Objective       To establish a risk prediction model for progression of microalbuminuria to proteinuria in diabetic nephropathy

and to evaluate its value for type 2 diabetes patients with early renal damage.       Methods       Totally 256 patients with confirmed type 2 diabetes who were inpatients or outpatients in our hospital from January to June 2012 were enrolled in this study. According to estimated glomerular filtration rate (eGFR) installments and urine albumin quantitative analysis

they were divided into proteinuria group (n=73) and early diabetic kidney damage group (n=183). The clinical data of the patients were recorded to analyze the main factors for the progression of diabetic nephropathy using the Exhaustive CHAID classification tree algorithm. The value of the established model was evaluated by the Risk statistics and index map.       Results       The model had 3 stratum and 11 nodes. There were 4 explanatory variables screened out in the model

that is

CysC level

history of hypertension

waist-hip ratio and albumin level. The most important risk factors were CysC level and waist-hip ratio. The risk value of misclassification probability of the model was 0.141

and the classification tree model fitted the actuality very well.       Conclusion      Classification tree model can not only properly predict the progression of microalbuminuria to proteinuria in diabetic nephropathy

but also reveal the complex interaction effects among the factors.