Search

This dissertation is comprised of three separate manuscripts. Each manuscript concerns questions of social structure in relation to health, morbidity, and mortality. The first examines the epidemiologic theories taught most frequently in core master of public health coursework and proposes curricular changes to better engage with the structural determinants of health. The second provides a broad outline of one structural epidemiologic theory-political economy of health theory-for a public health audience. The third proposes a framework for 'structurally competent' global health pedagogy by defining and describing individual competencies. Together, these manuscripts seek to contribute to structural approaches to public health theory and pedagogy. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com/en-US/products/dissertations/individuals.shtml.]

This dissertation revealed several problems on the analysis of influenza propagation by indicating how agent-based modeling can be employed to measure the effectiveness of control measures and assist in improving health policy to decrease absenteeism among elementary students. The primary question posed was as following: "What is the effect of hand hygiene on the possible incidence rates among school children?" After creating an agent-based model representing the influenza transmission dynamic, the incidence rates were calculated based on the hand-washing success rates. The statistical results from the simulation model were displayed in graphical format. Finally, the author addressed the issue of measuring validity of the model. The statistical analysis on absenteeism from influenza was performed using data on missed school days in classrooms in one of the local schools in Tippecanoe County, where students exercise hand washing with soap on a regular basis. The analysis also considered data on absenteeism among children who were not required to perform hand washing routinely. This agent-based simulation method is an innocuous and economical approach to model the propagation of respiratory diseases such as influenza. It enables the researcher to model individual behaviors and interaction among individuals and their environment. This feature enables the researcher to represent influenza transmission dynamic more realistically and to provide in-depth analysis to inquiries for epidemiologist and public health professionals. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com/en-US/products/dissertations/individuals.shtml.]

In many control systems changes in the dynamics occur unexpectedly or are applied by a controller as needed. The time at which a controller implements changes is not necessarily known a priori. For example, many manufacturing systems and flight operations have complicated control systems, and changes in the control systems may be automatically implemented as needed in response to possibly unexpected external factors affecting the operations of the systems. Public policy makers in health-care have to launch timely and cost effective policies to deal with disease epidemics long term and short term. In drug dosing strategy for HIV treatment how treatment interruption strategies stimulate the body's own immune response and also control cost of treatment are important. Modeling of such systems lead to discontinuities in the dynamics governing the evolutions of the systems. We consider control problems governed by differential equations with discontinuities in the states at a sequence of points. The differential equations are referred to as impulsive differential equations, and the control problems as impulsive control problems. We consider general deterministic and nondeterministic impulsive control problems. Then, we consider the analysis of models coming from HIV treatment strategies, production planning, and asset allocation. The first chapter is devoted to a study of the qualitative analysis of an HIV treatment model. In Chapter two we present a mathematical model of an impulsive control problem. In Chapter three we consider the HIV treatment model considered in Chapter one along with HIV-immune system model. In Chapter four we consider a general impulsive stochastic control problem. In Chapter five we consider an example of production planning model. In the last chapter we study an asset allocation model. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com/en-US/products/dissertations/individuals.shtml.]

Information technology (IT) use for epidemiological purposes in state public health departments has been documented only for a limited number of specific applications, leaving questions about its actual utilization and hindering IT's potential for information sharing. Communications, stages of change, and systems theories all influence the use of IT applications and were used as the theoretical framework for this study. An observational, cross-sectional design was used to survey 59 state public health departments to determine the current state of IT usage with barriers to implementation and timeliness of data availability. Survey responses were received from 17 departments. Chi square and Fisher's exact test were used to examine the relationship between IT usage and timeliness of data availability as well as barriers to implementation. Results showed IT applications are currently implemented for less than 50% of the epidemiological functions (Bioterrorism/Emergency Preparedness, Chronic Disease, Environmental Health, Infectious Disease, Injury, Maternal and Child Health, Occupational Health, and Oral Health) being performed across all departments and that there is a significant association between the use of IT applications and a reduction in the time to data availability for 6 of the 8 epidemiological functions. Chronic Disease and Environmental Health functions did not show a significant association. Budget and priorities in the department were found to be significant barriers to implementation of IT functions. The findings provide support for positive social change by identifying the benefit of IT usage to allow more timely data for epidemiological activities, which can improve health outcomes and quality of life for those served by the state public health departments. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com/en-US/products/dissertations/individuals.shtml.]

Background: Methicillin Resistant Staphylococcus aureus (MRSA) is a hardy and extremely virulent multidrug resistant organism that has been a major cause of hospital acquired infections ever since its discovery in the 1960's. It has severe consequences such as causing increased hospital length of stay, economic burden, morbidity, and mortality. MRSA prevention strategies have been advocated by national and international organizations which have been successful in reducing the burden of healthcare-associated MRSA. However, MRSA has been increasing in the community settings and this is an alarming and poorly understood rend because these infections occur in populations that have no known risk factors. In order to develop successful control strategies for this emerging threat. In order to develop successful control strategies for this emerging threat, it is important to understand the epidemiology, risk factors and links associated with community associated MRSA so that new and novel prevention strategies, using existing resources and cutting edge technology, can be developed. Methods: A cross sectional observational study design was used. The aims were accomplished by leveraging and utilizing the existing infrastructure of the OSUMC Information Warehouse, the Ohio State Health Network, and the OSUMC Microbiology Laboratory. Specific aim 1 was to develop an infection control collaborative and an innovative cross institutional platform, using existing information technology resources and infrastructure, for use as an electronic health information exchange between multiple hospitals spread across a large geographic area. Specific aim 2 was to estimate the proportion of community associated MRSA cases among all MRSA cases in rural community hospitals, and to identify the risk factors associated with community associated MRSA. Logistic regression was used to examine risk factors for community MRSA strain. Lastly, specific aim 3 was to identify patterns of intra-facility and inter-facility transmission of MRSA by evaluating epidemiological and geographic links via use of geographic information systems and social network analysis. Results: An infection control collaborative among the Ohio State University Medical Center and 7 rural outreach hospitals was formed with the successful development and deployment of a health information exchange for MRSA. Over a 1 year time period from March 2009-March 2010, 1024 MRSA isolates were collected, 625 from the OSUMC and 399 from the 7 outreach hospitals. The proportion of community associated MRSA in rural hospitals was 85%, while it was only 26% at the OSUMC (p-value < 0.001). A risk factor analysis found that age was a significant (p-value = <0.001) predictor of community associated MRSA (OR 0.92, 95% CI 0.90 - 0.94; p< 0.001). Rep-PCR testing found 75 distinct clusters. All clusters were examined for potential patterns of intra and/or inter-facility transmission. Two rep-PCR patters, 29 and 101, provide vivid illustrations of potential transmission links. Conclusions: Using surveillance to understand the transmission of MRSA is critical in infection prevention and control activities. Using health information technology is necessary to safely, effectively and efficiently treat patients and prevent outbreaks. By utilizing and enhancing existing resources, protocols, and information technology infrastructure health information exchanges can be developed. Novel methods including SNA and GIS should be further explored to provide an in-depth understanding of the transmission patterns of MRSA within and between facilities, regional geographical variations, and national trends. This increased understanding can then be used to plan and implement effective prevention programs that respond to local and regional trends. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com/en-US/products/dissertations/individuals.shtml.]

Timely identification and reporting of rare infectious diseases has important economic, social and health implications. In this study, we investigate how different stakeholders in the existing reporting system influence the timeliness in identification and reporting of rare infectious diseases. Building on the vision of the information supply chain (Marinos, 2005; Sun & Yen, 2005) and drawing upon co-ordination theory to investigate inter-organizational dependencies, this dissertation treats information processing and transfer as an information supply chain system whose key performance indicator is timeliness. Jajosky and Groseclose (2004) identified that information reporting lead time is related to the number of layers of reporting. In this dissertation, we look at three layers of reporting rare infectious diseases in the Ohio Disease Reporting System (ODRS), and identify factors that influence the delay in processing and reporting of these diseases. The three layers considered are those for which the county public health system is responsible for the preventive and control measures of any event of rare infectious diseases and is also responsible for entering the confirmatory information into the state reporting system i.e. ODRS. This dissertation investigates the rare infectious disease reporting system in a way different from traditional approaches. Our view of a reporting system is an information supply chain, just as any product supply chain, with different layers in reporting, in which exist interactions between the members (hospitals, laboratories and public health system). We no longer treat the rare infectious disease information supply chain system as point-to-point, but instead as layer-to-layer relationships and examine in detail the factors influencing the delay in these layers. We use simulation based modeling to represent in a more natural way the individual interactive entities in the information supply chain system and to investigate the lead times at various stages during the transfer of information between these entities. In this dissertation, we demonstrate that interactions between different stakeholders in the reporting system have plausible consequences on the information supply chain performance for managing rare infectious diseases. In particular, we looked at the lead times at various stakeholder sites including laboratories, hospitals and local health jurisdictions. Through the trace-driven simulation study, using data collected for last two years, we replicated the results in real life settings in a decentralized system of reporting and compared it with a centralized, actively monitored reporting system and confirmed that the proposed centralized reporting method ensures considerable reduction in lead time in an information supply chain system for managing rare infectious diseases. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com/en-US/products/dissertations/individuals.shtml.]

Deep penetration of personal computers, data communication networks, and the Internet has created a massive platform for data collection, dissemination, storage, and retrieval. Large amounts of textual data are now available at a very low cost. Valuable information, such as consumer preferences, new product developments, trends, and opportunities, can be found in this large collection of textual data. Growing worldwide competition, new technology development, and the Internet contribute to an increasingly turbulent business environment. Conducting surveillance on this growing collection of textual data could help a business avoid surprises, identify threats and opportunities, and gain competitive advantages. Current text mining approaches, nonetheless, provide limited support for conducting surveillance using textual data. In this dissertation, I develop novel text quantification approaches to identify useful information in textual data, effective anomaly detection approaches to monitor time series data aggregated based on the text quantification approaches, and empirical evaluation approaches that verify the effectiveness of text mining approaches using external numerical data sources. In Chapter 2, I present free-text chief complaint classification studies that aim to classify incoming emergency department free-text chief complaints into syndromic categories, a higher level of representation that facilitates syndromic surveillance. Chapter 3 presents a novel detection algorithm based on Markov switching with jumps models. This surveillance model aims at detecting different types of disease outbreaks based on the time series generated from the chief complaint classification system. In Chapters 4 and 5, I studied the surveillance issue under the context of business decision making. Chapter 4 presents a novel text-based risk recognition design framework that can be used to monitor the changing business environment. Chapter 5 presents an empirical evaluation study that looks at the interaction between news sentiment and numerical accounting earnings information. Chapter 6 concludes this dissertation by highlighting major research contributions and the relevance to MIS research. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com/en-US/products/dissertations/individuals.shtml.]

This dissertation has two areas of focus: learning and developmental disorders (LDDs) and polyfluoroalkyl chemicals (PFCs). Epidemiological and exposure assessment methods are applied to each. The first paper used geographic location as a surrogate for exposure and broadly assesses the effect of the environment, both physical and social, on LDD risk in a cohort of individuals born on Cape Cod, Massachusetts. Using generalized additive models and geographic information systems, maps of the risk of LDD in the study area were produced. Results suggested that children living in certain parts of upper Cape Cod may have been more likely to have an LDD. Observed variation, although not statistically significant, may be due to local differences in the social or physical environment. The second paper evaluated the associations between exposures to four PFCs and a specific LDD, attention deficit hyperactivity disorder (AMID) using data from the National Health and Nutrition Examination Survey (NHANES) 1999-2000 and 2003-2004 for children aged 12-15 years. Results were consistent with increased odds of ADHD in children with higher serum levels of perfluorooctane sulfonic acid (PFOS), perfluorooctanoic acid (PFOA, C8), perfluorononanoic acid (PFNA), and perfluorohexane sulfonic acid (PFHxS). As the sources of exposure to PFCs are not well characterized, the third paper included in this dissertation assessed the contribution of contaminated drinking water to the overall body burden of PFOA in Ohio and West Virginia communities contaminated by a fluoropolymer production facility. Standard regression approaches as well as pharmacokinetic models were used. In this population, PFOA contaminated drinking water was a significant contributor to serum levels. Regression methods and pharmacokinetic modeling produced similar estimates of the relationship. Together, these three papers make a significant contribution to the current literature on LDDs and PFCs. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com/en-US/products/dissertations/individuals.shtml.]

Dental caries is a disease process, one that will not be eliminated by tooth repair alone. Caries is the most prevalent disease of children and the primary reason for most restorative dental visits in both adults and children. A risk-based approach to managing caries targets those in greatest jeopardy for contracting the disease and provides evidenced-based decisions to treat current disease and prevent it in the future. This dissertation focuses on an approach to diagnosing and managing caries disease that holds promise of transforming the ways dentists treat this disease. This approach focuses on assessing the risk of caries and designing an individualized treatment plan that treats the disease in the least invasive way possible known as "Caries Management by Risk Assessment" or CAMBRA. Taken in total, the chapters presented in this dissertation address the related problems of disseminating information about CAMBRA and influencing both the practice of dentistry and the education of dentists. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com/en-US/products/dissertations/individuals.shtml.]

The study developed and implemented a questionnaire to examine misconceptions related to Acquired Immune Deficiency Syndrome (AIDS) among 57 college students enrolled in Psychology 101 courses at Essex Community College in Maryland. The questionnaire covered etiology and epidemiology of AIDS and the personal attitudes and experiences of the respondents. Two forms of the questionnaire were developed, each containing the same number of items, type of content, and level of difficulty. Subjects were exposed to a total of three administrations of the two questionnaires. Results indicated that more misconceptions were held about how one gets AIDS than about its impact on society. Subjects believed that infected patients should pay their own treatment costs. They also indicated that they would feel very uncomfortable around someone who had AIDS. Although there were no differences as a function of test form, time of administration had a significant effect on performance. An appendix contains copies of the questionnaires. Includes 41 references. (JDD)

Deer are susceptible to infection from various pathogens which are typically associated with livestock. Parasitic helminths, including gastro-intestinal nematodes, are of increasing economic concern in the livestock industry due to the growing issue of anthelmintic drug resistance. Deer can obtain nematodes with anthelmintic resistant alleles, and transmit these to new livestock hosts. Factors which impact a deer populations ability to transmit drug resistant nematodes, and contemporary methods for researching this phenomenon are identified, and the research aims of this thesis are determined (chapter 1). Further, it's found that deer which reside in landscapes associated with livestock farming, have significantly higher prevalence of the abomasal nematodes Teladorsagia circumcincta, Haemonchus contortus, Trichostrongylus axei and Ostertagia ostertagi (P = 0.01) as highlighted using a meta-analysis (chapter 2). All of these nematode species are prone to contain drug resistant alleles in livestock populations, and infections are often heavily driven by climate. Indeed, climate conditions are found to predict H. contortus infection in sheep and goats (R = 0.67, P = 0.02) as determined by a macro parasite reproductive quotient model (chapter 3). No such relationship was identified between climate and H. contortus or T. circumcincta infection in deer in Europe, indicating that other factors are more important in driving infection in wild hosts. As such, an Agent Based Model was developed for fallow (Dama dama) and roe deer (Capreolus capreolus), to examine the role of host ecology and landscape factors in determining their potential to transmit drug resistant nematodes to sheep pastures in fragmented and non-fragmented landscapes. Across all landscapes, roe deer deposited more eggs which developed to adult nematodes in sheep compared to fallow deer (P < 0.01). Lastly, livestock related abomasal nematodes including T. circumcincta were discovered in wild red deer (Cervus elaphus) on the Isle of Rona (chapter 5), which were in close proximity to pastures containing sheep. Further, the helminth fauna of wild fallow deer was identified in Northern Ireland, with common cervid nematodes being identified such as Spiculopteragia asymmetrica (chapter 5). This work indicates that livestock proximity, and access to livestock pasture, is a key factor influencing the helminth fauna of wild hosts and further, that deer helminth fauna could provide insights into their grazing behaviour which is also relevant to the epidemiology of other pathogens.

Cardiovascular diseases (CVD) are the leading causes of mortality around the world and disentangling cause and effect is central to better understanding and treating these diseases. While randomised clinical trials are the "gold standard" of assessing the effect of an intervention, some hypotheses cannot be feasibly tested in the randomised setting. In these cases, observational studies with appropriate methods of confounding adjustment can deliver reliable evidence concerning the association between an exposure and outcome. Indeed, trusted conventional statistical models guided by subject area experts for confounder selection have been used to estimate associations in many observational studies; however, in the observational studies for which confounding is unknown and/or the population suffers from complex illness, the conventional approaches render insufficiently adjusted estimates. In parallel, recently, there has been unprecedented access to nationally representative multimodal electronic health record (EHR) datasets and advances in statistical learning including "deep" machine learning, a form of machine learning that relies on automatic feature capture dissolving the need for expert-driven feature engineering. In this doctoral research, the aim was to develop a deep learning approach for causal inference on EHR. To do so in a structured way, the research was split into three investigations: 1) The development of a deep learning model for EHR data and assessment of risk prediction performance 2) Given the "black box" nature of deep learning modelling, the development of methods to explain the proposed model. 3) The derivation of a model for causal inference, and application of the models for association estimation in elderly/at-risk patient subgroups. The model, Bidirectional EHR Transformer (BEHRT) was created for EHR representation learning and risk prediction. The model outperformed several benchmarks for risk prediction on a variety of tasks including incident heart failure prediction. Furthermore, in the second work, explainability investigations yielded that the model captured validated factors of risk (e.g., hypertension, diabetes, and other diseases) and offered several more factors that could be potentially preventative of incident heart failure. Lastly, a derivation of BEHRT was developed for association estimation, Targeted-BEHRT, that fused advances in deep learning and semi- parametric statistics. The model demonstrated superior estimation abilities on several simulated data experiments, and was applied to better understand the effects of antihypertensives, blood pressure, and paracetamol on cardiovascular endpoints, mortality, and other outcomes in at-risk patients. Overall, the doctoral research has made advances in both methodological and clinical cardiovascular research. While the research focuses on developing methods for the study of cardiovascular diseases, the methods developed and tested have several important implications for epidemiological research in the observational setting at large. Especially in patient groups with pre-existing health issues, the causal models developed can be a more appropriate approach for association analysis than conventional statistical ones. In terms of clinical impact, the research has progressed our understanding of risk and protection in the context of CVD.

Background: Atrial fibrillation (AF) is the most common cardiac arrhythmia and is associated with a number of cardiovascular complications including stroke. While obesity has been shown to be associated with risk of AF, body-size and shape are heterogeneous and differ substantially by sex. Additionally, the independent causal relevance of related anthropometric measures remains uncertain. Pericardial adipose tissue (PAT) has been associated with risk of AF; however, the strength of the association between PAT and risk of AF is unclear. Adiposity has also been associated with inflammation; although, whether inflammation mediates associations between adiposity and risk of AF has not been established. Methods: Using data from UK Biobank (UKB) the independent associations between anthropometric measures and risk of AF were explored in men and women. A novel PAT area phenotype was developed in a sub-set of UKB participants using cardiac magnetic resonance imaging data and used to examine the association with risk of AF. Genetic scores for a number of anthropometric measures and inflammatory proteins were developed. The associations between genetically predicted anthropometric measures and both inflammation and risk of AF were explored. Mediation analyses used to determine the proportion of the effect of anthropometric measures on risk of AF mediated by a set of systemic inflammatory proteins. Results: Lean mass, fat mass, and waist circumference were all independently associated with risk of AF in both men and women. A novel PAT phenotype was developed in ~3,500 participants. Greater PAT area was shown to be associated with risk of AF and was comparable to other anthropometric measures. When adjusted for abdominal visceral adipose tissue, the association between PAT area and risk of AF remained similar. Genetically predicted anthropometric measures including body mass index, waist circumference, fat mass, and lean mass were found to be associated with risk of AF. A set of genetically predicted inflammatory proteins were shown to be associated with risk of AF but to only mediate a modest proportion of the effect of anthropometric measures on risk of AF. Conclusions: Anthropometric measures of both body size and body shape are independently causally associated with risk of AF in both men and women. PAT area was associated with risk of AF to a similar extent to other more easily obtained anthropometric measures. Genetically predicted systemic inflammatory proteins were associated with risk of AF but only mediated a modest proportion of the association between anthropometric measures and risk of AF.

Antimicrobial resistance (AMR) in food-producing animals is a key concern for global health and food security. To tackle this problem, we must first understand the risk factors for antimicrobial use and AMR development. This thesis presents new Bayesian models for examining risk factors for both antimicrobial use and resistance, and then applies these models to data from dairy farms in Southwest England, UK. Veterinary sales and milk recording data were used to investigate trends in antimicrobial sales to 124 dairy farms between 2010-18 from routinely collected data. A natural language processing algorithm was harnessed for semi-automated linking of sales data to antimicrobial product specifications. Significant reductions in sales of AMs were observed over the study period (41%). Predictive projective feature selection was employed to identify potential risk factors for antimicrobial sales and strong evidence was found for associations with five predictors which included antimicrobial purchase frequency and average parity. Secondly, a non-linear Bayesian model coupled to a microbiology technique for analysing proportional resistance was developed to enable the statistical modelling of environmental AMR load for the first time. This model was compared against existing methods and was shown to better incorporate uncertainty and reduce the biasing of risk factor estimates produced by the varying bacterial abundance in the samples. Using data from a previous study into antimicrobial resistance E. coli in the environment of dairy farms, the model was then used to investigate the effect of climate on antimicrobial resistance, finding evidence that both temperature and relative humidity, as well as the interaction between them, were associated with resistance to four of the antimicrobials tested. These findings improve our understanding of both antimicrobial use and resistance in UK dairy farming, and the flexible Bayesian modelling methods presented have the potential to underpin future research into AMR and its surveillance.

Hyperkalaemia is the term used to describe raised potassium level in the blood and, if severe, can constitute a medical emergency. Prevalence estimates have historically varied greatly depending on the population investigated with reports ranging from 1-10% in patients admitted to hospital. Management of severe hyperkalaemia (serum potassium >6.5 mmol/L) is typically undertaken in hospital with the use of intravenous insulin and dextrose as first line treatment, which requires regular monitoring, often requires repeat dosing and can expose patients to additional risk, morbidity and a prolonged length of hospital stay. Thus, there is an unmet need to develop a treatment regimen that both acts quickly and specifically enough to be of use in the acute setting but also which is safe and well-tolerated by patients. This thesis investigates the epidemiology and management of hyperkalaemia and explores clinical outcomes associated with hyperkalaemia with a view to enhancing the design of hyperkalaemia management trials by identifying clinically meaningful endpoints. This thesis was undertaken during the COVID-19 pandemic, which caused wide-spread disruption to non-COVID related research, and as a result also includes my work on the design and execution of the COVID-19 prophylaxis trial PROTECT-V. To improve understanding of the prevalence of hyperkalaemia we conducted the first systematic review (chapter two) of all reported observational studies describing the prevalence and/or incidence of hyperkalaemia. This work has provided a comprehensive published review on the prevalence and incidence of hyperkalaemia across a large number of relevant sub-groups and clarifies our understanding of risk-factors for developing hyperkalaemia. Pooled mean hyperkalaemia prevalence was 1.3% amongst the general population, 8.7% amongst adult inpatients and up to 20.7% in patients requiring renal replacement therapy. Another sub-group at risk of hyperkalaemia are users of medications that inhibit the renin-angiotensin-aldosterone system (RAASi) which are commonly prescribed as a cornerstone therapy to patients with chronic kidney disease or heart failure who already have an increased risk of hyperkalaemia. The prevalence of hyperkalaemia in RAASi users was 5.8% which rises to 12.2% amongst hospital inpatients taking RAASi. Hyperkalaemia can cause stoppages in RAASi therapy and the work described in chapter 3 which resulted in a publication, explores the adverse clinical impact of these stoppages in a large UK primary care dataset that was linked to secondary care records. The risk of hospitalisation, cardiac arrhythmia, heart failure hospitalisation and cardiac arrest were all higher amongst patients who suffered interruptions or cessation of their RAASi therapy. The fourth chapter reports the largest, published UK observational study to date of the emergency management of hyperkalaemia. Our work highlights that insulin dextrose treatment failed in over one third of patients and hypoglycaemia occurred in one in five patients. This work was used to provide clinically meaningful trial endpoints and guide the design of the HELP-K trial that is described in chapter 5. This trial aimed to provide contemporary evidence for the use of novel potassium-binding medications in the emergency setting to try and reduce the risk of treatment failure with insulin dextrose and to limit the associated morbidity of hypoglycaemia and prolonged length of hospital stay reported in chapter 4. Whilst there remains much work to be done to improve the evidence base for emergency hyperkalaemia management the work contained within this thesis helps improve the epidemiological understanding of hyperkalaemia in the real-world setting with contemporary knowledge that, I hope, will be useful to health-care providers, health policy makers and patients.

Physical activity in the form of moderate-to-vigorous physical activity (MVPA), sedentary behaviour in the form of recreational screen time (ST), and sleep, are individually associated with cardiovascular disease (CVD) risk factors in adolescents. Commonly, researchers investigate these three movement behaviours in relation to CVD risk factors in isolation. However, after the launch of the Canadian 24-Hour Movement Guidelines for children and youth, there have been calls to integrate all three 24-hour behaviours (PA, ST, & sleep). Furthermore, a scientific statement from the American Heart Association highlights the importance of cardiorespiratory fitness (CRF) as a measurement for health screening in children and adolescents as CRF is a strong predictor of CVD risk. Previous studies have individually explored how the 24-hour movement guidelines or CRF are associated with CVD risk factors in adolescents; however, which of the two (24-hour movement guidelines or CRF) is more strongly associated with CVD risk factors in adolescents is unknown. Currently, most studies have been conducted in North America and Europe, with the Middle East neglected, including Saudi Arabia, which has a very different lifestyle, environment, climate, school system, and culture compared to North America and Europe. Therefore, to address this gap in the literature in the context of Saudi Arabia, the studies included in this thesis aimed to investigate the prevalence of 24-hour movement guidelines and CRF and their relationship with CVD risk factors in adolescents from Saudi Arabia aged 12-15 years old. Chapter 4 narratively reviewed movement guidelines and CRF in relation to CVD risk factors in children and adolescents from the Gulf Cooperation Council Countries, which indicated gaps in the literature investigating the 24-hour movement guidelines. In addition, Chapter 4 highlighted that there is poor evidence for an association between single movement guidelines and CRF with multiple CVD risk factors; most studies have only examined associations with weight status. Chapter 5 examined the reliability and validity of using the Global School-Based Student Health Survey (GSHS) to assess the 24-hour movement guidelines in (n = 120; 50% females) adolescents aged 12-15 years old, from Saudi Arabia and found acceptable reliability and validity. In Chapter 6, the GSHS was used to examine the prevalence of the 24-hour movement guidelines and their associations with overweight/obesity in a sample of (n = 4224; 52% females) participants aged 12-15 years old. The results indicated that only 2.7% achieved the recommended 24-hour movement guidelines and highlighted weak, non-significant associations with the risk of overweight/obesity. In Chapter 7 a subsample of (n = 120; 50% females) participants aged 12-15 years old from the study in Chapter 6 underwent a 24-hour movement guidelines assessment using GENEActiv accelerometers (for MVPA and sleep) and the GSHS (for ST). In addition, CRF was assessed using the Progressive Aerobic Cardiovascular Endurance Run test, and fasting blood samples, blood pressure, and anthropometrics were collected. None of the adolescents in this subsample met all three of the 24-hour movement guidelines, and there were no favourable significant associations with individual or clustered CVD risk factors. In addition, few males (16%) and almost half of the female (46.3%) participants had a healthy CRF. However, in males only, having a CRF was significantly and negatively associated with waist circumference (WC) and clustered CVD risk. Collectively, the results presented in this PhD thesis contribute to enhancing the literature in four ways. Firstly, it demonstrates that there is a lack of studies on the 24-hour movement guidelines and CRF in relation to CVD risk factors for adolescents from the Gulf Cooperation Council countries. Secondly, it is the first study to examine a questionnaire for its reliability and validity to measure the 24-hour movement guidelines, which found an acceptable reliability and validity. Thirdly, it shows that via a questionnaire few adolescents (2.7%) from Saudi Arabia are meeting the 24-hour movement guidelines, which is not significantly associated with being at risk of overweight/obesity. Lastly, it shows that when using device-based accelerometry, none of the adolescents from Saudi Arabia are meeting the 24-hour movement guidelines, which is not significantly associated with individual or clustered CVD risk factors profiles in both sexes. Moreover, healthy CRF was not significantly associated with individual or clustered CVD risk factors in females, while in males, healthy CRF was only significantly and inversely associated with WC and clustered CVD risk factors. Based on the above findings, it is suggested that 24-hour movement guidelines are not favourable strongly associated with individual or clustered CVD risk factors in adolescents of both sexes from Saudi Arabia. In addition, in females it appears that there are no significant associations between CRF and individual or clustered CVD risk factors. Initiatives should be developed to encourage male adolescents from Saudi Arabia aged 12-15 years old to achieve a healthy CRF to reduce the risk of WC and clustered CVD risk factors. However, future longitudinal studies with larger sample sizes would enable further insight into the associations between 24-hour movement guidelines and CRF in relation to CVD risk factors in adolescents from Saudi Arabia and should provide specific recommendations for both sexes.

Background: Multimorbidity, a major global health concern nowadays, is increasing at an alarming rate worldwide, including in South Asian countries. Non-communicable diseases (NCDs) explain the majority of current global mortality and with an increasingly ageing population, NCD multimorbidity can overwhelm health systems worldwide. Although there is some evidence of several determinants and risks of multimorbidity from western populations, these cannot be generalised to apply to South Asian populations where the sociodemographics and lifestyles are quite different. Objectives: The main aims of this thesis are to 1) summarise the existing epidemiological evidence of multimorbidity in South Asia; 2) investigate the burden of multimorbidity and the cross-sectional association between multimorbidity and socioeconomic status; 3) identify the association between multimorbidity and adiposity measures; and 4) describe the association between multimorbidity and mental health. Methods: BELIEVE is a large-scale multicentre population-based prospective study from Bangladesh with a sample size of ~73,500, over three distinct sites covering urban, rural, and urban slum environments. BELIEVE has collected extensive baseline information on all participants' sociodemographic, economic, lifestyle and physical measures along with various health conditions. Therefore, it serves as a unique scientific resource to provide an indication of the burden and correlates of multimorbidity across a varied range of individuals in Bangladesh. Results: A systematic review demonstrated that there was a scarcity of published data on multimorbidity from South Asian countries. The retrieved literature indicated a wide range of prevalence estimates reflecting different study samples, methods, and definitions of multimorbidity. The most widely used was the World Health Organization's definition, namely "two or more diseases in an individual, definition was adopted for the BELIEVE study with the focus on any two diseases from a total list of 10. Using this definition, the overall prevalence of multimorbidity in BELIEVE was 16.8% with males having a higher prevalence (18%) than females (15%). In terms of site of residence, urban participants had the highest prevalence of 18% followed by 10% in the slum and 6% in the rural site. Cross-sectional analyses of the BELIEVE data revealed that multimorbidity was strongly associated with higher socioeconomic status based on the highest individual income (OR 2.05, 95% CI 1.71-2.47) and the highest wealth quintile (OR 2.38, 95% CI 2.19-2.60). The prevalence of overweight/obesity was very high (around 70%) in the multimorbid BELIEVE participants and there was a strong association between greater general and central measures of adiposity and increased likelihood of multimorbidity. Total body fat % (general adiposity) (OR 1.74, 95% CI 1.66-1.82) and waist circumference (central adiposity) (OR 1.70, 95% CI 1.65-1.75) showed the strongest association with multimorbidity. Possible depression and possible anxiety affected an estimated 16.8% and 11.7% of the BELIEVE population respectively, with higher prevalence among those with multimorbidity. When considered separately both possible depression and anxiety showed a moderate association with the risk of multimorbidity, but the odds ratio of the risk of having multimorbidity (OR 1.89, 95% CI 1.51-2.36) was greater for participants with co-occurring probable depression and anxiety. Conclusion: The present analyses represent the largest cross-sectional study on multimorbidity in a South Asian population and focus on a diverse range of individuals. Results confirmed some previous associations with the risk of multimorbidity found in western populations, but also revealed some novel insights into the association of socioeconomic characteristics, adiposity measures and mental health status with multimorbidity that may be specific to Bangladesh. Owing to the cross-sectional nature of the study, causality cannot be assumed. Nevertheless, the findings may indicate certain subgroups that would benefit from targeted public health intervention to reduce NCD impact. It is hoped that the findings of this study will stimulate further longitudinal research allowing causal links to be determined and to help reduce the rising burden of multimorbidity in Bangladesh and other South Asian countries.

The incidence of oesophageal adenocarcinoma (OAC) has rapidly increased, and its prognosis remains poor. Barrett's oesophagus (BO) is considered the precursor to OAC, however, BO is not apparent adjacent to tumour in nearly half of OAC patients. We have previously demonstrated that patients with BO-adjacent tumours (BO+ve OAC phenotype) have a favourable prognosis compared to those without evidence of BO adjacent to the tumour (BO-ve OAC phenotype). It has been suggested that the BO-ve OAC tumour phenotype may arise independently of BO. Recent experimental and computational studies show that all OACs likely arise from BO, even if this precursor lesion is not histologically apparent adjacent to the tumour. However, there is a lack of consolidated clinical, epidemiological and molecular data to examine this question. In this thesis, I used orthogonal approaches to examine the overlap between the BO+ve OAC and BO-ve OAC phenotypes and explain the aetiology and the observed altered prognosis. To achieve this, I assembled a large cohort (n=4,695) comprising BO+ve OAC cases (n=1,235), BO-ve OAC cases (n= 880), OAC cases with an unascertainable BO status (BO(?) OAC; n= 985), cancer-free BO cases (n=1,091) and reflux controls (n=554). A subset of the OAC cases (n=950) with available clinical, epidemiological and whole-genome sequencing data was also examined. There was little to no association between most of the 34 clinical and epidemiological factors and the OAC phenotypes. Weak associations were observed for cigarette smoking and gender with self-reported ever-smoking and female cases being more likely in the BO-ve OAC phenotype group relative to the BO+ve OAC phenotype group. However, tumour stage, lymph node spread and metastasis (TNM) was strongly associated with increased risk of BO-ve OAC. Higher TNM stage was strongly correlated with BO-ve OAC with an adjusted odds ratio of 2.4 (95% CI:1.8-3.3), 2.9 (95% CI:2.2-3.9) and 3.2 (95% CI: 1.7-5.9) for stages II, III and IV, compared to stage I. The improved survival associated with BO+ve OAC persisted in survival analyses adjusted for the tumour stage and location as well as the effects of smoking, obesity and heartburn (adjusted hazard ratio=0.88, 95% CI: 0.77-0.95, p=0.015). Of note, the BO-ve OAC phenotype was reported for 21 OAC cases with a history of undergoing endoscopic surveillance for BO. Seven different types of genomic alterations were examined in relation to the OAC phenotypes. BO+ve OAC tumours had a slightly higher tumour mutation load relative to BO-ve OAC tumours, which was not explained by the effects of ageing and cigarette smoking. There were no differences in the distribution of driver gene alterations, frequency of whole-genome doubling events, or rates of aneuploidy between the OAC phenotypes. Similarly, the prevalence of complex events such as breakage-fusion-bridges and extrachromosomal DNA did not differ according to OAC phenotype. Importantly, signature 17, which is shown to be preserved across the BO-OAC continuum, was equally enriched among the BO+ve OAC and BO-ve OAC tumours. This thesis presents the first comprehensive evaluation of epidemiological, clinical and molecular factors between the two defined OAC phenotypes. While certain epidemiological factors differ between the phenotypes, they do not explain the observed prognostic difference. The genomic landscapes of these tumour phenotypes were remarkably similar. Taken together, it is likely that all OACs arise from BO even if this precursor lesion is no longer apparent at the time of diagnosis or resection pathology. The advanced-stage tumours suggest that the precursor lesion is overgrown in BO-ve OACs. This work contributes evidence for screening strategies focused on identifying individuals with BO to reduce the public health burden of OAC.

Humeral shaft fractures comprise 1% of adult fractures, with isolated, closed injuries usually managed non-operatively in the first instance. However, there is limited data regarding the modern epidemiology of these injuries and understanding of longer-term patient-reported outcomes is incomplete. There is also uncertainty regarding a broader role for initial operative fixation, and studies comparing operative and non-operative management would inform discussions regarding the optimal treatment. Nonunion complicates 15-20% of non-operatively managed fractures, and methods for the early identification of patients at increased risk would help mitigate the morbidity associated with this complication. The aims of this thesis were to investigate the epidemiology, outcomes and predictors of nonunion following a humeral shaft fracture. A retrospective database of 898 patients with acute humeral shaft fractures sustained over a 10-year period was developed and analysed to determine the modern epidemiology and explore variation in patient and injury characteristics. All living, cognitively-intact patients with complete radiographic follow-up were sent a postal survey, to assess longer-term patient-reported outcomes and factors associated with an adverse outcome. Comparison of operative versus non-operative management was undertaken through a systematic review and meta-analysis of published prospective randomised controlled trials (RCTs), and then a single-centre prospective RCT (the HU-FIX Study). The primary outcome measure for these studies was the abbreviated (QuickDASH) or full Disabilities of the Arm, Shoulder and Hand (DASH) score. Secondary outcome measures included union/nonunion and health-related quality of life (HRQoL). Using EuroQol-5 Dimension (EQ-5D) data obtained during the longer-term outcomes study, a retrospective cost-utility analysis was performed to estimate the cost-effectiveness of routine operative fixation. Techniques for the early prediction of nonunion were investigated through a large retrospective study of baseline risk factors, a retrospective analysis of radiographic callus formation at six-weeks post-injury, and a prospective pilot study of freehand ultrasound assessment of humeral shaft fracture healing. The incidence of humeral shaft fractures was 12.6/100,000/year (typical fractures 11.2/100,000/year, pathological fractures 1.1/100,000/year, periprosthetic fractures 0.4/100,000/year). For typical fractures (n=798), peak incidence was bimodal in men and unimodal in older women (Type G distribution). Fractures involving the proximal- (31%) and middle-thirds (48%) were more likely to occur in older, female patients with comorbidities after a fall from standing (all p<0.001), suggesting these should be considered as fragility fractures. Retrospective analysis of 291 patients at a mean of five years (minimum one year) found that overall patient-reported outcomes were comparable to healthy, age-matched populations (mean QuickDASH 20.8, mean EQ-5D 0.730). However, union after nonunion surgery was independently associated with inferior upper limb function (QuickDASH difference 8.1, p=0.019) and HRQoL (EQ-5D difference -0.102, p=0.028) compared with union after initial operative/non-operative management. Meta-analysis of four previously published prospective RCTs (292 patients) found that surgery was associated with superior upper limb function at six months (DASH difference 7.6, p=0.01; Constant-Murley difference 8.0, p=0.003), with no difference at one year (DASH, p=0.30; Constant-Murley, p=0.33). No differences in HRQoL or pain scores were found at any timepoint. However, surgery was associated with a lower risk of nonunion (OR 0.13, p=0.004) and the number-needed-to-treat (NNT) to avoid one nonunion was seven. The HU-FIX Study included 64 patients randomly allocated to either operative or non-operative management. At the primary outcome timepoint (three months), the median DASH score for the operative group was superior (20.8 vs. 39.2, p=0.013), but this was not sustained at six months or one year. Overall, eight patients (12.7%) developed a nonunion, one (4%) managed operatively and seven (20%) managed non-operatively (OR 6.75, p=0.066). There were early advantages of operative management (in terms of shoulder/elbow motion, pain/satisfaction scores, HRQoL and return to sport), but no differences between the groups at one year. Retrospective cost-utility analysis of 215 patients at a mean of five years demonstrated that routine operative fixation, in order to reduce the rate of nonunion following non-operative management, was associated with additional treatment costs (£1,542/patient, compared with routine non-operative management) but conferred a potential EQ-5D benefit of 0.120/patient. The incremental cost-effectiveness ratio of routine fixation was £12,850/QALY gained, suggesting this strategy would be cost-effective according to the NICE threshold (<£20,000/QALY gained). Retrospective analysis of 662 patients found that pre-injury non-steroidal anti-inflammatory drugs (aOR 20.58, p=0.009) and glenohumeral arthritis (aOR 2.44, p=0.043) were associated with an increased risk of nonunion following non-operative management. However, non-operative management itself was associated with an increased risk of nonunion compared with initial surgery and was the strongest risk factor for this complication (aOR 9.91, p<0.001). Based upon these findings, five patients would need to undergo initial surgery to avoid one nonunion. Retrospective analysis of 60 patients led to the development of the Radiographic Union Score for HUmeral fractures (RUSHU), a novel assessment tool that demonstrated substantial inter-observer agreement (intraclass correlation coefficient [ICC] 0.79). At six weeks post-injury, a RUSHU<8 was highly predictive of nonunion (sensitivity 75%, specificity 80%, positive predictive value (PPV) 65%, negative predictive value (NPV) 86%). Based on a PPV of 65%, if all patients with a RUSHU<8 underwent fixation the NNT to avoid one nonunion would be 1.5. Moreover, selective fixation based on a RUSHU<8 would be associated with estimated cost savings of £415/patient (compared with routine non-operative management), as well as conferring an overall EQ-5D utility benefit. Prospective ultrasound assessment of 12 patients suggested this modality had substantial inter-observer reliability (ICC 0.76). Absence of sonographic callus (SC) at six weeks demonstrated sensitivity 50%, specificity 100%, PPV 100% and NPV 91% in nonunion prediction (overall accuracy 92%). Absence of sonographic bridging callus (SBC) at six weeks demonstrated sensitivity 100%, specificity 70%, PPV 40% and NPV 100% (overall accuracy 75%). Of three patients at risk of nonunion (RUSHU<8), one had SBC on six-week ultrasound (that subsequently united) and the others had non-bridging or absent SC (both developed nonunion). The association between proximal- and middle-third humeral shaft fractures and fragility highlights the importance of balancing the risks and benefits of surgery in this patient group. Nonunion impairs longer-term patient-reported function and HRQoL, even after successful nonunion surgery. Initial surgery may confer early functional benefits, but these are not sustained beyond six months. The most significant advantage of surgery may lie in the substantially lower risk of nonunion. Routine fixation for patients with humeral shaft fractures is cost-effective at five years, as the additional costs are offset by reducing the impact of nonunion on HRQoL. Strategies to target early operative intervention to patients at increased risk of nonunion represent a pragmatic option. Identifying patients using baseline characteristics alone appears challenging, but both radiographic (RUSHU) and ultrasound assessment at six weeks appear to be reliable and accurate in predicting subsequent nonunion risk.

Systemic Sclerosis (SSc) is a rare autoimmune disease. In this thesis we investigate the risk of cancer (the outcome of interest) in patients diagnosed with SSc (the exposure of interest) and compare this against the risk of cancer in people without SSc. A large UK primary care dataset provides the data for this study. This dataset contains 806 patients who were diagnosed with SSc over the chosen study period 1998 to 2018, forming the incident cohort group, termed 'SSc'. An additional 780 patients diagnosed with SSc prior to their entry to the UK database (which may be later than 1998) and cancer free at this time, form the prevalent cohort, potentially a valuable additional resource for analysis as its addition greatly increases sample size and length of follow-up. A pre-requisite for inclusion of this prevalent cohort is that it is consistent with the incident cohort and does not distort the study population. The thesis examines how to account for prevalent patients in a competing risk framework, including in the presence of informative censoring, and the issues that could be encountered by left truncated data with a long follow-up. Each patient diagnosed with SSc, in both incident and prevalent cohorts, is matched to 6 other patients who will be used as comparators. These matches form a group termed 'non-SSc' and this group enables the risk of cancer in patients without an SSc diagnosis to be assessed. For both SSc and non-SSc patients there is a competing risk, death, and the thesis considers methods of analysis in a competing risk setting. As common nonparametric methods of analysis are often limited by confounding, we investigate the parametric g-formula to permit causal interpretation, as opposed to an association, between SSc and cancer. The expansion of the parametric g-formula to include prevalent patients, including the derivation of a weighted form, is formulated. Methods to adjust for differences observed due to date of diagnosis of SSc are developed, with particular relevance to the potential adjustment of patients in the prevalent cohort to provide a better estimation of current risk. The thesis concludes that prevalent inclusion is often beneficial, following pre-analysis of differences between the incident and prevalent cohorts and use of the Cox model to test for temporal trends. Including prevalent patients in this analysis without temporal trends suggests that there is an increase in risk of cancer in those with SSc when compared to those without (approximate 1.25 times the risk), however once temporal trends are accounted for there is no longer a statistically significant difference in the risk of cancer between SSc patients and non-SSc patients. Due to the increased mortality of SSc patients, the cause-specific risk ratio between SSc and non-SSc patients is not significant at the 5% level and decreases over calendar time. The parametric g-formula method allowing for prevalent patients was used for this study but no change in outcome between this method and the prior nonparametric methods was observed, possibly due to the matching having already accounted for the majority of confounding. We recommend the parametric g-formula when there are differentiating covariate distributions between the exposed and unexposed. The results of this study will support those who are interested in the epidemiology of SSc, those who are considering inclusion of prevalent patients and how best to do this, and lastly those who may be interested in using the g-formula with left truncated data.

Introduction. There is a lack of research on endometriosis from the Eastern Mediterranean region and a complete absence of data from Northern Cyprus. It is unknown whether regional-specific factors effect endometriosis presentation and if risk factors differ. Methods. The Cyprus Women's Health (COHERE) Initiative is a cross-sectional study of 7,646 women living in Northern Cyprus. Baseline recruitment involved completing an expanded version of the World Endometriosis Research Foundation Endometriosis and Phenome Biobanking Harmonisation Project (WERF-EPHect) questionnaire, including validated tools. Clinical data was obtained from ultrasound scans attended by 668 women and information on diet was obtained from 513 women who completed a validated Food-Frequency-Questionnaire (FFQ). Associations with endometriosis were compared to a pain symptomatic and a pain asymptomatic control group. Results. Prevalence of endometriosis was 5.4% (95%Confidence Interval:4.9%-5.9%; n=410), including 15 incident cases identified during the clinical visit. Average age of endometriosis symptom onset was 25.9-years despite average age of first pain occurring at 16.2-years, giving a diagnostic delay of 9.3-years. Compared to asymptomatic controls, endometriosis cases were more likely to report anxiety and migraines (Odds Ratio=1.56; 95%CI:1.03-2.38 and OR=1.39; 95%CI:1.06-1.83, respectively). Physical health-related-quality-of-life was impaired in endometriosis cases compared to symptomatic controls (p=0.034). There was no significant difference in economic burden between endometriosis cases and symptomatic controls (Int $9,864; 95%CI: $8,812-$10,917 vs Int $10,917; 95%CI: $10,004-$10,855, respectively). Average cost for endometriosis cases peaked between 26-35-years-of-age and decreased after age 36 but remained high across the life-course for symptomatic controls. Participants with endometriosis were more likely to report reproductive, allergic, metabolic/obesity-related and cardiovascular diseases. Women who were iron deficient and vitamin-D deficient were more likely to report to have endometriosis compared to asymptomatic controls (OR:1.70; 95%CI:1.37-2.12 and OR:2.04; 95%CI:1.37-3.02, respectively). Discussion. For the first time, the epidemiology of endometriosis in Northern Cyprus has been investigated and presented, suggesting that burden of endometriosis is high in this population and the awareness is low. There are multiple opportunities to improve care for endometriosis patients and the findings on iron and vitamin-D deficiency may be unique risk factors for the region, warranting further investigation.

Immunity is both a lens to understand the ecology of adversarial host-pathogen interactions, but also a lever for clinical intervention in combating infectious disease. This thesis uses mathematical modelling to interrogate both the function and effectiveness of natural immune systems, and how human interventions like vaccination can be best deployed. A defence of the value of this scientific approach in overcoming the empirical and interpretative challenges for these topics is provided in chapter 1. The first two investigations consider the evolution and functional performance of natural immune systems. Chapter 2 proposes that immune adaptations can plausibly arise from Fisherian selection, and therefore could be maladaptive, and constructs a simple mathematical model of hosts and pathogens to examine this potential mechanism. Chapter 3 assesses the utility of a particular immune adaptation: fever. It models the impact of different proposed thermal strategies of fever in terms of suppressing pathogen temperature-dependent growth, and compares these to the calorimetric costs of heating. The second two investigations consider how artificial immunity is best deployed, focusing upon vaccination strategies in the COVID-19 pandemic. Chapter 4 considers the risk and benefit of very early emergency use of a vaccine, before its safety and efficacy is known, finding this balance can favour such use for many individuals in the early stages of the COVID-19 pandemic. Chapter 5 investigates another emergency strategy for multiple dose vaccines, prioritising individuals for first doses and the expense of postponing individuals receiving subsequent doses, and when this strategy would be beneficial for public health in terms of risk reduction, disease transmission, and earlier relaxation of non-pharmaceutical interventions. I conclude with a discussion of broader themes which span across these investigations, and suggestions for further research.

Colorectal, breast, and prostate cancer account for nearly a third of all diagnosed cancers, and some dietary factors may influence risk. Moreover, circulating insulin-like growth factor-I (IGF-I) concentrations have been positively associated with risks of colorectal, breast, and prostate cancer, and IGF-I might mediate some diet-cancer associations. This thesis investigates the associations of dietary factors, IGF-I concentrations, and IGF-I related cancers among participants in the UK Biobank cohort. Associations between intakes of food groups (438,453 participants) and macronutrients (11,815 participants) with IGF-I concentrations were explored. Intake of fish was positively associated with IGF-I concentrations, as was protein from dairy products; however, associations differed by dairy protein source in that milk and yogurt were positively associated with IGF-I concentrations whereas protein from cheese was not. Intakes of fibre and wholegrains were also positively associated with IGF-I concentrations. The association between protein from all dairy products, milk, and cheese and colorectal, breast, and prostate cancer was explored among ~114,000 participants. Protein from all dairy products and milk were inversely associated with colorectal cancer risk, whereas a borderline positive association was observed for prostate cancer risk for men in the highest quartile of intake compared to those in the lowest quartile. IGF-I concentrations did not appear to explain these associations. Prospective associations between carbohydrate types and sources and colorectal cancer were explored among ~114,000 participants. Non-free sugars and fibre from wholegrains were inversely associated with colorectal cancer risk. Evidence of heterogeneity by genetically predicted butyrate synthesis was observed; participants with high predicted butyrate and consuming higher wholegrains had a lower risk of colorectal cancer whereas no association was observed for those with low genetically predicted butyrate synthesis. Finally, diet groups in relation to all cancer and common cancer types were investigated in prospective analyses of ~472,000 participants. In comparison to regular meat-eaters, low meat-eaters had lower risk of all cancer and colorectal cancer, fish-eaters had a lower risk of all cancer and prostate cancer, and vegetarians had a lower risk of all cancer, postmenopausal breast cancer, which may be due to their lower body mass index, and prostate cancer. Differences in IGF-I concentrations or free testosterone did not appear to mediate these associations. In conclusion, research from this thesis supports dietary factors, such as dairy protein, being associated with IGF-I concentrations. However, whether these same dietary factors are positively associated with IGF-I related cancers remains unclear, suggesting other factors or mechanisms external to the IGF-I pathway. Potential novel findings suggest that wholegrain intake and colorectal cancer risk may vary by genetic factors for butyrate production and further research is needed to replicate and explore this further. Different diet groups may also be related to IGF-I related cancers; however, IGF-I concentrations did not mediate this association. Further research exploring mechanisms through and external to the IGF-I pathway is needed.

Background: Blood pressure lowering is an established strategy for preventing microvascular and macrovascular complications of diabetes, but its role in the prevention of diabetes itself is unclear. On the other hand, controversy exists as to whether the threshold of blood pressure for initiation of antihypertensive therapy should differ between people with and without type 2 diabetes. I aimed to integrate individual participant data from major randomised controlled trials and genetic data to fill these knowledge gaps. Objectives: This thesis sought to examine three main objectives: to investigate the effect of pharmacological blood pressure-lowering on the risk of new-onset type 2 diabetes; to investigate the separate effects of blood pressure-lowering drug classes on the risk of new-onset type 2 diabetes; to investigate the effect of pharmacological blood pressure-lowering treatment for the prevention of major cardiovascular disease in persons with and without type 2 diabetes. Methods: For the first and second objectives, I conducted a one-stage individual participant-level data meta-analysis of major randomised controlled trials using data from the Blood Pressure Lowering Treatment Trialists' Collaboration (BPLTTC). Analyses were complemented with Mendelian randomisation studies using naturally randomised genetic variants associated with systolic blood pressure and genetic variants in the gene that encodes the therapeutic targets of each drug class. For the third objective, I used one-stage individual participant-level data meta-analysis using the BPLTTC dataset. I expressed the treatment effect per 5 mmHg reduction in systolic blood pressure on the risk of developing a major cardiovascular event as the primary outcome, defined as the first occurrence of fatal or non-fatal stroke or cerebrovascular disease, fatal or non-fatal ischaemic heart disease, or heart failure causing death or requiring hospitalisation. Cox proportional hazard models, stratified by trial, were used to estimate hazard ratios (HRs) separately by type 2 diabetes status at baseline, with further stratification by baseline categories of systolic blood pressure (in 10 mmHg increments from < 120 mmHg to ≥170 mmHg). Results: For the first and second objectives, blood pressure-lowering treatment was found to reduce the risk of diabetes by 11% (hazard ratio per 5 mmHg lower systolic blood pressure 0.89 [95% confidence interval [CI] 0.84 to 0.95]). Similarly, in the Mendelian randomisation study, each 5 mmHg genetically influenced lower systolic blood pressure was associated with an 11% lower risk of diabetes (odds ratio 0.89 [95% CI 0.86 to 0.93]). Evidence from genetic data and trials was also consistent in that angiotensin-converting enzyme inhibitors and angiotensin receptor blockers reduced the risk of diabetes, and beta-blockers increased this risk. There was no effect for calcium channel blockers and findings for thiazide diuretics were inconsistent. For the third objective, over 4.2 years median follow-up (IQR 3.0 to 5.0), a 5 mmHg reduction in systolic blood pressure decreased the risk of major cardiovascular events in both groups, but with a weaker relative treatment effect in participants with type 2 diabetes (HR 0.94 [95% CI 0.91 to 0.98]) compared with those without type 2 diabetes (0.89 [0.87 to 0.92]; p for interaction=0.001). However, absolute risk reductions did not differ substantially between people with and without type 2 diabetes (absolute risk reduction -1.54 [95% CI -2.04 to -1.04] in people with diabetes and -1.61 [-1.86 to -1.36] in people without diabetes, p for interaction =1). We found no reliable evidence for heterogeneity of treatment effects by baseline systolic blood pressure in either group. Conclusions: Blood pressure lowering is an effective strategy for the prevention of new-onset type 2 diabetes. Established pharmacological interventions, however, have qualitatively and quantitively different effects on diabetes, likely due to their differing off-target effects, with angiotensin-converting enzyme inhibitors and angiotensin II receptor blockers having the most favourable outcomes. Additionally, although the relative beneficial effects of blood pressure reduction on major cardiovascular events were weaker in participants with type 2 diabetes than in those without, absolute effects were similar. The difference in relative risk reduction was not related to the baseline blood pressure or allocation to different drug classes. Therefore, the adoption of differential blood pressure thresholds, intensities of blood pressure lowering, or drug classes used in people with and without type 2 diabetes is not warranted.

Non-equilibrium systems, or active systems, span virtually all disciplines of science. While the underlying processes driving these systems out of equilibrium (may that be thermal, chemical or another type of equilibrium) may be entirely unrelated, it is often the case that the same or a similar mathematical model is capable of describing either of the systems. The particular active systems we are concerned with here are (a) the microscopic interactions of active matter in colloidal physics, and (b) the epidemiological dynamics of infectious diseases in human populations. Aside from their manifestly out-of-equilibrium character, both of these systems comprise fundamentally stochastic processes. Thermal fluctuations at energy-scales comparable to the ones of colloidal dynamics lead to Brownian motion in the former. The latter is governed by both pathogen- and data-specific noise. Remarkably, applying a functional central limit theorem, we find that to linear order both of these active systems can be described by the same process - the Ornstein-Uhlenbeck process. This is despite the fact that colloidal physics is concerned with microscopic scales, while epidemiological models operate on the scale of human populations. Using this, we apply methods from Bayesian inference to quantify the respective underlying uncertainties from stochastic trajectories alone. In the following, we briefly introduce the two parts of this thesis on active systems: Active matter and Epidemiology. Abstract Active matter consists of mesoscopic self-propelled units, active particles, which on an individual level can metabolise energy into systematic movement. The mechanics and statistical mechanics of a suspension of active particles are determined by the traction (force per unit area) on their surfaces, arising from long-ranged hydrodynamic interactions at low Reynold's number. Here, we present a solution of the direct boundary integral equation for the traction on spherical active particles interacting with each other and their surroundings. For a single particle away from any boundaries this solution is exact, as both single- and double-layer integral operators can be simultaneously diagonalised in a basis of irreducible tensor spherical harmonics. The solution, thus, can be presented as an infinite number of linear relations between the harmonic coefficients of the traction and the velocity at the boundary of the particle. These generalise Stokes laws for the force and torque. In confinement, the dynamics of an active particle are uniquely defined by its mobility matrices, fully characterising hydrodynamic interactions due to the geometry of the system alone, and so-called propulsion tensors, arising from activity. Using Jacobi's method, we analytically derive iterative solutions for these quantities, once again starting from the direct boundary integral formulation of Stokes flow. We exemplify our results by explicitly providing the dynamics of an axisymmetric squirmer in the vicinity of a plane interface. For the computationally efficient simulation of many-body systems we develop a novel linear solver, based on a Krylov subspace method, that retains the true many-body character of hydrodynamic interactions. Finally, at non-zero temperature we consider thermal fluctuations of the suspending fluid, introducing stochasticity to the model above. We review and further develop a fast Bayesian method for parameter inference and model selection for a class of active systems that can be approximated by a stationary Ornstein-Uhlenbeck process. Such systems are said to be in a non-equilibrium steady state. A key aim of this study is to understand how one may distinguish between reversible and irreversible dynamics from information contained in stochastic trajectories alone. Abstract Epidemiological data is beset by uncertainties about the underlying epidemiological processes, and the surveillance process through which the data is acquired. We present a Bayesian inference methodology that quantifies these uncertainties, for epidemics that are modelled by non-stationary, continuous-time Markov population processes. The efficiency of the method derives from a functional central limit theorem approximation of the likelihood, akin to van Kampen's system size expansion in physics, and valid for large well-mixed populations. The Ornstein-Uhlenbeck process arising from this approximation can once again be studied using Bayesian inference, including maximum a posteriori parameter estimates, computation of the model evidence, and the determination of parameter sensitivities via the Fisher information matrix. Our methodology is implemented in PyRoss, an open-source platform for analysis of epidemiological compartment models.

Adults with hearing loss are at a greater risk of additional chronic health conditions and poorer psychosocial wellbeing than those without. Many of these conditions (e.g., dementia, diabetes, depression) are potentially modifiable via interventions that target health behaviours such as physical activity. Adults with hearing loss are an under researched population who face additional barriers to engaging in protective health behaviours like physical activity. However, there is a paucity of research assessing how behavioural interventions could mitigate the onset of these conditions in individuals with hearing loss across the adult lifespan. This thesis is comprised of two aims. First, the thesis aims to assess the associations between hearing and different health domains in a cohort of adults from 18-87 years, including an exploration of the potential mediating pathways between hearing and memory problems (Chapter 2), an investigation of the impact of the mode of cognitive assessment delivery on cognitive performance (Chapter 3), and an assessment of the associations between hearing and health domains using behavioural measures (Chapter 4). Second, the thesis aims to further explore the relationship between hearing loss and physical activity in middle- and older- adults. First a longitudinal analysis from the English Longitudinal Study of Ageing was conducted to explore the possibility of reverse causality (Chapter 5), followed by a systematic review of existing behavioural interventions to increase physical activity in this population (Chapter 6), and a qualitative evaluation of the barriers and facilitators to physical activity using semi-structured interviews (Chapter 7). This thesis highlights that a holistic approach to hearing healthcare is necessary to improve health outcomes for adults with hearing loss across the lifespan. Novel findings suggest that behavioural interventions to improve physical activity should be tailored to the needs of adults with hearing loss, as they experience several hearing loss specific barriers to physical activity. Addressing the lack of physical activity in middle- and older- adults with hearing loss may be critical to support healthy ageing and longevity.

Replacing motorised transport with active travel, such as walking or cycling for transport, has multiple environmental and health co-benefits. However, with the large variety of intervention forms that may be implemented to shift population levels of active travel, such as modifying the built environment or implementing policies that restrict the use of motorised vehicles, it is difficult to know which should be prioritised. It is thus important to understand which types of interventions are most effective at encouraging active travel uptake. This includes identifying whether there are mechanisms or underlying processes common to successful active travel interventions (i.e., intervention functions), or whether differences in intervention design or the specific components included (i.e., intervention form) can influence intervention effectiveness. Thus, this thesis aims to identify, evaluate, and understand the key characteristics of effective interventions to encourage shifts towards active travel. The first part of my research involves a systematic review comparing the effectiveness of carrot, stick, and a combination of the two, or carrot-and-stick transport interventions. Interventions with a stick strategy, such as those with functions involving taking away road space or leading to decreases in convenience, are less widely assessed in the literature compared to those with a carrot strategy, such as those aiming to increase access and safety for active travel. The findings also suggest that stick and carrot-and-stick interventions are more effective at either reducing driving or increasing active travel than carrot-only interventions. I also find that interventions seeking to influence driving behaviour using financial means or active travel by modifying access, safety, and space are more impactful than others. Having identified that there is a need to increase the evidence base on interventions that include a stick strategy, I evaluate three interventions with stick or carrot-and-stick strategies. The first is a study evaluating the London Ultra Low Emission Zone, an area-based vehicle charging scheme. I use data from the Children's Health in London and Luton study, a parallel cohort study of primary school children. Children living further from school and exposed to the intervention had a higher likelihood to switch from inactive to active modes of travelling to school than those who were unexposed. The second evaluation study examines the introduction of low traffic neighbourhoods (LTNs) in Southwark, London. LTNs use physical barriers to restrict traffic from passing through residential streets while allowing pedestrians and cyclists to access the street. Moreover, other minor streetscape improvements to pedestrian infrastructure were made, and altogether, the interventions constitute carrot-and-stick strategies. Findings reveal that while traffic volumes and speeds decreased in some of the neighbourhoods, there were no corresponding increases in walking and cycling. The third evaluation study involves investigating an intervention with carrot and carrot-and-stick strategies. I use interrupted times series analyses of routinely collected cycle count data to assess how cycling levels changed following cycle lane expansions in Paris and Lyon in France. Although there were no discernible level or trend changes in cycling counts, cycling increased more in Paris compared to Lyon. Given that various cycling infrastructure designs were introduced in both cities, the final study explores how to advance our understanding of whether differences in cycling infrastructure design can influence intervention effectiveness. I develop a virtual street audit using Google Street View to characterise infrastructural changes made due to the French cycling lane interventions, from which I derive function scores. I find that cycling infrastructure forms that involves stick strategies, such as taking traffic lanes and parking space from drivers, have the highest associations with increases in cycling levels. Correspondingly, functions aiming to increase safety or space were found to be among the most impactful. This thesis adds to the literature by not only identifying common characteristics of effective active travel interventions, such as whether it includes a stick strategy, but also finding that these are relatively under-represented in the literature. My findings also suggest there may be a hierarchy of active travel needs that may be important to address when developing interventions, such as increasing safety and space for pedestrians and cyclists. To further increase intervention effectiveness, the intervention form, or the way interventions are designed and implemented, should align with their core functions and adapted based on individual contexts. Further research is needed taking consideration of intervention form, function, and contextual factors into account to strengthen and inform public health policy and practice.

Background: Multimorbidity is the presence of two or more chronic diseases in one person, and in this thesis, cardiovascular multimorbidity (CVM) refers to the co-occurrence of two or more cardiovascular disease subtypes. Multimorbidity is a major public health concern in the UK and CVM is a key component. However, there is currently limited large-scale, prospective evidence on risk factors for CVM. Aim: This thesis aims to provide a broad overview of CVM, its component conditions and potential risk factors in UK women. Methods: This was a population-based prospective cohort study of 1.2 million women aged 50-64 years at recruitment in 1996-2001. Using linked hospital admission and death records up to 31 December 2016, each participant was followed-up for nineteen subtypes of incident cardiovascular disease. CVM was defined either by counting the number of cardiovascular diseases in one person or as specific pairs of diseases. Established cardiovascular risk factors, socioeconomic, and reproductive factors were investigated. Main results: More than 40% of UK women with any incident cardiovascular disease developed CVM (≥2 cardiovascular diseases). The most common components of CVM were ischaemic heart disease (in 55% of CVM cases), atrial fibrillation (46%), heart failure (30%), and other arrhythmias (28%). Smoking, obesity, diabetes, hypertension, and hyperlipidaemia was each associated with 1.7-2.5 times the risk of CVM. Taller height was associated with higher CVM risk, which appeared to be driven by the association between height and atrial fibrillation. There was evidence of somewhat higher CVM risks associated with lower educational attainment and higher area deprivation. However, 80% or more of each of the observed associations was accounted for by four health behaviours measured at recruitment: smoking, alcohol consumption, physical activity, and BMI. Similarly, over 75% of each of the associations between reproductive factors (age at menarche, parity, and breastfeeding) and CVM was accounted for by those four factors. Conclusion: Roughly 1 in every 2.5 UK women with incident cardiovascular disease had CVM. Established cardiovascular risk factors were also independent risk factors for CVM. Much of the difference in CVM risk by levels of socioeconomic and reproductive factors could be attributed to known modifiable cardiovascular risk factors.

Background: Adenoviruses are an endemic cause of human infection with 104 types across 7 species currently identified. Clinical syndromes vary but infection in transplantation can be severe. Deficiencies in current classification methods are apparent as the number of novel types continue to grow. Whole genome sequencing is employed in investigation of pathogenicity, drug resistance, transmission dynamics and nosocomial outbreak investigation but is rarely used in human Adenoviruses. Bioinformatic processing of sequencing data is particularly challenging as no commercial solutions or best practice guidelines currently exist. This project aimed to develop a method for direct enrichment of Adenovirus DNA from clinical samples for whole genome sequencing alongside generation of a validated open access pipeline for data analysis to update current knowledge of Adenovirus epidemiology. Method: A custom Agilent SureSelectXT RNA oligonucleotide Target Enrichment system was designed for adenovirus enrichment direct from faecal, blood and respiratory tract secretions for subsequent illumina sequencing. Minor modifications were made to the SureSelectXT low input protocol to improve suitability for non-human applications. A bespoke three stage Unix coded bioinformatics pipeline was constructed and validated through direct comparison of open-source tool performance. Illumina sequencing data from multiple Adenovirus types was used to determine suitability of quality trimming parameters, genome assembly and variant calling tool for inclusion. Genomic similarity of isolates was assessed at a whole genome and classical typing gene level phylogenetically and through pairwise distance models to assess current adenovirus epidemiology. Results: Adenovirus whole genomes could be enriched directly from various clinical sample matrices with a success rate of 94%. Sequence data quality trimming using Trimmomatic did not impact whole genome sequence generation regardless of parameters used. No significant difference was observed in genome assembly using BWA-MEM and Bowtie2 however variant calling tool performance was more inconsistent with LoFreq reporting the most variants. Starting concentration of Adenovirus DNA heavily influenced enrichment success where Adenovirus species or type did not. Species C Adenovirus most commonly found at 49% closely followed by B and A. Genetic differences between Adenoviruses of the same type were not always apparent from classical typing gene assessment but required whole genome analysis. Higher sequence divergence between Adenovirus types than previously observed was found during assessment of the hexon, penton and fiber genes Conclusions: This modified custom SureSelectXT protocol is a robust method for generating Adenovirus whole genome sequences from clinical samples and suitable bioinformatic pipelines can be generated using only open-source tools. Adenovirus epidemiology in the North West has remained largely unchanged however phylogenetic analysis has highlighted deficiencies in current typing methods and confirms the need for whole genome analysis in outbreak investigation.

While fungal diseases are a growing global problem, there remains a paucity of epidemiological data in most developing countries. Hence, this thesis aimed to investigate the epidemiology, immunology, and aetiology of fungal allergic diseases in Zimbabwe. This study demonstrated that the burden of fungal diseases in Zimbabwe is high (14%), with tinea capitis being the most prevalent condition. The country's background sensitivity to fungal allergens and the effect of host-related factors was unknown. Therefore, this thesis further investigated the prevalence of fungal sensitisation and reactivity among Zimbabwean children and how the gut mycobiome composition affected these, whilst considering the effects of host age, gender, and Schistosoma haematobium infection status. The prevalence of fungal sensitisation to ≥1 fungal species was 96%, and the metagenomic analysis of the gut mycobiome showed that the mycobiome comprised < 1% of the sequenced gut microbiome. There was no association between the mycobiome and fungal reactivity, or the host factors studied. Interestingly, an increase in Aspergillus, Tricholoma, and Periglandula abundance was associated with schistosome infection. Due to the high prevalence of fungal sensitisation, this thesis further identified and characterised fungal proteins that were immunoreactive against serum samples from fungal-sensitised children. Furthermore, the utility of a specific Aspergillus fumigatus allergen (Asp f 2) peptide in the differential diagnosis of fungal allergy was evaluated. As a result, novel immunogens from fungi were discovered, potentially increasing the number of known fungal allergens. The Asp f 2 peptide was demonstrated to be an inadequate indicator for diagnosing fungal allergy in the population. Taken together, the findings of this thesis add to the ongoing global discussions about the burden of fungal diseases. Furthermore, it provides novel data on several important aspects of fungal allergy that must be evaluated and validated since they may have implications for the care of allergic individuals.

The courts have become an increasingly important venue for efforts to combat the impacts of climate change. Claims have brought on a range of legal grounds, but many climate lawsuits are based on the existence of a causal relationship between emissions of greenhouse gases and impacts that occur as a result. Such claims may be made to establish emitters' liability for climate change impacts, or to demonstrate that contributions to climate change cause harm, and therefore that greenhouse gas emissions should be curbed. Establishing causation in climate litigation requires synthesis of climate science and legal arguments. This thesis evaluates how legal arguments can align with developments in climate change attribution science, a subfield of climate science that can assess the contribution of anthropogenic influence to climate-related impacts. It then considers how attribution-science evidence has been used in past lawsuits, and responds to emerging scientific questions regarding the effects of climate change that have arisen in the courts. Legal scholars have identified attribution science as holding the potential to establish causation in the courts, but the analyses described in section 4 of this thesis found that the evidence presented in climate lawsuits often lags developments in climate science and that, in some cases, this may have hindered findings of specific causal relationships between greenhouse gases and their consequences. Responding to these findings, sections 5 and 6 report analyses that advance the state of knowledge in areas of attribution science that hold particular legal relevance. Section 5 studied anthropogenic influence on the glacial lake outburst flood hazard from Laguna Palcacocha in the Peruvian Andes, in the context of ongoing litigation (Lliuya v. RWE). It concluded that 85-105% (5-95% confidence interval) of the observed industrial-era warming in the region surrounding Laguna Palcacocha was attributable to anthropogenic climate change and that the overall retreat of the Palcaraju glacier since 1880 is entirely attributable to this warming. Moreover, it is virtually certain (>99% probability) that the observed glacier retreat cannot be explained by natural variability alone. Laguna Palcacocha has expanded in the wake of this glacier retreat, substantially increasing the outburst flood hazard to the downstream city of Huaraz. Finally, this thesis responds to the increased use of human rights-based arguments in climate litigation and develops methods that extend attribution science analyses from quantifying the effect of climate change on physical hazards to their impacts on health. These methods address limitations in existing approaches, including the representation of the climate-health relationship, choices in calculating counterfactual temperatures, differing approaches for assessing long-term trends and individual events, and estimating the effects of adaptation. The approach described in this section is then applied to quantify heat-related mortality attributable to climate change in the Canton of Zürich (Switzerland) over 1969-2018. We found a substantial burden of heat-related mortality attributable to climate change both within and outside of heatwaves. Overall, 1,7000 heat-related deaths were attributed to climate change, and adaptation was found to have avoided at least a further 700 deaths. The thesis spans climate science, glaciology, epidemiology, and legal analysis. Altogether, this research explains how science can strengthen legal causation arguments in climate litigation, and develops methods and analyses to this end. Scientific evidence has a central role to play in developing well-grounded climate lawsuits, and methodological developments can resolve questions raised in court.

Healthcare is increasingly digitised and recorded, producing huge volumes of 'routinely- collected' data. This data is increasingly viewed as a valuable resource for the improvement of healthcare, although a truly unifying framework to translate data into knowledge and thence change is lacking. From an initial introduction to key concepts and a critique of competing methodologies this thesis identifies a 'systems-approach' as a potential framework to guide data analysis, design, implementation, and overarching project management. The thesis provides evidence for the potential of a systems-approach through two main avenues, a systematic review of the medical improvement literature and the reporting and reflection on two data-intensive improvement initiatives pertaining to the perioperative care of older surgical patients. Four key stages of these case studies are used to determine the role of a systems-approach in the use of routinely-collected healthcare data, with the selection of these phases of work informed by a detailed review of the improvement and epidemiological literature. Evidence for the benefits of a systems-approach is advanced from reflections on project set-up, database construction, development of a causal-model, the design of covariates to capture complex system events (such as integrated working), and the translation of statistical results towards real-world solutions. These results provide evidence that a systems-approach appears to benefit work conducted at a 'project' and 'task' level, can increase transparency in key phases of a data-science workflow, aid in the identification of risks that could lead to biased results, and that an under- standing of the data-generating system appears crucial to understand causal relationships that may be captured in statistical results. These conclusions, as well as the results of each key project phase, form the clear original contributions of this work. The thesis concludes by positing a model of data-intensive improvement as incorporating iterative phases of systems-supported data-science, data- informed design, and data-evidenced implementation which is ready for evaluation in future studies.

Existing evidence on healthy dietary patterns suggests that they are modestly inversely associated with incidence of common non-communicable diseases, and in particular cardiometabolic diseases. Trials are rarely feasible to assess causality of these relationships which motivates leveraging observational data with improved methodological approaches. In this PhD thesis, I attempted to address the common limitation in nutritional epidemiology of subjective assessment of adherence to dietary patterns. Chapter 2: I conducted a systematic review of the effects of Mediterranean diet interventions on nutritional biomarkers. I identified 29 trials reporting on 25 biomarkers eligible for meta-analysis (5-18 studies available per biomarker). Circulating carotenoids, vitamin C and fatty acids emerged as candidate biomarkers of compliance. Effect sizes were mostly small which likely reflected the multifaceted nature of whole diet interventions and insufficient validity of single analytes as biomarkers of adherence to the Mediterranean dietary pattern. Chapter 3: I used the EPIC-InterAct case-cohort study which measured nutritional biomarkers at scale (~13,000 subcohort participants and ~9,000 incident type 2 diabetes cases) to evaluate the utility of combining them into biomarker scores predictive of adherence to dietary patterns, and to test their associations with incidence of type 2 diabetes. The available biomarkers included circulating carotenoids, vitamins C and 25(OH)D, fatty acid profiles, iron status biomarkers and cations. The dietary patterns of interest were the Mediterranean diet, alternative Healthy Eating Index-2010 and the Dietary Approaches to Stop Hypertension. The analyses showed modest correlations of the biomarker scores with their respective dietary patterns (r ~0.3) and statistically significant inverse associations with disease risk (hazard ratios ~0.8 per standard deviations of the scores). Chapter 4: I established a collaboration with one of the randomised trials identified in Chapter 2, the MedLey trial, to address the limitations of internal derivation and validation of the biomarker scores. It compared the effects of a partial-feeding Mediterranean diet intervention with continuation of habitual diet in Australia on circulating carotenoids and fatty acids, 29 of which overlapped with those available in EPIC-InterAct. Using end-of-trial biomarker concentrations as predictors of the randomised assignment (n = 128), I developed a biomarker score which discriminated well between the trial arms (C-statistic = 0.88). It was robustly inversely associated with incidence of type 2 diabetes in EPIC-InterAct (hazard ratio 0.71 per standard deviation; 95% confidence and prediction intervals: 0.65-0.77 and 0.55-0.91). Chapter 5: I additionally used the combined InterAct-MedLey data to test generalisability of biomarker scores of the Mediterranean diet. I derived a series of biomarker scores predictive of self-reported adherence to the Mediterranean diet in EPIC-InterAct countries, the MedLey trial baseline sample, and non-InterAct participants of the EPIC-Norfolk cohort (~5,000 with relevant biomarkers). Controlling for multiple testing, values of 8/13 biomarker scores were higher in the Mediterranean diet intervention than the control group of the MedLey trial, and 10/13 scores were inversely associated with incidence of type 2 diabetes in EPIC-InterAct. Chapter 6: In the EPIC-Norfolk study, I investigated the impact of expanding a base set of predictors from circulating carotenoids and fatty acids with additional groups of nutritional biomarkers (urinary sodium, potassium and sugars, urinary and serum phytoestrogens, circulating vitamin C, iron status biomarkers, cations and stable isotopes) or using metabolomics on (i.) the correlations between self-reported Mediterranean diet and its biomarker scores and (ii.) the associations between the biomarker scores and incident cardiovascular disease, cancer, type 2 diabetes and mortality (n range ~500-11,000). The base set biomarker score had a moderate cross-validated correlation with self-report (r = 0.40) and the performance was similar or decreased with inclusion of additional nutritional biomarkers (r range: 0.30-0.41) and modestly improved with metabolomics (r = 0.46). Biomarker scores were inversely associated with disease and mortality outcomes, and use of different sets of biomarkers modified the strength of association for few diet-disease associations. Inverse associations were notably stronger for type 2 diabetes (hazard ratio ~0.80 per standard deviation of biomarker scores) than for other outcomes (range ~0.90-0.95). Chapter 7: I conducted an outcome-wide analysis of 27 incident noncommunicable diseases in the EPIC-Norfolk study using as exposures the biomarker scores derived throughout the thesis (n range ~7,000-11,000) and dietary self-report of the Mediterranean diet (n ~22,000). Controlling for multiple testing, inverse associations were robustly detected across ≥2 of 4 methods of exposure assessment for type 2 diabetes, chronic obstructive pulmonary disease, and heart failure. At the nominal α = 0.05, corresponding relationships were identified for ischaemic heart disease, renal disease, oesophageal and stomach cancers, and cataracts. This PhD identified combinations of nutritional biomarkers as plausible biomarkers of the Mediterranean diet for application in epidemiological investigations. These findings contribute towards development of methods of objective assessment of diet and strengthen the evidence on the inverse relationship between the Mediterranean diet and cardiometabolic disease.

More time in moderate-to-vigorous physical activity (MVPA) is associated with lower cardiovascular disease (CVD) risk. However, most of the 24-hour day is spent in other behaviours, including light physical activity, sedentary behaviour, and sleep. Device-based movement measurement in large prospective cohorts now enables detailed investigation of 24-hour movement behaviours. However, traditional approaches to classifying behaviours in device data, based on 'cut-points', can be crude and sometimes inaccurate. Furthermore, approaches to associate 24-hour movement behaviours with disease risk do not always account for how behaviours make up the 24-hour day. While biases are known to affect analyses, they are rarely assessed quantitatively. This work therefore aimed to investigate effects of device-measured 24-hour movement behaviours on risk of incident CVD, addressing challenges around behaviour classification, analysis of 24-hour movement behaviours, and interpretation. This work developed machine-learning-based classification methods to classify free-living wrist-worn accelerometer data as sleep, sedentary behaviour, light physical activity behaviours, or moderate-to-vigorous physical activity behaviours. The classifier used a Random Forest with a Hidden Markov model for time smoothing. It was developed using free-living data from a study of 152 participants who had a camera-and diary-based ground-truth measured alongside accelerometry (the CAPTURE-24 study). In Leave-One-Participant-Out analysis, this method classified free-living movement behaviours with a mean accuracy of 88% (87%, 89%) and a mean Cohen's kappa of 0.80 (0.79, 0.82). UK Biobank (UKB) is a population-based prospective cohort of middle-to-older aged adults, in which∼100,000 participants wore a wrist-worn Axivity AX3 accelerometer for a 7-day period between 2013 and 2015. The machine-learning model was used to classify movement behaviours of UKB participants, and classification patterns by time of day and by participant characteristics showed good face validity. UK Biobank participants' 24-hour movement behaviours were associated with risk of incident cardiovascular disease (ICD-10 codes I20-25, I60-69) using a multivariable-adjusted Cox regression model. To account for the compositional nature of 24-hour movement behaviour data, it was modelled using a Compositional Data Analysis approach. Over a median of 6.2 years' follow-up, there were 4,105 incident CVD events among 87,498 participants. Reallocating time from other behaviours to MVPA was associated with a lower risk of incident CVD. Reallocating time from sedentary behaviour to other behaviours was also associated with lower risk. For example, compared to the average movement behaviour composition in this cohort, reallocating 20 minutes/day to MVPA from all other behaviours proportionally was associated with a Hazard Ratio (HR) of 0.91 (0.90, 0.93), while reallocating 1 hour/day to sedentary behaviour from all other behaviours proportionally was associated with a HR of 1.05 (1.03, 1.07). A range of approaches were used to quantitatively assess the impact of information bias, selection bias, reverse causality, and residual/unmeasured confounding on results. Adjusting for within-person variation in measurements modestly de-attenuated some results and little affected others. Although there were clear selection effects from UK Biobank into the accelerometry sub-study, this did not appear to substantially affect analytic results. Findings on reverse causality were mixed, with little difference by time period but different findings on sedentary behaviour by (cardiovascular) health status. There was some evidence residual/unmeasured confounding affected MVPA results, but it did not entirely explain findings. This work contributes to emerging evidence on device-measured 24-hour movement behaviours and risk of incident cardiovascular disease. It both supports existing guidelines encouraging adults to engage in regular moderate-to-vigorous physical activity and limit sedentary time, replacing it with physical activity of any intensity, and supports the development of future guidelines and interventions targeting behaviours throughout the 24-hour day. This work also supports future research, particularly by developing and making available a machine-learning-based classifier with good performance, and by using emerging methods for associating movement behaviours with incident disease and for assessing bias.

Plant diseases pose a significant threat to food security, and fungal plant pathogens can be particularly damaging due to enormous evolutionary potential. Two major control strategies for fungal pathogens include chemical control with fungicides and use of disease-resistant cultivars. Both strategies are threatened by pathogen evolution leading to fungicide resistance and loss of cultivar efficacy ('cultivar breakdown'). Both fungicide resistance and cultivar breakdown can be characterised as qualitative or quantitative, depending on the genetic basis for resistance/breakdown. Mathematical models have an important role in understanding fungicide resistance and cultivar breakdown, and modelling studies offer numerous advantageous over experimental studies. We present a model of qualitative fungicide resistance and one of quantitative fungicide resistance and cultivar breakdown. The model of qualitative resistance extends a model from the literature, while the quantitative resistance model is entirely novel. Both models are parameterised to address Septoria, the most prevalent disease of wheat. Although qualitative resistance has been widely studied, many single-site fungicides (i.e. those challenged by qualitative resistance) face widespread resistance, meaning that quantitative resistance is increasingly important. However, the mechanisms underlying quantitative resistance/breakdown are more complex and suitable data for model fitting is harder to source. We present the first model of quantitative resistance to be fitted to field data for both fungicide resistance and cultivar breakdown. Fungicide mixtures can help delay resistance. Although many studies focus on mixtures of fungicides to which resistance is qualitative, the optimal strategy is not characterised if the initial resistance frequencies (RFs) to two fungicides in a mixture differ. Past work showed that equal selection for single resistant strains in the first year was the optimal strategy when initial RFs were equal but did not consider when the initial RFs differ. We show that this strategy is often sub-optimal if the initial RFs differ and present an alternative strategy based on equalising RFs in the breakdown year. We test the robustness of this strategy to changes in parameters controlling fungicide efficacy and pathogen epidemiology, including between-season pathogen sexual reproduction. Previous modelling studies of Septoria neglected pathogen sexual reproduction, for simplicity, despite evidence that sexual reproduction is an important part of Septoria's life cycle. We use the quantitative resistance model to determine how the number of fungicide applications per year affects resistance development, disease severity and yield. We consider how the optimal strategy varies depending on the time-frame of interest, before exploring how cultivar and fungicide control can be optimally combined to control the pathogen whilst minimising degradation of host and fungicide. We explore the effect of fungicide dose and compare the recommendations from the quantitative resistance model to those from the qualitative resistance model. Most fungicide resistance studies consider strategies which are fixed in time - i.e. rely on the same tactic used every year until failure. We present a flexible approach to optimise time-variable strategies, based on dynamic programming. This allows us to address an otherwise computationally infeasible problem to find the optimal strategy when time-variable strategies are permitted. We compare the improvement offered by the optimal time-variable strategy to the optimal fixed-time strategy.

Whilst the spread of plant disease depends strongly on biological factors controlling transmission, epidemics clearly also have a human dimension. Disease control depends on decisions made by individual growers, who are in turn influenced by a broad range of factors. Despite this, human behaviour has rarely been included in plant epidemic models. This thesis focusses on addressing this oversight by developing combined epidemic and economic models of disease spread. We use simple continuous-time models of disease spread which we couple with behavioural models which set the management programme of the growers for the next growing season. Our models are rooted in game theory, with growers making strategic decisions based on the expected profitability of different control strategies. In the first instance, we compare different versions of this behavioural model, which differ in terms of the information used by the growers to assess profitability. We investigate these models in the context of Cassava Brown Streak Disease (CBSD) and its management via the use of clean seed systems (CSS). We find that both the information used by growers to assess profitability and the perception of economic and epidemiological parameters influence long-term participation in the CSS. Over-estimation of infection risk leads to lower participation in the CSS, as growers perceive that paying for the CSS will be futile. Additionally, though the CSS can reduce the burden of disease, and allow a scenario in which all growers control, disease is not eliminated from the system. For the remainder of this thesis, we use one behavioural model to investigate the deployment of crop that is either resistant or tolerant to Tomato Yellow Leaf Curl Virus (TYLCV). We find that when growers used resistant crop, higher yields were achieved by both controllers and non-controllers, though widespread use of resistant crop was not achieved. The use of tolerant crop reduced the yields for non-controllers, but generally benefited its users, inducing a positive feedback loop that resulted in a high proportion of growers using tolerant crop. By extending this TYLCV model to allow a three-way choice of tolerant, resistant, and unimproved crop, we see again how growers prefer tolerant crop. However, these responses can be manipulated by changing the cost of each crop type through subsidisation schemes. To do this, we consider the efforts of a ``social planner" who moderates the price of crops. We find that subsidising tolerant crop costs the social planner more in subsidies, as its use encourages selfishness and widespread adoption. Subsidising resistant crop, however, increased the use of resistant crop, again enabling higher yields across the community of growers. Many of our results obtained were robust to spatial and stochastic effects. Some differences arose when growers narrowed their information sources to only consider those growers whose fields are in close proximity to their own, as this allowed assessments of profitability to be based on local disease pressures. In this thesis, we show how simple models of grower behaviour can be incorporated into both deterministic and spatial-stochastic models of disease spread. Understanding the influence of economic and epidemiological factors, as well as the feedback loops induced by different control mechanisms, on these behaviours can help to promote better outcomes for growers.

Research into dog health has historically relied on small scale cross-sectional studies or specialised medical and clinical data, which are subject to bias and are difficult to generalise to the wider canine population. The digital era presents an opportunity to collect sources of Big Data for health surveillance and research, defined as data that is high volume, velocity or variability. Data science techniques have made accessing, managing and analysing such datasets more achievable. Large scale cohort studies are needed to estimate the incidence of disease and to identify factors associated with long-term canine health. This project was primarily based on dog owner questionnaires from Dogslife, an internet-based cohort of Labrador Retrievers in the UK set up in 2010. In this thesis, I designed data cleaning methods for Dogslife and validated some of them on veterinary and human medical records and investigated the epidemiology of canine health using Dogslife data, Google Trends and 16S ribosomal RNA gene sequencing data derived from canine faecal samples. A decision-making algorithm for identifying, correcting or removing implausible values in growth measurements was designed and tested in combination with five different data cleaning methods, which were then applied to five datasets. The algorithm was most effective in combination with non-linear mixed effects models and increased the average sensitivity and specificity of the models alone by 7.68% and 0.42% respectively. This method was adaptable and had several useful functions including allowing for individual growth trajectories, preserving data where possible and removing duplications. A vomiting outbreak was evident in UK dogs between December 2019 and March 2020 in data from Dogslife and Google Trends search queries. The odds of a vomiting incident being reported to Dogslife was 1.51 (95% CI: 1.24 - 1.84) in comparison to the same time period in previous years (December to March, 2010 to 2019). Dogslife data identified risks for a dog experiencing a vomiting episode and differences in owner-decision making when seeking veterinary attention for vomiting during the outbreak. Compared with previous years (March 23rd to July 4th, 2010 to 2019), the COVID-19 restrictions study period (March 23rd to July 4th 2020) was associated with owners reporting increases in their dogs' exercise and worming and decreases in insurance, titbit-feeding and vaccination. Odds of owners reporting that their dogs had an episode of coughing (0.20, 95% CI: 0.04 - 0.92) and that they took their dogs to a veterinarian with an episode of any illness (0.58, 95% CI: 0.45 - 0.76) were lower during the COVID-19 restrictions compared to before. A longitudinal sub-study of Dogslife Labrador Retriever puppies was designed to investigate associations between environmental and health factors and the development of the canine microbiome. When their puppies were three to four, seven and 12 months of age, owners submitted digestive health questionnaires and faecal samples from their puppies, which were used to produce 16S ribosomal RNA gene sequencing data. Dogs' faecal microbiota were successfully characterised for each wave of sample collection at the different dog ages. The largest source of variation in the composition of dogs' microbiomes was explained by differences between individual dogs, explaining approximately 50%. Additional associations were found between age, sex, coat colour, UK geographical region, household type, coprophagia, contact with other animals, recent antibiotic use and recent diarrhoea and various differences in the diversity and composition of the microbiome. Owner-derived data can be used alongside other sources of Big Data and provides valid and valuable information for the surveillance of veterinary health that contains detail about environmental factors not typically present in medical records or clinical studies. Such information is becoming easier to handle and analyse with the use of data science techniques. Furthermore, cohort studies can be used for the recruitment of participants to sub-studies that aim to answer a specialised question, such as microbiome research.

Glioblastoma is the most common malignant primary brain tumour with poor survival of under one year. The current standard of care treatment offers modest survival benefits, and treatment options for patients with glioblastoma have not changed since 2005. Incomplete understanding of glioblastoma tumour biology presents challenges in drug development for glioblastoma, but inefficiencies resulting from suboptimal methodology may also contribute to the lack of effective treatment. Systematic evaluation of translational studies and analyses of population-based data can optimise research strategy to improve survival for all patients with glioblastoma. This thesis aims to apply epidemiological and health data science approaches to glioblastoma research data for improving translational research and population-level survival. The first chapter of the thesis discusses epidemiological and clinical aspects of glioblastoma in the context of all brain tumours and outlines the challenges and opportunities for research. Chapter 2 defines the problem by summarising the longer-term (≥2 years) survival of people diagnosed with glioblastoma in a systematic review and meta-analysis of 63 population-based cohort studies. The next chapter (Chapter 3) examines how changes in diagnostic practice can affect brain tumour incidences and how this may explain population-level survival described in the previous chapter. The following two chapters (Chapters 4 and 5) are systematic reviews that evaluated reporting standards, consistencies and analytic approaches of preclinical glioblastoma studies to identify methodological limitations precluding clinical translation. Chapter 6 is a cohort study of 414 consecutive patients with glioblastoma in Southeast Scotland that investigated how well clinical trial findings apply to real-world clinical cohorts concerning the modification of treatment effect of chemotherapy by a molecular marker. Chapter 7 presents a matched cohort study comparing the risks of cardiovascular events between brain tumour patients and the general population in Wales. The final chapter summarises the findings presented in this thesis and discusses their implications for future research. The programme of work in this thesis emphasises the need for transparency and consistency in translational glioblastoma research, and it demonstrates the value of routine healthcare data to understand population-level survival for people with glioblastoma.

Although medical devices (MDs) and surgery have been part of medicine since its start, the methods for their study in epidemiology have not kept pace with the developments in pharmacoepidemiology. The recent increased attention from society to the potential harms of unsafe implants has led to improved legislation that call for a much closer evaluation of MDs. However, the evaluation of MDs and surgery presents unique caveats and challenges, such as the ascertainment of the indication of the surgery or how to include surgeon characteristics in the analysis, that can greatly influence outcomes. To advance research on how one can overcome these challenges, I tested several methods. I used Propensity Score (PS) methods, namely PS matching, stratifications and inverse probability weighting (IPW); and Instrumental Variables (IVs), based on surgeon and hospital preference in the study of the effectiveness and safety of partial vs total knee replacement and evaluated them by comparing observational results to those from a randomised controlled trial. I applied the target trial framework to the study of the timing of tracheostomy in patients with COVID-19. I further studied the safety of knee replacements using the self-controlled case series method. I studied the potential heterogeneity by subgroups (on high-risk patients, by gender, by age, and by deprivation) on the safety and effectiveness of knee replacements. I explored the effect of surgical volume on knee replacement outcomes. I finally used simulation studies to examine the flaws and challenges of preference-based IV. I found that PS stratification and IPW may be able to minimise confounding and bias in the study of partial vs total knee replacement, provided that the study is carefully designed to consider both patient and surgeon characteristics. I also showed how preference-based IVs may not be fit for purpose for this use case. Future research is needed to examine whether my findings are generalisable to other settings and whether other IVs can be used safely in MD and surgical research.

Estimating the rate of nucleotide and amino acid substitutions is fundamental to our understanding of how fast species evolve. It also enables us to reconstruct their natural history. The development of sophisticated statistical methods and software in molecular sequencing analysis has made it easier for biologists to test and compare various evolutionary hypotheses and infer evolutionary parameters. However, the use of these software packages as a 'blackbox' without paying close attention to the underlying assumptions can lead to spurious rate estimations across different timescales which often result in erroneous inferences of virus evolutionary histories. In this thesis, I develop quantitative methods to understand biological mechanisms that can alter the molecular evolutionary rate of viruses over time and investigate the implications of having a time-dependent rate of evolution for reconstructing the phylogenetic history of viruses. I also explore alternative methods to estimating the time of origin and reconstructing the evolutionary trajectory of viruses using mobility data, national vital statistics, and serology surveys. I show how three evolutionary processes of rapid within-host evolution during chronic infections, purifying selection, and site saturation can determine the changes in evolutionary rate of viruses over different timescales, from a few weeks to millions of years. I also leverage excess mortality and seroprevalence data to correct for under-reporting of COVID-19-related deaths, reconstruct early transmission dynamics of SARS-CoV-2, and estimate the start of the Iranian epidemic as well its growth rate in the absence of high-quality genome sequence and province-level epidemiological data. This work will build towards the goal of making more accurate estimations of the timescale of phylogenetic histories and the rate at which species evolve. It also further strengthens the conceptual link between phylogenetics, population genetics, and molecular evolution.

Recent large cohort studies follow participants for tens of years to collect their medical histories, allowing longitudinal analysis of many complex diseases. These datasets are typically accompanied by genotype data, whose effects on diseases are typically estimated as static values without reference to potential age varying effects. In this work we developed methods to study the relationship between genetic variations and complex diseases as a function of age. In the first step, we developed methods that focus on a single disease at a time, which we applied to infer the age-varying genetic risk profiles for common complex diseases. When applying the methods to 24 common diseases in the UK Biobank, we found a common decreasing pattern of genetic risk effect for several complex diseases, after using a frailty model to adjust for unobserved risk factors such as environmental factors. In the second step, we developed methods that look at many common diseases simultaneously, which we applied to infer age-dependent multi-morbidity profiles under a low-rank representation framework. We analysed the shared genetic variations underlying clusters of diseases as well as the genetic heterogeneity among subtypes of diseases implied by the comorbidity profiles. When jointly modelling 349 Phecodes that have more than 1000 occurrences in the UK Biobank, we identified several disease topics that contain genetically- correlated diseases. When assigning incident diseases in the UK Biobank to these topics, several complex diseases are stratified into subtypes that have different genetic profiles. Our results suggested that genetic risks are modulated by contexts including age and comorbid diseases. To better predict disease incidences and identify biological pathways, genetic information should be modelled jointly with other information available in biobank data sets.

Endemic and emerging viruses are essential to study in order to understand their prevalence and disease association within Northern Ireland's livestock and wildlife as, they may be considered an important risk factor to food security, food safety and public health. In recent decades, the prevalence of viral zoonotic pathogens transmitted from animals to humans is currently creating an increasing socio-economic impact. Accordingly, routine pathogen molecular epidemiology surveillance is the key to the identification and mitigation of new and endemic viruses but is often lacking with undiagnosed diseases not being thoroughly investigated. This thesis focuses on four research areas that fall within the scope of molecular detection and investigation of emerging and endemic viruses in Northern Ireland's livestock and wildlife including; the study of novel porcine parvoviruses (PPV) coinfection status with porcine circovirus 2 (PCV2) and statistical disease association, hepatitis E virus (HEV) prevalence, enteric virome of clinically diseased animals and the prevalence and genotyping of rotavirus (RV) by utilising real-time PCR, Sanger sequencing, Illumina second generation sequencing and bioinformatic analysis. The recently identified novel PPVs 2-4's clinical significance is poorly understood. Pig endemic diseases for example porcine respiratory disease complex and porcine circovirus associated disease are caused initially by primary viral infections, however disease association by coinfections with novel parvoviruses and porcine circovirus 2 remains to be determined. The analysis of archival samples from clinically diseased pigs' indicates that PPV2 is an emerging pathogenic agent, and is statistically associated with porcine respiratory disease complex in pigs. The emerging HEV is ubiquitous in pig herds worldwide, and HEV genotypes 3 and 4 are zoonotic viruses transmitted mainly as a foodborne disease leading to subclinical to acute infections in humans. A HEV molecular epidemiology study on the prevalence of HEV through an abattoir study and the dynamics of the HEV infection in a longitudinal study were completed. The enteric microbiome is a complex community of bacteria, viruses, fungi and protozoa. This a new area of research where limited studies exists on the enteric virome of livestock and wildlife. Studies of the virome will aid in the understanding of; viral taxonomic identification, viral prevalence, virus genetic and phenotypic diversity, coinfections and virus-bacteriome and host interactions modelling. Next generation viral metagenomics is a new tool for monitoring and detecting new emerging viruses and will help to fill the knowledge gap of the mammalian virome composition. Rotavirus group A (RVA) is the major cause of acute diarrhoea and severe dehydration in children and young mammalian host species. Infection of neonate and young animals may lead to death, resulting in economic losses for the farmer. A rotavirus molecular epidemiological surveillance study was conducted mainly on clinically diseased Northern Ireland animals by Sanger sequencing for RVA detection, NGS whole genome amplification and phylogenetic analysis. The study revealed a high level of rotavirus group A phylogenetic diversity and demonstrates the importance of accumulating knowledge of disease and molecular epidemiological surveillance coupled with viral metagenomics for understanding the viruses circulating in Northern Ireland livestock and wildlife.

Non-tuberculous mycobacteria (NTM) are an emerging group of drug-resistant and opportunistic pathogens. Case numbers of non-tuberculous mycobacterial pulmonary disease (NTM-PD) are rising rapidly around the world, particularly in high-income countries. There is a growing cohort of susceptible patients, typically with pre-existing lung disease and/or impaired immunity, for whom NTM-PD carries a poor prognosis. Mycobacterium avium complex (MAC) is the leading bacterial cause and survives by subverting lung macrophages in susceptible patients. Current treatment strategies are unsatisfactory, consisting of protracted courses of multiple antimicrobial drugs which are poorly tolerated and carry a high failure rate of 40-60% within 5 years. Three research priorities that are regarded as important to improving outcomes for patients with NTM-PD and are the focus of this thesis: 1) detailed epidemiological characterisation; 2) clinically-relevant disease modelling; 3) novel therapeutic strategies. Chapter 3 describes the epidemiological profile of NTM-PD in Northern Ireland. NTM isolates from patient specimens has risen more than 3-fold in the last 12 years. This is driven by the isolation of MAC from the pulmonary specimens of older people, particularly males over 50 years old. These findings warrant urgent attention into improving the recognition of NTM-PD in at-risk patients. Chapter 4 describes the characterisation of a clinically-relevant in vitro model of MAC infection using primary human macrophages (monocyte-derived macrophages, MDMs). Important outcomes such as bacterial uptake, intracellular proliferation, pro-inflammatory cytokine response are variably dependent on macrophage donor and multiplicity of infection (MOI). An MOI of 1 appears to achieve a rate of bacterial uptake that correlates with significant levels of intracellular proliferation and pro-inflammatory cytokine response, but without significant cytotoxicity. Yet, there remain further variations in bacterial uptake and proliferation between MAC strains (laboratory and clinical strains) which will necessitate the use of multiple strains in testing candidate therapies. The focus of the thesis is reporting the testing of mesenchymal stromal cells (MSCs) in cellular (chapter 5) and animal (chapter 6) models of MAC-PD. MSCs) have antimicrobial and immunomodulatory properties that are being investigated for many infectious diseases, but their potential in MAC infection remains unknown. I hypothesised that MSCs can directly kill MAC, and/or promote their clearance by macrophages. MSCs had no direct activity against MAC but inhibited their intracellular growth in human MDMs by 30% over 72 hours (p < 0.05). This activity was dependent on MSC COX-2 activity and mediated through MSC secretion of PGE2. Activation of PI3K in infected MDMs by MSC PGE2 appeared to be important for inhibiting intracellular growth for some but not all donors. Two doses of 1 million human bone marrow MSCs reduced pulmonary counts of MAC in mice with chronic infection by 20% (p < 0.05) but did not reduce bacterial dissemination to liver and spleen. These findings warrant further investigation into the downstream mechanisms of MSCs in promoting macrophage activity against intracellular MAC. I discuss the next steps required to prepare MSCs for human trials as a potential adjunct therapy for MAC pulmonary disease.

This thesis investigated the epidemiology of injuries in professional Rugby Union and the methods in which they are analysed and reported. Furthermore, the thesis explored the development and application of a player-specific method of analysing injuries. Chapter 3 investigated match injuries within the Welsh regional professional Rugby Union teams, demonstrating a higher incidence than previously reported in professional Rugby Union (99.1 injuries/1000 match hours). In addition, the highest proportion of injuries in matches were sustained during the tackle event (50-63%). Whilst this study followed the recommendations of the consensus statement, Chapter 4 aimed to identify whether research in elite or professional sports with published consensus statements also follow their respective recommendations. Chapter 4 demonstrated that there remain inconsistencies with regards to reporting injuries, identifying that the pooling of injury data across individuals remains a common issue within research. As this method is recommended by the consensus statement on data collection and analysis procedures within Rugby Union, Chapter 5 explored a player-specific approach to analysing and reporting injuries. The consensus statement method of calculating team-level exposure, using the standard number of players in a team and standard match length, was compared against the use of global positioning systems (GPS) to identify if the consensus statement provides an adequate method of analysing injury incidence. Interestingly, there were differences between the injury incidence calculated using standard match length and the injury incidence using GPS-derived exposure hours (59.5 vs 95.7 injuries/1000 match hours, respectively). While the team-level injury incidence was influenced by the number of players providing consent, the study demonstrated that team-level injury incidence does not reflect the variation in injury rates at a player-specific level, with 94% of players falling outside of the 95% confidence intervals for the team-level injury incidence. In addition, an alternative to injury incidence statistic was explored through analysing the probability of injury. This identified that when players at the same incidence are exposed to higher match hours, they have a higher probability of incurring multiple injuries. The player-specific methods of analysis from this Chapter were applied in Chapter 6, where the injury rates and mechanisms of starter and replacement players were analysed. Chapter 6 identified that starters had a higher injury incidence than replacement players (80.8 vs 57.2 injuries/1000 match hours, respectively). When accounting for the number of replacements used in a match, the injury rate per player exposure did not change for each number of replacements used. However, the number of replacement players had a significant effect on the number of match injuries, where injuries increased by 12% per replacement used (p = 0.037). This increase may, in part, be due to a small number of matches using less than seven replacements (20%). Therefore, further analysis accounting for the replacement time-in-game was also implemented, showing a non-significant 1% increase in injury for every 10 minutes of replacement player time-in-game (p = 0.099). Chapter 6 also identified that the tackle event was the mechanism responsible for the highest proportion of injuries for both starter and replacement players. The propensity for injury was similar for tackles involving two starters or two replacements, with a higher injury propensity only shown when an injured starter was tackling a replacement ball carrier. Starter and replacement players exhibited different characteristics during an injury inciting tackle, specifically when the injured player was making a tackle. Replacement players maintained a lower body position, predominantly using a shoulder tackle and contacting the upper leg of the ball carrier. In contrast, starters demonstrated a higher body position, contacting the ball carriers head whilst in an upright position. This thesis demonstrated that whilst consensus statements are important for the consistent definition and data collection associated with injury surveillance, there remains inaccuracies and inconsistencies with the way data is analysed and reported. The lack of research reporting the subsequent injuries fails to consider the potential for multiple injury occurrence at a player specific level. Furthermore, the continued pooling of data across individuals within a team (team-level injury incidence) fails to account for the variation in injury rates at a player-specific level. This is emphasised further when player-specific analysis is applied to starter and replacement players. The application of a player-specific analysis demonstrated that current regulations associated with the use of replacement players within matches is appropriate. Importantly, however, the analysis of tackle characteristics at a player-specific level demonstrated differences in injury inciting tackle characteristics between players. Where possible, methods such as GPS-derived exposure and probability analysis should be incorporated within injury surveillance to provide a more comprehensive player-specific analysis of injury that can aid in injury management and improve player welfare. In addition, injury risk mitigation strategies associated with the tackle should incorporate a player-specific approach, specifically considering the type of player (i.e. starter or replacement player) used within matches.

The research in this thesis was conducted in collaboration with Somerset NHS Foundation Trust, and the University of Exeter, during the COVID-19 pandemic. Prior to COVID-19 this research was directed at determining whether predictive models of chronic disease, developed from routinely collected clinical data, could be safely deployed into the clinical workflow. As part of this we published a novel safety framework for data driven clinical decision support. When the COVID-19 pandemic emerged, the priorities of the NHS changed, shifting towards the acute care of unwell patients. Adjusting for this, we re-oriented the research to answer the many questions about the impact of COVID-19 on the NHS. Our study on the Alpha variant was one of the earliest to show increased severity. Our identification of outbreaks of the Delta variant, and its rapid growth, was at the forefront of decision making in local NHS trusts, NHS England, Public Health England, the UK Health Security Agency, and ultimately the UK Government. As we move forward from the acute phase of the pandemic, understanding the longer term impact on chronic health management, and as a co-morbidity for chronic disease will become more of a priority for the NHS. A continued collaboration between the University of Exeter, Somerset NHS Foundation Trust and the UK Health Security Agency is well placed to address this need.

The livestock industry relies heavily upon anthelmintics for the treatment of parasitic nematode diseases. This has resulted in widespread resistance to multiple anthelmintic classes across several parasitic genera. In response to this spread in resistance and subsequent decrease in efficacy there has been increased interest in the use of combined treatment. It is hoped that this combined treatment approach will reduce resistance and increase overall efficacy. However there is little known on how treatments interact with one another and how this impacts implementation. The systematic review (chapter 1) shows a distinct lack of knowledge into large livestock species such as the equines and bovines. Certain combinations such as anthelmintic and plants have a small body of research with conflicting information on how the treatments interact with one another. The current research shows the interaction between anthelmintic and plants to be positive for gastro-intestinal nematodes (chapter 2) and negative for ascarids (chapter 3), specifically Parascaris equorum. The direct impact between treatments has also been investigated, with a negative direct interaction being found between condensed tannins and Duddingtonia flagrans (chapter 4). The GLOWORM-FL model was adapted to predict the larval burdens for plant and D. flagrans in various European regions (chapter 5), this study found there to be clear differences in how combined treatments are targeted which can change dependent on both geographical region and temporally dependent on abiotic factors. The current body of work tries to determine interaction between nematode control but relies heavily on in vitro experimentation. These results may not transfer directly to in vivo field trials of these interactions. There is clear evidence that the treatment interactions need to be considered and optimum application may change dependent on region, time of year and presence or absence of interaction.

This thesis concerns the principled use of mathematical models for informing control interventions in the context of an infectious disease outbreak context. We address two main related questions. First, how can interventions be evaluated and compared in a principled manner using a dynamic (time-varying), mathematical model? The second question pertains to structural identifiability, a prerequisite for successful parameter inference in dynamic models: how does the type of model output influence the structural identifiability of dynamic, mathematical models for infectious disease transmission? We address the first question using two outbreaks of infectious disease - hepatitis A and African swine fever (ASF) - as illustrative examples, and the second question by conducting structural identifiability analysis of various ordinary differential equation (ODE) model versions. Links to all code and files needed to reproduce the results in this thesis have been provided. Our contributions are detailed in three papers: a published article and two manuscripts, now summarized. Chapter 3: Dankwa et al. (2021) developed a dynamic, deterministic model to explain transmission in the 2017-2019 hepatitis A virus (HAV) outbreak in Louisville, Kentucky, US, among persons experiencing homelessness or who use drugs, known to be at a high risk of HAV infection. With the model, alternative vaccination scenarios were examined for effectiveness and cost, and an estimate for the critical vaccination threshold required for herd immunity in this population was derived. Prior to this study, no such estimate had been obtained for this population anywhere in the US. Chapter 4: Dankwa et al. (2022b) developed a dynamic, stochastic, spatial model to explain ASF virus (ASFV) transmission among wild boar and domestic pig herds, and to evaluate alternative outbreak management measures. The model was developed such that it could be refined to account for more outbreak data and hence it is suitable for real- time outbreak analysis. This study is relevant, given the ongoing ASF epidemic in Europe, and the scarcity of ASFV transmission models which account for transmission at the wildlife-livestock interface, despite the evidence of such transmission in Europe. Chapter 5: Dankwa et al. (2022a) conducted structural identifiability analysis of unknown parameters, including initial conditions, of 26 ODE model versions to demonstrate the influence of the type of model output(s) on models' structural identifiability. Data types such as incidence and prevalence, typically encountered in disease surveillance, were studied as model outputs. This analysis emphasizes the importance of a careful consideration of model outputs prior to performing inference with transmission models.

Complete and accurate health information systems are necessary inputs for effective health policy. Although many countries maintain civil registration and infectious disease surveillance systems, variation in data completeness often impedes analysis of aggregated health records at the spatial level where policymaking occurs, discouraging greater investment in these systems. This thesis aims to integrate health information systems into local health decision-making using spatial modelling approaches. In four case studies, I introduce a class of spatial statistical models based on incomplete vital and health surveillance records that offer insights into the health of a country that would be impossible to derive from other sources. I demonstrate how civil registration and vital statistics data (CRVS) can be synthesised with supplementary spatial data sources to estimate both neonatal mortality and CRVS completeness by municipality across Mexico. This spatial modelling strategy can be applied to a wide array of health outcomes, including infectious diseases. I demonstrate how an analogous model can combine data from a tuberculosis (TB) prevalence survey and TB case notifications to estimate TB prevalence across Uganda. Complete registration of births and deaths ensures that all citizens receive the same legal and health protections. I take a holistic approach to analyse India's three health surveillance systems in relation to the Indian National Health Plan's child survival goals. The COVID-19 pandemic has highlighted gaps in health surveillance capacity worldwide, including in high-income countries. In Italy, I develop a small-area excess mortality model to estimate the number of misdiagnosed COVID-19 deaths during the first six months of the pandemic. This analysis reveals important information about the mortality dynamics of the pandemic across sub-populations of Italy. The results, limitations, and conclusions of these case studies are discussed with recommendations for how these findings influence our understanding of health information systems and implications for greater integration between health surveillance data and policy.

Congenital heart disease (CHD) refers to the collection of diseases involving a problem with the structure of the heart present at birth. CHDs are a significant cause of infant morbidity and mortality. The causes of CHD are largely unknown. Observational epidemiological studies have reported links between a wide range of maternal lifestyle factors during pregnancy and CHD in the offspring. However, the causal relevance of these is unclear. This PhD thesis explores relationships between maternal pregnancy exposures and offspring CHD by employing a range of epidemiological techniques using multiple independent data sources. Chapter 1 introduces CHDs, summarises the literature in relation to maternal pregnancy exposures and offspring CHD, puts forward the case for using pregnancy metabolomics and outlines epidemiological methods that could help improve causal inference. Chapter 2 describes cases of congenital anomalies and CHD in the Avon Longitudinal Study of Parents and Children (ALSPAC) cohort. In Chapter 3, I used parental negative exposure control analyses to explore the effects of maternal pre- pregnancy body mass index (BMI), and pregnancy smoking and alcohol on offspring CHD. I found that maternal pregnancy smoking may increase offspring CHD risk via intrauterine mechanisms but did not find evidence to suggest maternal overweight or obesity increase risk. Results for alcohol were inconclusive. In Chapter 4, I used Mendelian randomisation (MR) to explore the effects of the same exposures studied in Chapter 3. I found no robust evidence of an effect for maternal BMI or smoking on offspring CHD using MR. Using a genetic risk score of drinks per week, there was some evidence of a potential causal effect for maternal alcohol intake on offspring CHD. In Chapter 5, I examined the relationship of maternal gestational mass spectrometry-derived metabolites with offspring CHD using multivariable regression and MR analyses. I found evidence that amino acid metabolism during pregnancy, androgenic steroid metabolites, and levels of succinylcarnitine could be important contributing factors to CHD. In Chapter 6, I explored the relationship of maternal gestational nuclear magnetic resonance-derived metabolic traits with offspring CHD. I replicated the findings for amino acids seen in Chapter 5 and found evidence of potential effects for some fatty acid and very low-density lipoprotein traits, albumin, and citrate. Chapter 7 provides an overview of the primary findings for analyses included in each chapter along with strengths and limitations. It then considers the public health and clinical implications of my findings and provides recommendations for future research.