Study of risk assessment in diabetic retinopathy patients based on discriminant analysis

Abstract

Objective To investigate the clinical application value in the risk assessment of diabetic retinopathy using a receiver operating characteristic curve and discriminant analysis.Methods This was a cross-sectional study.The outcomes of correlated clinical and biochemical examinations were obtained in 250 cases with type 2 diabetes mellitus (100 cases with retinopathy and 150 cases without retinopathy as the control).Standard measurements were obtained with an ophthalmoscope and fundus fluorescein angiography.A receiver operating characteristic (ROC) curve was used to assess their value in the diagnosis of diabetic retinopathy and formulate a diagnostic discrimination model with Bayes's discriminant analysis.The correlated indexes of diabetic retinopathy were endothelin-1,vascular endothelial growth factor,magnesium,correlation indexes from renal and pancreas function,hemorheology indexes,and blood fat.Results ROC analysis showed that the diagnostic reliability of ET-1 was the best indicator in the mild stages of NPDR and PDR (AUCmild NPDR=0.742,AUCPDR=0.857); the diagnostic reliability of urinary albumin was the best indicator in the moderate and severe stages of NPDR (AUCmoderate and severe NPDR=0.742). The samples rechecked by the model showed a coincidence rate of 96.4％ in the mild stage of NPDR,and 100％ in the moderate and severe stages of NPDR and PDR.Conclusion The diagnostic discrimination model may be used to screen and monitor the high-risk group with diabetic retinopathy in type 2 diabetes mellitus. Key words: Diabetic retinopathy ; Receiver operating characteristic analysis ; Discriminant analysis ; Risk assessment