Scholarlink

Hypertension is a major global health problem linked to heart disease, stroke, and premature death, affecting more than one billion adults worldwide. Its complex interplay of physiological and behavioural determinants, combined with infrequent clinical visits and reliance on self-reported readings, makes its control challenging and contributes to its growing prevalence. Digital health technologies, including wearable devices, mobile health (mHealth) apps, and artificial intelligence (AI), have emerged to enable continuous health monitoring and adaptive feedback. However, most existing apps track only isolated determinants (e.g., sleep or physical activity) and lack explainability, thereby limiting accurate risk assessment and clinical trust. To address these gaps, we present HyperCare – a novel, AI-driven, personalized, and adaptive persuasive technology that integrates multimodal sensing of health determinants, persuasive strategies (PS), and explainability to prevent and manage hypertension. The technology continuously collects multimodal data, including blood alcohol content, blood glucose, body weight, sleep patterns, and activity level (step count). These data are then automatically analyzed using a retrieval-augmented generation (RAG)-based and explainable large language model (LLM) to recommend evidence-based interventions in real time while operationalizing 11 PS. Next, we conduct an expert evaluation with clinicians using 20 well-established heuristics across three hypertension-risk scenarios. The clinicians rate HyperCare with mean scores of 100% for clinical relevance, 98.9% for transparency and explainability, 100% for persuasion, 100% for ethics, and 100% for usability. Strong inter-rater agreement is observed (Krippendorff’s α = 0.92, 95% CI [0.75, 1]; Cohen’s κ = 0.90, p < .001), confirming the system’s overall effectiveness. These findings demonstrate the potential of integrating multimodal sensing, PS, and explainable AI into a technology (HyperCare) for continuous and personalized hypertension prevention and management.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

HyperCare: An AI-Driven, Personalized, and Adaptive Persuasive Technology for Continuous Hypertension Prevention and Management

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Josteve Adekanbi,
  • Japheth Kimeu,
  • Gladwin Irudayaraj,
  • Olumide Thomas Adeleke,
  • Ibukun Okunade,
  • Rita Orji,
  • Oladapo Oyebode

摘要

Hypertension is a major global health problem linked to heart disease, stroke, and premature death, affecting more than one billion adults worldwide. Its complex interplay of physiological and behavioural determinants, combined with infrequent clinical visits and reliance on self-reported readings, makes its control challenging and contributes to its growing prevalence. Digital health technologies, including wearable devices, mobile health (mHealth) apps, and artificial intelligence (AI), have emerged to enable continuous health monitoring and adaptive feedback. However, most existing apps track only isolated determinants (e.g., sleep or physical activity) and lack explainability, thereby limiting accurate risk assessment and clinical trust. To address these gaps, we present HyperCare – a novel, AI-driven, personalized, and adaptive persuasive technology that integrates multimodal sensing of health determinants, persuasive strategies (PS), and explainability to prevent and manage hypertension. The technology continuously collects multimodal data, including blood alcohol content, blood glucose, body weight, sleep patterns, and activity level (step count). These data are then automatically analyzed using a retrieval-augmented generation (RAG)-based and explainable large language model (LLM) to recommend evidence-based interventions in real time while operationalizing 11 PS. Next, we conduct an expert evaluation with clinicians using 20 well-established heuristics across three hypertension-risk scenarios. The clinicians rate HyperCare with mean scores of 100% for clinical relevance, 98.9% for transparency and explainability, 100% for persuasion, 100% for ethics, and 100% for usability. Strong inter-rater agreement is observed (Krippendorff’s α = 0.92, 95% CI [0.75, 1]; Cohen’s κ = 0.90, p < .001), confirming the system’s overall effectiveness. These findings demonstrate the potential of integrating multimodal sensing, PS, and explainable AI into a technology (HyperCare) for continuous and personalized hypertension prevention and management.