Validation of a Diagnostic Support System for Diabetic Retinopathy Based on Clinical Parameters

Abstract

To validate a clinical decision support system (CDSS) that estimates risk of diabetic retinopathy (DR) and to personalize screening protocols in type 2 diabetes mellitus (T2DM) patients. Methods: We utilized a CDSS based on a fuzzy random forest, integrated by fuzzy decision trees with the following variables: current age, sex, arterial hypertension, diabetes duration and treatment, HbA1c, glomerular filtration rate, microalbuminuria, and body mass index. Validation was made using the electronic health records of a sample of 101,802 T2DM patients. Diagnosis was made by retinal photographs, according to EURODIAB guidelines and the International Diabetic Retinopathy Classification. Results: The prevalence of DR was 19,759 patients (19.98%). Results yielded 16,593 (16.31%) true positives, 72,617 (71.33%) true negatives, 3165 (3.1%) false positives, and 9427 (9.26%) false negatives, with an accuracy of 0.876 (95% confidence interval [CI], 0.858-0.886), sensitivity of 84% (95% CI, 83.46-84.49), specificity of 88.5% (95% CI, 88.29-88.72), positive predictive value of 63.8% (95% CI, 63.18-64.35), negative predictive value of 95.8% (95% CI, 95.68-95.96), positive likelihood ratio of 7.30, and negative likelihood ratio of 0.18. The type 1 error was 0.115, and the type 2 error was 0.16. Conclusions: We confirmed a good prediction rate for DR from a representative sample of T2DM in our population. Furthermore, the CDSS was able to offer an individualized screening protocol for each patient according to the calculated risk confidence value.

A Classification Tree Model with Optical Coherence Tomography Angiography Variables to Screen Early-Stage Diabetic Retinopathy in Diabetic Patients

“…is finding might suggest that a long DM duration does not necessarily accelerate DM development because DM development is also closely related to blood glucose control and other factors [18]. Moreover, the percentage of patients with a DM duration of not less than 18.5 was small, and the difference may not be obvious.…”

“…In previous studies on DR screening [18][19][20], demographical information, disease information, and biochemical indicators were major variables in DR prediction models. In our study, DM duration was included in the classification tree model but not included in the logistic regression model because of a collinearity between DM duration and BCVA existed in the regression analysis.…”

“…Therefore, a classification tree model seems to be more appropriate because it can keep more clues and evidence to support a diagnosis. Moreover, in other studies on DR screening [ 18 , 19 ], the risk variables for the occurrence of DR were diverse. The alterations among these models or studies imply that the construction of a statistical model may be affected by a series of factors.…”

“…Ethnicity may have contributed to the difference. DM duration, insulin treatment, HbA1c level, and microalbuminuria were major risk contributors in a Spanish study [ 18 ], while the presence of hypertension, DM family history, and low frequency of physical activity were risk factors in a Danish study [ 22 ]. Besides, our study was a hospital-based study, and the included patients might have more serious and complicated conditions than patients in communities or regions outside hospitals.…”

“…Shen and his colleagues [ 23 ] proposed a model using the XGB-Stacking algorithm based on improved backward search, and the highest accuracy was 83.95%. Similarly, in Romero–Aroca's study [ 18 ], a clinical decision support system based on a fuzzy random forest was utilized and an accuracy of 87.6% was obtained. Although proper models have been established using advanced algorithms, these models solely focus on the systemic conditions of T2DM patients.…”