Using AI to explore cause and effect in large healthcare datasets

This early-stage study will explore whether artificial intelligence (AI) can help our researchers uncover the relationships between cause and effect in large healthcare datasets. Large healthcare datasets bring together information from thousands or even millions of people, allowing researchers to identify patterns, understand the causes of disease, and evaluate how effective treatments are in real-world settings.

The team behind this project will analyse routine health records to gain a better understanding of what influences people’s health and how different treatments affect outcomes. Their research will focus on two areas:

• Type 2 diabetes management
• Stroke prevention

Using machine learning, the team will look for potential instrumental variables linked to treatments.

In health research, instrumental variables help researchers estimate cause-and-effect relationships. For example, a doctor might prefer to prescribe a particular medicine, and this might influence which treatment a patient receives. It should not, however, directly affect the patient’s health outcome.

Project aims

We want to:

• Demonstrate that using AI in this way is feasible
• Identify technical challenges
• Find new instrumental variables for studying cause-and-effect relationships in health data

What we hope to achieve

We hope this initial study will directly support the development of future studies.

Identifying new instrumental variables using AI will allow our researchers to investigate treatments for type 2 diabetes and stroke prevention.