Blab AI
AI Tools Promise Better Heart Screening, but May Raise Costs, Study Finds
Artificial‑intelligence models are poised to reshape how doctors spot and treat heart disease, but new research warns that the gains may come at a higher price.
In two separate studies, researchers demonstrated that a computer‑trained algorithm can read the humble electrocardiogram (EKG) to uncover hidden structural heart problems, while a deep‑learning model can flag heart‑failure patients who would most benefit from implantable cardioverter‑defibrillators (ICDs). Both innovations promise to expand the pool of patients eligible for expensive imaging and device implantation.
EchoNext, developed by scientists at New York Presbyterian Hospital and Columbia University Irving Medical Center, was published in a 2025 Nature Medicine paper. The model achieved a diagnostic accuracy of 77.3 % in detecting structural heart disease on a standardized EKG dataset, surpassing the 64 % accuracy of an average cardiologist. By parsing waveform nuances and tabular features, EchoNext flags subtle signs of valve dysfunction or early heart‑failure risk.
The algorithm will soon be accessible through OpenEvidence, a clinical decision‑support platform that blends large‑language‑model reasoning with a curated database of high‑impact medical literature. Founded in 2022 and valued at $12 billion in early 2026, OpenEvidence serves more than 10 000 hospitals and medical centers. Integrating EchoNext into its workflow will give clinicians who rarely interpret EKGs for structural disease an automated risk score that can prompt a follow‑up echocardiogram.
A separate study, appearing in Nature, trained a deep‑learning algorithm to identify a biomarker that predicts which heart‑failure patients face life‑threatening arrhythmias. The model was built on EKGs and mortality data from Sweden and then validated with data from the United States and Taiwan. In the Swedish cohort, the biomarker outperformed left‑ventricular ejection fraction alone in selecting patients who would benefit from an ICD, potentially improving prioritization of this life‑saving device.
Both papers underscore the expanding role of AI in cardiovascular care. EchoNext shows how a low‑cost, widely available test can uncover disease that might otherwise go unnoticed until symptoms appear. The ICD biomarker demonstrates how machine learning can refine risk stratification beyond traditional clinical metrics.
However, the authors caution that these tools are unlikely to curb overall healthcare spending. By identifying more patients who qualify for expensive imaging or device implantation, the models may drive up utilization. The studies did not include cost‑effectiveness analyses, and the researchers noted that further work is needed to assess economic impact.
EchoNext received its first FDA clearance for AI detection of hidden heart disease on June 23 2026. The approval followed a series of studies that demonstrated the model’s safety and effectiveness. While the clearance marks a milestone for AI‑driven diagnostics, it also highlights the regulatory scrutiny that accompanies clinical deployment.
The two AI tools illustrate a broader trend: the integration of machine‑learning models into clinical decision‑support systems. By providing clinicians with evidence‑based risk scores, these systems aim to improve diagnostic accuracy and patient outcomes, but the potential for increased utilization and cost remains a key consideration for payers, providers, and policymakers.
At present, EchoNext is available through OpenEvidence, while the ICD biomarker remains in the research phase. Future steps include prospective validation studies, health‑economic evaluations, and integration into electronic health‑record workflows. The next wave of AI‑driven cardiac tools will likely focus on balancing clinical benefit with cost containment.
In sum, the two studies demonstrate that AI can enhance cardiac screening and treatment decisions. EchoNext offers a way to detect structural heart disease from an EKG, while the ICD biomarker refines risk stratification for device implantation. Both tools are expected to broaden access to advanced diagnostics but may also raise healthcare utilization and spending. Continued research and careful implementation will determine how these technologies shape cardiovascular care in the coming years.
In two separate studies, researchers demonstrated that a computer‑trained algorithm can read the humble electrocardiogram (EKG) to uncover hidden structural heart problems, while a deep‑learning model can flag heart‑failure patients who would most benefit from implantable cardioverter‑defibrillators (ICDs). Both innovations promise to expand the pool of patients eligible for expensive imaging and device implantation.
EchoNext, developed by scientists at New York Presbyterian Hospital and Columbia University Irving Medical Center, was published in a 2025 Nature Medicine paper. The model achieved a diagnostic accuracy of 77.3 % in detecting structural heart disease on a standardized EKG dataset, surpassing the 64 % accuracy of an average cardiologist. By parsing waveform nuances and tabular features, EchoNext flags subtle signs of valve dysfunction or early heart‑failure risk.
The algorithm will soon be accessible through OpenEvidence, a clinical decision‑support platform that blends large‑language‑model reasoning with a curated database of high‑impact medical literature. Founded in 2022 and valued at $12 billion in early 2026, OpenEvidence serves more than 10 000 hospitals and medical centers. Integrating EchoNext into its workflow will give clinicians who rarely interpret EKGs for structural disease an automated risk score that can prompt a follow‑up echocardiogram.
A separate study, appearing in Nature, trained a deep‑learning algorithm to identify a biomarker that predicts which heart‑failure patients face life‑threatening arrhythmias. The model was built on EKGs and mortality data from Sweden and then validated with data from the United States and Taiwan. In the Swedish cohort, the biomarker outperformed left‑ventricular ejection fraction alone in selecting patients who would benefit from an ICD, potentially improving prioritization of this life‑saving device.
Both papers underscore the expanding role of AI in cardiovascular care. EchoNext shows how a low‑cost, widely available test can uncover disease that might otherwise go unnoticed until symptoms appear. The ICD biomarker demonstrates how machine learning can refine risk stratification beyond traditional clinical metrics.
However, the authors caution that these tools are unlikely to curb overall healthcare spending. By identifying more patients who qualify for expensive imaging or device implantation, the models may drive up utilization. The studies did not include cost‑effectiveness analyses, and the researchers noted that further work is needed to assess economic impact.
EchoNext received its first FDA clearance for AI detection of hidden heart disease on June 23 2026. The approval followed a series of studies that demonstrated the model’s safety and effectiveness. While the clearance marks a milestone for AI‑driven diagnostics, it also highlights the regulatory scrutiny that accompanies clinical deployment.
The two AI tools illustrate a broader trend: the integration of machine‑learning models into clinical decision‑support systems. By providing clinicians with evidence‑based risk scores, these systems aim to improve diagnostic accuracy and patient outcomes, but the potential for increased utilization and cost remains a key consideration for payers, providers, and policymakers.
At present, EchoNext is available through OpenEvidence, while the ICD biomarker remains in the research phase. Future steps include prospective validation studies, health‑economic evaluations, and integration into electronic health‑record workflows. The next wave of AI‑driven cardiac tools will likely focus on balancing clinical benefit with cost containment.
In sum, the two studies demonstrate that AI can enhance cardiac screening and treatment decisions. EchoNext offers a way to detect structural heart disease from an EKG, while the ICD biomarker refines risk stratification for device implantation. Both tools are expected to broaden access to advanced diagnostics but may also raise healthcare utilization and spending. Continued research and careful implementation will determine how these technologies shape cardiovascular care in the coming years.
