Derivation and validation of a simple risk score for undiagnosed diabetes for Tanzania and other African populations

Abstract

Background: Diabetes prevalence is increasing worldwide. Increased burden of diabetes and recent developments in treatment and prevention of diabetes and cardiovascular complications present opportunities for screening of people at risk of diabetes in order to implement disease-modifying intervention and prevent long-term complications. Aim: The aim of this study was to develop a simple inexpensive score for identifying individuals with undiagnosed diabetes in the African context. Methods: A population based sample of 5193 individuals aged 15 years and above from diabetes surveys in Tanzania, Senegal and Guinea was used to develop the score, the derived score was then validated in populations from South Africa, Guinea and Tanzania. New cases of diabetes were defined using fasting glucose measurements. Binary logistic regression model coefficients were used to assign individual scores for the predictor variables in the model. Results: Age, hypertension and waist circumference were the variables included in the final model. The model has an area under the ROC curve (AUC) of 0.83 (95% CI 0.82 to 0.84). A meta-analysis of applying the score at individual country data yielded a summary ROC curve with an AUC of 0.8 (95% CI 0.74-0.85) and an inconsistency score (I²) of 0%. The performance of the newly derived risk score in the validation samples was comparable to the performance in derivation study population with Area under the ROC curve ranging from 0.7 to 0.82. Conclusion: Presented in this thesis is the first ever diabetes risk score derived from Africa. It is a simple inexpensive tool for identifying individuals with undiagnosed diabetes in African settings. Further work is needed to externally validate the score in other populations