Monitoring for heart failure has moved out of the clinic and into the patient’s home. Connected scales, blood pressure monitors, and wearables now send readings to care teams every day. Artificial intelligence (AI) reviews that stream of data and separates routine readings from early warning signs.
A 2023 study in Scientific Reports tested one version of this approach. Researchers paired Internet of Things (IoT) sensors with an AI model built to flag heart problems in near real time. Since then, newer trials have tested AI-guided monitoring for heart failure with real patients. This update covers the original study and what the research has shown since.
The need is large. Nearly 6.7 million U.S. adults age 20 and older have heart failure, according to the CDC. Heart failure was listed on 452,573 death certificates in 2023. That is 14.6% of all deaths in the country.
Why Monitoring for Heart Failure Matters Between Visits
Monitoring for heart failure works because the condition usually warns before it becomes a crisis. Fluid builds slowly. Weight creeps up. Blood pressure and heart rate shift, and a patient may feel more tired or short of breath for days before calling anyone.
Those changes are easy to miss when a patient sees a cardiologist every few months. Readmission numbers show the cost of that gap. Researchers running the LINK-HF2 trial note that about 20% of heart failure patients return to the hospital within 30 days of discharge. By 90 days, the figure is about 30%.
The American Heart Association asks patients to weigh themselves every day. A gain of more than 2 to 3 pounds in 24 hours, or 5 pounds in a week, is a reason to contact the care team. Remote patient monitoring (RPM) turns that advice into a reading a nurse can see the same morning.
For a real-world example, see how a French care team used remote patient monitoring for heart failure to end a cycle of repeat hospital stays.
The Original Study on AI and IoT Monitoring for Heart Failure
The 2023 Scientific Reports paper proposed a low-cost system for monitoring for heart failure, aimed partly at regions with limited healthcare infrastructure. It had three parts. Wearable IoT sensors collected patient data, a cloud platform stored it, and an AI model called ET-CNN analyzed it. Physicians could then view results remotely and step in when a reading looked dangerous.
How the ET-CNN Model Works
ET-CNN combines two methods. An Extra Trees classifier works alongside an 8-layer convolutional neural network (CNN). The model reviewed 13 clinical signals, including blood pressure and cholesterol, and sorted each record as normal or abnormal.
The system protected data with encryption and access controls. ZigBee and Firebase handled authentication and authorization. The design was modular, so developers could add sensors or security protocols later.
What the Results Showed
The team trained and tested the model on 300 patient records, split 70/30. ET-CNN reached 95.24% accuracy on this dataset. It outperformed nine other machine learning methods, including RNN, LSTM, Random Forest, AlexNet, and VGG-16. Precision, recall, and F1 scores each reached 97%.
The researchers then tested ET-CNN on a 303-record dataset from the Cleveland Clinic Foundation, published in the UCI Machine Learning Repository. Accuracy rose to 97.18%. Training also took up to 50% less time than other deep learning approaches.
Limits Worth Noting
Those results are strong, and they deserve context. Both datasets were small and historical. The model predicted risk from stored records instead of guiding care for patients living at home. High accuracy on a benchmark dataset does not always hold up in a busy cardiology practice.
That gap is why the newer trials matter. They test whether AI changes what clinicians do, and whether patients stay out of the hospital as a result.
New Research on AI-Guided Monitoring for Heart Failure
Since 2023, research on monitoring for heart failure has shifted from prediction toward action. The newer studies ask a harder question. Can AI help a care team adjust treatment in time to keep a patient home?
TURING-HF: Turning Home Blood Pressure Into Treatment Advice
Cedars-Sinai published the TURING-HF trial in JACC: Heart Failure in 2025. The study enrolled 50 patients with heart failure with reduced ejection fraction (HFrEF). The team built software called HF-AI, trained on Cedars-Sinai heart failure patient data from 2020 to 2022. It follows national and international heart failure guidelines.
HF-AI reviewed home blood pressure readings and recommended medication changes. Clinicians then approved or adjusted each suggestion. The study team reported that this let them fine-tune medications sooner, with the goal of keeping more patients out of the hospital.
This matters because of guideline-directed medical therapy (GDMT). Guidelines recommend four drug classes for HFrEF, but many patients never reach target doses. Safe titration depends on frequent blood pressure and heart rate data. Home readings supply that data between visits.
ASSIST-HF SIRIO: A Virtual Assistant After Discharge
A randomized pilot trial published in JACC tested a generative AI virtual assistant for GDMT after an HFrEF hospitalization. At 12 weeks, patients in the assistant group had better medication optimization across all drug classes. They also had lower NT-proBNP levels, a blood marker of heart strain, and fewer hospitalizations.
The researchers have since described a larger trial, ASSIST-HF, that uses nonphysician staff supported by the AI tool. If it holds up, this model could help practices without enough cardiologists to manage every titration visit.
LINK-HF2: Predicting Readmission About a Week Early
Earlier work from the LINK-HF research group found that machine learning analysis of continuous wearable data predicted heart failure rehospitalization about one week before it happened. LINK-HF2 is a pragmatic randomized trial built on that finding.
Every patient wears a disposable sensor after discharge. In one group, clinicians receive the AI notifications. In the other group, they do not. The design will show whether acting on the alerts changes outcomes, which is the question earlier telemonitoring trials struggled to answer.
Speech Analysis and Other New Signals
Some signals come from sources other than a cuff or a scale. The HearO system uses a smartphone app that analyzes a patient’s voice for signs of fluid buildup before decompensation. A 2025 paper in Heart Failure Reviews describes a multicenter randomized trial of 1,437 patients comparing this approach with usual care. The authors estimate that heart failure affects more than 64 million people worldwide, so a low-cost signal like speech could reach many of them.
In Italy, the SMART-CARE study is evaluating AI-enabled monitoring with CE-certified wearable devices. Its main outcome is time to first hospitalization.
What the Evidence Says About Monitoring for Heart Failure So Far
A December 2025 review in Current Heart Failure Reports examined non-invasive remote monitoring, wearables, and AI in heart failure. It focused on three questions: whether monitoring prevents hospitalizations, whether it is cost-effective, and whether AI makes it easier to run.
Across the recent studies, several patterns stand out:
- Action drives results. Readings help patients when a clinician reviews them and adjusts care. Some early telemonitoring trials, such as BEAT-HF in 2016, showed limited benefit, and the Heart Failure Reviews authors suggest those studies may have tracked the wrong signals.
- Medication optimization is a strong use case. Both TURING-HF and ASSIST-HF SIRIO used home data to move patients toward guideline-recommended doses faster.
- Prediction is getting earlier. LINK-HF showed that continuous data can signal trouble about a week ahead, which gives teams time to intervene.
- Alert volume needs a plan. A 2025 Netherlands Heart Journal editorial on implanted cardiac devices warned about “alert pollution.” The authors suggested AI could filter out irrelevant alerts and flag urgent ones first.
Devices Used in Monitoring for Heart Failure at Home
Most programs for monitoring for heart failure start with a small set of devices. Each one captures a different part of the patient’s condition.
- Weight scale. Daily weight is the most direct home signal of fluid retention.
- Blood pressure monitor. Blood pressure and heart rate trends guide medication titration, as TURING-HF showed.
- Pulse oximeter. Oxygen saturation can point to worsening breathing or fluid in the lungs.
- Symptom check-ins. Patient-reported shortness of breath, swelling, and sleep changes add context to vital signs.
- Wearables and newer sensors. Continuous patches and speech tools are still being tested in clinical trials.
Device simplicity affects whether patients keep taking readings. Many heart failure patients are in their later years and may not want to manage apps, passwords, or Wi-Fi. Tenovi offers standalone cellular devices with their own SIMs, plus Bluetooth devices that connect through the Tenovi Cellular Gateway. Patients plug in the Gateway once, and readings transmit automatically.
What Care Teams Should Plan For
Technology is one part of a program. The workflow around it decides whether patients benefit. Before launching, care teams should plan for the following:
- Personal alert thresholds. The care team should set weight and blood pressure limits for each patient, not only default values.
- A named reviewer. Someone should own the daily review of readings and know when to escalate.
- Easy-to-use devices. Equipment should work out of the box, with no pairing or setup for the patient.
- A medication pathway. Teams need a clear process for dose changes that start from home data.
- Alert management. Programs should decide early how to sort urgent alerts from routine ones, so staff can focus on patients who need a call.
Research on heart failure remote monitoring and readmissions shows what these programs can do when the pieces work together.
Frequently Asked Questions About Monitoring for Heart Failure
1) What is monitoring for heart failure?
Monitoring for heart failure is the regular tracking of weight, blood pressure, heart rate, oxygen levels, and symptoms. At home, connected devices send these readings to a care team. The team watches for changes that suggest the patient’s condition is getting worse.
2) Which readings matter most in monitoring for heart failure?
Daily weight is often the first sign of fluid buildup. The American Heart Association advises patients to contact their care team after a gain of more than 2 to 3 pounds in 24 hours or 5 pounds in a week. Blood pressure and heart rate matter for medication titration.
3) Can AI predict a heart failure hospitalization?
Early research is promising. The LINK-HF study found that machine learning analysis of wearable data predicted rehospitalization about one week in advance. Larger randomized trials, such as LINK-HF2, are testing whether acting on these predictions improves outcomes.
4) Does AI replace the care team?
No. In the trials discussed here, AI flags risks or suggests medication changes, and clinicians make the final decision. AI can reduce the time staff spend sorting routine readings, which frees them for patient care.
5) Who benefits most from remote patient monitoring for heart failure?
Patients recently discharged after a heart failure hospitalization often benefit most, because readmission risk is highest in the first 90 days. Patients whose medications are still being titrated also benefit. Frequent home readings help clinicians adjust doses safely.
6) How does Tenovi support monitoring for heart failure?
Tenovi supports more than 60 connected devices, including cellular-connected scales, blood pressure monitors, and pulse oximeters. Readings reach the care team automatically through the Tenovi Cellular Gateway or through standalone cellular devices. Patients do not need an app, Bluetooth pairing, or Wi-Fi.
Understanding Monitoring for Heart Failure
Monitoring for heart failure gives care teams a view of a patient’s condition between office visits. The 2023 Scientific Reports study showed that an AI model could classify heart risk with high accuracy on historical data. Newer trials have taken the next step. TURING-HF and ASSIST-HF SIRIO used AI with home readings to speed up medication optimization. LINK-HF2 and the HearO trial are testing whether early AI alerts keep patients out of the hospital. The evidence points in one direction. Home readings, reviewed by a clinician and paired with a clear plan, give patients with heart failure a better chance of staying well at home.
Tenovi is a healthcare IoT company that provides cellular-connected remote patient monitoring and remote therapeutic monitoring infrastructure. RPM software companies, chronic care management companies, telehealth platforms, health systems, payers, and pharmacies use Tenovi to deliver proactive care at home. To learn how Tenovi can support your heart failure program, contact us for a free demo and consultation.
