalemia. , 2015, The New England journal of medicine.

[20]  M. Sood,et al.  Emergency management and commonly encountered outpatient scenarios in patients with hyperkalemia. , 2007, Mayo Clinic proceedings.

[21]  M S Pepe,et al.  Phases of biomarker development for early detection of cancer. , 2001, Journal of the National Cancer Institute.

[22]  C. Wingo,et al.  Hyperkalemia: a potential silent killer. , 1998, Journal of the American Society of Nephrology : JASN.

[23]  K. Fitch,et al.  The Clinical and Economic Burden of Hyperkalemia on Medicare and Commercial Payers. , 2017, American health & drug benefits.

[24]  A. Laupacis,et al.  Rates of Hyperkalemia after Publication of the Randomized Aldactone Evaluation Study , 2004 .

[25]  M. Zhan,et al.  The frequency of hyperkalemia and its significance in chronic kidney disease. , 2009, Archives of internal medicine.

[26]  M. Raebel,et al.  Diabetes and Drug-Associated Hyperkalemia: Effect of Potassium Monitoring , 2010, Journal of General Internal Medicine.

[27]  K. Bhaskaran,et al.  Adherence to guidelines for creatinine and potassium monitoring and discontinuation following renin–angiotensin system blockade: a UK general practice-based cohort study , 2017, BMJ Open.

[28]  C. Kovesdy Management of hyperkalaemia in chronic kidney disease , 2014, Nature Reviews Nephrology.

[29]  Jesús Carlos Pedraza Ortega,et al.  Location of mammograms ROI's and reduction of false-positive , 2017, Comput. Methods Programs Biomed..

[30]  S. Brunelli,et al.  Association between Serum Potassium and Outcomes in Patients with Reduced Kidney Function. , 2016, Clinical journal of the American Society of Nephrology : CJASN.

[31]  J. Souverijn,et al.  Multitarget stool DNA testing for colorectal-cancer screening. , 2014, The New England journal of medicine.

[32]  M. Drazner,et al.  2013 ACCF/AHA guideline for the management of heart failure: a report of the American College of Cardiology Foundation/American Heart Association Task Force on Practice Guidelines. , 2013, Journal of the American College of Cardiology.

[33]  B. Surawicz Relationship between electrocardiogram and electrolytes. , 1967, American heart journal.

[34]  Michael B. Gravenor,et al.  Assessment of Remote Heart Rhythm Sampling Using the AliveCor Heart Monitor to Screen for Atrial Fibrillation: The REHEARSE-AF Study , 2017, Circulation.