3-005 Assessing the predictive capabilities of an AI-device in heart failure: a heart failure nurse's perspective from a UK multi-centre trial
No Thumbnail Available
Authors
Matthews, Iain
Issue Date
2025
Type
Article
Language
en
Keywords
heart failure , monitoring device , artificial intelligence (AI) , peripheral oedema
Alternative Title
Abstract
Introduction Heart failure (HF) remains a major challenge in healthcare, with hospital readmissions often driven by late detection of worsening symptoms. Traditional self-monitoring methods, such as daily weighing, rely on patient adherence, which can be inconsistent. This study explores an AI-powered device designed for remote, home-based monitoring of peripheral oedema, a key indicator of HF deterioration.1 The device captures real-time data without requiring patient engagement, alerting healthcare teams or caregivers when early signs of worsening HF are detected. This passive, patient-friendly monitoring approach may improve patient safety and reduce the burden on clinical teams. 2 3
This observational study aimed to assess the AI device’s ability to predict HF events through changes in peripheral oedema, comparing its effectiveness to standard weight monitoring. 4 5 Key objectives included evaluating the data's frequency, reliability, and clinical relevance and determining the lead time for early intervention before HF-related deterioration.
Methods Between February 2020 and June 2022, 26 patients across five NHS Trusts in the UK were approached by nursing teams for enrolment in the study and were monitored using both the AI device and Bluetooth-enabled weighing scales. While weight readings were visible to patients, the AI device collected foot volume data passively without requiring any user interaction. Clinical records were later analysed to assess how well each method identified HF deterioration.
Results Among the 26 monitored patients, seven HF events occurred across six individuals, including one HF-related death. The AI device provided an average lead time of 13 days (range: 8–19 days) before hospitalisation for patients enrolled at least two weeks before an alert was triggered. When all five HF events detected by the device were considered, the lead time averaged 8 days (range: 1–19 days).
In contrast, weight monitoring was highly inconsistent, with scales recording only 1.1 days [0.3–4.1] of data per week, compared to the 5.9 days [4.9–6.6] of weekly data from the AI device. Notably, weight-based monitoring failed to predict any HF-related hospitalisations in this study.
Patient adherence to traditional weight monitoring was poor, whereas the AI device required no action from patients, making it a practical tool for nursing teams managing HF care remotely. At the end of the study, 82% of surviving participants (18 of 22) chose to keep the device. As of January 2024, five participants were still actively using the device. This high level of patient acceptability suggests that a non-intrusive, passive monitoring approach may improve long-term management with HF management.
Conclusion This study demonstrates that an AI-enabled monitoring device offers a more reliable, patient-friendly alternative to weight-based self-monitoring in HF. Providing advanced warning of clinical deterioration could help nurses and care teams intervene earlier, improving patient outcomes and reducing hospital readmissions. As remote monitoring becomes a larger part of HF care, nurses will play a key role in integrating new technologies into routine practice.
Further research is underway to validate the device’s impact on HF management, particularly in reducing hospitalisations. While the findings suggest promising benefits, a clear demonstration of hospitalisation reduction is essential before significant investment in time and resources is made to integrate new technology into clinical practice. Ensuring cost-effectiveness and clinical utility will be critical in determining its broader adoption in HF care.
Description
Citation
Hann K, Dewhurst M, Matthews I, et al (2025). '3-005 Assessing the predictive capabilities of an AI-device in heart failure: a heart failure nurse’s perspective from a UK multi-centre trial', Heart. 111 (A104-A106)
Publisher
License
Journal
Heart