HDR UK Big Data for Complex Diseases PhD Programme

The PhD Projects

• Characterising a high-risk Type 2 Diabetes phenotype using big data - the role of comorbidity and ethnicity

  Key details

  • Hosted by: University of Bristol
  • Lead supervisor: Dr Rachel Denholm, Bristol Medical School
  • Duration: Four years
  • Stipend: UKRI stipend rates apply
  • No international fee waivers
  • Start date: October 2023

  Project summary

  Recent research indicates that about a third of people initially diagnosed with Type 2 Diabetes (T2D) have very high levels of blood glucose at diagnosis (severe hyperglycaemia, characterised by high HbA1c – a measure of average blood glucose in the preceding three months); this appears particularly common in people of African/ African Caribbean ethnicity. However, it is unclear whether this subtype is indicative of inequalities in early diagnosis of T2D, including whether certain patients are less likely to seek treatment; the impact of existing illnesses, or a very aggressive form of T2D.

  This project will use national data from general practices andhospitals to identify a group of people with T2D, characterised by glycaemic status at diagnosis, and investigate therole of patient’s medical history, such as existing health conditions, and demographic characteristics, like ethnicity, inexplaining differences in HbA1C levels at diagnosis. We will also investigate the long-term consequences of high HbA1C levels at diagnosis on the risk of cardiovascular conditions and cancer, compared to those with lower levels, and the role of pre-existing health conditions and ethnicity.

  Findings from the study will help identify whether the severely hyperglycaemic form of T2D is a distinct, severe form of the disease, or a consequence of inequalities in early diagnosis and treatment or the effect of the complex inter­ relationships with other long-term conditions.

  Suitability and eligibility

  This project would suit students with a degree in a biomedical, mathematics, data science or related quantitative discipline (with some familiarity with statistical analysis) and an interest in the application of statistical methods incardiometabolic research. Some post-graduate research experience would be an advantage

  Experience in the analysis of routine health data, the use of statistical software such as R, or a programming language such as Python would be an advantage.

  Click here for further details and to access the application form.

  • If you have any questions contact r.denholm@bristol.ac.uk
• Advancing primary prevention strategies for complex diseases

  Key details

  • Hosted by: University of Cambridge
  • Lead supervisor: Angela Wood, Department of Public Health and Primary Care
  • Duration: Three years
  • Stipend: UKRI stipend rates apply
  • No international fee waivers
  • Start date: Flexible for right candidate

  Project summary

  It is far better for people and cheaper for health services to prevent or delay development of complex diseases, such as cardiovascular disease and cancer, than to treat patients after they become unwell. Knowing about the future risk of complex diseases can help people and health professionals caring for them make decisions about how risk best be managed through life-style changes and/or treatments.

  At present, a wide array of tools are used to predict a person’s future risk of single diseases based on particular risk factors (eg, age, blood pressure, body mass index) measured at a single point in time. Two opportunities are being missed. First, because these tools target single diseases, the opportunity to identify people who are at risk of multiple complex diseases, and often interrelated, is being missed. Second, decisions based on measures at a single time point do not take into account important fluctuations and changes in risk over time.

  There is an urgent need to combine risk prediction tools and decision making for multiple complex diseases together, and to extend these tools for monitoring of risk over a patients’ lifetime.

  The overarching goal of this research is to enhance the prevention of multiple complex diseases by developing and evaluating risk prediction tools that monitor a person’s risk of multiple diseases over time and to translate the risk prediction tools into primary prevention strategies.​

  Eligibility and suitability

  Applicants need a good first degree and/or master’s degree or equivalent experience in one of the following subjects: statistics, epidemiology, biostatistics, mathematics, data science, informatics.

  They will need to be committed to open source, reproducible, research and able to work accurately, with attention to detail. Excellent written and verbal communication skills and organisational and time management skills are also essential.

  Strong data manipulation and analysis experience/skills would be beneficial including:

  • scripting skills in writing re-usable code in at least one programming language (eg, SQL, Python/PySpark)
  • skills in at least one statistical software package (e.g. R, Stata).
  • experience of analysing person-level data, ideally from electronic health records.

  Click here for further details and to access the application form.

  • If you have any questions contact  amw79@medshcl.cam.ac.uk
• Expediting the diagnosis of cancer and other diseases where early diagnosis can improve clinical outcomes in patients presenting to a GP with new symptoms

  Key details

  • Hosted by: UCL
  • Lead supervisor: Georgios Lyratzopoulos, Deptartment of Behavioural Science and Health, Institute of Epidemiology and Healthcare (IEHC), UCL
  • Duration: Three years (four years part time)
  • Stipend: UKRI stipend rates apply
  • No international fee waivers
  • Start date: Preferred start October 2023

  Project summary

  Improving medical diagnosis is a priority for the NHS and health systems across the world. However, the diagnosisof cancer and several other diseases where early diagnosis can improve clinical outcomes is often delayed.

  Meticulous analysis of electronic health records data can provide information to support doctors in deciding when to order tests or refer a patient to a specialist, but such data remain under-explored. The proposed PhD will systematically examine the predictive value for consequential diseases of one or more non-specific symptoms (e.g. loss of weight, abdominal pain, or abdominal bloating). Incorporating information from additional presenting features (e.g. common blood test results, symptom combinations, or comorbidity) will also be explored.

  The supervisory team includes healthcare epidemiologists; data scientists; statisticians; data managers and academic GPs. Rich linked data (~2.7M records) are available to the student at the outset to support the project. Example prior work includes: Phenotyping electronic health records at scale (PMID:31329239); assessing risk of different diseases (PMID:34339405); examining the predictive values of symptom combinations (PMID:36702593); formally identifying diseases with greatest earlier diagnosis potential (doi.org/10.3399/BJGP.2023.0044)

  The project will improve the assessment of risk of underlying consequential illness in primary care, thereby improving clinical outcomes and patient experience and helping to update national clinical practice guidelines.

  Eligibility and suitability

  This project requires applicants with high academic potential, with affinity for coding and analysis of complex data, fluency in statistical analysis software and relational databases, and the ability to be trained in statistical techniques. Excellence in written communication is essential. Interest is welcome from candidates who are:

  • working as data analysts (e.g. post-master’s) seeking to study for a PhD
  • or completing a master’s degree in a quantitative field (e.g. data science, statistics)
  • or completing undergraduate studies in a quantitative discipline and have an exceptional academic record.

  Research experience in analysing large longitudinal health datasets such as administrative data, electronic health records and disease registries would be an advantage as would:

  • experience of study design and planning of analysis for successful completed/published research projects
  • relevant analytical work experience in academic research departments, or Public Health England/NHS, or relevant sectors of the health industry.

  Click here for further details and to access the application form.

  • If you have any questions contact y.lyratzopoulos@ucl.ac.uk
• Repurposing and enriching cardiovascular risk prediction model to identify people at risk of cancer

  Key details

  • Hosted by: UCL
  • Lead supervisor: A Floriaan Schmidt, Institute of Cardiovascular Science, University College London
  • Duration: Three years
  • Stipend: UKRI stipend rates apply
  • International fee waivers may be available for the right candidate
  • Start date: Flexible for the right candidate

  Project summary

  Risk-stratified management of cardiovascular disease (CVD), where people without established disease receive preventative interventions and monitoring based on their 10-years predicted risk, has been highly successful to ensure healthcare resources are allocated to those most likely to benefit.

  Through the development of tumour-specific medicines, cancer has been on the forefront of personalised medicine. Preventative strategies for cancer have, however, focussed on costly on-size-fits­-all screening programmes, irrespective of the individual cancer risk. While there have been attempts to prioritise screening and prevention strategies to high risk individuals, such as using prostate-specific antigen, there is a general sparsity of risk-based approached for early detection and prevention of cancers in clinical practice.

  Despite distinct pathways of disease development, and depending on the type of cancer (e.g. lung cancer, colorectal cancer), CVD risk factors also contributed to the development of cancer. For example, smoking and alcohol use are important risk factors for both diseases, as well as sedentary lifestyle, poor diet and environmental factors such as air pollution. Given the interrelation between risk factors for CVD and risk factors for cancer, as well as the high clinical uptake of risk prediction tools for CVD, we wish to explore to what extent established models for CVD prediction can be repurposed and enriched to consider the onset of cancer.

  Eligibility and suitability

  Applicants will need:

  • A background in statistics/epidemiology/computer science or related disciplines.
  • An interest in developing their knowledge of statistics and/or machine learning.
  • The ability to work collaboratively with a team of clinical and statistical experts.

  Experience would also be valuable in:

  • working computationally with tabular data, and healthcare data in particular, is appreciated.
  • programming languages such as Python or R
  • working genomic, metabolomics, or proteomics.

  Click here for further details and to access the application form.

  • If you have any questions contact amand.schmidt@ucl.ac.uk
• BLOod Test Trend for cancEr Detection (BLOTTED): an observational and prediction model development study using English primary care electronic health records data
  • Hosted by: University of Oxford
  • Lead supervisor: Brian Nicholson, Nuffield Department of Primary Care Health Sciences
  • Duration: Three years
  • Stipend: UKRI stipend rates apply
  • No international fee waivers
  • Start date: Flexible for the right candidate

  Project summary

  The PhD student will lead research exploring trends in blood tests for early detection of cancer and assess how the presence of CVD influences trends and cancer risk. Analyses will be conducted using linked national electronic health record data. The student will be responsible for developing and validating prediction models, with a particular focus on methods to utilise longitudinal data. The student will be a member of the multidisciplinary Cancer Research, Medical Statistics, and Cardiovascular Groups at the University of Oxford and Epidemiology of Cancer Healthcare Outcomes Group at University College.

  Eligibility and suitability

  The successful applicant will need:

  • to be educated to degree level (minimum 2:1 at undergraduate level) or equivalent experience
  • experience in conducting/supporting medical research
  • an understanding of basic/fundamental statistics/data science
  • to be motivated and interested in cancer/CVD research, primary care, and linked electronic health data
  • the ability to communicate research findings to a multidisciplinary group
  • the ability to plan, implement and deliver programmes of work
  • high proficiency in the English language.

  It will be an advantage to have:

  • experience in analysing longitudinal data from primary care
  • experience in working in a cancer/CVD-related position
  • familiarity with analytic software, such as Stata, R, or Python
  • experience in developing/validating risk prediction models.

  Click here for further details and to access the application form.

  • If you have any questions contact pradeep.virdee@phc.ox.ac.uk
• Modelling the Impact of Diagnostic Pathways in Cancer and Cardiovascular Disease

  Key details

  • Hosted by: University of Swansea
  • Lead supervisor: Dr Rhiannon Owen, Swansea University Medical School
  • Duration: Three years
  • Stipend: UKRI stipend rates apply
  • No international fee waivers
  • Start date: Flexible for the right candidate

  Project summary

  Many patients who are diagnosed with cancer and/or cardiovascular disease (CVD) receive their diagnosis in Accident and Emergency (A&E). These patients tend to have more severe disease than those who receive a diagnosis from their general practice (GP).

  This PhD project will develop mathematical models to predict what the benefits of receiving an earlier diagnosis (via their GP) would have been for both patients and the NHS. With the frequent co-existence of cancer and CVD, this presents an important opportunity to improve population health, and reduce potential health inequalities, by improving the diagnostic pathway of both diseases.

  The project will specifically explore the epidemiology of cancer and/or CVD diagnoses, especially with respect to where, and for whom, such diagnoses are made, using population-scale data including 2.9 million individuals in the Secure Anonymised Information Linkage (SAIL) Databank Wales Multimorbidity e-Cohort.

  The project will use statistical modelling and machine learning techniques to predict the impact of in-hospital diagnoses for cancer and/or CVD on patient-relevant and NHS outcomes, especially with regards to life-expectancy and quality of life. In doing so, it will provide both a benchmark against which existing diagnostic/pathway initiatives can be evaluated, as well as identifying potential inequalities and predicting the impact of new areas of system development to improve patient outcomes.

  Eligibility and suitability

  Applicants will need an MSc in Statistics/Biostatistics or Epidemiology/Health Data Science (with a strong analytical component) plus programming and data analysis skills/experience in R and/or Python.

  Experience of analysing large-scale linked electronic health record data Knowledge of Bayesian methods would be an advantage.

  Click here for further details and to access the application form.

  • If you have any questions contact R.K.Owen@Swansea.ac.uk or Keith.Abrams@warwick.ac.uk