Artificial Intelligence approach to improving blood pressure treatment

Hypertension is the leading cause of premature death worldwide, but the prevalence of hypertension control remains low. Physician capacity (number of physician clinic visits available for hypertension care) is a critical rate-limiting, determinant of hypertension control resulting in a major care-gap in hypertension management in all regions of the world. Artificial Intelligence Clinical Decision Support Systems (AI-CDSS) for Hypertension holds considerable potential to improving hypertension management but require rigorous evaluation before assimilation into routine clinical practice.

Objective: To develop and evaluate an AI-CDSS for Hypertension management. Populations: Training: SPRINT trial (n=9361) and External Validation: ACCORD trial (n=4733) and UK Biobank/ General Practice Data for Planning and Research (GPDPR)(n=222,063).

Methods:

Development: Create a clinical decision support system using a machine learning approach (XGBoost) and a deep learning approach (Feed-forward neural network).

Validation: Externally validate in a clinically distinct RCT cohort (ACCORD trial) and a real-world population (UK Biobank/GPDPR) to ensure generalisability.

Explainability: Use state of the art explainability methods including Local interpretable model-agnostic explanations (LIME), SHapley Additive exPlanations (SHAP) and Diverse Counterfactual Explanations (DiCE) to aid both clinical and patient interpretation of the treatment decision.

Safety: Develop an out-of-distribution detector and uncertainty estimate to operate in series with AI-CDSS for Hypertension.

Human Factors: Analyse the perceptions of doctors and persons with hypertension using validated tools. Conduct usability tests and use this usability information to improve the doctor-computer interaction.

Prospective Validation: Conduct a prospective comparative evaluation between the AI-CDSS for Hypertension, General Practitioners, and Hypertension Specialists.

Impact: Development and validation of an AI-CDSS could revolutionise the management of hypertension, and form a template for similar approaches to managing other risk factors (e.g., hypercholesterolemia and hyperglycaemia). This research proposal will also facilitate my transition towards research independence and development as an emerging leader in AI-guided cardiovascular medicine. 

Award Date
22 September 2023
Award Value
€711,222.63
Principal Investigator
Dr Conor Judge
Host Institution
University of Galway
Scheme
HRB Postdoctoral Fellowships: CSF 2023