Spotting a warning sign and preventing a hospital stay turned out to be two different steps. Worth reading for anyone with heart failure, or caring for someone who has it, who relies on home monitoring to raise the alarm.
On 9 October 2026 the journal JACC: Heart Failure published the results of LINK-HF2, a randomized trial run at five US Veterans Affairs medical centres. It enrolled adults at discharge after a hospital stay for acute heart failure, the condition in which the heart cannot pump enough blood for the body's needs. Every participant wore a multisensor patch on the chest for up to 90 days. A personalized model learned each person's usual cardiorespiratory and activity signals and flagged deviations from them. In the intervention group those alerts reached clinicians, who followed a structured response algorithm; the control group received standard care. The trial was open-label, with blinded assessment of outcomes.
The alerts arrived, and clinicians acted within a day
The analysis included 171 participants: 80 in the intervention group and 91 in the control group. The median participant wore the patch on 72.4% of follow-up days. Clinicians responded to the notifications they received within a median of one day, and never later than five. A qualitative analysis found high acceptance among patients and engagement among clinicians; the main barriers were fitting the alerts into routine clinic work and uncertainty about what to do in response.
Compared with controls, the intervention group had fewer episodes of physiological instability detected by the algorithm (hazard ratio 0.76; 95% CI 0.60–0.97). This is an implementation outcome, not proof that each individual response by a clinician prevented a deterioration.
Readmissions stayed about where they were
Readmission for heart failure occurred in 13 of 80 people with alerts sent to clinicians (16.3%) and in 17 of 91 in standard care (18.7%): hazard ratio 0.90, 95% CI 0.43–1.84, P = 0.76. A combined measure of unplanned readmission for any cause, emergency department visits and death did not differ either (hazard ratio 1.03; 95% CI 0.63–1.70).
An interval stretching from 0.43 to 1.84 is compatible with a sizeable benefit, with no effect and with harm. With 171 people, the trial could not tell these apart, and the authors treat both clinical endpoints as exploratory. Their own reading is the most useful line in the paper: accurate detection of physiological risk may not be enough to change clinical risk unless it is paired with a treatment response that matches the underlying cause and that clinicians trust.
The first study showed the signal predicts; this one tested whether acting on it helps
The idea was tested earlier. In the first LINK-HF study, published in Circulation: Heart Failure in 2020, 100 participants with a mean age of 68.4 years, 98% of them men, wore a similar patch for up to three months after a heart failure admission. The analytics detected early signs of an upcoming heart failure readmission with 76–88% sensitivity and 85% specificity, a median of 6.5 days before the person returned to hospital. It was a prediction study: it tested how well the signal forecast readmission, not whether changing care improved outcomes.
LINK-HF2 asked the harder question: does acting on the warning keep people out of hospital? In this small trial, the answer was not yes. Prediction and prevention turned out to be two separate steps, and the second depends on what clinicians can actually change once an alert arrives.
How far the result reaches
This was a feasibility trial at veterans' hospitals, designed mainly to show whether the approach could work in practice, not to prove an effect on outcomes. It was open-label, so patients and doctors knew who received alerts. The trial tested one patch, one algorithm and one response protocol; it does not show whether the findings apply to consumer watches or other monitoring systems. The trial was funded by a Veterans Health Administration research grant, and six of the authors are employees of Prolaio Inc, a Chicago company. A larger trial powered for clinical outcomes would be needed for a firm answer.
If you or someone close to you lives with heart failure and uses home monitoring, what an alert should trigger and how the treatment plan changes are questions to discuss with your doctor.
This material is for information only and does not replace a consultation with a doctor. Nothing in it is a recommendation to start, stop or change treatment.
Sources
Sideris K, et al. Artificial Intelligence-Enabled Detection and Management of Physiological Instability in Heart Failure: Results of the Randomized LINK-HF2 Trial. JACC Heart Fail. 2026. PMID 42853165
Stehlik J, et al. Continuous Wearable Monitoring Analytics Predict Heart Failure Hospitalization: The LINK-HF Multicenter Study. Circ Heart Fail. 2020. PMID 32093506
LINK-HF2 — Remote Monitoring Analytics in Heart Failure. ClinicalTrials.gov. NCT04502563
Articles in this section are educational and are not medical advice, a diagnosis, or a prescription. Consult a qualified professional before acting on anything you read here.
Read next
ResearchMedicine and treatmentHearing-aid users died less often in large cohorts. Here is the catch
Pooled population data from four countries add weight to the idea that hearing belongs to long-term health, not just comfort. Worth reading for anyone putting off a hearing test or a prescribed aid, with a plain account of what this kind of data can and cannot show.
6 min read
ResearchMedicine and treatmentShorter radiotherapy after mastectomy held up in a phase 3 trial
Breast cancer radiotherapy, fatty liver disease, the quality of screening colonoscopy and bone health in older women: what four studies from mid-October found and where each result stops.
7 min read
ResearchMedicine and treatmentClots in the brain's veins: a large registry compares blood thinners
A rare form of stroke leaves patients with months of blood thinners and a choice of tablet. A registry of nearly 2,500 people adds data to that choice and shows how far evidence without randomization can take it.
4 min read