Autonomous artificial intelligence increases screening and follow-up for diabetic retinopathy in youth: the ACCESS randomized control trial

In the paper

“The regulatory approval of the first diagnostic autonomous AI system for diabetic eye exams was based on a pivotal trial against a prognostic standard, i.e., patient outcome24, showing its safety, efficacy, and lack of racial and ethnic bias for diagnosing DED in adults with diabetes with 87% sensitivity and 91% specificity24.”

What the sources say

• Disputed

  Split into 2 atoms — 0 confirmed, 1 partially supported, 1 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.

  Checked by a reasoning model — may be wrong.

  cites [24] Abramoff, M. D., Lavin, P. T., Birch, M., Shah, N. & Folk, J. C. Pivotal trial of an autonomous AI-based diagnostic system for detection of diabetic retinopathy in primary care offices. NPJ Digit. Med. 1, 39 (2018). sourceExperimental — accuracy not yet measured

In the paper

“Some of the participants in this study were familiar with autonomous AI diabetic eye exams from our prior study and may have been more accepting of participation in the current study30.”

What the sources say

• Supported

  The Abstract identifies the cited work as a prior study in which youth received autonomous AI diabetic eye exams. The citation identifies that prior study; the suggestion that familiarity may have increased acceptance in the current study is the citing authors’ own inference. (LLM, conf=0.86)

  cites [30] Wolf, R. M. et al. The SEE Study: safety, efficacy, and equity of implementing autonomous artificial intelligence for diagnosing diabetic retinopathy in youth. Diabetes Care 44, 781–787 (2021). source

In the paper

“This care gap is a major source of health disparity, with racial and ethnic minorities, and under-resourced communities, having worse outcomes, and a disproportionately higher prevalence of DED2,5–9.”

What the sources say

• Partly supported

  The Results report higher prevalence of diabetic retinopathy and vision-threatening diabetic retinopathy among non-Hispanic Black and Mexican American individuals with diabetes than among non-Hispanic White individuals (page not provided). The Comment suggests that differences in screening and treatment may contribute, but does not establish the care gap as a major cause. The paper does not report DED prevalence or show worse outcomes in under-resourced communities. (LLM, conf=0.86)

  cites [2] Zhang, X. et al. Prevalence of diabetic retinopathy in the United States, 2005-2008. JAMA 304, 649–656 (2010). sourceExperimental — accuracy not yet measured
• Partly supported

  The Introduction reports higher diabetes risk, complications, and mortality among racial and ethnic minority and low-income populations. The Health Care and Diabetes section documents disparities in care quality, including dilated eye examinations. Neither passage establishes that the care gap causes a higher prevalence of DED in those groups. (LLM, conf=0.86)

  cites [5] Hill-Briggs, F. et al. Social determinants of health and diabetes: a scientific review. Diabetes Care 44, 258–279 (2020). sourceExperimental — accuracy not yet measured
• ?Not checked — no full text of the sourcecites [6] Effect of intensive diabetes treatment on the development and progression of long-term complications in adolescents with insulin-dependent diabetes mellitus: Diabetes Control and Complications Trial. Diabetes Control and Complications Trial Research Group. J. Pediatr. 125, 177–188 (1994). source
• ?Not checked — no full text of the sourcecites [7] Benoit, S. R., Swenor, B., Geiss, L. S., Gregg, E. W. & Saaddine, J. B. Eye care utilization among insured people with diabetes in the U.S., 2010-2014. Diabetes Care 42, 427–433 (2019). source
• Partly supported

  The Conclusions (Summary; page unavailable) report lower eye-examination rates among racial and ethnic minority patients with diabetes, and the Introduction reports higher diabetic-retinopathy prevalence in some minority groups. The paper does not establish that the care gap causes worse outcomes in under-resourced communities, and it does not address DED prevalence if DED means dry eye disease. Its Study Limitations state that the cross-sectional analysis could not establish causality. (LLM, conf=0.86)

  cites [8] 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). sourceExperimental — accuracy not yet measured
• Partly supported

  The Introduction and §3.5.2 report worse diabetes-related outcomes and reduced eye-care access among underserved and minority populations. Section 3 reports disproportionately higher prevalence of diabetic retinopathy in some ethnic minority groups. The paper addresses diabetic retinal disease, not DED, so it does not support the claimed higher prevalence of DED. Page numbers were not provided. (LLM, conf=0.88)

  cites [9] Patel, D. et al. Social determinants of health and impact on screening, prevalence, and management of diabetic retinopathy in adults: a narrative review. J. Clin. Med. 11. https://doi.org/10.3390/jcm11237120 (2022). sourceExperimental — accuracy not yet measured

In the paper

“the majority of the 34 million people with diabetes in the US have a DED screening care gap, due to a lack of access, and education around the need for a diabetic eye exam3,4.”

What the sources say

• Partly supported

  The Discussion (Vision Health Disparities and Public Health Implications; page not provided) identifies access and awareness barriers to eye care and says challenges remain in getting people at risk screened for diabetic retinopathy. It does not establish that a majority of the 34 million US residents with diabetes have a diabetic-eye-exam screening gap, or that lack of access and education explain such a gap in that population. (LLM, conf=0.86)

  cites [3] Zhang, X. et al. Vision health disparities in the United States by race/ethnicity, education, and economic status: findings from two nationally representative surveys. Am. J. Ophthalmol. 154, S53–62.e51 (2012). sourceExperimental — accuracy not yet measured
• ?Not checked — no full text of the sourcecites [4] Hale, N. L., Bennett, K. J. & Probst, J. C. Diabetes care and outcomes: disparities across rural America. J. Community Health 35, 365–374 (2010). source

In the paper

“Autonomous AI allows real-time, point-of-care diagnosis, whether in the primary care, endocrine, or other outpatient setting, can be integrated into the diabetes care workflow, and was included in the ADA standards of care for DED screening in adults as of 202036.”

What the sources say

• Partly supported

  Split into 2 atoms — 1 confirmed, 1 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.

  cites [36] American Diabetes Association. Microvascular complications and foot care: standards of medical care in diabetes-2020. Diabetes Care 43, S135–S151 (2020). sourceExperimental — accuracy not yet measured

In the paper

“it is important to consider pigmentation of the retina that could introduce bias in an AI system for the diabetic eye exam43.”

What the sources say

• ?Could not verify

  The Development phase discusses how unrepresentative training data can introduce bias, and the Access and monitoring phases discuss bias in AI-assisted diabetic-retinopathy care. Neither passage addresses retinal pigmentation as a source of bias. Searches for retinal pigmentation, pigment, and retinal-image characteristics found no passage making that connection. (LLM, conf=0.73)

  cites [43] Abramoff, M., Tarver, E. M., Loyo-Barrios, N., Trujillo, S., Char, D. Considerations for addressing bias in artificial intelligence for health equity. NPJ Digit Med 6, 170 (2023). sourceExperimental — accuracy not yet measured

In the paper

“However, the 100% care-gap closure rate by autonomous AI in this racially and ethnically diverse pediatric population could have the potential to reduce health disparities for vision loss from diabetes, furthering health equity, and deserves further study42,43.”

What the sources say

• Partly supported

  The passage defining health equity and health disparities says that reducing disparities is evidence of progress toward health equity. It does not address autonomous AI, a 100% care-gap closure rate, or whether that intervention could reduce disparities in diabetes-related vision loss among children. (LLM, conf=0.85)

  cites [42] Braveman, P. What are health disparities and health equity? We need to be clear. Public Health Reports (Washington, D.C.: 1974) 129, 5–8 (2014). sourceExperimental — accuracy not yet measured
• Partly supported

  The Introduction says AI systems may help address health disparities and cites work on autonomous AI for diabetic retinopathy in youth. It does not report a 100% care-gap closure rate or establish that rate in the racially and ethnically diverse pediatric population described in the claim. (LLM, conf=0.84)

  cites [43] Abramoff, M., Tarver, E. M., Loyo-Barrios, N., Trujillo, S., Char, D. Considerations for addressing bias in artificial intelligence for health equity. NPJ Digit Med 6, 170 (2023). sourceExperimental — accuracy not yet measured

In the paper

“This randomization sequence was created by a statistician unaffiliated with the study to ensure masking to the randomization scheme and was implemented by REDCAP’s randomization software based on the participant’s location52,53.”

What the sources say

• Partly supported

  The Phase 5 section says REDCap added a participant-randomization module, supporting the identification of REDCap as randomization software. It does not describe this study’s sequence, an unaffiliated statistician, masking, or randomization based on participant location; searches for those details found no supporting passage. No page number is available in our copy. (LLM, conf=0.88)

  cites [52] Harris, P. A. et al. The REDCap consortium: building an international community of software platform partners. J. Biomed. Inform. 95, 103208 (2019). sourceExperimental — accuracy not yet measured
• ?Could not verify

  The Methods section describes REDCap as a system for collecting and managing research data, but does not describe a randomization module or location-based assignment. Searches for a randomization sequence, masking, and assignment by location found no passage addressing those details or the statistician’s role. (LLM, conf=0.76)

  cites [53] Harris, P. A. et al. Research electronic data capture (REDCap)–a metadata-driven methodology and workflow process for providing translational research informatics support. J. Biomed. Inform. 42, 377–381 (2009). sourceExperimental — accuracy not yet measured

In the paper

“The autonomous AI system (IDx-DR, Digital Diagnostics, Coralville, Iowa, USA) for diagnosing diabetic eye disease (DED) was US FDA De Novo authorized (“FDA approval”) in 2018 for adults with diabetes15.”

What the sources say

• ?Could not verify

  The Methods section, “Diabetic Eye Disease” (page not provided), describes fundus photography graded at a centralized reading center, not the IDx-DR system. Searches for IDx-DR, FDA De Novo authorization, and autonomous AI found no passage in our copy addressing the claimed 2018 authorization for adults with diabetes. (LLM, conf=0.77)

  cites [15] Bjornstad, P. et al. Long-term complications in youth-onset type 2 diabetes. N. Engl J. Med. 385, 416–426 (2021). sourceExperimental — accuracy not yet measured

In the paper

“The system diagnoses specific levels of diabetic retinopathy and diabetic macular edema (Early Treatment of Diabetic Retinopathy Study level 35 and higher, clinically significant macular edema, and or center-involved macular edema)32,33, referred to as “referable DED”34, that requires further management or treatment by an ophthalmologist or retina specialist.”

What the sources say

• Partly supported

  Split into 2 atoms — 1 confirmed, 1 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.

  cites [32] Verbraak, F. D. et al. Diagnostic accuracy of a device for the automated detection of diabetic retinopathy in a primary care setting. Diabetes Care 42, 651–656 (2019). sourceExperimental — accuracy not yet measured
• ?Could not verify

  Split into 2 atoms — 0 confirmed, 0 partially supported, 0 disputed, 0 misattributed, 2 unverifiable. Expand the claim to see per-atom verdicts.

  cites [33] Abràmoff, M. D. et al. Foundational considerations for artificial intelligence using ophthalmic images. Ophthalmology 129, e14–e32 (2022). sourceExperimental — accuracy not yet measured
• ?Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  cites [34] NCQA. HEDIS Measurement Year 2020 and Measurement Year 2021. Technical Specifications for Health Plans. 2L (NCQA, 2020).

In the paper

“In prior studies of youth with diabetes, we demonstrated that the diagnosability of autonomous AI in youth was 97.5%, with 85.7% sensitivity and 79.3% specificity in detecting DED, with no difference in diagnosability across demographic groups30.”

What the sources say

• Partly supported

  The Results report 97.5% diagnosability, 85.7% sensitivity, and 79.3% specificity in the youth cohort, with no significant effect of race, ethnicity, age, or sex on diagnosability. The sensitivity and specificity, however, are specifically for detecting more-than-mild diabetic retinopathy, not diabetic eye disease broadly. (LLM, conf=0.84)

  cites [30] Wolf, R. M. et al. The SEE Study: safety, efficacy, and equity of implementing autonomous artificial intelligence for diagnosing diabetic retinopathy in youth. Diabetes Care 44, 781–787 (2021). sourceExperimental — accuracy not yet measured

In the paper

“Commonly reported barriers to screening include miscommunication regarding the need for a diabetic eye exam, time for an additional doctor’s visit, and transportation barriers17,18.”

What the sources say

• ?Not checked — no full text of the sourcecites [17] Thomas, C. G. et al. Racial/ethnic disparities and barriers to diabetic retinopathy screening in youths. JAMA Ophthalmol 139, 791–795 (2021). source
• Partly supported

  The Results report missed work or school (20.7%), not thinking an exam was necessary (7.7%), lack of time among other barriers (1.8%), and transportation difficulties (2.2%). The Discussion says lack of a provider recommendation was not common in this population. The paper therefore supports time-related and transportation barriers, but does not establish miscommunication about the need for an exam as a commonly reported barrier. (LLM, conf=0.88)

  cites [18] Bruggeman, B. et al. Barriers to retinopathy screening in youth and young adults with type 1 diabetes. Pediatr. Diabetes 22, 469–473 (2021). sourceExperimental — accuracy not yet measured

In the paper

“In this study, the participant’s eyes were not pharmacologically dilated, as pilot studies found that pharmacologic dilation is unnecessary in youth30.”

What the sources say

• Partly supported

  The Research Design and Methods section says, “No pupil dilation was used,” and the paper states that digital fundus photography can be performed without dilation (Abstract; page unavailable). Our copy does not say that pilot studies established pharmacologic dilation was unnecessary in youth, so the stated reason is unsupported. (LLM, conf=0.86)

  cites [30] Wolf, R. M. et al. The SEE Study: safety, efficacy, and equity of implementing autonomous artificial intelligence for diagnosing diabetic retinopathy in youth. Diabetes Care 44, 781–787 (2021). sourceExperimental — accuracy not yet measured

In the paper

“However, only 35-72% of diabetic youth undergo recommended screening exams, with even higher care gap rates in minority and lower socioeconomic background youth10,17.”

What the sources say

• Partly supported

  The Discussion reports that 42.2% of youths with type 2 diabetes and 64.9% with type 1 diabetes received an eye examination within six years of diagnosis, and identifies lower examination rates among Black, Latino, and less affluent youths. It also cites rates of 35% and 70–72% from prior studies, but the 35% figure measures referrals rather than completed exams, while the 70–72% figure measures examinations around a transition in care rather than adherence to recommended screening. No page numbers are available in our copy. (LLM, conf=0.86)

  cites [10] Wang, S. Y. et al. Ophthalmic screening patterns among youths with diabetes enrolled in a large US managed care network. JAMA Ophthalmol 135, 432–438 (2017). sourceExperimental — accuracy not yet measured
• ?Not checked — no full text of the sourcecites [17] Thomas, C. G. et al. Racial/ethnic disparities and barriers to diabetic retinopathy screening in youths. JAMA Ophthalmol 139, 791–795 (2021). source

In the paper

“while this particular autonomous AI has been shown not to have racial bias”

What the sources say

• ?Could not verify

  The 'Measuring bias in AI systems' section discusses testing diagnostic accuracy across subgroups, and the References list a trial of an autonomous diabetic-retinopathy AI. Neither passage says that this particular AI was tested and found free of racial bias. Searches for 'no racial bias' and race-specific validation results found no such finding in our copy. (LLM, conf=0.72)

  cites [43] Abramoff, M., Tarver, E. M., Loyo-Barrios, N., Trujillo, S., Char, D. Considerations for addressing bias in artificial intelligence for health equity. NPJ Digit Med 6, 170 (2023). sourceExperimental — accuracy not yet measured

In the paper

“We have also shown that this system can be implemented in a multidisciplinary diabetes clinic and has the potential to increase DED screening rates in underserved youth, while also being cost-savings to patients and caregivers30,31.”

What the sources say

• Supported

  The Research Design and Methods describe implementation in a multidisciplinary pediatric diabetes clinic. The Conclusions report improved screening rates, identify minority and lower-income youth as less likely to be screened, and state that AI offers time and cost savings to patients. The paper does not separately quantify caregiver savings, but its discussion of time off work for additional eye-care visits supports the claim’s qualified reference to potential caregiver savings. (LLM, conf=0.88)

  cites [30] Wolf, R. M. et al. The SEE Study: safety, efficacy, and equity of implementing autonomous artificial intelligence for diagnosing diabetic retinopathy in youth. Diabetes Care 44, 781–787 (2021). source
• Partly supported

  The Discussion (page unavailable) says point-of-care screening can be performed in a diabetes clinic, potentially increasing screening and sparing families another visit. The Results (page unavailable) find patient cost savings only when adherence to conventional eye examinations reaches at least 23%; at baseline, modeled patient payments were higher with AI screening. The paper does not establish implementation in a multidisciplinary clinic or specifically assess underserved youth. (LLM, conf=0.85)

  cites [31] Wolf, R. M., Channa, R., Abramoff, M. D. & Lehmann, H. P. Cost-effectiveness of autonomous point-of-care diabetic retinopathy screening for pediatric patients with diabetes. JAMA Ophthalmol. 138, 1063–1069 (2020). sourceExperimental — accuracy not yet measured

In the paper

“The CONSORT requirements for RCT were followed51.”

What the sources say

• Supported

  The citation identifies the CONSORT guidance the citing authors say they followed; it is not offered as independent evidence of their compliance. The Intent of CONSORT 2010 section identifies this paper as the statement containing the 25-item checklist and flow diagram for reporting randomised controlled trials. (LLM, conf=0.88)

  cites [51] Schulz, K. F., Altman, D. G. & Moher, D. CONSORT 2010 statement: updated guidelines for reporting parallel group randomised trials. BMJ (Clinical Research ed.) 340, c332 (2010). source

In the paper

“The baseline data in this study confirmed these racial, ethnic, and socioeconomic gap closure disparities, similar to other studies demonstrating the wide range of disparities associated with social determinants of health in diabetic retinopathy screening9,41.”

What the sources say

• Supported

  Split into 2 atoms — 2 confirmed, 0 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.

  cites [9] Patel, D. et al. Social determinants of health and impact on screening, prevalence, and management of diabetic retinopathy in adults: a narrative review. J. Clin. Med. 11. https://doi.org/10.3390/jcm11237120 (2022). source
• Partly supported

  Split into 2 atoms — 1 confirmed, 0 partially supported, 0 disputed, 0 misattributed, 0 unverifiable, 1 not checked. Expand the claim to see per-atom verdicts.

  cites [41] Lock, L. J., Channa, R., Brennan, M. B., Cao, Y. & Liu, Y. Effect of health system on the association of rurality and level of disadvantage with receipt of diabetic eye screening. BMJ Open Diabetes Res. Care 10. https://doi.org/10.1136/bmjdrc-2022-003174 (2022). sourceExperimental — accuracy not yet measured

In the paper

“Diabetic eye exam completion rate was significantly higher (100%, 95%CI: 95.5%, 100%) in the intervention group (n = 81) than the control group (n = 83) (22%, 95%CI: 14.2%, 32.4%)(p < 0.001).”

• Also stated in: § Primary outcome · plain

  “in the intervention arm, 81/81 (100%) participants completed their diabetic eye exams, so the primary care gap closure rate was 100% (95%CI: 96%, 100%). All images were diagnostic for the AI system (output was either “DED present” or “DED absent”). In the control arm, 18/82 completed the diabetic eye exam within 6 months, so the primary care gap closure rate was 22% (95%CI: 14%, 32%).”

In the paper

“Participants reported a high level of satisfaction with autonomous AI, 92.5% were satisfied with the length of time it took to complete the exam, and 96% were satisfied with the experience.”

In the paper

“Those without a prior eye exam were more likely to be Black (p = <0.001), Hispanic (p = 0.02), have household income of less than $50,000 (p = 0.005), Medicaid insurance (p < 0.001), parental education of up to high school completion (p = 0.02), and a shorter duration of diabetes (p < 0.001) in univariate analysis.”

In the paper

“Regarding disparities in ever having a prior diabetic eye exam, 79% of participants reported having had such an exam at some time in the past (but per inclusion criteria, not in the prior 6 months) and 35/164 (21%) did not (i.e., had a care gap).”

In the paper

“In the control arm, 18 participants visited the ECP, for a follow-through completion rate of 22% (95%CI: 14%, 32%) and none had DED.”

In the paper

“The difference of 78% (95% CI: 69%, 87%) in gap closure between control and intervention groups was statistically significant (p < 0.001).”

In the paper

“The difference of 42% (95%CI: 21%, 63%) in follow-through completion rates between control and intervention groups was significant (p < 0.001).”

In the paper

“Of those who were in the intervention arm, 85% reported they would choose to do the AI-based eye exam in the future, and only 57%, an ECP-based diabetic eye exam (Table 4).”

In the paper

“There were differences in diabetes type, race, parental education, and diabetes management characteristics between the two recruitment sites, but glycemic control (measured by hemoglobin A1c) was similar across sites.”

In the paper

“However, in the intervention arm, follow-through was significantly higher at 64%, as participants with “DED present” results were immediately informed of the diagnosis and provided education.”

In the paper

“Limitations of the present study are that the autonomous AI used is not FDA-cleared for use in ages 21 and under”

In the paper

“To test this hypothesis, we designed ACCESS (AI for Childrens’ diabetiC Eye examS Study), a pre-registered, rigorously designed randomized control trial (RCT)”

• Also stated in: § Discussion · plain

  “Strengths of the present ACCESS study are its rigorous hypothesis-testing, pre-registered RCT design”

In the paper

“A mixed multilevel multivariable logistic regression analysis, using the site as a nesting level to account for the clustering of observations, was performed in order to examine the relationship between demographic characteristics and the odds of having a previous diabetic eye exam amongst the entire study cohort, adjusting for known covariates associated with DED and site.”

In the paper

“At the time of the DSMB meeting on 9/16/2022, all participants had already been enrolled in the trial, and the DSMB determined that there were no safety concerns and the study should continue to completion.”

In the paper

“Since randomization and study visits occurred at the same time there was little risk of attrition and thus the sample size was not expanded to account for attrition.”

In the paper

“the tenets of the Declaration of Helsinki were followed”

In the paper

“To prevent selection bias and ensure sample size balance between the groups and sites, stratified randomization (by site) was used”

In the paper

“an independent Data Safety and Monitoring Board was established”

• Also stated in: § Data governance · plain

  “an independent Data Safety and Monitoring Board was established to protect the interests of study participants and to preserve the integrity and credibility of the study data”

In the paper

“ACCESS is a hypothesis-driven, pre-registered, prospective parallel, RCT with a 1:1 allocation ratio”

In the paper

“All parties were masked to the allocation until the participant was randomized in the study, and then all parties were unmasked.”

In the paper

“To ensure safety and that no cases of disease would be missed, all images were also overread by a board-certified retina specialist.”

In the paper

“With this autonomous AI system, a medical diagnosis is made independently by the system without human oversight.”

In the paper

“The study was approved by the Johns Hopkins IRB”

In the paper

“as well as the scripted clinical education given to participants above the standard of care in the control arm, to maximize the likelihood of care gap closure in the control arm.”

In the paper

“potentially improving visual outcomes in this vulnerable population and advancing health equity in youth with diabetes”

In the paper

“participants were randomized in permutated block schedules of 4 and 6”

In the paper

“If the ETDRS level is 20 or lower and no macular edema is present, appropriate management is to retest in 12 months35.”

What the sources say

• ?Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  cites [35] Center for Medicare and Medicaid Services. Proposal to Establish Values for Remote Retinal Imaging (CPT code 92229). Pages 56ff (2021).

In the paper

“There were no changes in trial design or outcome after trial commencement.”

In the paper

“The American Diabetes Association (ADA) and American Academy of Ophthalmology (AAO) recommend DED screening in T1D within 3–5 years of diagnosis and age greater than 11 years, and in T2D at the time of diagnosis in youth16.”

What the sources say

• ?Not checked — no full text of the sourcecites [16] American Diabetes Association. Children and adolescents: standards of medical care in diabetes-2021. Diabetes Care 44, S180–s199 (2021). source

In the paper

“The IDx-DR autonomous AI is not labeled for youth <22 years, as currently no autonomous AI for DED has been cleared for a pediatric population.”

In the paper

“ACCESS was pre-registered on ClinicalTrials.gov (NCT05131451).”

In the paper

“Within each block, participants were randomized with a 1:1 allocation ratio to the control group and intervention group.”

In the paper

“the pre-specified primary outcome was defined as the proportion of participants who completed a documented diabetic eye exam in each arm (“primary care gap closure rate”)”

In the paper

“The pre-specified secondary outcome (“follow-through completion rate”) was defined as the proportion of participants who completed follow-through at the ECP in each arm.”