Artificial intelligence and advanced statistical methods for predictive modelling in lifestyle improvement and noncommunicable disease prevention

Non-communicable diseases (NCDs), driven by aging populations and behavioral risk factors like poor diet and tobacco use, pose a major global health threat. Innovative prevention strategies are essential to reduce their rising impact on mortality and morbidity. AI and digital health technologies show great promise for NCD prevention, offering accurate risk prediction and personalized interventions. Advances in data-driven methods enhance health monitoring and can significantly support efforts to reduce the global NCD burden. This thesis demonstrates how the Food4HealthyLife (F4HL) project contributes to advancing personalized disease prevention. Led by the University of Bergen in collaboration with the University of Bologna and others, the project aims to develop predictive models for personalized risk estimation based on dietary habits. It also highlights how artificial intelligence, particularly natural language processing (NLP), can enhance these models. The project is structured in three parts: 1. Identification and synthesis of epidemiological evidence for expanding the F4HL suite. Key findings include links between food groups and mortality, diabetes outcomes, and identifying harmful dietary habits in Italians. 2. Validation and development of NLP-based tools to optimize the appraisal of scientific evidence. Key results include the development of TextAlchemy for extracting data from articles and the validation of ASReview’s effectiveness for screening in epidemiological umbrella reviews. 3. Investigation of the habits and attitudes of citizens regarding the use of digital health technologies (DHTs) and their impact on interactions between patients and healthcare professionals (HCPs). Findings reveal age and education influence DHT use and perceptions, with country-specific patterns. Citizen-HCP interactions remain limited but show potential for growth, as many users express interest in more frequent engagement. This work supports integrating AI into research and healthcare, aiming to build an evidence-based suite for personalized lifestyle recommendations. With professional support, its adoption could reduce the global burden of unhealthy habits and NCDs.