CoChat
Sign Up
Cited by 318ECG Monitoring and AnalysisPotassium and Related DisordersMachine Learning in Healthcare

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

Open access
Galloway, Conner D., Valys, Alexander V., Shreibati, Jacqueline B., Treiman, Daniel L., Petterson, Frank L., Gundotra, Vivek P., Albert, David E., Attia, Zachi I. +6 more·2019·JAMA Cardiology
DOI 10.1001/jamacardio.2019.0640PMC PMC6537816
Open this paper in CoChat — free accountOpen PDF

Abstract

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.

AI snapshot

Written from the paper’s full text.

TL;DR

A deep-learning model using 2 or 4 ECG leads detected hyperkalemia in patients with chronic kidney disease, with 2-lead AUCs of 0.853 to 0.883 across three Mayo Clinic validation sets. At a 90% sensitivity operating point, specificity was 54.7% to 63.2% and negative predictive value was 99.0% to 99.6%.

Key findings

  • For 2-lead ECG (leads I and II), the model achieved AUCs of 0.883 (95% CI, 0.873-0.893) in Minnesota, 0.860 (95% CI, 0.837-0.883) in Florida, and 0.853 (95% CI, 0.830-0.877) in Arizona; 4-lead AUCs were 0.880 to 0.901.
  • At the high-sensitivity operating point, 2-lead sensitivity/specificity were 90.2%/63.2% in Minnesota, 91.3%/54.7% in Florida, and 88.9%/55.0% in Arizona, corresponding to negative predictive values of 99.0% to 99.6%.
  • At the equal sensitivity/specificity operating point, 2-lead sensitivity/specificity were 79.9%/81.3% in Minnesota, 80.5%/75.2% in Florida, and 78.1%/75.3% in Arizona, with negative predictive values of 98.6% to 99.4%.
  • False-positive results reached up to 42% at the high-sensitivity point, while false-negative results were less than 1% of all test results; among 58 false-negative patients reviewed, median blood potassium was 5.9 mEq/L and 52% had a subsequent potassium measurement below 5.5 mEq/L within 8 hours.

Methods

  • Retrospective study using a deep convolutional neural network with 11 layers (10 convolutional plus a fully connected softmax layer), trained on 1,576,581 ECGs from 449,380 patients at Mayo Clinic, Rochester, Minnesota, from 1994 to 2017.
  • The model was trained with 2 leads (I and II) or 4 leads (I, II, V3, and V5) to detect serum potassium ≥5.5 mEq/L and validated on 61,965 ECG-potassium pairs from patients with stage 3 or greater CKD across Mayo Clinic sites in Minnesota, Florida, and Arizona; serum potassium was drawn within 4 hours after the ECG.
  • Performance was assessed by AUC and by sensitivity and specificity at two operating points—equal sensitivity and specificity, and 90% sensitivity—with serum potassium as the reference standard; validation focused on ECGs with potassium ≤5.3 mEq/L or ≥5.7 mEq/L.

Limitations

  • The paper states that data were retrospective and lacked clinical data such as medication use and medical history; the model was developed and validated using 12-lead ECG data in clinic/hospital settings, so prospective testing is needed for home or ambulatory ECG inputs and to determine whether it improves detection, care, and outcomes.
  • The paper states that the DLM does not analyze renal function, that false-negative results may cause false reassurance, and that low specificity with up to 42% false-positive results may cause anxiety and inconvenience, with risk increasing over repeated testing; it also notes insufficient repeated measures to develop a personalized model.
  • Evident from the text: validation excluded ECGs with potassium levels between 5.3 and 5.7 mEq/L to avoid misclassification near the threshold, and a sensitivity analysis including all potassium values decreased AUC by 0.02 to 0.04, suggesting performance may be lower in an unselected screening population.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1001/jamacardio.2019.0640

Cites

37 of 37 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

0 of 26 cited sources checked · 0 supported · 0 flagged

  • Findingown data

    Using the operating point with equal sensitivity and specificity, for ECG leads I and II, the DLM's sensitivity was 80.5% (95% CI, 75.4%-84.9%) in Florida.

    “Using the operating point with equal sensitivity and specificity, for ECG leads I and II, the DLM's sensitivity and specificity were 79.9% (95% CI, 77.6%-82.0%)and 81.3% (95% CI, 80.9%-81.6%) in Minnesota; 80.5% (95% CI, 75.4%-84.9%)and 75.2% (95% CI, 74.0%-76.3%) in Florida; and 78.1% (95% CI, 72.7%-82.9%)and 75.3% (95% CI, 74.1%-76.4%) in Arizona.”
    Quote in the paperSubstantial claim
  • Findingown data

    The algorithm's sensitivity for validation at the high sensitivity operating point was 90.2% (95% CI, 88.4%-91.7%) in Minnesota.

    “The algorithm's sensitivity and specificity for validation were 90.2% (95% CI, 88.4%-91.7%)and 63.2% (95% CI, 62.7%-63.6%) in Minnesota; 91.3% (95% CI, 87.4%-94.3%)and 54.7% (95% CI, 53.4%-56.0%) in Florida; and 88.9% (95% CI, 84.5%-92.4%)and 55.0% (95% CI, 53.7%-56.3%) in Arizona.”
    Quote in the paperSubstantial claim
  • Findingreasoning

    Given the less than 5% prevalence of hyperkalemia, these findings correspond to a negative predictive value of 98.6% to 99.4% across the validation data sets.

    “Given the less than 5% prevalence of hyperkalemia, these findings correspond to a negative predictive value of 98.6% to 99.4% across the validation data sets.”
    Quote in the paperSubstantial claim
  • Findingown data

    Up to 42% of all test results were false-positive.

    “Between 50% and 70% of patients in the validation data sets did not have hyperkalemia predicted by the DLM, with less than 1% of all test results being false-negative; on the other hand, up to 42% of all test results were false-positive.”
    Quote in the paperSubstantial claim
  • Findingown data

    Using the operating point with equal sensitivity and specificity, for ECG leads I and II, the DLM's specificity was 75.3% (95% CI, 74.1%-76.4%) in Arizona.

    “Using the operating point with equal sensitivity and specificity, for ECG leads I and II, the DLM's sensitivity and specificity were 79.9% (95% CI, 77.6%-82.0%)and 81.3% (95% CI, 80.9%-81.6%) in Minnesota; 80.5% (95% CI, 75.4%-84.9%)and 75.2% (95% CI, 74.0%-76.3%) in Florida; and 78.1% (95% CI, 72.7%-82.9%)and 75.3% (95% CI, 74.1%-76.4%) in Arizona.”
    Quote in the paperNot a substantial claim
  • Classified as probably not a substantial claim. 20
  • The quote only partly states this claim. 1
  • May combine two claims. 16

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    Feature visualization on the mean ECG beat sometimes showed expected changes in QRS and T-wave morphology.

    “To try to understand how the DNN builds up its understanding of images for hyperkalemia over many layers, we performed feature visualization on the mean ECG beat [33][34][35] (eFigure 3 in the Supplement); sometimes, expected changes in QRS and T-wave morphology were seen.”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites [33]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites [34]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites [35]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • The quote only partly states this claim.

    Findingown data

    The AUC was 0.02 to 0.03 higher using ECG leads I, II, V3, and V5 than using leads I and II.

    “The AUC was 0.02 to 0.03 higher using ECG leads I, II, V3, and V5.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingnone given

    The DLM is a screening test with low specificity, with upwards of 42% false-positive results.

    “Finally, the DLM is a screening test with low specificity, with upwards of 42% false-positive results”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    The prevalence of hyperkalemia in the Arizona validation data set was 4.6% (n = 270 of 5855).

    “Fewer than 3% in Minnesota had hyperkalemia (n = 1282 of 50 099) compared with 4.6% (n = 270 of 5855) to 4.8% (n = 287 of 6011) in Florida and Arizona, respectively.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    The prevalence of hyperkalemia in the Minnesota validation data set was fewer than 3% (n = 1282 of 50,099).

    “Fewer than 3% in Minnesota had hyperkalemia (n = 1282 of 50 099) compared with 4.6% (n = 270 of 5855) to 4.8% (n = 287 of 6011) in Florida and Arizona, respectively.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    The prevalence of hyperkalemia in the Florida validation data set was 4.8% (n = 287 of 6011).

    “Fewer than 3% in Minnesota had hyperkalemia (n = 1282 of 50 099) compared with 4.6% (n = 270 of 5855) to 4.8% (n = 287 of 6011) in Florida and Arizona, respectively.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingown data

    Hyperkalemia ECGs were more likely to be recorded in patients who were older and men and who had stage 3 or greater CKD.

    “Hyperkalemia ECGs were more likely to be recorded in patients who were older and men and who had stage 3 or greater CKD.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    Between 50% and 70% of patients in the validation data sets did not have hyperkalemia predicted by the DLM, with less than 1% of all test results being false-negative.

    “Between 50% and 70% of patients in the validation data sets did not have hyperkalemia predicted by the DLM, with less than 1% of all test results being false-negative; on the other hand, up to 42% of all test results were false-positive.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Hyperkalemia is potentially life threatening for the 35 million American adults with chronic kidney disease (CKD).

    “H yperkalemia is potentially life threatening for the 35 million American adults with chronic kidney disease (CKD). 1”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [1]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • May combine two claims.

    Premisecitation

    Guideline-directed potassium monitoring is severely underused; up to one-third of patients taking RAAS inhibitors receive no monitoring.

    “However, guidelinedirected potassium monitoring 5,6 is severely underused: up to one-third of patients taking RAAS inhibitors receive no monitoring, and only 10% receive potassium monitoring before and after RAAS inhibitors are started.7,8”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites [7]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites [8]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • May combine two claims.

    Interpretationcitation

    A heightened fear of MRA-associated hyperkalemia and hyperkalemia-associated mortality and the complex potassium monitoring requirements for MRA initiation and titration are 2 potential reasons for underuse.

    “A heightened fear of MRA-associated hyperkalemia and hyperkalemia-associated mortality 31 and the complex potassium monitoring requirements for MRA initiation and titration 6 are 2 potential reasons for underuse.”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites [31]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites [6]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    The DLM does not offer an analysis of renal function.

    “The DLM does not offer an analysis of renal function.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationreasoning

    The diagnosis of hyperkalemia, particularly outside of the hospital or clinic, is challenging because patients are often asymptomatic, and guideline-directed blood potassium monitoring is severely underperformed.

    “the diagnosis of hyperkalemia, particularly outside of the hospital or clinic, is challenging because patients are often asymptomatic, and guideline-directed blood potassium monitoring is severely underperformed.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

    Interpretationreasoning

    In other high-probability visualizations, QRS and T-wave features were not present because there are features the DNN extracts that surpass human visualization.

    “In other high-probability visualizations, QRS and T-wave features were not present because there are features the DNN extracts that surpass human visualization.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data

    The sample size for each validation data set was determined to be 6133 ECGs based on the maximal sample size needs of the AUC.

    “The sample size for each validation data set was determined to be 6133 ECGs and was based on the maximal sample size needs of the area under the receiving operating characteristic curve (AUC) (eMethods in the Supplement).”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Previous approaches have used standard regression models to estimate potassium and have highlighted the importance of T-wave width, T-wave amplitude, and descending T-wave slope as predictors of hyperkalemia.

    “Previous approaches have used standard regression models to estimate potassium and have highlighted the importance of T-wave width, T-wave amplitude, and descending T-wave slope as predictors of hyperkalemia.16,32”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [16]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites [32]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Classified as probably not a substantial claim. May combine two claims.

    Premisecitation

    In patients with CKD, hyperkalemia is common and life threatening.

    “In patients with CKD, hyperkalemia is common and life threatening.23,24”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites [23]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites [24]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

    Interpretationreasoning

    An ECG-derived potassium level may be more germane to health and arrhythmia risk than the blood test.

    “an ECG-derived potassium level may be more germane to health and arrhythmia risk than the blood test.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Interpretationreasoning

    Patients who most likely require an assessment of renal function are those with possible hyperkalemia, for whom the envisioned screening test would require a confirmatory blood test that typically includes renal function in addition to potassium level.

    “However, the patients who most likely require an assessment of renal function are those patients with possible hyperkalemia, for whom the envisioned screening test would require a confirmatory blood test that typically includes renal function in addition to potassium level.”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: Hyperkalemia is often present in the setting of acute CKD.

  • May combine two claims.

    Premisecitation

    Hyperkalemia causes cardiotoxic effects and is associated with a defined series of ECG abnormalities including peaking of T waves, QRS prolongation, and PR shortening.

    “Hyperkalemia causes cardiotoxic effects and has been associated with a defined series of ECG abnormalities, including peaking of T waves, QRS prolongation, and PR shortening.9”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [9]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • May combine two claims.

    Premisecitation

    In patients with heart failure, MRAs reduce cardiovascular mortality, but only a small fraction (less than one-third of eligible patients) receive them at hospital discharge.

    “In patients with heart failure, MRAs reduce cardiovascular mortality, but only a small fraction (less than one-third of eligible patients) receive them at hospital discharge.30”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [30]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationreasoning

    False-positive results may cause anxiety and inconvenience for patients.

    “which may cause anxiety and inconvenience for patients.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Treatments for hyperkalemia are effective and readily available, including patiromer and zirconium cyclosilicate.

    “Although treatments for hyperkalemia are effective and readily available, 2 5 including the US Food and Drug Administration-approved potassium binders patiromer 26 and zirconium cyclosilicate, 27”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites [25]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites [26]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites [27]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S2 (checkable). May combine two claims.

    Interpretationreasoning

    A DLM that enables noninvasive potassium monitoring could assuage the fears of health care clinicians and enable more patients with CKD with heart failure to receive target doses of MRAs and other guideline-recommended medications, potentially improving care and outcomes.

    “A DLM that enables noninvasive potassium monitoring could assuage the fears of health care clinicians and enable more patients with CKD with heart failure to receive target doses of MRAs and other guideline-recommended medications, potentially improving care and outcomes.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    The risk of hyperkalemia in CKD is substantially elevated due to kidney dysfunction and the recommendation that these patients be treated with RAAS inhibitors that inhibit renal potassium secretion.

    “Owing to the effects of kidney dysfunction on potassium homeostasis, as well as the recommendation that these patients be treated with renin-angiotensin-aldosterone system (RAAS) inhibitors that preserve renal and cardiac function but also inhibit renal potassium secretion, the risk of hyperkalemia in CKD is substantially elevated.2”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [2]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Premisenone given

    To our knowledge, this is the first deep-learning approach to evaluation of potassium levels from the ECG.

    “To our knowledge, this is the first deep-learning approach to evaluation of potassium levels from the ECG.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationreasoning

    The risk of false-positive results will increase over time with repeated testing.

    “The risk of falsepositive results will also increase over time with repeated testing.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    The 1-day odds of mortality are up to 13 times higher for patients with CKD with outpatient serum potassium levels of at least 6.0 mEq/L compared with patients without CKD with potassium levels of less than 5.5 mEq/L.

    “The 1-day odds of mortality are up to 13 times higher for patients with CKD with outpatient serum potassium levels of at least 6.0 mEq/L compared with patients without CKD with potassium levels of less than 5.5 mEq/L.3”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [3]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    On average, patients with CKD spend up to 9% of their time with serum potassium levels 5.5 mEq/L or greater.

    “On average, patients with CKD spend up to 9% of their time with serum potassium levels 5.5 mEq/L or greater. 2”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [2]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • May combine two claims.

    Premisecitation

    Hyperkalemia is often asymptomatic and associated with cardiac arrhythmias and death.

    “When present, hyperkalemia is often asymptomatic and associated with cardiac arrhythmias and death.3”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [3]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

30 of 30

  • The paper relies on this without support.

    The serum potassium level measured within 4 hours after the ECG reflects the potassium level at the time of the ECG.

    “All ECGs within 4 hours before a serum potassium draw were identified.16”

  • The paper relies on this without support.

    The Gaussian process used to estimate potassium level at the time of ECG in the development data set accurately estimates the true potassium level.

    “Gaussian process to estimate potassium level at time of ECG”

  • The paper relies on this without support.

    Hyperkalemia is often present in the setting of acute CKD.

    “Hyperkalemia is often present in the setting of acute CKD.”

  • The paper relies on this without support.

    ECG-based tests are not susceptible to mechanical, temperature, contamination, or other potential errors associated with processing blood.

    “The ECGbased tests are not susceptible to mechanical, temperature, contamination, or the other potential errors associated with processing blood.”
4 of 4

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • Potassium models may prioritize urgent laboratory testing without replacing serum measurement.

    “potassium models may prioritize urgent laboratory testing without replacing serum measurement [6,15].”

    From Deep Learning Applied to 12-Lead ECGs for Detection of Structural, Metabolic and Systemic Disease (2026) · cites this paper as [6]

    The Discussion describes the ECG model as a noninvasive screen for hyperkalemia, and the Limitations discuss subsequent laboratory testing after a possible positive result. This supports using the model alongside, rather than in place of, serum testing. The paper does not specifically say that model results should prioritize urgent laboratory tests. (LLM, conf=0.82)

    Experimental — accuracy not yet measured

  • ECG waveforms contain latent signatures of metabolic abnormalities.

    “Emerging evidence suggests that ECG waveforms contain latent signatures of ventricular dysfunction [4], structural heart disease [5], metabolic abnormalities [6], and broader physiologic characteristics such as biological age [7] and future mortality risk [8].”

    From Deep Learning Applied to 12-Lead ECGs for Detection of Structural, Metabolic and Systemic Disease (2026) · cites this paper as [6]

  • ECG waveforms could support noninvasive screening for a metabolic abnormality ordinarily diagnosed through laboratory testing.

    “This study expanded AI-ECG by suggesting that ECG waveforms could support noninvasive screening for a metabolic abnormality ordinarily diagnosed through laboratory testing.”

    From Deep Learning Applied to 12-Lead ECGs for Detection of Structural, Metabolic and Systemic Disease (2026) · cites this paper as [6]

  • A deep-learning model using only leads I and II detected hyperkalemia with AUC values ranging from 0.853 to 0.883 across validation sites in patients with renal disease.

    “The model was trained on more than 1.5 million ECGs from over 449,000 patients and, using only leads I and II, detected hyperkalemia with AUC values ranging from 0.853 to 0.883 across validation sites [6].”

    From Deep Learning Applied to 12-Lead ECGs for Detection of Structural, Metabolic and Systemic Disease (2026) · cites this paper as [6]

  • The reduced-lead hyperkalemia model demonstrated that clinically relevant metabolic information can be retained using only two or four ECG leads.

    “The reduced-lead hyperkalemia model discussed earlier demonstrated that clinically relevant metabolic information can be retained using only two or four ECG leads [6].”

    From Deep Learning Applied to 12-Lead ECGs for Detection of Structural, Metabolic and Systemic Disease (2026) · cites this paper as [6]

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

  • Reserved cohort is not 10% of patients

    Minor

    The stated 109 860 reserved patients are 13.9%, not 10%, of the 787 661 patients.

    Reported 10 · recomputed 13.947625

  • Development cohort is not 60% of patients

    Minor

    The stated 449 380 patients are 57.1%, not 60%, of the 787 661 patients.

    Reported 60 · recomputed 57.052463

  • Abstract reverses the hyperkalemia threshold

    Minor

    The abstract says the model detects potassium levels of 5.5 mEq/L or less, whereas the Methods define hyperkalemia as at least 5.5 mEq/L.

  • Arizona White percentage is inconsistent with its count

    Minor

    The table reports 5271 White patients among 5855 Arizona patients as 9.0%; the proportion is 90.0%.

    Reported 9 · recomputed 90.025619

  • Florida Black percentage is inconsistent with its count

    Minor

    The table reports 650 Black patients among 6011 Florida patients as 1.8%; the proportion is 10.8%.

    Reported 1.8 · recomputed 10.813509

  • Florida and Arizona hyperkalemia rates are assigned to the wrong sites

    Minor

    The Results assign 4.6% to Florida and 4.8% to Arizona, respectively. The table assigns 4.8% to Florida and 4.6% to Arizona.

6 of 6

Cites

Flagged first, then by citations

Research works printed but not linked: 1

  1. Erhan, 2009, Visualizing Higher-Layer Features of a Deep Network: Technical Report 1341

Cited by

Sorted by citations

Related

Authors
Galloway, C. D., Valys, A. V., Shreibati, J. B., Treiman, D. L., Petterson, F. L., Gundotra, V. P., Albert, D. E., Attia, Z. I., Carter, R. E., Asirvatham, S. J., Ackerman, M. J., Noseworthy, P. A., Dillon, J. J., & Friedman, P. A.
Affiliations
  • AliveCor Inc, Mountain View, California
  • Mayo Clinic (US)
Corresponding author
Friedman, Paul A.
ORCID
Venue
JAMA Cardiology
Publisher
American Medical Association (AMA)
Published
2019-05-01
Volume
4
Issue
5
Pages
428
ISSN
2380-6583 (ISSN-L)
DOI
10.1001/jamacardio.2019.0640
PMC
PMC6537816
OpenAlex
W2928542791
Cited by
318
References
37
Citations by year
2026: 23 · 2025: 45 · 2024: 53 · 2023: 53 · 2022: 54 · 2021: 45 · 2020: 34 · 2019: 10
FWCI
23.27
Type
journal article
Language
en
Open access
Locations
  • JAMA Cardiology (publishedVersion)
  • PubMed (publishedVersion)
  • Europe PMC (PubMed Central) (submittedVersion)
  • PubMed Central (submittedVersion)
Topics
  • Health Sciences › Medicine › Cardiology and Cardiovascular Medicine › ECG Monitoring and Analysis
  • Health Sciences › Medicine › Pulmonary and Respiratory Medicine › Potassium and Related Disorders
  • Physical Sciences › Computer Science › Artificial Intelligence › Machine Learning in Healthcare
Keywords
hyperkalemia, electrocardiogram, deep convolutional neural networks, chronic kidney disease, serum potassium, non-invasive screening, arrhythmia risk, model validation, sensitivity and specificity, clinical decision support, arrhythmia prevention
MeSH terms
Deep Learning, Aged, Aged, 80 and over, Algorithms, Arrhythmias, Cardiac, Artificial Intelligence, Electrocardiography, Female, Humans, Hyperkalemia, Male, Mass Screening, Middle Aged, Retrospective Studies, Sensitivity and Specificity, Prevalence, Neural Networks, Computer, Renal Insufficiency, Chronic

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-27.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-27.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review