Description
The National Heart, Lung, and Blood Institute (NHLBI), of the National Institutes of Health (NIH), convened the "Transforming Hypertension Diagnosis and Management in the Era of Artificial Intelligence (AI)" workshop to discuss gaps and opportunities in leveraging AI technologies for hypertension diagnosis and management. Participants included healthcare practitioners, researchers, industry partners, and policymakers with expertise in adult and pediatric cardiology, hypertension, AI, health disparities, clinical trials, and outcomes research. The virtual workshop objectives were to:
- Assess the potential of AI technologies in improving the diagnosis and management of hypertension, aligned with patient-centric goals for enhanced quality and length of life.
- Explore novel AI-based therapeutic approaches and new information about social determinants of health (SDOH) and resource allocation.
- Examine the ethical, legal, and social implications of AI adoption in hypertension diagnosis and management.
Background
Hypertension, or high blood pressure (BP), is a significant public health concern in the United States, affecting approximately 45% of U.S. adults. Hypertension is a leading risk factor for cardiovascular diseases and stroke, and its prevalence has been increasing over time. Disparities in hypertension prevalence and control are evident among different racial, ethnic, and socioeconomic groups, with SDOH playing a crucial role in these disparities.
Current work in AI and machine learning (ML) has shown promise in various aspects of health, including disease prediction, diagnosis, and personalized treatment plans. The application of AI in hypertension prevention and management is a growing area of interest, with potential benefits such as improved BP measurement accuracy, better risk assessment, and tailored treatment strategies. The workshop was divided in the following themes addressing gaps and opportunities and to identify high priorities questions, current barriers and research opportunities.
Session 1: Bridging the Communication Gap Between Population-Health, Clinical Medicine, and Engineering
Gaps and Opportunities:
Gaps:
- Communication gaps between population health, clinical medicine, and engineering in relation to AI. Differences in language, culture, and priorities can lead to misunderstandings and misinterpretations, resulting in ineffective technologies and inadequate patient care.
- Comprehensive and representative datasets are essential for training AI models, but biases in large datasets can limit the effectiveness of AI in addressing the diverse patient needs.
- Users’ expectations of AI systems may not match the current state of the technology, and efforts should be made to build trust and establish clinically driven expectations regarding AI's role in healthcare.
- Regulatory challenges related to remote patient monitoring and cuffless BP devices and AI need to be addressed.
- Healthcare professionals and patients need education about applied AI technologies in their healthcare, including increased quantitative and statistical methods.
- Integrating AI algorithms into clinical workflows requires careful consideration to ensure that they are used effectively and efficiently.
- Ethical principles, data security, and respecting individuals' choices regarding data sharing need to be considered when implementing AI technologies in hypertension research.
Opportunities:
- Machine learning can be used to optimize treatment paths and improve hypertension management.
- New models of care can be developed that empower individuals and improve communication between patients, providers, and the AI system.
- Generative models, such as large language models, can be used to process complex data and improve electronic health record (EHR) usability for hypertension applications.
- Bidirectional collaboration between clinicians and engineers is required for better outcomes in AI-driven hypertension management.
- External validation can help ensure the effectiveness of AI technologies in addressing the diverse needs of patients.
- AI/ML health tools can be used to advance equity in healthcare including hypertension since they can analyze large volumes of diverse healthcare data and provide evidence-based decision support for healthcare professionals.
- Rigorous design in pragmatic trials is essential to test devices and digital health tools at scale.
Session 2: AI and Blood Pressure Measurement, Hypertension Risk, and Blood Pressure Control
Gaps and Opportunities:
Gaps:
- Lack of standardization and gaps in knowledge regarding several recent novel technologies for BP measurement.
- Challenges such as bias and lack of validation of novel technologies need to be addressed in AI approaches for hypertension detection and diagnosis.
- Challenges in data quality when integrating wearable devices for continuous monitoring, including diagnostic definitions/processes when using these new measurement devices.
Opportunities:
- The potential of novel technologies such as cuffless BP measurement devices using photoplethysmography (PPG), pulse arrival time (PAT), and multi-sensor approaches including AI methods.
- The potential of AI-driven models for personalized treatment plans and identifying individual risk factors and treatment modifiers for hypertension.
Session 3: Making a Dent: Improving Hypertension Management Using AI Methods
Gaps and Opportunities:
Gaps:
- Lack of infrastructure, interoperability, and content for AI-enabled technology (e.g., healthbots) to support self-management of hypertension.
- Limited precision and accuracy of AI technology may lead to limited readability.
- Need for evaluating AI tools in a robust way and developing better methods to optimize clinical benefit in hypertensive patients.
- Limited data inputs and data bias for people with hypertension who have never seen a clinician or are not in a study.
- Difficulty in finding signals of response and solutions due to large biases and confounding real-world data.
Opportunities:
- Using data collection and analysis to subcategorize patients into different clusters could help identify biomarkers for medication distribution or for tailoring different exposures for the patients.
- AI can be used to predict which individuals are at risk for developing hypertension and other diseases.
- Using AI to personalize the benefit and risks of treatment strategies in hypertension.
- AI can improve early diagnosis and management in clinical practice and help in risk predictions for heart failure, mild cognitive impairment, and dementia.
- AI risk scores can help identify those at the highest risk and improve shared decision-making between clinicians and patients.
Session 4: Real World Implementation Challenges and Issues for AI and Hypertension
Gaps and Opportunities:
Gaps:
- Limited support for AI integration in EHRs, especially in low-resource settings, potentially exacerbating health disparities between the rich and the poor.
- Difficulty in integrating various devices and software that relates to hypertension into a cohesive system for patient care.
- Overburdened health-care providers and staff with data input, updates, and ensuring interoperability.
- Regulatory compliance challenges, which can prevent the implementation of AI systems.
- Accuracy and non-random biases in remote patient monitoring and cuffless BP devices.
Opportunities:
- Equitable delivery of advanced, AI-driven care by leveraging and advancing interoperability standards, with a focus on hypertension-specific requirements.
- Direct patient engagement through patient-facing apps, providing guidance and information to patients.
- Data-driven, individualized patient care, considering variations in patient profiles and needs.
- Providing AI-based decision support for titration of antihypertensive medication.
- Ensuring accuracy and calibration in emerging technologies, such as cuffless BP devices, to maintain clinical value.
- Addressing racial and gender biases and other potential discrimination issues in AI and ML technologies by calibrating and validating across diverse and comprehensive populations.
- Upholding ethical principles, data security, and respecting individuals' choices when developing and implementing AI and ML tools.
- Collaborating with legal experts to navigate rules and regulations related to data usage in health care.
- Focusing on health equity by providing solutions that are accessible, affordable, and efficient across various settings and populations.
Conclusions:
The workshop identified key gaps and opportunities in leveraging AI technologies for hypertension diagnosis and management. Participants emphasized the importance of interdisciplinary collaboration; data collection, management, and integration; testing AI models; addressing SDOH factors; ensuring ethical AI use; and overcoming real-world implementation challenges.