The number of people living in the UK with diabetes in all its forms is approaching 5 million. The expected lifetime of a diabetic patient is significantly lower than for a non-diabetic patient of similar age and other conditions. The analysis of diabetes-related mortality is complicated by the fact that patients rarely die from diabetes directly – instead, diabetic people have increased risk of death from other diseases. In the last decade, we have seen significant advances in treatments for Type 2 diabetes, which seem to reduce the mortality risk for many patients. However, the effect of these drugs is not yet fully understood, because many of the risk estimates are derived from data which are 10 or more years old. This project aims to develop a deeper understanding of the risks associated with Type 1 or Type 2 diabetes, analyse the risk of medical complications, and the impact of chronic conditions on life expectancy and disability. The effect of recent improved treatments will also be investigated. As a result, we will produce, for all age ranges, mortality tables at a granular level for lives with and without diabetes. We will also produce morbidity tables at a granular level for inceptions of diabetes. It is hoped that the research undertaken in this project will help to support individuals with diabetes. In particular, this will improve access to insurance products for people with diabetes.
This study investigates mortality in individuals with type 1 and type 2 diabetes compared to matched controls without diabetes. Participants will be matched on presence of risk factors, such as age, smoking status, alcohol consumption, body mass index, and other diseases. We will pay special attention to significance of risk factors, separately and in combination. We will also use the control to assess risk factors for progression to type 2 diabetes. Finally, we investigate trends as to how diabetes-related mortality and risk factors have changed over the past 10 years.
We will use a matched cohort study design with three cohorts: Type 1 diabetes, Type 2 diabetes, and controls. After collection of relevant data from CPRD, we will check for outliers, and correct inconsistences. The treatment of missing values requires their classification according to the modern typology of ‘dark data’, finding missing values patterns and corresponding data imputation is the next step. For hypothesis testing, we will use chi-square test, Fisher’s exact test, t-test, KS test, Mann Whitney U test, Yates test, and other tests according to tested hypothesis. For dimensionality reduction and extraction of important features we will combine classical principal component analysis and factor analysis with manifold learning, principal graphs and filament extraction. Some of these technologies were developed with active participation of our team. Original software libraries for these technologies are developed, tested, and applied by different institutions for biomedical data analysis (Institute Curie, Paris, Harvard Medical School, etc.).
We will use a combination of analytical approaches: the descriptive modelling approach will be used to determine the most significant risk factors, while predictive modelling approach will be used to construct the mortality tables.
HES APC and ONS data will be utilised to assess hospital admissions and mortality trends in the three cohorts.
Outcomes to be measured:
Death;
Incidence of type 2 diabetes.
The short answer is: in the mortality analysis the outcome to be measured is death, while in the morbidity analysis the outcome to be measured is probability of developing type 2 diabetes.
In more details, in the mortality analysis, what we will measure/estimate is the probability of death for a diabetes patient and how this probability is affected by factors such as age, gender, smoking status, and all other covariates and risk factors listed in Section "Exposures, Outcomes and Covariates". Also, we will measure/estimate the probability of a non-diabetic person to become diabetic within the next n years, and how this probability is affected by the risk factors.
Bogdan Grechuk - Chief Investigator - University of Leicester
Bogdan Grechuk - Corresponding Applicant - University of Leicester
Alexander Gorban - Collaborator - University of Leicester
Clare Gillies - Collaborator - University of Leicester
Evgeny Mirkes - Collaborator - University of Leicester
Francesco Zaccardi - Collaborator - University of Leicester
Kamlesh Khunti - Collaborator - University of Leicester
HES Admitted Patient Care;ONS Death Registration Data;Patient Level Index of Multiple Deprivation;Practice Level Index of Multiple Deprivation