Predicting the epidemic: a study of diabetes risk profiling in a multi-ethnic inner city population

Abstract

Type 2 diabetes has increased in prevalence globally in recent years, mainly due to obesity. Many other risk factors are well known. Identifying those at high risk of type 2 diabetes may guide targeted interventions aimed at reducing risk. Type 2 diabetes risk prediction is a complex science. The first half of this thesis presents a quantitative and qualitative systematic review of 145 risk prediction models and scores. Many are available; few are usable in real life clinical practice. Seven have high potential to be used with routine data (such as electronic primary care records). The second half of this thesis describes the use of one of the risk prediction scores locally, the QDScore, on a dataset of 519,288 electronic primary care records in East London, UK to calculate the ten year risk of developing type 2 diabetes. Ten percent of the population were at high risk (defined as a ten year risk of greater than 20%). Ethnicity and deprivation were key factors responsible for increasing risk, and there was overlap with cardiovascular morbidity. A sub-section of these data were mapped to explore the feasibility of using geospatial mapping to convey the risk of non-communicable disease in a public health setting. Previous research has focussed on targeting individuals with pre-diabetes (e.g. Impaired Fasting Glucose) and screening for undiagnosed diabetes. Going a step further back and identifying those at risk of type 2 diabetes is theoretically possible due to the wide availability of prediction algorithms, and such an approach is potentially achievable locally using electronic primary care records. This produces important descriptive data 3 to aid the interventions of general practitioners, public health specialists and urban planners. Future research should focus on interventions which reduce risk of type 2 diabetes in otherwise healthy adults.