ISSUE 026 · Oct 28, 2026 · 8 min read

What if a heart-failure alert reached same-day decongestion—not just a dashboard?

Connect personalized post-discharge deterioration signals to same-day clinical review and label-concordant home subcutaneous diuresis for eligible patients—then test days alive and out of hospital and safety, not alert accuracy alone.

01Clinical problem · Evidence

An alert can fire while congestion remains untreated.

Soon after heart-failure discharge, recurrent fluid retention can fall between symptoms and scheduled care. In LINK-HF2, AI detected physiological instability and clinicians responded within a median of one day, yet rehospitalization, emergency visits, and death were not reduced.

02Technology stack · Proposal

Join sensing, human judgment, and home checks into a same-day pathway.

When a multisensor patch deviates from a patient-specific baseline, the HF team confirms symptoms, weight, blood pressure, medicines, and alternative diagnoses by phone, video, or home visit. Only eligible patients proceed—after renal and electrolyte checks and a prescription—to label-concordant subcutaneous furosemide, with next-day response and safety review.

03Research question · Hypothesis

Does closing the action loop increase time safely spent outside hospital?

Compared with AI alerts plus structured clinical review alone, does adding same-day home decongestion for eligible patients increase 90-day days alive and out of hospital without more acute kidney injury, electrolyte disorders, hypotension, or inappropriate treatment?

04First study · Proposal

Thirty hospitals, 1,200 patients, three arms, 90 days.

Randomize adults discharged after acute HF to (1) usual care with sensor data masked, (2) AI alerts plus same-day clinical review, or (3) alerts plus review and, when eligible, home subcutaneous diuresis. The primary efficacy outcome is 90-day days alive and out of hospital.

A safety composite of death, KDIGO stage 2/3 acute kidney injury, severe hyponatremia or hypokalemia, and symptomatic hypotension uses a 3-percentage-point absolute noninferiority margin. Also measure HF readmission, ED use, alert-to-treatment time, KCCQ, patient and staff burden, and cost.

05Defensible data · Hypothesis

The moat is why treatment happened—and what followed—not a risk score.

Link anomaly → symptoms, weight, blood pressure, and labs → clinician decision or override → dose completion → urine and weight response → renal, electrolyte, and clinical events. Evaluate model changes only against this full event stream and subgroup safety.

06Venture thesis · Assumption

Start with HF programs, ACOs, and Medicare Advantage—the product is a 90-day care episode.

Bundle sensing, same-day nurse and physician review, home-lab operations, drug-device logistics, and safety reporting into a 90-day episode fee. Value is fewer hospital days, revisits, and staff hours—not alert volume.

Add outcomes-based pricing only after clinical benefit and net savings are demonstrated; avoid per-alert or per-dose incentives.

Key numbers

The delivery gap between two clinical trials

171 people

Analyzed in the LINK-HF2 wearable-AI alert trial

HR 1.03

LINK-HF2 composite clinical events—no difference versus usual care

+4.0 days

Thirty-day days-alive-and-out-of-hospital difference in SUBCUT-HF II

14% vs 6%

HF readmission in SUBCUT-HF II—imprecise but important numerical signal

Evidence and interpretation

Detection and delivery are each feasible—the combined pathway is unproven

Established: in the 2026 randomized LINK-HF2 trial, the wearable-AI intervention reduced physiological instability but, among 171 analyzed participants, did not reduce HF rehospitalization (HR 0. 90, 95% CI 0.

43–1. 84) or the composite of rehospitalization, ED use, or death (HR 1.

03, 95% CI 0. 63–1.

70). In SUBCUT-HF II, an early-discharge pathway using wearable subcutaneous furosemide increased 30-day days alive and out of hospital by 4.

0 days, but HF readmission was numerically higher at 14% versus 6%, with wide uncertainty. The FDA label authorizes FUROSCIX for edema in chronic HF or CKD but not chronic use, and requires monitoring of electrolytes, BUN, and creatinine plus training to respond to incomplete-dose alarms.

Still an assumption: no evidence yet shows that combining these components improves outcomes or cost. That is the research and venture hypothesis for a new randomized trial.

LINK-HF2 randomized trial LINK-HF2 registry NCT04502563 SUBCUT-HF II randomized trial SUBCUT-HF II registry NCT05419115 FDA FUROSCIX prescribing information (July 2026) Real-world pathway study NCT06979726

Failure conditions and safety, ethics, and regulatory risks

Clinical and business failure

Do not scale if improvement in 90-day days alive and out of hospital is under three days or the upper confidence bound for the safety-risk difference exceeds three percentage points. Reject the business case if staff time exceeds 30 minutes per patient-month or avoided cost does not cover the service.

Inappropriate automation

AI never determines diagnosis, dose, or prescribing. A clinician reviews alternatives such as acute coronary syndrome, arrhythmia, infection, or pulmonary embolism, and every dose requires a separate prescription and eligibility check.

Renal, electrolyte, and blood-pressure harm

Check weight, blood pressure, creatinine, sodium, and potassium before and after dosing. Hold treatment or escalate to in-person care for hypotension, acute kidney injury, or severe electrolyte abnormalities, and review ototoxicity risk and interacting medicines.

Device and incomplete-dose failure

Adhesion loss, fluid contact, or movement can interrupt delivery. Train patients and caregivers to respond to alarms; if completion is unconfirmed, never recommend a repeat dose automatically—route it to the clinical team.

Equity, adherence, and surveillance burden

Audit alert performance and outcomes by skin tone, age, sex, rhythm, renal function, housing, and digital access. Never interpret nonwear or disconnection as clinical stability; include device, connectivity, and caregiver support in the intervention.

Combination-product and algorithm regulation

Current evidence concerns separate sensor alerts and a drug-device product—not an authorized autonomous closed loop. Adaptive thresholds, automated recommendations, new treatment populations, and model updates may require medical-device, drug-label, investigational, and change-control review.

Primary sources

References

Established facts are separated from research and venture hypotheses that still require validation.

01
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. Published online October 9, 2026.
Analyzed 171 patients across five VA centers. Physiological instability decreased, but HF rehospitalization and the composite clinical outcome did not.DOI 10.1016/j.jchf.2026.103422
02
ClinicalTrials.gov. LINK-HF2: Multisensor Non-invasive Remote Monitoring for Prediction of Heart Failure Exacerbation. NCT04502563.
Registry record for the LINK-HF2 randomized design, intervention, and outcomes.NCT04502563
03
Campbell RT, Osmanska J, Docherty KF, et al. Early Discharge Strategy Including Subcutaneous Furosemide in Hospitalized Heart Failure. J Am Coll Cardiol. 2026;88(15):1781–1793.
Randomized 172 patients across 20 UK hospitals. The early-discharge arm gained 4.0 days alive and out of hospital at 30 days, while HF readmission was 14% versus 6%.PMID 42872757 · DOI 10.1016/j.jacc.2026.07.047
04
ClinicalTrials.gov. Subcutaneous Furosemide in Acute Heart Failure II (SUBCUT-HF II). NCT05419115.
Prospective registration for the early-discharge strategy using wearable subcutaneous furosemide.NCT05419115
05
U.S. Food and Drug Administration. FUROSCIX (furosemide injection) prescribing information. Revised July 2026.
Label for intermittent use in edema due to chronic HF or CKD in adults. It specifies 80 mg over five hours, monitoring, training, and response to incomplete-dose alarms.NDA 209988 · revised 07/2026
06
ClinicalTrials.gov. Gator SCALES-WHF: Subcutaneous Ambulatory Lasix Experience Study in Worsening Heart Failure. NCT06979726.
An observational evaluation of subcutaneous furosemide in a real-world worsening-HF pathway; it is not evidence for the proposed AI-triggered randomized pathway.NCT06979726

All issues