Causal Representation Learning for Decision Support Systems: Non-Communicable Diseases (NCDs) Management Empowerment via Interventions

Project Overview

There is a call for more preventive action in public health interventions. In their "2030 agenda for sustainable development" the World Health Organization has specifically addressed the need for preventing noncommunicable diseases. While modern public health care originally emerged in response to infectious disease epidemics, public health now includes noncommunicable diseases.

The World Health Organization defines noncommunicable diseases (NCDs) as chronic conditions caused by a combination of genetic, physiological, environmental and behavioural factors which can affect all age groups and countries. The main types of noncommunicable diseases include cardiovascular diseases, cancers, chronic respiratory diseases and diabetes.

The management of NCDs represents an increasing burden on health care systems in Europe and worldwide. They are on the rise, and represent 86% of all deaths in Europe. Furthermore, across Europe there are major disparities in life expectancy with differences between EU member states of up to 7.7 years for women, and 11 years for men. Health care systems also have to deal with multimorbidity, due to both lifestyle choices and the inherent health issues of an aging population.

AI in Healthcare

The health sector has been described as a key application area for Artificial Intelligence (AI) because of the richness and complexity of health data and the wish to improve the quality and efficiency of health care. There has as yet been very little research into the use of AI applications for patients or healthy citizens, although the technology may allow for a much more individual and person-centred approach regarding medical decision support. The core challenge is that most AI-based predictions lack explanations that allow for intuitive interpretation which is often needed to support decision making.

The rise of AI has led to a growing interest in developing more explainable models, especially causal models in healthcare. There is a growing demand for AI techniques that not only deliver reliable results, but are also trustworthy, transparent, interpretable, and explainable to human experts. To achieve this goal, methods and models are required to replicate and interpret the machine decision-making process, as well as to simulate and explain the learning and knowledge extraction processes.

Research Goals

• To develop novel structure learning algorithms to advance our understanding of the complex causal relationships underlying NCDs
• Enhance personalised treatment plans which could optimize resource allocation, and minimize the burden of NCDs on global health
• Create interpretable AI models that provide actionable insights for healthcare practitioners
• Develop frameworks for causal inference in longitudinal health data