Methods to explore risk factors in complex health conditions

Abstract

Understanding how risk factors relate to complex health outcomes is a key goal in health research. This thesis investigates two such conditions: Irritable Bowel Syndrome (IBS) and frailty. Both conditions are marked by multifactorial origins, inconsistent definitions, and diverse behavioural and biological contributors. Demographics and geography add further layers of complexity. Despite their differences, IBS and frailty are examined using a shared analytical framework and population-based datasets. IBS, a Disorder of Gut-Brain Interaction, affects gastrointestinal function and presents symptoms like pain, bloating, constipation, and diarrhoea. Its variability and sensitivity to dietary triggers make it analytically challenging. Frailty, defined as heightened vulnerability to adverse health outcomes, lacks a standard definition and is often assessed using biomarkers. These biomarkers, however, show high variability, raising concerns about their reliability. To navigate these complexities, the thesis applies advanced modelling techniques: regression, Principal Component Analysis (PCA), Latent Class Analysis (LCA), and Model-Based Recursive Partitioning (MOB). Results show that LCA improves interpretability in IBS research by grouping dietary patterns in symptom-relevant ways. MOB further identifies subtle variations in diet-IBS links. In contrast, frailty’s association with biomarkers yields less clarity, suggesting that biological data alone cannot capture its multidimensional nature. Yet, MOB proves useful again, revealing nuanced relationships across frailty’s biological, functional, and social domains.This thesis originated from a research interest in elucidating the associations between complex health conditions and their underlying risk factors through the application of diverse analytical methodologies. Complex health conditions are inherently multifactorial, arising from the dynamic interplay of biological, behavioural, demographic, and environmental determinants. Contemporary health research increasingly relies on multivariate datasets encompassing heterogeneous populations and a broad spectrum of risk factors, necessitating sophisticated analytical approaches to disentangle these intricate relationships. To address these challenges, this thesis draws upon two distinct datasets, each offering valuable insights into different dimensions of health. The first dataset comprises cross-national data on Disorders of Gut-Brain Interaction (DGBIs), with a particular emphasis on dietary patterns, while the second focuses on older adults and includes detailed biological markers. The use of these datasets facilitates the investigation of two separate yet methodologically compatible topics: the association between frailty and biomarkers, and the relationship between Irritable Bowel Syndrome (IBS), one of the DGBIs, and dietary factors. Despite the clinical and conceptual differences between frailty and IBS, both conditions exemplify the analytical complexity posed by multifactorial health outcomes. The Rome Foundation Global Epidemiology Study (RFGES) and the English Longitudinal Study of Ageing (ELSA) provide robust platforms for such inquiry, offering extensive data on sociodemographic characteristics, self-reported symptoms, biomarkers, and dietary intake. The analytical framework employed in this thesis includes regression modelling, dimensionality reduction techniques such as Principal Component Analysis (PCA), clustering, and a further technique: Model-Based Recursive Partitioning (MOB). These methods enable the identification of associations, latent patterns, and subgroup heterogeneity, thereby advancing both epidemiological understanding and methodological innovation in the study of complex health conditions.