Background Type 2 diabetes is common, costly and often goes unrecognised fo
r many years. When patients are diagnosed, the majority exhibit associated
tissue damage or established cardiovascular risk. Evidence is accumulating
that earlier detection and management of diabetes and related metabolic abn
ormalities may be beneficial. We aimed to develop and evaluate a score base
d on routinely collected information to identify people at risk of having u
ndetected diabetes.
Methods A population-based sample of 1077 people, aged 40 to 64 years, with
out known diabetes, from a single Cambridgeshire general practice, underwen
t clinical assessment including an oral glucose tolerance test. In a separa
te 12-month study, 41 practices in southern England reported clinical detai
ls of patients aged 40 to 64 years with newly diagnosed Type 2 diabetes. A
notional population was created by random selection and pooling of half of
each dataset. Data were entered into a regression model to produce a formul
a predicting the risk of diabetes. The performance of this risk score in de
tecting diabetes was tested in an independent, randomly selected, populatio
n-based sample.
Results Age, gender, body mass index, steroid and antihypertensive medicati
on, family and smoking history contributed to the score. In the test popula
tion at 72% specificity, the sensitivity of the score was 77% and likelihoo
d ratio 2.76. The area under the receiver-operating characteristic curve wa
s 80%.
Conclusions A simple score, using only data that are routinely collected in
general practice, can help identify those at risk of diabetes. This score
could contribute to efficient earlier detection through case-finding or tar
geted screening. Copyright (C) 2000 John Wiley & Sons, Ltd.