42 Use of a home monitoring application to develop and test ai-driven, personalized heart failure care at home – initial experience of the stabilise-hf study and 'passion-HF' decision engine. (6th October 2022)

• Record Type: Journal Article
• Authors: Taha, S; Barrett, M; Murphy, M; McDonald, K
• Is Part Of: Heart. Volume 108(2022)Supplement 3
• Page Start: A35
• Page End: A36
• Publication Date: 2022-10-06
• Subjects: Heart -- Diseases -- Treatment -- Periodicals; Cardiology -- Periodicals; 616.12
• DOI: 10.1136/heartjnl-2022-ICS.42
• Languages: English
• ISSNs: 1355-6037

Abstract

Abstract : Introduction: The demand for scheduled outpatient care is increasing worldwide, with heart failure (HF) care posing a particularly challenging problem. Telemonitoring and eHealth strategies may provide a solution to the increasing burden of care by shifting care away from specialist and tertiary services, and move care nearer to the patient. SanaCoach HF is an existing basic telemonitoring application ('app'), currently in clinical use in the Netherlands, which allows patients to submit measurements such as weight, blood pressure and symptom scores for review by a supervising clinical team. In our study, we applied a basic therapeutic algorithm (based on the ESC HF guidelines) to see if automated titration of medications could be integrated into this platform and perform routine aspects of HF care without need for healthcare provider (HCP) input. Methods: Each patient was instructed to input vital signs, symptom scores, and weight through their use of the 'SanaCoach HF' app. Further parameters such as cardiac imaging data, baseline medications and laboratory results were manually inputted by local HCPs. Patients submitted a 'monitoring session' for evaluation every 2 weeks. Based on a predefined therapeutic algorithm, including above measurements and patient symptoms, a background application (PASSION Graphic User Interface) made a recommendation on either 1.Disease-modifying therapy change a.Addition of new agent b.Increase in dose of existing agent 2.No medication change a.Contact HCP for discussion Results: Data from the first 3 months of the ongoing STABILISE-HF study were included, comprising 21 patients and 45 'sessions'. On review of these sessions, 35 (77.8%) of the reviewed sessions were deemed appropriate responses, and 10 (22.2%) were deemed to not be the ideal response. Of these 10 'inappropriate' responses, 4 were merely because the patient had reached optimal medical therapy, which is not accounted for in our model ( figure 1 ). Table 1, below, demonstrates a summary of the outputs generated by the AI model. Interestingly, there is a relatively high prevalence of a recommendation for increasing mineralocorticoid receptor antagonist (MRA) dose (26.7% of all outputs), consistent with published literature indicating undertreatment with these agents in real world data. 20% of all recommendations are to commence a sodium-glucose co-transporter 2 inhibitor (SGLT2i), reflecting the novelty of this agent and the fact that stable outpatients were recruited at routine visits. Conclusion: In this early experience of eHealth supported heart failure home monitoring, our model generated a correct suggested therapy change in 77.8% of cases, suggesting that home titration using a closed/self-sufficient system may potentially safely replace healthcare provider-guided medication titration. Enriching the dataset further with more clinical data will allow refinement of the algorithm and hopefully allow the creation of a comprehensive 'Doctor-at-home' system.