Clinical Burden and Predictors of Diabetic Retinopathy in Conflict-Affected Yemen: A Cross-Sectional Study

Background The ongoing humanitarian crisis in Yemen has severely disrupted healthcare delivery for patients with diabetes mellitus (DM). This study characterized the severity distribution of diabetic retinopathy (DR) among referred patients and explored the relative contribution of disease duration versus current glycemic control in predicting proliferative DR (PDR) in a cohort with chronically poor metabolic control. Methods A hospital-based cross-sectional analysis was conducted at Al-Thawra General Hospital, Sana’a. Patients with type 2 diabetes mellitus (T2DM) referred for ophthalmic evaluation underwent a comprehensive fundus examination. DR was staged using the International Clinical Diabetic Retinopathy (ICDR) Scale. Binary logistic regression was used to identify independent predictors of PDR. Receiver operating characteristic (ROC) curves were constructed, and DeLong’s test was used to compare the discriminatory performance of the significant predictors. Results Among 202 referred patients (mean age 64.6 ± 13.3 years; 96.0% with HbA1c > 7.0%), 66.3% had vision-threatening DR (Severe NPDR, PDR, or DME). The cohort included PDR (27.2%), Severe NPDR (27.7%), Moderate NPDR (19.8%), Mild NPDR (13.9%), and isolated DME (11.4%). In the univariate analysis, PDR was significantly associated with older age (p = 0.009), longer diabetes duration (p = 0.001), and higher HbA1c levels (p = 0.036). However, in the multivariate model, only diabetes duration remained an independent predictor (adjusted OR: 1.072 per year; 95% CI: 1.008–1.140, p = 0.027), while HbA1c did not (p = 0.505). The model explained 11.0% of the variance in PDR status (Nagelkerke R²). ROC analysis showed that diabetes duration (AUC = 0.643, 95% CI: 0.563–0.722) and HbA1c (AUC = 0.603, 95% CI: 0.515–0.691) had statistically similar discriminatory performance (DeLong’s test: p = 0.400). Stratified analysis identified males aged 16–25 years with diabetes as having the highest PDR rates (44.0%). Conclusions In this conflict-affected cohort with near-universal poor glycemic control, cumulative disease duration was a more robust independent predictor of PDR than a single cross-sectional HbA1c measurement. These findings support the use of duration-based risk stratification for DR screening in humanitarian settings where laboratory resources are scarce. Prospective studies are required to confirm these observations.