Cambridge Diabetes Risk Score
Most diabetes risk scores are validated against an outcome of clinically diagnosed diabetes rather than using a design in which the whole study population undergoes the diagnostic test. This leads to bias, as the variables included in risk scores are related to the patient characteristics likely to encourage doctors to undertake diagnostic testing. The Cambridge Diabetes Risk Score is validated in populations that underwent diagnostic testing to reduce this bias.
Please fill out required fields.
From the Creator
Why did you develop the Cambridge Diabetes Risk Score? Was there a particular clinical experience or patient encounter that inspired you to create this tool for clinicians?
In the mid-1990s, it was believed that 50% of type 2 diabetes remained undiagnosed, people could have diabetes for up to 12 years before they were diagnosed, and [they] often presented with complications at the point of diagnosis. At the time, I was working on a trial among patients with newly-diagnosed type 2 diabetes. The data collected by practice nurses for this trial suggested to me that these newly-diagnosed patients exhibited characteristics that might help us to identify people with undiagnosed disease. Furthermore, these simple characteristics were either readily available in the medical records or could be easily obtained without the need for expensive medical consultations or blood tests.
What pearls, pitfalls and/or tips do you have for users of the Cambridge Diabetes Risk Score? Do you know of cases when it has been applied, interpreted, or used inappropriately?
The Cambridge diabetes risk score was originally designed to identify people at high risk of having undiagnosed diabetes. However, it also performs reasonably well at predicting who will develop diabetes and cardiovascular disease, and experience premature mortality. As with most such scores, it is based on a logistic regression equation, hence the score, from 0 to 1, does not by itself provide a measure of absolute risk over a 10 year period, unlike, for, example the Framingham or QRisk cardiovascular scores. So a diabetes risk score of 0.7 does not imply a 70% lifetime or 10 year risk of diabetes.
What recommendations do you have for doctors once they have applied the Cambridge Diabetes Risk Score? Are there any adjustments or updates you would make to the score based on new data or practice changes?
The score helps to stratify the population so that subsequent testing or preventive interventions can be targeted at those at greatest risk. It can be used as the first step in a population screening or prevention programme. Depending on available resources, those in the higher distribution of the risk score might be offered a diagnostic blood test for diabetes such as glycated hemoglobin (HgbA1c) and advice about how to reduce their risk.
How do you use the Cambridge Diabetes Risk Score in your own clinical practice? Can you give an example of a scenario in which you use it?
In the English NHS Health Checks and Diabetes Prevention Programmes, a similar score developed in Leicester around 10 years after the one that I originally reported is now being used to identify those at high risk of diabetes.
According to the score, steroid use increases risk of T2DM. Is this because steroid-induced diabetes was included with T2DM in your analysis?
We did not distinguish between the different potential underlying causes of diabetes. In our original analysis, the prescription of steroids was significantly associated with the presence of diabetes. Presumably, this is because the underlying indication for the prescription, or the steroids themselves increase the risk of diabetes. We included any variable that increased the predictive utility of the model, irrespective of whether we understood the underlying mechanism. Unsurprisingly, the obvious known risk factors such as age and body mass index were the variables that most strongly predicted the presence of diabetes.
Any other research in the pipeline that you’re particularly excited about?
We are using similar approaches to develop and validate simple risk scores for cancers such as colorectal and prostate. As with diabetes and cardiovascular disease we are finding that these scores perform surprisingly well and usually rather better than genetic tests.
About the Creator
Simon J. Griffin, DM, is professor of general practice at the University of Cambridge, Group Leader in the MRC Epidemiology Unit and an assistant general practitioner at Lensfield Medical Practice in Cambridge, UK. He leads the Prevention of Diabetes and Related Metabolic Disorders Programme. Professor Griffin’s research interests include prevention and early detection of chronic conditions such as diabetes.
To view Dr. Simon J. Griffin's publications, visit PubMed