Coronavirus (Covid-19)

Advice for people with diabetes and their families

Savefor later Page saved! You can go back to this later in your Diabetes and Me Close

Predicting Type 1 diabetes with machine learning

Project summary

Many people with type 1 diabetes are diagnosed when they have diabetic ketoacidosis (DKA) – a serious and sometimes fatal condition involving consistently high blood glucose levels. Using machine learning, Julia Townson wants to develop a tool to spot people with undiagnosed type 1 diabetes much earlier. Early diagnosis has many benefits, including helping people produce their own insulin for longer, and preventing DKA.

Background to research

Having undiagnosed type 1 diabetes can lead to diabetic ketoacidosis (DKA) – a serious condition arising from having consistently high blood glucose levels. It can often lead to hospitalisation, with some cases leading to coma or death. DKA can be avoided, but some people don’t realise they have type 1 diabetes until they are very unwell with DKA. This can be particularly dangerous for children.

In a previous research project, Julia Townson and her team collected GP records from children both with and without type 1 diabetes. The team found that children who developed type 1 diabetes were more likely to have had a GP appointment in the 12 months before being diagnosed. They were also more likely to have had a GP appointment six months before their diagnosis for urinary infections, bedwetting or urinating a lot. 

Research aims

Building on her previous research, Julia Townson wants to create a predictive tool to identify children who might have undiagnosed type 1 diabetes. Using machine learning, the team will develop the tool to detect patterns in children’s GPs appointment records before they were diagnosed with Type 1 diabetes.

The team will test how well the tool works using health records from hospitals and GP practices in Wales, to see if it can correctly identify children who are already diagnosed with type 1 diabetes. They will then test the tool again on a different set of data, using similar information from hospital and GP records in England.

Potential benefit to people with diabetes

If successful, this research could lead to a tool to ensure children with type 1 diabetes are diagnosed as early as possible. An early diagnosis, and therefore earlier insulin treatment, is beneficial for two reasons: it may help to protect a child’s own ability to produce insulin and reduce the risk of developing serious DKA.

Brand Icons/Telephone check - FontAwesome icons/tick icons/uk