The American Diabetes Association stated in their 2025 guidelines that retinopathy status should be assessed among patients with diabetes who undergo an intensification of glucose-lowering therapies, including glucagon-like peptide-1 receptors, because of the potential for initial worsening of retinopathy.1 Although ophthalmic screening guidelines for patients with type 2 diabetes mellitus (T2DM) are well-established, screening rates remain low, suggesting there may be underrecognized barriers that contribute to low adherence. A 2-center, retrospective, observational study by the University of California quantified these barriers and identified strong predictors that could inform clinicians in streamlining diabetes-associated eye disease care continuum.2
The Bottom Line
“Interventions should include automated electronic health record prompts; point-of-care screening methods, such as teleophthalmology integration; additional care coordination resources; and targeted health education on the importance of annual screenings for patients and primary care physicians.” — Ayati et al. Diabetes Care. 2025.
The study evaluated electronic health records of patients with T2DM who had a primary care visit between January 2020 through December 2022, the latter serving as the index date, to assess 4 outcomes: screening referral, screening visit completion, receipt of diagnosis for diabetic retinopathy within a year, and treatment within 2 years. Those who had prior diabetes-associated eye disease, recent eye screenings, or present with significant comorbidities were excluded.2
A total of 8240 patients were identified across University of California, Irvine (UCI; n=5648) and University of California, San Francisco (UCSF; n=2592). Patient demographics differed across the 2 study sites, with patients at UCI being younger (53.8 years; standard deviation [SD]12.1) compared with patients at UCSF (57.1 years; SD, 11.9; standardized mean difference [SMD]0.275). Patients at UCI had a higher proportion with prior history of non-diabetes-related eye diseases (n=2591; 45.9%) compared with patients at UCSF (n=885; 34.1%; SMD, -0.241).2
Screening differences between the study sites
Data pooled from both sites revealed that only 16% of patients underwent a diabetes-associated eye disease screening, despite the 3-fold higher proportion of patients who received a screening referral within 1 year of the index date (43%). More patients at UCSF received a screening visit referral and completed the screening appointment than patients at UCI (54% vs 37% and 21% vs 13%, respectively). However, the number of patients who received a diagnosis for diabetes-associated eye disease and received treatment for it were similar across both UCI and UCSF (5% vs 4% and 1% vs ~1%, respectively).2
Insurance coverage influenced odds of screening at 1 year
Both nonclinical and clinical factors were evaluated as predictors of a screening completion at 1 year. In a pooled data analysis, patients enrolled in Medicare had the highest odds of completing a screening visit at 1 year (OR, 2.6; 95% CI, 2.3-2.9; P<.001), followed by commercial insurance (OR, 1.6; 95% CI, 1.4-1.8; P<.001), and Medicaid (OR, 0.5; 95% CI, 0.4-0.5; P<0.001). Patients who self-paid had the lowest odds (OR, 0.1; 95% CI, 0.0-0.1; P<.001). The type of provider that patients saw also affected their odds in screening completion at 1 year. Patients who saw an internist were at highest odds of completing their screening at 1 year (OR, 2.8; 95% CI, 2.5-3.2; P<.001) compared to patients who saw a primary care provider (OR, 1.6; 95% CI, 1.4-1.8; P<.001) or family medicine physician (OR, 0.3; 95% CI, 0.2-0.3; P<.001).2
The strongest predictor of screening visit completions?
Most clinical factors, such as hemoglobin A1c (HbA1c) level, body mass index (BMI), and comorbidities demonstrated comparable associations with a patient’s completion of screening visit (OR, 0.9; 95% CI, 0.9-1.0; P<.001; OR, 1.0; 95% CI, 1.0-1.0; P<.001; and OR, 1.1; 95% CI, 1.1-1.2; P<.001, respectively). These associations were generally consistent between the 2 study sites.2
However, receipt of a screening referral within 1 year and a history of other eye diseases were the 2 strongest predictors with large variability between the 2 study sites. Overall, patients who received a screening referral within 1 year had nearly 57-fold higher odds (OR, 56.7; 95% CI, 42.1-76.4; P<.001) of screening completion, while patients who had a history of other eye diseases had nearly 7-times higher odds (OR, 6.5; 95% CI, 5.6-7.4; P<.001). When analyzing data specific to each study site, UCSF patients who had received a screening referral within 1 year and a prior history of other eye diseases had nearly 22-fold higher odds (OR, 21.9; 95% CI, 15.1-31.9; P<.001) and nearly 3-fold higher odds (OR, 2.7; 95% CI, 2.2-3.3; P<.001) of completing their screening visit, respectively. On the other hand, UCI patients had 126.7-times (95% CI, 75.7-212.1; P<.001) and 40.7-times (95% CI, 28.0-59.3; P<.001) higher odds, respectively, of completing their screening visit at 1 year.2
Taken together, these findings mirrored current literature, which demonstrates low adherence to recommended eye screening among patients with T2DM, particularly among underserved communities. Variability in the strengths of identified predictors—specifically referral and eye disease history—between the 2 sites, were likely influenced by the difference in socioeconomic distribution between the surrounding areas of each site. UCI is a center that serves a significantly higher proportion of underserved populations, while UCSF is a tertiary institution with considerably more resources.2
Interventions to improve outcomes may be multifaceted
This study leveraged real-world data that gave a glimpse of real patient and provider behaviors as well as associated factors that may influence their odds to complete a diabetic eye screening; however, considering its retrospective design, there are limitations inherent to the study, including potential for misclassification of relevant information in a patient’s health records that could skew measured outcomes. Additionally, approximately 15% of patients were screened outside of the 2 study sites, which could have resulted in underestimation of true screening rates. Nevertheless, these findings could inform intervention development to improve screening rates and, subsequently, patient outcomes. The authors posited, “Interventions should include automated electronic health record prompts; point-of-care screening methods, such as teleophthalmology integration; additional care coordination resources; and targeted health education on the importance of annual screenings for patients and primary care physicians.”2
Published: February 05, 2026
Christella Gordon-Kim is a medical writer with a special interest in neurodegeneration and oncology. She earned her PhD in biomolecular chemistry at Emory University.
