Artificial Intelligence Prediction Tool for Prediabetes and Type 2 Diabetes Risk Stratification Using Computed Tomography Scans: An Early Economic Evaluation

Speaker(s)

Altayeb R1, Mavrogiannis M2, Patel P3, Dicken L4, Sohan Y4, Greenwood JP5, Kardos A6, Antoniades C3, Rose J7, Bajre M8
1Health Innovation Oxford and Thames Valley, Reading, UK, 2University of Oxford, Oxford, OXF, UK, 3University of Oxford, Oxford, Oxfordshire, UK, 4Caristo Diagnostics Limited, Oxford, Oxfordshire, UK, 5Leeds Teaching Hospitals NHS Trust, Leeds, West Yorkshire, UK, 6Milton Keynes University Hospital NHS Trust, Buckingham, Buckinghamshire, UK, 7Health Innovation Oxford and Thames Valley, Oxford, Oxfordshire, UK, 8Health Innovation Oxford and Thames Valley, Oxford, OXF, UK

Presentation Documents

OBJECTIVES: Early economic evaluation was performed to compare the indicative cost savings associated with an AI-powered risk stratification tool developed to detect pre-diabetes and type 2 diabetes (T2D) by identifying adipose tissue inflammation using routine computed tomography (CT) scans.

METHODS: The AI risk stratification tool can be integrated into existing clinical pathways to provide additional analysis of current routine CT scans from any manufacturer’s platform. A scenario-based hypothetical decision analysis model was developed to evaluate the tool's potential impact on the diagnosis of diabetes and associated complications. The scenarios tested were the use of AI in providing valuable insights for risk prediction leading to a potential reduction in the prevalence of diabetes by facilitating earlier identification of pre-diabetic cases and, thus, enabling timely intervention through programs such as the NHS Diabetes Prevention Program (NHS DPP), and the reduction in the number of patients developing diabetes complications through earlier intervention and management. A comprehensive scoping review identified direct costs within the NHS associated with pre-diabetes, diabetes management, and its complications. The model considered direct healthcare costs along with the AI implementation costs, disease prevalence, and the DPP uptake rates.

RESULTS: The analysis estimated potential cost savings from reduced incident diabetes and its complications. The results indicated an annual cost saving of £3,234 per identified prediabetic case reverted to normoglycemia and £2,548 per identified prediabetic case prevented from developing complications. In a population of 10,000, with a projected 7% reduction in diabetes cases and a 20% reduction in complications due to diabetes, the model suggests that the proposed AI tool applied to routine CT scans for pre-diabetes identification may potentially save £341,734 annually in the NHS.

CONCLUSIONS: This early-stage economic analysis indicates that implementing the new AI risk stratification tool for identifying pre-diabetes could result in cost savings for the NHS.

Code

EE312

Topic

Economic Evaluation, Medical Technologies, Methodological & Statistical Research

Topic Subcategory

Cost-comparison, Effectiveness, Utility, Benefit Analysis, Diagnostics & Imaging

Disease

Cardiovascular Disorders (including MI, Stroke, Circulatory), Diabetes/Endocrine/Metabolic Disorders (including obesity)