Predicting diabetes-related complications with machine learning techniques

Supervisors

• Dr Sharmin Shabnam (ss1279@le.ac.uk)
• Dr Francesco Zaccardi  
• Professor Kamlesh Khunti

Project description

Diabetes mellitus is characterised by chronic hyperglycaemia, which is associated with a higher risk of cardiovascular complications. Regular monitoring, management, and control of risk factors, such as glycated haemoglobin (HbA1c), blood pressure, and lipids within the recommended range, are critical in helping individuals maintain overall health and well-being and reducing the risk of further complications. Continuous monitoring generates a large amount of intra-individual longitudinal observations of blood glucose levels which can be used to track disease progression and predict diabetes-related complications. 
Recently, the rapid development of machine learning methods has resulted in their applications in various areas of healthcare-related research. This PhD project aims to apply different statistical models and machine learning (ML) algorithms (including classification and regression trees, support vector machines (SVM), k-nearest neighbour, gradient boosting machines, and supervised principal component analysis) to predict various diabetes-related complications and develop a risk stratification system that categorises patients with diabetes into risk groups, enabling personalised interventions and treatment plans to mitigate complications. The post holder will undertake different statistical analyses using the Clinical Practice Research Datalink (CPRD) database, which includes anonymized patient data from a network of GP practices across England, to identify key features (i.e., age, gender, ethnicity, and diabetes duration) which contribute to the risk of diabetes complications.

The student will be embedded within a team of experts in clinical diabetes, epidemiology, data science, and statistics, and receive training in a broad range of ML and statistical methods used to investigate cross-sectional and longitudinal real-world data, as well as methods for prognostic research (development and validation of predictive models) using ML and statistical modelling approaches.

The Ph.D. project will be integrated into a vibrant postgraduate research community within the Real-World Evidence Unit and the Diabetes Research Centre, University of Leicester, and help advance the aims of the National Institute of Health and Care Research Leicester Biomedical Research Centre (BRC) and East Midlands Collaboration for Leadership in Applied Health Research and Care (ARC).

References

• Zhang, L., Shang, X., Sreedharan, S., Yan, X., Liu, J., Keel, S., ... and He, M. (2020). Predicting the development of type 2 diabetes in a large Australian cohort using machine-learning techniques: longitudinal survey study. JMIR medical informatics, 8(7), e16850.
• Lai, H., Huang, H., Keshavjee, K., Guergachi, A., and Gao, X. (2019). Predictive models for diabetes mellitus using machine learning techniques. BMC endocrine disorders,19(1), 1-9.
• Gray, L. J., and Khunti, K. (2013). Type 2 diabetes risk prediction—Do biomarkers increase detection? Diabetes research and clinical practice, 101(3), 245-247.
• Abbasi, A., Peelen, L. M., Corpeleijn, E., Van Der Schouw, Y. T., Stolk, R. P., Spijkerman, A. M., ... and Beulens, J. W. (2012). Prediction models for risk of developing type 2 diabetes: systematic literature search and independent external validation study. BMJ 345.
• Collins, G. S., Mallett, S., Omar, O., and Yu, L. M. (2011). Developing risk prediction models for type 2 diabetes: a systematic review of methodology and reporting. BMC medicine,9(1), 1-14.