Using UK Biobank data for England to understand health disparities in obesity and also the relationship between obesity and COVID-19

Abstract

This thesis examines spatial disparities in obesity and the relationship between obesity and COVID-19 by using UK Biobank cohort data of half a million participants aged 40-69 across England. Employing multilevel models and also methods of machine learning, it investigates whether predictors of obesity vary spatially within London or regionally across cities in England. Contrary to our expectations, the models suggest limited geographical variation with the direction and magnitude of the effects of the main risk factors, as well as the predictive power, broadly stable (spatially stationary) across cities. Furthermore, the impact of the main risk factors on obesity remains spatially stationary within the neighbourhoods of London city when the study area is limited to London. The thesis also investigates the spatial-temporal relationships between obesity and various individual COVID-19 risks, including COVID-19 infections, severe infections and mortality, during three waves of the pandemic outbreak in England. Participants who are overweight or obese are associated with a higher risk of infection by COVID-19, severe COVID-19 and mortality, especially in Wave 2 of the pandemic (from October 2020 to February 2021). Multilevel random slope models, allowing the effects of obesity to vary across cities in England, illustrate that the effect of obesity or being overweight on the risks of COVID-19 has little geographical variation for the Biobank participants. The cross-level interaction between neighbourhood poverty levels and obesity is not statistically significant, further suggesting that the risks of COVID-19 are nearly the same if participants are obese, regardless of where they live. This lack of geographical variation questions the significance of geographical context on obesity and its associations between obesity and COVID-19. The lack of evidence of contextual effects may be attributed to the 'healthy volunteer' effect. This phenomenon suggests that individuals who volunteer to take part in surveys tend to be healthier than those who do not. Overall, the thesis re-evaluates and questions the significance of geographical context on obesity and associations between obesity and COVID-19, enriches the literature with the adoption of machine learning methods for spatial analysis and examines the potential spatial varying effect from individual-level participation rather than using traditional area-based explorations.

Date of Award	11 Jan 2024
Original language	English
Awarding Institution	University of Bristol
Supervisor	Richard J Harris (Supervisor) & Emmanouil Tranos (Supervisor)