Joint Modelling of Incidence of Diabetic retinopathy and Change in fasting blood sugar among diabetes mellitus patients at referral hospitals in Amhara region, Ethiopia

Abstract
Background
Diabetes mellitus is a global public health problem characterized by an elevated level of blood glucose, and fasting blood sugar (FBS) is an important test. It helps track progress and can delay or prevent diabetes-related complications. Over time, uncontrolled diabetes can damage the blood vessels in the eyes and cause permanent vision loss. In resource-limited settings like Ethiopia, joint modeling that assesses the association between longitudinal biomarkers and the incidence of diabetic retinopathy (DR) is rarely considered. Therefore, this study was aimed to assess the incidence and predictors of DR and its associations with FBS.
Methods
A multicenter retrospective follow-up study was conducted at a referral hospital in the Amhara region, Ethiopia. A total of 462 newly diagnosed DM patients were selected using a computer-generated, simple random sampling technique. The proportional hazard assumption was checked for the survival sub-model, and for the longitudinal sub-model, the normality assumption was checked. Then the joint model with time-dependent lagged parameterizations was fitted. For the joint model, the goodness of fit was checked using the Cox-Snell residual test, and model comparison and selection were made based on AIC and BIC values. Finally, the hazard ratio with a 95% confidence interval (CI) with a corresponding P-value <0.05 was used.
Results: A total of 462 diabetes patients were followed for 23111.00 person-months of observation. Overall, 54 patients developed diabetic retinopathy, with an incidence rate of 2.33 per 1000 person-months with a 95% CI: [1.78, 3.05]. Rural residence (AHR = 2.21, 95% CI: [1.21, 4.05]), having hypertension (AHR = 3.01, 95% CI: [1.85, 6.53]), and duration of DM >5 years (AHR = 2.28, 95% CI: [1.91, 5.15]) were significant predictors of DR. The time-dependent lagged value of fasting blood sugar change was significantly associated with an incident of DR (AHR = 4.20, 95% CI [1.62, 10.85]).

Conclusion: The incidence of diabetic retinopathy was relatively high when compared with previous similar studies in Ethiopia. Besides, rural residence, having hypertension, a duration of DM greater than five years, and longitudinal changes in FBS had significant associations with incidents of DR.
Key words:  Diabetic retinopathy, fasting blood sugar, diabetes mellitus, joint modeling, Ethiopia.