Diabetes is a common condition that causes a person's blood sugar level to become too high. There are many different forms of diabetes. It is important to get the diagnosis right to ensure patients get the right treatment. The most well-known forms are Type 1 diabetes, which needs life-long treatment with insulin injections, and Type 2 diabetes, which is usually treated with diet or tablets. There are also rarer types such as Maturity Onset Diabetes of the Young (MODY; a genetic form of diabetes) and Type 3c (where the pancreas, which produces insulin, becomes damaged). There are clear treatments that work for MODY patients, but for Type 3c, there is very little guidance.
Getting the correct diagnosis can be challenging. Around 7-15% of people with diabetes may be misdiagnosed with the wrong type. For people who actually have MODY or Type 3c, around 80% are misdiagnosed with other forms of diabetes.
Approaches, such as clinical calculators (https://www.diabetesgenes.org/exeter-diabetes-app/) have been developed to help with diabetes classification, but these are mostly used by diabetes specialists. Their utility in GP practice has not been assessed.
This project aims to explore how different approaches for improving diabetes classification could work by testing them in GP records, and will complement another project working with GPs to see how these approaches could be implemented in real life. Better approaches to diagnosing diabetes subtypes will have clear benefits for patients in ensuring they get the right diagnosis, and therefore the right treatment and management for their condition.
Background: There are a number of different diabetes subtypes. Ensuring patients receive the correct diagnosis is essential to them receiving appropriate treatment and care. However, misclassification is estimated to occur in 7-15% of cases, and is an even greater problem for rarer subtypes (e.g. MODY is estimated to be misdiagnosed in ~80% of cases). In line with the 2022 NICE Recommendations for Research, we have been developing approaches that use routinely available clinical features for distinguishing between different subtypes of diabetes, including clinical criteria and prediction models. These approaches work well in research data, but, to date, we have not assessed their performance in primary care.
Aim: We will analyse CPRD to explore the utility of features and models for misclassification of diabetes in primary care, and to explore their potential impact
Study population: All patients with diabetes from 2004 to date
Exposures: Clinical features routinely recorded in primary care e.g. demographics, blood test results
Outcomes: Diabetes subtype and measures relating to diabetes care e.g. glycaemic response, treatment, healthcare utilisation
Data sources: CPRD Aurum, Hospital-episode statistics, ONS death data
Methods: Analysis will be largely descriptive exploring coding of the data to enable prediction models to be implemented and examining the potential impact of running the models e.g. proportions flagged as misclassified by different approaches and the characteristics and clinical outcomes of these individuals.
Intended benefits: This work will complement an NIHR funded project where we will work with GPs to test out how the clinical models could be implemented as automated searches of their patient records to flag patients who may be misclassified and need further investigation. Helping patients get the right diagnosis for their diabetes will ensure they get the optimal treatment and management for their condition.
Diabetes subtype - as coded by the GP and as predicted by the different classification approaches.
We will also explore the potential impact of misclassification by comparing the outcomes in those flagged as misclassified to those not flagged as misclassified including:
- Glycaemic control (HbA1c);
- Healthcare utilisation (GP visits, hospitalisation);
- Treatment (e.g. time to requiring insulin, number of tablets);
- Development of acute and chronic diabetes related complications
Beverley Shields - Chief Investigator - University of Exeter
Katherine Young - Corresponding Applicant - University of Exeter
Andrew Hattersley - Collaborator - University of Exeter
Andrew McGovern - Collaborator - University of Exeter
Angus Jones - Collaborator - University of Exeter
John Dennis - Collaborator - University of Exeter
Julieanne Knupp - Collaborator - University of Exeter
Nicholas John Thomas - Collaborator - University of Exeter
Pedro Cardoso - Collaborator - University of Exeter
Rhian Hopkins - Collaborator - University of Exeter
Timothy McDonald - Collaborator - University of Exeter
Trevelyan McKinley - Collaborator - University of Exeter
HES Admitted Patient Care;ONS Death Registration Data;Patient Level Index of Multiple Deprivation