CoChat
Sign Up
Cited by 3Retinal Imaging and AnalysisOphthalmology and Visual Health ResearchRetinal Diseases and Treatments

Autonomous AI-assisted diabetic retinopathy screening at primary care is associated with increased presentation to eye care by at risk patients

Open access
Leong, Ariel, Wolf, Risa M., Channa, Roomasa, Wang, Jiangxia, Lehmann, Harold, Abramoff, Michael D., Liu, T. Y. Alvin·2026·npj Digital Medicine
DOI 10.1038/s41746-026-02460-5PMC PMC13077085
Open this paper in CoChat — free accountOpen PDF

Abstract

Adult patients with diabetes (n = 3745) seen at Johns Hopkins Medicine primary care sites were referred to the Wilmer Eye Institute either based on a primary care provider referral or autonomous AI diagnostic result (referral was made after a positive or non-diagnostic result). An inverse-probability-weighted regression, which incorporated propensity score matching on social determinants of health and relevant clinical variables, showed that implementation of an autonomous AI-assisted diabetic screening program in a primary care clinic was associated with increased presentation to eye care specialist care by African-Americans (p = 0.02). This is significant because African-Americans have traditionally been less likely to undergo annual screening exams and more likely to present with more severe forms of diabetic retinopathy (DR). The results suggest a potential association between office-based, AI-assisted DR screening and improved downstream ophthalmic access for African-American patients. However, given that the analysis was exploratory, this association should be interpreted cautiously and further validated.

AI snapshot

Written from the paper’s full text.

TL;DR

In a retrospective study of 3,745 patients with diabetes referred from Johns Hopkins Medicine primary care to the Wilmer Eye Institute, autonomous AI-assisted diabetic retinopathy screening was associated with increased downstream eye care presentation among African-American patients. The association was exploratory and non-causal, with no significant difference for Medicaid coverage.

Key findings

  • Among referred patients, the AI group (n = 393) was more likely to be African-American than the eye care provider referral group (n = 3,352): 64.9% vs 44.4%, p < 0.001.
  • After inverse-probability-weighted regression with propensity score matching, the AI group remained more likely to be African-American (OR = 1.15, 95% CI: 1.02–1.29, p = 0.022), while there was no significant difference in probability of Medicaid coverage (OR = 0.97, 95% CI: 0.93–1.02, p = 0.245).
  • The AI group was also more likely to speak English as a preferred language (97.7% vs 91.9%, p < 0.001), have hypertension (89.6% vs 82.6%, p = 0.001), and have chronic kidney disease (26.2% vs 20.9%, p = 0.017), and less likely to be Hispanic or Latino (3.3% vs 8.6%, p < 0.001).
  • The paper reports the association as exploratory and hypothesis-generating, with no causal inference.

Methods

  • Retrospective case-control study of patients with diabetes receiving primary care at Johns Hopkins Medicine and referred to the Wilmer Eye Institute between August 2020 and September 2022; patients were excluded if their primary care visit was more than 12 months before the Wilmer visit.
  • Referral pathways were eye care provider (ECP) referral by a primary care provider, or autonomous AI pathway after AI-assisted DR screening at primary care yielded a positive or non-diagnostic result.
  • Comparisons used Mann-Whitney U, Student’s t, and chi-squared tests; inverse-probability-weighted regression and propensity score matching adjusted for demographic and systemic health variables, with robust standard errors clustered at the site level and Bonferroni correction yielding a p-value cutoff of ≤0.025.

Limitations

  • Stated by the paper: retrospective design; single healthcare system serving a predominantly urban population; inability to account for patients who sought ophthalmologic care outside Johns Hopkins; unknown proportion of eligible patients who declined AI-assisted DR screening; differences in PCP referral behavior; small Medicaid numbers limiting meaningful inference; and possible partial attribution to office-based screening independent of whether autonomous AI was involved.
  • Evident from the text: the analysis is described as exploratory and hypothesis-generating, so no causal inference can be drawn and the association should be interpreted cautiously and further validated.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1038/s41746-026-02460-5

Cites

31 of 31 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

35 of 53 cited sources checked · 3 supported · 1 flagged · accuracy not yet measured

  • Premisecitation

    In the previous study's subgroup analysis, the improvement in adherence to annual DR evaluation was particularly prominent in African-Americans and patients covered under Medicaid.

    “In our sub-group analysis, the improvement was particularly prominent in African-Americans and patients covered under Medicaid.”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites [28] Huang, J. J. et al. Autonomous artificial intelligence for diabetic eye disease increases access and health equity in underserved populations. Npj Digit. Med. 7, 196 (2024). source
  • Premisecitation

    In a previous study, deployment of autonomous AI technology was associated with improved adherence to annual DR evaluation on a population level.

    “In our previous study28, we demonstrated that the deployment of this autonomous AI technology was associated with improved adherence to annual DR evaluation on a population level.”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites [28] Huang, J. J. et al. Autonomous artificial intelligence for diabetic eye disease increases access and health equity in underserved populations. Npj Digit. Med. 7, 196 (2024). source
  • Premisecitation

    After rigorous validation on safety and efficacy, autonomous AI-enabled DR detection could be more sensitive than examinations by retinal specialists.

    “After rigorous validation on safety and efficacy, it was also shown that autonomous AI-enabled DR detection could be more sensitive than examinations by retinal specialists25”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites [25] Lim, J. I. et al. Artificial Intelligence Detection of Diabetic Retinopathy: Subgroup Comparison of the EyeArt System with Ophthalmologists’Dilated Examinations. Ophthalmol 3, 100228 (2023). source

      The Discussion reports that EyeArt was more sensitive than retina specialists for detecting more-than-mild diabetic retinopathy against a photographic reference standard. The paper describes a prospective pivotal trial and FDA clearance, but the cited passages do not establish the claimed rigorous validation of safety; the sensitivity finding is also specific to more-than-mild disease. (LLM, conf=0.85)

      Experimental — accuracy not yet measured
  • Premisecitation

    Patients covered under Medicaid insurance are at higher risk for poor visual outcomes.

    “Patients diagnosed with kidney disease, hypertension, and covered under Medicaid insurance are at higher risk for poor visual outcomes13–24.”
    Quote in the paperNot a substantial claim8 of 10 sources checked
    • cites [13] Yau, J. W. Y. et al. Global Prevalence and Major Risk Factors of Diabetic Retinopathy. Diabetes Care 35, 556–564 (2012). source

      The conclusion (p. 8) identifies hypertension as a risk factor for vision-threatening diabetic retinopathy, but it does not address Medicaid coverage. Searches for Medicaid insurance and insurance coverage found no passage linking Medicaid enrollment to poor visual outcomes. (LLM, conf=0.76)

      Experimental — accuracy not yet measured
    • cites [14] Gallego, P. H., Craig, M. E., Hing, S. & Donaghue, K. C. Role of blood pressure in development of early retinopathy in adolescents with type 1 diabetes: prospective cohort study. BMJ 337, a918 (2008). source

      The Clinical characteristics section reports an association between blood pressure and retinopathy in adolescents with type 1 diabetes, but does not address insurance coverage. Searches for Medicaid, insurance coverage, and socioeconomic status found no passage linking Medicaid to visual outcomes in our copy; page numbers were not provided for these passages. (LLM, conf=0.74)

      Experimental — accuracy not yet measured
    • cites [15] TODAY Study Group Development and Progression of Diabetic Retinopathy in Adolescents and Young Adults With Type 2 Diabetes: Results From the TODAY Study. Diabetes Care 45, 1049–1055 (2021). source

      The Risk Factors for Progression of DR section reports associations for hypertension and kidney disease, but does not address Medicaid coverage. Searches for Medicaid, insurance coverage, and socioeconomic status found no passage linking Medicaid insurance to retinopathy progression or poor visual outcomes. (LLM, conf=0.75)

      Experimental — accuracy not yet measured
    • cites [16] Al-Rubeaan, K. et al. Diabetic retinopathy and its risk factors in a society with a type 2 diabetes epidemic: a Saudi National Diabetes Registry-based study. Acta Ophthalmol. (Copenh.) 93, e140–e147 (2015). source

      The Materials and Methods describe a Saudi diabetes-registry cohort, and the Discussion says the study lacked information on the degree of visual impairment and blindness. Searches for Medicaid and insurance status found no passage assessing coverage or its association with visual outcomes. (LLM, conf=0.78)

      Experimental — accuracy not yet measured
    • cites [17] Nittala, M. G., Keane, P. A., Zhang, K. & Sadda, S. R. Risk Factors For Proliferative Diabetic Retinopathy In A Latino American Population. RETINA 34, 1594 (2014). source

      The Data Collection section lists the variables studied but does not include insurance coverage. The Results section reports risk factors for proliferative diabetic retinopathy, not a relationship between Medicaid coverage and poor visual outcomes. A search for Medicaid, insurance coverage, and visual outcomes found no passage addressing the attributed claim. (LLM, conf=0.75)

      Experimental — accuracy not yet measured
    • cites [18] Pavkov, M. E., Harding, J. L., Chou, C.-F. & Saaddine, J. B. Prevalence of Diabetic Retinopathy and Associated Mortality Among Diabetic Adults With and Without Chronic Kidney Disease. Am. J. Ophthalmol. 198, 200–208 (2019). source
    • cites [19] Pearce, I., Simó, R., Lövestam-Adrian, M., Wong, D. T. & Evans, M. Association between diabetic eye disease and other complications of diabetes: Implications for care. A systematic review. Diabetes Obes. Metab. 21, 467–478 (2019). source

      The Introduction discusses diabetic retinopathy, vision loss, and delayed screening, but does not link Medicaid coverage to visual outcomes. Searches of our copy for Medicaid, insurance coverage, and payer status found no passage addressing that association. (LLM, conf=0.75)

      Experimental — accuracy not yet measured
    • cites [20] Hammes, H.-P. et al. Risk Factors for Retinopathy and DME in Type 2 Diabetes—Results from the German/Austrian DPV Database. PLOS ONE 10, e0132492 (2015). source

      The Results passages identify hypertension and albuminuria as risk factors for retinopathy or diabetic macular edema, but do not address Medicaid coverage. Searches of our copy for Medicaid, insurance coverage, and payer status found no passage linking coverage to visual outcomes. (LLM, conf=0.76)

      Experimental — accuracy not yet measured
    • cites [21] Romero-Aroca, P. et al. Relationship between diabetic retinopathy, microalbuminuria and overt nephropathy, and twenty-year incidence follow-up of a sample of type 1 diabetic patients. J. Diabetes Complications 26, 506–512 (2012). source
    • cites [22] Figueroa, A. E., Roy, N., Millan-Ferro, A., Silva, P. S. & Rosas, S. E. Association of Diabetic Retinopathy with Chronic Kidney Disease Progression in Latinos with Type 2 Diabetes. Ethn. Dis. 33, 9–16 (2023). source

      The Introduction (page not provided) discusses diabetic retinopathy and vision loss, while the Discussion (page not provided) addresses retinopathy as a marker of kidney-function decline. Searches for “Medicaid,” insurance coverage, and visual outcomes found no passage linking Medicaid coverage to poor visual outcomes. (LLM, conf=0.76)

      Experimental — accuracy not yet measured
  • Findingown data

    Among patients with diabetes referred to the Wilmer Eye Institute for DR evaluation, the AI group was more likely to be African-American than the ECP group (OR = 1.15, 95% CI 1.02-1.29, p = 0.022).

    “Of the patients referred to the Wilmer Eye Institute for DR evaluation, the AI group was more likely to be African-American (OR = 1.15, 95% CI: 1.02, 1.29, p = 0.022), and there was no significant difference in the probability of being covered by Medicaid (OR = 0.97, 95% CI: 0.93, 1.02, p = 0.245).”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: The propensity score model adjusted for all relevant confounders, so there is no unmeasured confounding between the AI and ECP groups.
  • Classified as probably not a substantial claim. 18
  • May combine two claims. 5

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    African-American patients are 1.6 times more likely to present with vision-threatening diabetic retinopathy as compared to non-Hispanic Whites.

    “For example, African-American patients are three times more likely to present with diabetic macular edema (DME), 1.6 times more likely to present with vision-threatening diabetic retinopathy, and 4 times more likely to present with vision loss as compared to non-Hispanic Whites9–12.”
    Quote in the paperNot a substantial claim2 of 4 sources checked
    • cites [9] Coney, J. M. & Scott, A. W. Racial disparities in the screening and treatment of diabetic retinopathy. J. Natl. Med. Assoc. 114, 171–181 (2022). source
    • cites [10] Varma, R. et al. Prevalence of and risk factors for diabetic macular edema in the United States. JAMA Ophthalmol 132, 1334–1340 (2014). source

      The Results report higher odds of diabetic macular edema among non-Hispanic black participants than non-Hispanic white participants (OR 2.64). Our copy does not report a 1.6-fold comparison for vision-threatening diabetic retinopathy; searches for that term and racial comparison found no supporting passage. (LLM, conf=0.76)

      Experimental — accuracy not yet measured
    • cites [11] Lundeen, E. A. et al. Prevalence of Diabetic Retinopathy in the US in 2021. JAMA Ophthalmol 141, 747–754 (2023). source

      The Results report age- and sex-standardized vision-threatening diabetic retinopathy prevalence of 8.66% among non-Hispanic Black individuals and 3.55% among non-Hispanic White individuals. The paper describes Black prevalence as 155.50% higher, not 1.6 times as high. It measures prevalence among people with diabetes, not likelihood of presenting with the condition. (LLM, conf=0.90)

      Checked by a reasoning model — may be wrong.

      Experimental — accuracy not yet measured
    • cites [12] Muñoz, B. et al. Causes of Blindness and Visual Impairment in a Population of Older Americans: The Salisbury Eye Evaluation Study. Arch. Ophthalmol. 118, 819–825 (2000). source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Patients diagnosed with kidney disease are at higher risk for poor visual outcomes.

    “Patients diagnosed with kidney disease, hypertension, and covered under Medicaid insurance are at higher risk for poor visual outcomes13–24.”
    Quote in the paperNot a substantial claim8 of 10 sources checked
    • cites [13] Yau, J. W. Y. et al. Global Prevalence and Major Risk Factors of Diabetic Retinopathy. Diabetes Care 35, 556–564 (2012). source

      The conclusion (p. 8) identifies hypertension, hyperglycemia, and dyslipidemia as risk factors for diabetic retinopathy, but does not identify kidney disease. Searches of our copy for kidney disease, renal disease, and nephropathy found no passage linking a kidney-disease diagnosis to poor visual outcomes. (LLM, conf=0.76)

      Experimental — accuracy not yet measured
    • cites [14] Gallego, P. H., Craig, M. E., Hing, S. & Donaghue, K. C. Role of blood pressure in development of early retinopathy in adolescents with type 1 diabetes: prospective cohort study. BMJ 337, a918 (2008). source

      The Longitudinal analysis and Systolic and diastolic blood pressure and retinopathy sections report that higher albumin excretion predicted retinopathy in adolescents with type 1 diabetes. That supports a kidney-related association, but the study does not establish that patients diagnosed with kidney disease generally have a higher risk of poor visual outcomes; its measured outcome was retinopathy, not vision loss. (LLM, conf=0.84)

      Experimental — accuracy not yet measured
    • cites [15] TODAY Study Group Development and Progression of Diabetic Retinopathy in Adolescents and Young Adults With Type 2 Diabetes: Results From the TODAY Study. Diabetes Care 45, 1049–1055 (2021). source

      The Risk Factors for Progression of DR section reports that kidney disease was associated with a two- to fourfold higher likelihood of diabetic retinopathy progression in the TODAY cohort. This supports a higher risk of worsening retinal disease, but the study concerns young people with type 2 diabetes and does not establish poorer visual outcomes for patients with kidney disease generally. Page numbers were not available in our copy. (LLM, conf=0.86)

      Experimental — accuracy not yet measured
    • cites [16] Al-Rubeaan, K. et al. Diabetic retinopathy and its risk factors in a society with a type 2 diabetes epidemic: a Saudi National Diabetes Registry-based study. Acta Ophthalmol. (Copenh.) 93, e140–e147 (2015). source

      The Results report that nephropathy was associated with a significantly higher risk of diabetic retinopathy among Saudi adults with type 2 diabetes. However, the Discussion says the study lacked data on the degree of visual impairment and blindness, so it does not directly establish a higher risk of poor visual outcomes. (LLM, conf=0.85)

      Experimental — accuracy not yet measured
    • cites [17] Nittala, M. G., Keane, P. A., Zhang, K. & Sadda, S. R. Risk Factors For Proliferative Diabetic Retinopathy In A Latino American Population. RETINA 34, 1594 (2014). source

      The Results report that high serum creatinine was associated with proliferative diabetic retinopathy (PDR), and that patients with PDR had lower visual acuity. But the study assessed creatinine rather than a kidney-disease diagnosis, and it did not directly test whether diagnosed kidney disease predicted poor visual outcomes. (LLM, conf=0.82)

      Experimental — accuracy not yet measured
    • cites [18] Pavkov, M. E., Harding, J. L., Chou, C.-F. & Saaddine, J. B. Prevalence of Diabetic Retinopathy and Associated Mortality Among Diabetic Adults With and Without Chronic Kidney Disease. Am. J. Ophthalmol. 198, 200–208 (2019). source
    • cites [19] Pearce, I., Simó, R., Lövestam-Adrian, M., Wong, D. T. & Evans, M. Association between diabetic eye disease and other complications of diabetes: Implications for care. A systematic review. Diabetes Obes. Metab. 21, 467–478 (2019). source

      The Results section reports that, among patients with diabetes, nephropathy was a risk factor for diabetic retinopathy, including severe forms. That supports an increased risk of some adverse eye outcomes, but the review does not establish a general increase in poor visual outcomes or visual-acuity loss among all patients diagnosed with kidney disease; it also reports mixed findings for diabetic macular oedema. (LLM, conf=0.86)

      Experimental — accuracy not yet measured
    • cites [20] Hammes, H.-P. et al. Risk Factors for Retinopathy and DME in Type 2 Diabetes—Results from the German/Austrian DPV Database. PLOS ONE 10, e0132492 (2015). source

      The Results report that micro- and macroalbuminuria were associated with higher risks of retinopathy and diabetic macular edema, and the Discussion calls albuminuria a strong risk factor for severe diabetic retinopathy. The study concerns patients with type 2 diabetes and measures retinal disease, so it does not establish the broader claim for all patients diagnosed with kidney disease or all poor visual outcomes. (LLM, conf=0.84)

      Experimental — accuracy not yet measured
    • cites [21] Romero-Aroca, P. et al. Relationship between diabetic retinopathy, microalbuminuria and overt nephropathy, and twenty-year incidence follow-up of a sample of type 1 diabetic patients. J. Diabetes Complications 26, 506–512 (2012). source
    • cites [22] Figueroa, A. E., Roy, N., Millan-Ferro, A., Silva, P. S. & Rosas, S. E. Association of Diabetic Retinopathy with Chronic Kidney Disease Progression in Latinos with Type 2 Diabetes. Ethn. Dis. 33, 9–16 (2023). source

      Table 2 (Materials and Methods; page unavailable) shows that CKD was more common among Latino patients with more severe diabetic retinopathy. The Discussion explains that advanced retinopathy can lead to vision loss, but the study did not measure visual outcomes or establish that a kidney disease diagnosis predicts them. (LLM, conf=0.84)

      Experimental — accuracy not yet measured

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    African-American patients are three times more likely to present with diabetic macular edema (DME) as compared to non-Hispanic Whites.

    “For example, African-American patients are three times more likely to present with diabetic macular edema (DME), 1.6 times more likely to present with vision-threatening diabetic retinopathy, and 4 times more likely to present with vision loss as compared to non-Hispanic Whites9–12.”
    Quote in the paperNot a substantial claim2 of 4 sources checked
    • cites [9] Coney, J. M. & Scott, A. W. Racial disparities in the screening and treatment of diabetic retinopathy. J. Natl. Med. Assoc. 114, 171–181 (2022). source
    • cites [10] Varma, R. et al. Prevalence of and risk factors for diabetic macular edema in the United States. JAMA Ophthalmol 132, 1334–1340 (2014). source
    • cites [11] Lundeen, E. A. et al. Prevalence of Diabetic Retinopathy in the US in 2021. JAMA Ophthalmol 141, 747–754 (2023). source

      The Introduction (page unavailable) identifies DME as one form of vision-threatening diabetic retinopathy (VTDR). The Results (page unavailable) report higher VTDR prevalence among Black than White people with diabetes, but do not report a separate DME comparison or support the claimed threefold difference in DME. (LLM, conf=0.82)

      Experimental — accuracy not yet measured
    • cites [12] Muñoz, B. et al. Causes of Blindness and Visual Impairment in a Population of Older Americans: The Salisbury Eye Evaluation Study. Arch. Ophthalmol. 118, 819–825 (2000). source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    African-American patients with diabetes are less likely to undergo annual screening exams for DR.

    “African-American patients with diabetes are less likely to undergo annual screening exams for DR4–7.”
    Quote in the paperNot a substantial claim1 of 4 sources checked
    • cites [4] Brown, G. C. Vision and quality-of-life. Trans. Am. Ophthalmol. Soc. 97, 473–511 (1999). source
    • cites [5] Heisler, M., Smith, D. M., Hayward, R. A., Krein, S. L. & Kerr, E. A. Racial Disparities in Diabetes Care Processes, Outcomes, and Treatment Intensity. Med. Care 41, 1221 (2003). source
    • cites [6] Shi, Q., Zhao, Y., Fonseca, V., Krousel-Wood, M. & Shi, L. Racial Disparity of Eye Examinations Among the U.S. Working-Age Population With Diabetes: 2002–2009. Diabetes Care 37, 1321–1328 (2014). source

      The Methods define receipt of an eye examination as at least one dilated examination in a year, and the Results report lower rates for minority patients than for non-Hispanic white patients. However, the study groups minorities together rather than reporting an African-American-specific finding, and its sample is limited to working-age adults with diabetes (18–64 years). (LLM, conf=0.88)

      Experimental — accuracy not yet measured
    • cites [7] An, J., Niu, F., Turpcu, A., Rajput, Y. & Cheetham, T. C. Adherence to the American Diabetes Association retinal screening guidelines for population with diabetes in the United States. Ophthalmic Epidemiol 25, 257–265 (2018). source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    African-American patients are 4 times more likely to present with vision loss as compared to non-Hispanic Whites.

    “For example, African-American patients are three times more likely to present with diabetic macular edema (DME), 1.6 times more likely to present with vision-threatening diabetic retinopathy, and 4 times more likely to present with vision loss as compared to non-Hispanic Whites9–12.”
    Quote in the paperNot a substantial claim2 of 4 sources checked
    • cites [9] Coney, J. M. & Scott, A. W. Racial disparities in the screening and treatment of diabetic retinopathy. J. Natl. Med. Assoc. 114, 171–181 (2022). source
    • cites [10] Varma, R. et al. Prevalence of and risk factors for diabetic macular edema in the United States. JAMA Ophthalmol 132, 1334–1340 (2014). source

      The Results report higher odds of diabetic macular edema among non-Hispanic blacks than non-Hispanic whites (OR 2.64), not a fourfold difference in vision loss. The Conclusion and Comment mention vision loss as a possible consequence of DME but provide no racial comparison for that outcome (pages not provided). (LLM, conf=0.78)

      Experimental — accuracy not yet measured
    • cites [11] Lundeen, E. A. et al. Prevalence of Diabetic Retinopathy in the US in 2021. JAMA Ophthalmol 141, 747–754 (2023). source

      The Results report higher vision-threatening diabetic retinopathy prevalence among Black than White people with diabetes (8.66% versus 3.55%), but do not report a racial comparison for vision loss. A search for the claimed fourfold vision-loss difference found no such estimate in our copy; vision-threatening retinopathy is not the same outcome as vision loss. (LLM, conf=0.76)

      Experimental — accuracy not yet measured
    • cites [12] Muñoz, B. et al. Causes of Blindness and Visual Impairment in a Population of Older Americans: The Salisbury Eye Evaluation Study. Arch. Ophthalmol. 118, 819–825 (2000). source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    African-American patients who do undergo screening are more likely to present with more severe forms of DR.

    “For African-American patients who do undergo screening, they are more likely to present with more severe forms of DR8.”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites [8] Huang, B. B., Radha Saseendrakumar, B., Delavar, A. & Baxter, S. L. Racial Disparities in Barriers to Care for Patients With Diabetic Retinopathy in a Nationwide Cohort. Transl. Vis. Sci. Technol. 12, 14 (2023). source

      The Introduction reports lower screening adherence among Black patients and a greater risk of DR-related visual impairment, but neither establishes the severity of DR when patients present for screening. The Results examine barriers to care rather than DR severity at screening; searches for severity and presentation at screening found no passage addressing the claim. (LLM, conf=0.73)

      Experimental — accuracy not yet measured

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Patients diagnosed with hypertension are at higher risk for poor visual outcomes.

    “Patients diagnosed with kidney disease, hypertension, and covered under Medicaid insurance are at higher risk for poor visual outcomes13–24.”
    Quote in the paperNot a substantial claim8 of 10 sources checked
    • cites [13] Yau, J. W. Y. et al. Global Prevalence and Major Risk Factors of Diabetic Retinopathy. Diabetes Care 35, 556–564 (2012). source

      The conclusion (p. 8) identifies hypertension as a risk factor for diabetic retinopathy, including sight-threatening forms. The study concerns people with diabetes and retinopathy outcomes; it does not establish higher risk of poor visual outcomes for patients with hypertension generally. (LLM, conf=0.87)

      Experimental — accuracy not yet measured
    • cites [14] Gallego, P. H., Craig, M. E., Hing, S. & Donaghue, K. C. Role of blood pressure in development of early retinopathy in adolescents with type 1 diabetes: prospective cohort study. BMJ 337, a918 (2008). source

      The Discussion reports that higher blood pressure predicted early retinopathy in adolescents with type 1 diabetes, and the Cox proportional hazard regression section reports higher retinopathy risk at or above the 90th blood-pressure centile. The paper does not establish the broader claim that patients diagnosed with hypertension have worse visual outcomes generally: it studied blood-pressure measurements and early retinopathy in a specific population, not hypertension diagnoses or vision loss. Page numbers were not provided in the copy. (LLM, conf=0.87)

      Experimental — accuracy not yet measured
    • cites [15] TODAY Study Group Development and Progression of Diabetic Retinopathy in Adolescents and Young Adults With Type 2 Diabetes: Results From the TODAY Study. Diabetes Care 45, 1049–1055 (2021). source

      The Risk Factors for Progression of DR section reports that hypertension was associated with a two- to fourfold higher likelihood of retinopathy progression in participants with youth-onset type 2 diabetes. The claim extends that finding to patients generally and to poor visual outcomes broadly; the paper measured retinopathy progression, and hypertension was not a significant predictor in its multivariate analysis. (LLM, conf=0.85)

      Experimental — accuracy not yet measured
    • cites [16] Al-Rubeaan, K. et al. Diabetic retinopathy and its risk factors in a society with a type 2 diabetes epidemic: a Saudi National Diabetes Registry-based study. Acta Ophthalmol. (Copenh.) 93, e140–e147 (2015). source

      The Results and Discussion identify hypertension as associated with increased risk of diabetic retinopathy in the studied diabetes population; Table 3 reports a multivariate-adjusted odds ratio of 1.44 (95% CI 1.25–1.66). But the Discussion says the study lacked information on the degree of visual impairment and blindness, so it does not establish higher risk for poor visual outcomes more broadly. Page numbers were not provided in our copy. (LLM, conf=0.86)

      Experimental — accuracy not yet measured
    • cites [17] Nittala, M. G., Keane, P. A., Zhang, K. & Sadda, S. R. Risk Factors For Proliferative Diabetic Retinopathy In A Latino American Population. RETINA 34, 1594 (2014). source

      The Results report that hypertension was associated with higher odds of proliferative diabetic retinopathy (PDR) in this Latino diabetic cohort (OR 1.64). The paper describes PDR as a cause of visual impairment, but the cited finding concerns risk of PDR, not a demonstrated association between hypertension and poor visual outcomes generally. (LLM, conf=0.86)

      Experimental — accuracy not yet measured
    • cites [18] Pavkov, M. E., Harding, J. L., Chou, C.-F. & Saaddine, J. B. Prevalence of Diabetic Retinopathy and Associated Mortality Among Diabetic Adults With and Without Chronic Kidney Disease. Am. J. Ophthalmol. 198, 200–208 (2019). source
    • cites [19] Pearce, I., Simó, R., Lövestam-Adrian, M., Wong, D. T. & Evans, M. Association between diabetic eye disease and other complications of diabetes: Implications for care. A systematic review. Diabetes Obes. Metab. 21, 467–478 (2019). source

      The Results section (page unavailable) reports that diabetic retinopathy was more prevalent in people with diabetes who had orthostatic hypertension. The Introduction (page unavailable) explains that retinopathy can cause vision loss, but the review does not establish that patients diagnosed with hypertension generally have a higher risk of poor visual outcomes. (LLM, conf=0.76)

      Experimental — accuracy not yet measured
    • cites [20] Hammes, H.-P. et al. Risk Factors for Retinopathy and DME in Type 2 Diabetes—Results from the German/Austrian DPV Database. PLOS ONE 10, e0132492 (2015). source

      The Results report that blood pressure above 140/80 mmHg was associated with retinopathy and diabetic macular edema in patients with type 2 diabetes; the Discussion describes blood pressure as a risk factor for these complications. The claim is broader: the paper does not establish risk for poor visual outcomes generally or specifically assess patients with a hypertension diagnosis. (LLM, conf=0.84)

      Experimental — accuracy not yet measured
    • cites [21] Romero-Aroca, P. et al. Relationship between diabetic retinopathy, microalbuminuria and overt nephropathy, and twenty-year incidence follow-up of a sample of type 1 diabetic patients. J. Diabetes Complications 26, 506–512 (2012). source
    • cites [22] Figueroa, A. E., Roy, N., Millan-Ferro, A., Silva, P. S. & Rosas, S. E. Association of Diabetic Retinopathy with Chronic Kidney Disease Progression in Latinos with Type 2 Diabetes. Ethn. Dis. 33, 9–16 (2023). source

      Table 1 (Materials and Methods; page unavailable) shows higher mean systolic blood pressure in groups with more severe diabetic retinopathy, with a between-group P value of .02. But the paper does not show that patients diagnosed with hypertension have a higher subsequent risk of poor visual outcomes; its main outcome is kidney-function decline. (LLM, conf=0.79)

      Experimental — accuracy not yet measured

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    As compared to the ECP group, the AI group was more likely to speak English as a preferred language (p < 0.001).

    “As compared to the ECP group, the AI group was more likely to be to be African-American (p < 0.001), speak English as a preferred language (p < 0.001), have hypertension (p = 0.001), and have chronic kidney disease (p = 0.017), while less likely to be Hispanic or Latino (p < 0.001) and married (p < 0.001).”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    As compared to the ECP group, the AI group was less likely to be married (p < 0.001).

    “As compared to the ECP group, the AI group was more likely to be to be African-American (p < 0.001), speak English as a preferred language (p < 0.001), have hypertension (p = 0.001), and have chronic kidney disease (p = 0.017), while less likely to be Hispanic or Latino (p < 0.001) and married (p < 0.001).”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationown data

    Implementation of an autonomous AI-assisted diabetic screening program in a primary care clinic is associated with increased presentation to eye care specialist care and improved downstream ophthalmic access for African-American patients.

    “An inverse-probability-weighted regression, which incorporated propensity score matching on social determinants of health and relevant clinical variables, showed that implementation of an autonomous AI-assisted diabetic screening program in a primary care clinic was associated with increased presentation to eye care specialist care by African-Americans (p = 0.02).”
    • Also stated in: § Discussion · hedged
      “After accounting for multiple comparisons, our data suggests that there is an association between office-based, AI-assisted DR screening and improved downstream ophthalmic access for African-American patients.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: The propensity score model adjusted for all relevant confounders, so there is no unmeasured confounding between the AI and ECP groups.
    • Rests on an assumption: Referral to the Wilmer Eye Institute indicates presentation to eye care specialist care.

  • May combine two claims.

    Study design and conductown data

    The overidentification test for covariate balance between the AI and non-AI sites was not significant, suggesting no significant differences in covariates, including race, between the two groups of sites after applying propensity score weights.

    “The overidentification test for covariate balance between the AI and non-AI sites was not significant, suggesting that there were no significant differences in the covariates, including race, between the two groups of sites, after applying the propensity score weights.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: The propensity score model adjusted for all relevant confounders, so there is no unmeasured confounding between the AI and ECP groups.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductnone given

    The retrospective case-control study was approved by the Johns Hopkins Medicine Institutional Review Board and adheres to the Declaration of Helsinki; informed consent was waived by the ethics committee because the research was classified as quality improvement.

    “This retrospective case-control study was approved by the Johns Hopkins Medicine Institutional Review Board (IRB # IRB00269812) and adheres to the Declaration of Helsinki. The need for informed consent was waived by the ethics committee due to the classification of the research as quality improvement.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data

    The small number of Medicaid patients in both groups likely limited meaningful inference.

    “The small number of Medicaid patients in both groups likely limited meaningful inference.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim. May combine two claims.

    Study design and conductreasoning

    Given that the analysis was exploratory, no causal inference can be drawn and the association should be interpreted cautiously and further validated.

    “However, given that the analysis was exploratory, no causal inference can be drawn and this association should be interpreted cautiously and further validated.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Study design and conductnone given

    Propensity score weights were calculated using all non-outcome systemic health and demographic variables listed in Table 1.

    “Propensity score weights were calculated using all non-outcome systematic health and demographic variables listed in Table 1.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: The propensity score model adjusted for all relevant confounders, so there is no unmeasured confounding between the AI and ECP groups.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationnone given

    The association seen in the study could be partially attributed to the fact that office-based screening was performed, independent of whether autonomous AI was involved or not.

    “Also, the association seen in our study could be partially attributed to the fact that office-based screening was performed, independent of whether autonomous AI was involved or not.”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: The observed association is not entirely due to office-based screening independent of whether autonomous AI was involved.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductnone given

    The study was limited by its retrospective nature and inclusion of patients managed at a single healthcare system that serves a predominantly urban population.

    “Our study was limited by its retrospective nature and inclusion of patients managed at a single healthcare system that serves a predominantly urban population.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Utilization of autonomous AI for DR testing has been scaling rapidly.

    “and as a result utilization has been scaling rapidly26,27.”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [26] Wu, K. et al. Characterizing the Clinical Adoption of Medical AI Devices through U.S. Insurance Claims. NEJM AI 1, AIoa2300030 (2023). source
    • cites [27] Abramoff, M. D., Dai, T. & Zou, J. Scaling Adoption of Medical AI — Reimbursement from Value-Based Care and Fee-for-Service Perspectives. NEJM AI 1, AIpc2400083 (2024). source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductreasoning

    Bonferroni correction was applied to account for multiple comparisons, resulting in a p-value cutoff of ≤0.025.

    “Bonferroni correction was applied to account for multiple comparisons, resulting in a p-value cutoff of ≤0.025.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Study design and conductreasoning

    Robust standard errors with clustering at the site level were calculated to account for correlation within individual clinic sites.

    “Robust standard errors with clustering at the site level were calculated to account for correlation within individual clinic sites.”
    Quote in the paperSubstantial claim
20 of 20

  • The paper relies on this without support.

    The propensity score model adjusted for all relevant confounders, so there is no unmeasured confounding between the AI and ECP groups.

    “These variables were included as covariates in the statistical models described below to control for as many confounders as possible.”

  • The paper relies on this without support.

    Referral to the Wilmer Eye Institute indicates presentation to eye care specialist care.

    “we identified 3745 subjects referred to the Wilmer Eye Institute for DR evaluation from primary care sites across the Johns Hopkins Medicine health system.”

  • The paper relies on this without support.

    Patients who sought ophthalmologic care outside Johns Hopkins did not differ systematically between the AI and ECP groups.

    “Other limitations include the inability to account for patients who were managed at primary care clinics within Johns Hopkins Medicine but sought their ophthalmologic care outside of Johns Hopkins”

  • The paper relies on this without support.

    Differences in PCP referral behavior between AI and non-AI sites do not explain the observed association.

    “differences in PCP referral behavior.”

  • The paper relies on this without support.

    The observed association is not entirely due to office-based screening independent of whether autonomous AI was involved.

    “Also, the association seen in our study could be partially attributed to the fact that office-based screening was performed, independent of whether autonomous AI was involved or not.”
5 of 5

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • AI-enabled screening may be particularly valuable for rural and remote populations, low-income patients, older adults, and people who face long travel distances or limited specialist availability.

    “It may be particularly valuable for rural and remote populations, lowincome patients, older adults, and people who face long travel distances or limited specialist availability (28)(29)(30).”

    From From validation to public health impact: implementing equitable AI-enabled diabetic retinopathy screening (2026) · cites this paper as [30]

    The Discussion (page not provided) suggests AI-assisted diabetic retinopathy screening may improve access for African-American patients and reports that a previous study found improved screening adherence among Medicaid-covered patients. It does not establish particular value for rural or remote populations, older adults, or people facing long travel distances or limited specialist availability; the study served a predominantly urban population, and its current Medicaid analysis found no significant difference. (LLM, conf=0.86)

    Experimental — accuracy not yet measured

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

  • Marital-status comparison reverses the table's direction

    Minor

    The text says AI-group patients were less likely to be married, but Table 1 reports a higher married percentage in the AI group than in the ECP group (58.3% versus 47.8%).

1 of 1

Cites

Flagged first, then by citations

Cited by

Sorted by citations

Authors
Leong, A., Wolf, R. M., Channa, R., Wang, J., Lehmann, H., Abramoff, M. D., & Liu, T. Y. A.
Affiliations
  • Johns Hopkins University (US)
  • University of Wisconsin–Madison (US)
Corresponding author
Liu, T. Y. Alvin
ORCID
Venue
npj Digital Medicine
Publisher
Springer Science and Business Media LLC
Published
2026-03-05
Volume
9
Issue
1
ISSN
2398-6352 (ISSN-L)
DOI
10.1038/s41746-026-02460-5
PMC
PMC13077085
OpenAlex
W7133518808
Semantic Scholar
ebc1656dea3177019253f7fee3e1c33349092146
Cited by
3
References
26
Citations by year
2026: 3
FWCI
14.13
Type
journal article
Language
en
Open access
Licence
Locations
  • npj Digital Medicine (publishedVersion, cc-by-nc-nd)
  • PubMed (publishedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
  • PubMed Central (submittedVersion, cc-by-nc-nd)
Funders
Research to Prevent Blindness Career Development Award
Topics
  • Health Sciences › Medicine › Radiology, Nuclear Medicine and Imaging › Retinal Imaging and Analysis
  • Health Sciences › Medicine › Ophthalmology › Ophthalmology and Visual Health Research
  • Health Sciences › Medicine › Ophthalmology › Retinal Diseases and Treatments
Keywords
diabetic retinopathy, AI-assisted screening, primary care, access to eye care, African American patients, health disparities, diabetes screening, ophthalmology referral, social determinants of health

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-13.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-13.

Cited sources are checked by gpt-6-sol.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review