In the prospective randomized LINK-HF2 study of 171 adults discharged after acute heart failure, AI-enabled wearable alerts integrated into clinician follow-up did not reduce HF rehospitalization versus standard care: 16.3% vs 18.7% (HR 0.90, 95% CI 0.43-1.84; P = 0.76). The composite of unplanned rehospitalization, ED visits, and all-cause mortality also was unchanged (HR 1.03, 95% CI 0.63-1.70; P = 0.89), despite lower AI-detected physiological instability in the intervention arm (HR 0.76, 95% CI 0.60-0.97; P = 0.025).

Why It Matters To Your Practice

• AI monitoring was feasible in postdischarge HF care, with median adherence on 72.4% of follow-up days and clinician response to alerts within a median of 1 day.

• But better signal detection did not translate into fewer clinical events, underscoring the gap between identifying risk and changing outcomes.

Clinical Implications

• Do not assume that adding AI alerts alone will reduce rehospitalizations after HF discharge.

• If you deploy AI monitoring, success likely depends on having a trusted, specific response pathway for what to do when an alert fires.

• Workflow integration and uncertainty about how to respond were key implementation barriers for clinicians.

Insights

• LINK-HF2 was a prospective, randomized, open-label, blinded-endpoint trial across 5 VA medical centers.

• Participants in both groups wore a continuous multisensor patch for up to 90 days; only the intervention arm triggered clinician notifications based on a personalized similarity-based model.

• Patients and clinicians generally accepted the system, suggesting implementation is possible even if outcome benefit remains unproven.

The Bottom Line

• For clinicians, this study is a reminder that accurate AI risk detection is not enough on its own.

• Until alert-linked actions are better aligned with HF pathophysiology and care workflows, AI monitoring may improve surveillance more than patient outcomes.