Abstract Background & objectives: One of the most important complications of diabetes is diabetic retinopathy. It is the leading cause of blindness in persons aged 20–74 years. The identification of risk factors associated with the development of retinopathy is important for early screening and intervention. The study attempts to compare of generalized additive models (GAM) and generalized linear models (GLM) for estimating the retinopathy risk factors for diabetic patients. Material & Methods: This cross-sectional study was carried out on 367 diabetic patients. Variables include age, duration of diabetes, BMI, hemoglobin A1C, cholesterol, and systolic blood pressure. All participants were checked to determine existence of retinopathy by eye examination. Fitting GLM and GAM were done with R software. Results: 120 cases (33%) were retinopathy patient and 247 cases (67%) were not. Results of the generalized linear model showed factors had effects on retinopathy including: duration of diabetes (p<0/001), hemoglobin A1C (p<0/001) and systolic blood pressure (p=0/018).while GLM can in addition to determination of retinopathy risk factors, identify nonlinear relationship between variables. Conclusion: To determine retinopathy risk factors, GLM is only able to discover linear relationship between variables. While GAM increases the quality of predicting response variable with more information of the data relationships. This quality of predicting has been show by decrease of mean square error and Hosmer Lemeshow test.
Rights and permissions | |
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. |