Ecological studies of Clostridioides difficile and COVID-19 infection with the application of space-time risk models

Student thesis: Doctoral Thesis

Abstract

Infectious diseases continue to pose major global health threats. With the recent devastation from the COVID-19 pandemic and growing concerns of healthcare-associated infections (HAIs), there is a worldwide requirement for stringent techniques to monitor and understand the key drivers for infections. Infectious diseases have an inherent spatial dimension due to the contagious nature of viruses and bacteria. This thesis aims to explore the use of spatial and spatio-temporal techniques applied to infections, specifically Clostridiodies difficile infection (CDI) and COVID-19, to identify risk factors at an ecological population-based level. A mixture of open-sourced and routinely collected data, at different spatial scales, were used to understand the surveillance capacities of observational public health data. Antimicrobial prescribing and stewardship have been a global focus in the last decade as concerns have grown with emergent novel antibiotic-resistant infections. CDI has been shown to have a well-defined association with certain broad-spectrum antibiotic classes and other environmental factors, however, there is a gap in the literature aiming to understand these relationships ecologically and spatially. The main focus of this thesis was to use spatio-temporal models to investigate spatial risk factors of CDI incidence, such as GP antimicrobial prescribing, in Scotland and Wales. Similar spatial techniques were then applied to investigate the spatial distribution of COVID-19 testing during the first wave of the 2020 epidemic in Scotland. The relevant spatial and spatio-temporal models applied throughout this thesis were initially discussed in Chapter 2. The spatial distribution of Scottish GP antibiotic prescribing rates, from 2016 to 2018, was investigated in Chapter 3 using spatial point-location correlation methods. Risk factors of increased GP antibiotic prescribing were explored, showing GP practice demographic information as key drivers of increased antibiotic prescribing. These analyses were followed by an exploration of Scottish CDI incidence data, from 2014 to 2018, at a small areal level (intermediate zones (IZ)), to understand spatial auto-correlation and temporal trends of CDI incidence in Chapter 4. Population demographic risk factors, as highlighted in the literature, were obtained at the same spatial scale and assessed as ecological risk factors of CDI incidence using conditional autoregressive (CAR) models. The next phase of this thesis then combined the previous two analyses, introducing a multi-level spatial problem, which aimed to explore central risk factors of CDI that were not available at the same spatial scale in Chapter 5. Spatial interpolation methods were applied to manipulate GP antibiotic prescribing point-location data and areal-unit cattle density data to match the CDI incidence at an IZ spatial scale. These data could then be explored as ecological risk factors of CDI incidence, carrying forward the previously defined CAR model from Chapter 4 and adjusting for demographic confounders. Welsh CDI incidence and primary care antibiotic prescribing data offered the opportunity to compare between two countries in the UK. The retrospective ecological study in Chapter 6 used aggregated disease surveillance data to understand the impact of total and high-risk Welsh GP antibiotic prescribing on total and stratified inpatient/noninpatient CDI incidence. Location and health board information were anonymised preventing a formal spatial analysis, however, the results were comparable to previous chapter findings and supported the hypothesis of an increased risk of CDI incidence reflected in GP antibiotic prescribing rates, particularly high-risk antibiotics, and population demographics. Finally, at the beginning of the COVID-19 pandemic, it became evident that the methodologies applied in this thesis could support the investigation of the spread of COVID-19 infections. The work presented in Chapter 7 aimed to explore how best to capture spatial patterns of community COVID-19 infection by conducting a spatiotemporal analysis on three data streams { positive test rates, relevant NHS24 calls and COVID Symptom Study (CSS) predicted cases, to assess which was best for early disease surveillance. Results showed both sources to identify similar trends of COVID-19 and gold-standard testing data, particularly when used in parallel. This thesis has provided new insights into the associated risks between CDI incidence and GP antibiotic prescribing in Scotland and Wales, demonstrating the capabilities of open-source and routinely collected public health data when applied in a spatial framework. These results support the requirement of stringent measures to reduce antibiotic prescribing in the community. It also highlights the beneficial use and suitability of analysing infectious disease data with spatial techniques to address gaps in the literature to understand population-based risk factors of disease. There is a strong argument for future research into methods of analysing multi-level spatial data, particularly in the application of observational public health data.
Date of Award31 Mar 2022
Original languageEnglish
Awarding Institution
  • University Of Strathclyde
SponsorsUniversity of Strathclyde & EPSRC (Engineering and Physical Sciences Research Council)
SupervisorChris Robertson (Supervisor) & Kimberley Kavanagh (Supervisor)

Cite this

'