Design of a Pragmatic Randomized Controlled Trial for Comparing the Michigan Screening and Intervention for Glaucoma and Eye Health and Telemedicine Protocol with Usual Care for Disease Detection Authors: Paula Anne Newman-Casey, Maria A Woodward, Leslie M Niziol, Ming-Chen Lu, Suzanne Winter, Mildred Silva-Zuccaro, Kristen Nwanyanwu, David Bush, Clarence Pierce, Amanda K Bicket, Michele Heisler, David W Hutton, Hyungjin Myra Kim, Angela R Elam Published: 2026 ABSTRACT PurposeTo detail the methodology for the Michigan Screening and Intervention for Glaucoma and Eye Health through Telemedicine Program Pragmatic Trial (MI-SIGHT 2) a follow-up to the original MI-SIGHT program that showed feasibility under ideal conditions and efficacy without a comparator.Patients and MethodsA pragmatic randomized controlled trial will compare rate of eye disease detected using the technology-based approach of the MI-SIGHT 2 program (intervention) vs standard optometric exam (control) in a Federally Qualified Health Center (FQHC) (n=900). Inclusion criteria are age ≥ 18 years and no acute eye symptoms. Exclusion criteria are pregnancy or cognitive impairment. The MI-SIGHT 2 intervention leverages technologies including autorefraction, pachymetry, optical coherence tomography, and external and fundus photography, in addition to care navigation support to access low-cost eyeglasses through an on-line retailer and support attendance at recommended follow-up. The primary outcome is detected eye disease as measured by an in-person optometrist exam (control) or by optometrist interpretation of the MI-SIGHT 2 technology-based protocol (intervention). The eye diseases targeted for detection include glaucoma/suspected glaucoma, visually significant cataract, and diabetic retinopathy. The secondary outcome is vision-related quality of life ascertained by survey six weeks after the initial study visit. The exploratory outcomes are program satisfaction and recommended follow-up completion. A cross-over design also will investigate within-person inconsistencies in diagnosis between arms by having participants undergo both screening approaches in randomized order. We will also determine accuracy of diagnoses by each protocol compared to a gold-standard diagnosis determined and adjudicated by glaucoma specialists. Two glaucoma specialist physicians will review all technology-based exams to assess the presence of eye diseases; a third glaucoma specialist physician will adjudicate when there is disagreement.ConclusionThis pragmatic trial will generate evidence to inform best practices for glaucoma and eye disease screening at community health centers. Introduction Three of the leading causes of preventable vision loss, beyond not having eyeglasses, are glaucoma, cataract, and diabetic retinopathy, underscoring the need for holistic disease screening. People who have decreased utilization of eye disease screening contribute to increased rates of visual impairment and blindness, leading to increased economic burden.1–12 The financial burden of vision loss to the US in 2004 was $35.4 billion in direct costs and productivity losses and an additional $13.7 billion in federal government costs;13 identifying and treating eye disease early could mitigate these costs. Federally Qualified Health Centers (FQHCs) provide health care to publicly insured Americans in urban and rural communities. FQHCs provide primary care to populations with high levels of poverty,14 and are thus providing care to people at high risk of underutilizing eye care.2 However, only 3% of FQHC patients receive any form of eye care.14 Furthermore, eye disease screening remains non-standardized both in FQHCs and in primary care practices in general. There is a critical need for evidence-based screening protocols to detect treatable eye diseases in FQHCs. In a community-academic partnership between the University of Michigan and an FQHC in Michigan, we created and piloted the CDC-funded Michigan Screening and Intervention for Glaucoma and eye Health through Telemedicine (MI-SIGHT) Program in 2021.15 MI-SIGHT identified eye diseases that had been undiagnosed until patients participated in the program. Over three years of running the MI-SIGHT program at the FQHC, we detected five times the national rate of vision impairment, three times the national rate of glaucoma, and 1.5 times the national rate of diabetic retinopathy.16,17 These findings were similar to another CDC-funded study evaluating a telemedicine-based eye disease screening protocol in FQHCs in rural Alabama.18 MI-SIGHT relied on an academic medical center’s technology-based screening protocols and grant-funded equipment not typically used in FQHCs (eg, optical coherence tomography, fundus photography, autorefraction, etc) to improve disease detection. Also, it funded academic physicians off-site to interpret the screening results and make care recommendations. The program leveraged outreach, low-cost online glasses and health care navigation to engage people in screening and appropriate medical treatment. The new research objective is to evaluate the real-world effectiveness and potential for sustainability of MI-SIGHT’s technology-based eye disease screening when run locally by an FQHC, and to compare it to usual care at the FQHC. In this new program, MI-SIGHT 2, optometrists and ophthalmic technicians employed by the FQHC will run the program and bill for services via a financially sustainable model for reimbursement in current payment systems. FQHC ophthalmic technicians capture digital ophthalmic data, and FQHC optometrists review the data asynchronously. In MI-SIGHT 2, ophthalmic technicians also function as health care navigators to help patients purchase low-cost glasses on-line and provide support to increase attendance at recommended follow-up care visits. The objectives of this proposed research are to: (1) investigate the effectiveness of MI-SIGHT 2 in identifying eye disease compared to usual care (comprehensive in-person eye exam) at an FQHC with eye care services; and (2) investigate the accuracy, sensitivity, and specificity of both the technology-based protocol and standard optometric exam compared to gold standard evaluation by glaucoma specialist to identify areas for protocol improvement. These objectives will be achieved by conducting a Type 1 effectiveness-implementation hybrid randomized pragmatic clinical trial19 with a crossover design (n=900) to evaluate for detection of the following leading causes of preventable blindness: (1) glaucoma/suspected glaucoma; (2) visually significant cataract; (3) diabetic retinopathy. Costs will be assessed to inform future cost-utility analyses. This manuscript will detail the protocol for this trial. Methods Study Overview We hypothesize that the technology-based MI-SIGHT 2 protocol, compared to usual care, will identify higher percentages of these three leading causes of blindness: (1) glaucoma/suspected glaucoma; (2) visually significant cataract; and (3) diabetic retinopathy. In both arms, the primary eye disease detection outcomes will be obtained at the participants’ initial visit (Figure 1). We also hypothesize that MI-SIGHT 2 will improve the patient-reported outcomes of vision-related quality of life (secondary outcome) and satisfaction with glasses and care (exploratory outcome) as access to low-cost glasses will be provided. These measures will be obtained in person, by e-mail or by telephone six weeks after the initial visit (Figure 1). For participants recommended for follow-up with a specialist, follow-up completion will be compared between arms up to 18 months after the initial visit. Figure 1Schematic design of initial randomization and primary, secondary, and exploratory outcomes data collection points.The flowchart illustrates the process of eligible participant randomization into ′Control First′ and ′Intervention First′ groups. For ′Control First′, participants have an appointment with an eye doctor, receive glasses prescription, undergo primary outcome assessment and a brief interview. They then crossover to a technology-based screening with a technician. For ′Intervention First′, participants start with a technology-based screening, receive glasses prescription and order low cost glasses, undergo primary outcome assessment, and then a remote assessment by the eye doctor . After a 4-5 week washout period, participants crossover to an in-person primary outcome assessment. If no pathology is found, a 10-minute appointment with an eye doctor follows, including picking up glasses and a brief interview. If pathology is found, a 30-minute appointment is conducted. Both paths lead to a 6 week visit where surveys assessing VRQOL and satsifaction are conducted. . The timeline includes Visit 1, Visit 2 and follow-up at week 6 in person or over the phone.A flowchart of participant randomization and follow-up in a study protocol. Schematic design of initial randomization and primary, secondary, and exploratory outcomes data collection points. The trial design is a prospective Hybrid Type 1 crossover trial where participants are randomized to either the technology-based MI-SIGHT 2 protocol (intervention) or the standard optometric exam (control arm) first. For the control arm, the participants will have a standard dilated eye examination with the optometrist who will record the disease outcomes. Immediately following this, control arm participants will be crossed over to intervention arm and complete the MI-SIGHT 2 technology-based protocol with an ophthalmic technician. Within 4–5 weeks of the first visit, the optometrist will review the technology-based exam results and record the primary outcomes for disease detected. For the intervention arm, the participant will undergo the MI-SIGHT 2 technology-based protocol with the ophthalmic technician, and the optometrist will review the testing within a week and document disease detected. The participant will be scheduled to cross over to in-person biometric examination in two to three weeks after their initial visit, and eye disease outcomes will be recorded. The optometrist conducts both the standard in-person examination and the review of the technology-based protocol (Figure 1). Paired (within-participant) agreement in diagnosis between intervention and control protocols will be assessed. We will also determine the accuracy of diagnoses by each protocol compared to a gold-standard diagnosis determined and adjudicated by glaucoma specialists. Two glaucoma specialist physicians will review all technology-based exams to assess the presence of eye diseases; a third glaucoma specialist physician will adjudicate when there is disagreement. This review will form a gold standard from which to assess the accuracy of the screening protocols deployed in the FQHC and inform protocol adjustments. Our study design is a Type 1 effectiveness-implementation hybrid randomized pragmatic clinical trial assessing effectiveness and implementation outcomes.19,20 It is pragmatic as it evaluates the effect of a new screening paradigm implemented by an FQHC, rather than by an academic medical center.21 The program and research will be guided by an Advisory Board (AB).22 The AB is comprised of patients, optometrists, ophthalmologists, ophthalmic technicians, administrators, and leaders from the FQHC alongside academic clinicians and researchers. The AB will meet quarterly to guide the planning and implementation of the trial and evidence-based eye disease screening at the FQHC. We will use the Practical, Robust, Implementation and Sustainability Model/Reach, Effectiveness, Adoption, Implementation and Maintenance (PRISM/RE-AIM) Model23,24 to quantitatively and qualitatively assess the best strategies to implement the most effective form of screening identified in the trial. With input from key stakeholders, we will create a local implementation blueprint and a toolkit of materials and protocols to inform implementation at other FQHCs. The trial will be conducted in accordance with the Declaration of Helsinki; it has ethics approval from the Institutional Review Board (HUM00267229) granted by the University of Michigan, and is registered at clinicaltrials.gov (NCT06934577). Capacity Building To prepare for the trial, we will build FQHC optometrist capacity to interpret the MI-SIGHT 2 technology-based protocols. The university physicians will create training materials with the FQHC optometrist.25–27 A grading manual for each screening test will be created. A training set of 100 participant screening records will be created from the initial MI-SIGHT cohort of participants, half with and half without ocular pathology. Before the trial begins, any participating FQHC optometrist will grade random sets of 10 participant records weekly and will discuss the results with glaucoma and cornea specialists until competence is met (ie, ≥80% concordance with specialist). Participants and Recruitment We will enroll 900 participants over 18 months from the Hamilton Community Health Center, an FQHC in Flint, Michigan. We will include current Hamilton Community Health Center patients and new patients to Hamilton who agree to a dilated eye exam, are ≥ 18 years of age, and speak English, Spanish, Arabic, Chinese, French, Hindi, Finnish, Albanian, Russian or Tagalog. Exclusion criteria include anyone with: (1) significant eye pain, (2) sudden decrease in vision within one week, (3) binocular diplopia, (4) cognitive impairment (defined as the inability to give an ocular or social history), or (5) pregnancy. Anyone with an acute eye problem will be scheduled for an examination directly with the optometrist and will not be included in the trial. We will use pro-active and passive recruitment methods that were used successfully in the initial MI-SIGHT study.28 For pro-active recruitment, we will obtain a list of patients from Hamilton’s electronic health record (EHR) with higher odds of eye disease: (1) those with diabetes and no annual eye exam, or (2) those over 40 years old and no eye exam in the past two years. Identified patients will be contacted via phone by the study team, who will describe the study and assess inclusion/exclusion criteria and interest. We will also advertise via flyers. The flyers will encourage people ≥40 years old, who have diabetes, or a family history of glaucoma to call the FQHC to participate. Passive recruitment includes offering patients who call for an eye care appointment the opportunity to participate if they meet inclusion/exclusion criteria. We will give study materials to each participant to pass on to friends and family, as word-of-mouth greatly contributed to enrollment in the initial MI-SIGHT program (40.5% of people enrolled). Flyers and study materials will be in English and Spanish. Members of the study team are fluent in both English and Spanish. Interpretation services will be available if a Spanish-speaking team member is not present or if a participant speaks a language other than Spanish or English. Randomization Participants will be randomized to either the MI-SIGHT 2 program (intervention) or standard optometric exam (control). Randomization will be stratified by disease risk in four strata (diabetes [yes vs no] and age [<40 vs ≥40 years old]) to prevent potential imbalances between arms across known prognostic variables. We will additionally use block randomization with random block sizes of 2 and 4. Varying the block size lowers the likelihood of knowing the group assignment of the next eligible participant. Randomization will be performed using Randomize.net software. Retention No retention strategies are necessary for the primary outcome, which is assessed at the initial visit. Intervention group participants receive a $10 incentive when they complete their first visit with the ophthalmic technician. A $40 incentive payment is provided following survey completion after the optometrist visit in both arms (first visit for control patients and second visit for intervention patients). Because the intervention group participants need to return for a second visit to see the optometrist, they receive text and phone reminders prior to their visit for confirmation. In both groups, vision-related quality of life and program satisfaction will be ascertained in person, or via email, mail or phone six weeks after participants’ initial visit; $10 will be given as an incentive to complete the week 6 research assessment. Transportation assistance is provided for all participants, as needed, for their eye clinic appointments as is customary for all Hamilton appointments. Medical transportation services are available through both Medicaid and the Flint Mass Transit Authority Your Ride program. Participants in the intervention arm have the opportunity to order low-cost glasses during their first visit with the ophthalmic technician. Control arm participants will have the opportunity to see an ophthalmic technician to order low-cost glasses after completing the six-week measures assessing vision-related quality-of-life as we hypothesize that having access to low-cost, fashionable glasses will increase vision-related quality-of-life in the intervention group. Baseline Data Collection Participants will complete electronic informed consent at the start of the initial study visit via REDCap. A research associate will then administer a baseline survey that includes vision-related quality of life, demographic data, a brief health history and referral source. Surveys are administered using a computer tablet, in English or Spanish; the font can be enlarged or surveys read out loud. Then, the research assistant will log into Randomize.net, confirm participant eligibility, and obtain the treatment assignment designating the participant to initial visit type. Intervention Arm, MI-SIGHT 2 Technology-Based Protocol First visit: Participants randomized to the intervention arm complete technology-based screening tests with the ophthalmic technician. The ophthalmic technician will use a protocol and technology-based equipment to collect the following data and enter it immediately into the EHR and picture archiving system (PACS), making them digital: (1) health history; (2) presenting visual acuity; (3) autorefraction (Marco Ophthalmic, ARK 1s Autorefractor and Keratometer, Jacksonville, FL, USA) and refinement with subjective refraction; (4) best corrected visual acuity; (5) central corneal thickness (DGH Technology, Pachmate, Exton, PA, USA); (6) inter-pupillary distance (Essilor International, Essilor Digital Pupilometer, Charenton-le-Pont, Île-de-France, France); (7) pupillary response, anterior chamber angle assessment, extraocular motility and alignment by penlight; and (8) intraocular pressure (iCare USA, iCare IC200 tonometer, Raleigh, NC, USA). Participants without a narrow angle by penlight29 and intraocular pressure < 30 mmHg are dilated with 0.5% tropicamide; others are not dilated to reduce the risk of angle closure.30 The risk of angle closure with all types of pharmacologic dilation is estimated to be 1/7,070 at a possible upper estimate and 1/25,755 as a lower estimate,31 whereas it is estimated that 38% of ocular pathology is missed when the eye is not dilated.32 The technician will obtain fundus photos (focusing on the disc, macula, and superotemporal arcade33–35) and optical coherence tomography (OCT) of the macula and retinal nerve fiber layer (Topcon Healthcare, Inc, Topcon Maestro 2, La Jolla, CA, USA). Additionally, the technician will obtain a Humphrey Visual Field (HVF, 24–2, SITA Fast) that will be reviewed after the technology-based protocol to enable the remote glaucoma specialists to assess for presence and severity of glaucoma as part of the gold standard evaluation. If urgent conditions are identified, the participant is moved to a clinician’s adjacent exam room for immediate evaluation. Participants receive free sunglasses after dilation during this first visit. Additionally, the ophthalmic technician will guide patients through the process of selecting and ordering low-cost glasses (starting at $12 for single vision lenses and $35 for bifocals from zennioptical.com; other online glasses retailers are available if needed). Whether glasses are ordered will be recorded. Participants will make their 2–3 week optometrist appointment (second visit) while at their initial visit, leaving adequate time for the glasses to arrive through the online retailer. Optometrist review: An FQHC optometrist will asynchronously review participant technology-based screening data within a week of the initial visit and confirm the prescription. Second visit: The optometrist will complete a biomicroscope exam and give recommendations for treatment and follow-up care, if needed. The participant will also be provided with written documentation of the diagnoses and recommendations from the remote glaucoma specialists. If ordered, participants will receive glasses and the technicians will fit these. Technicians will act as care navigators to provide education, assist in scheduling follow-up care if recommended, and connect with social work, if needed.36 Control Arm, Usual Care, Standard Optometric Exam The NIH Pragmatic Model for Comparator Selection recommends choosing a comparator that “best serves the primary purpose of the trial”.37 The clinically relevant alternative to MI-SIGHT 2 is “usual care” with on-site FQHC eye clinicians, considered the highest level of care.38,39 As 91% of FQHC eye clinicians are optometrists, FQHC “usual care” follows American Optometric Association guidelines.40,41 Participants in the control arm will have one mandatory in-person study visit, which will include: an in-person comprehensive eye exam by the optometrist with assessment of health history, visual acuity, pupil examination, ocular motility and alignment, intraocular pressure, and biomicroscopy with additional tests as deemed necessary but excluding technology-based testing.41 Participants will receive a prescription for glasses, if needed, free sunglasses after dilation and recommendations for follow-up care. After completion of the optometrist exam, all participants will also receive the technology-based screening assessment and a Humphrey Visual Field (HVF, 24–2, SITA Fast) by an ophthalmic technician at their initial visit to complete the crossover design. The technology-based screening assessments will be batched for the optometrist to interpret in four to five weeks to minimize recall bias. Masking and Gold Standard Diagnosis for Accuracy Assessment Participants and the eye care providers cannot be masked to arm assignment, but the statistician assessing the main outcome measure will be masked. A second statistician, not masked to arm assignment, will tabulate weekly reports of adverse events by arm for review by the medical monitor. Additionally, to mitigate the risk of bias from the FQHC optometrist serving as outcome assessor for both arms, all participants will be crossed-over to receive the alternate exam type, so the optometrist will assess disease detected in all participants using both data modalities. To assess the accuracy of optometric exam using either the MI-SIGHT 2 technology-based protocol or standard exam techniques, we will establish a gold standard diagnosis utilizing the technology-based screening data and visual field data. Two University of Michigan (UM) glaucoma specialists masked to optometrist diagnosis and to each other’s assessment will establish a gold-standard diagnosis by grading each technology-based exam and assessing for the presence/absence of three leading causes of preventable blindness: (1) glaucoma/suspected glaucoma; (2) visually significant cataract (defined as the presence of cataract without other pathology and best corrected visual acuity ≤20/40); and (3) diabetic retinopathy. They will then interpret the visual field and comment as to whether the diagnosis is: (1) no glaucoma, (2) glaucoma suspect, (3) definite glaucoma, (4) indeterminate glaucoma. Additionally, recommended follow-up specialty and time frame will be recorded. These final diagnoses along with the recommended follow-up specialty and time frame will be compared between the two glaucoma specialists. Agreement for time frame is considered any recommendation within one category of the other (eg, 4–6 months and 7–9 months are still considered agreement, see Table 1). If there is any disagreement, a third UM glaucoma specialist will adjudicate by reviewing the screening data in a masked fashion, without reading the other glaucoma specialist assessments. The gold standard diagnosis is the agreed upon diagnosis by 2/3 specialists. If there is any disagreement about non-dichotomous outcomes such as the recommended follow-up specialty, or time frame, or final glaucoma status, the three glaucoma specialists will review the case and come to a consensus. These gold standard diagnoses will be compared to the optometrist’s first set of diagnoses (first period, before cross-over) from either the in-person standard exam or the technology-based protocol.Table 1Categories for Recommended Follow UpRisk groupTime frameDM w/o DR or w/mild NPDR1 yearLower risk, age-based Age <405–10 years Age 40–493–4 years Age 50–642–3 years Age >651–2 yearsHigher risk, diagnosis-basedWithin 2 weeksWithin 3–4 weeks1–3 months4–6 months7–9 months10–12 monthsAbbreviations: DM, Diabetes mellitus; w/o, without; DR, diabetic retinopathy; w, with; NPDR, non-proliferative diabetic retinopathy. Table 1Categories for Recommended Follow UpRisk groupTime frameDM w/o DR or w/mild NPDR1 yearLower risk, age-based Age <405–10 years Age 40–493–4 years Age 50–642–3 years Age >651–2 yearsHigher risk, diagnosis-basedWithin 2 weeksWithin 3–4 weeks1–3 months4–6 months7–9 months10–12 monthsAbbreviations: DM, Diabetes mellitus; w/o, without; DR, diabetic retinopathy; w, with; NPDR, non-proliferative diabetic retinopathy. Categories for Recommended Follow Up Abbreviations: DM, Diabetes mellitus; w/o, without; DR, diabetic retinopathy; w, with; NPDR, non-proliferative diabetic retinopathy. Week 6 Data Collection and Follow-Up Data Collection At six weeks (with a four week-window), participants will complete a survey to assess the secondary and exploratory outcome measures. Participants will have the option of completing the survey in-person, over email, or over the phone. The survey will include measurements of the secondary outcome measure, vision-related quality-of-life (VRQOL), which will be assessed using the 9-item National Eye Institute Visual Function Questionnaire (NEI VFQ-9).42 The survey will also measure the exploratory outcome measures: the 7-item Vision-related assessment of Quality of Life (AQoL)-7D; whether glasses were purchased and whether participants were satisfied with their glasses; and participant satisfaction with care during the eye diseases screening process overall. Participant satisfaction with care will be measured with seven items from the Consumer Assessment of Healthcare Providers and Systems 4.0,43–46 used by the Center for Medicare and Medicaid Services (CMS) to assess patient satisfaction with healthcare providers and three questions used in the MI-SIGHT observational study.17,47 At 18 months, we will ascertain whether participants completed their recommended follow-up after screening among those participants for whom follow-up was recommended within one year, allowing a six-month window for completion. Completion will be ascertained by reviewing the EHR or calling participants, if needed. Data and Safety Monitoring All participants identified with an urgent condition (eg, intraocular pressure ≥35 mmHg) in the intervention arm will be seen immediately by the FQHC optometrist, rather than at a second visit. The optometrist will review and interpret screening data and enter it into the EHR and research database immediately and will provide a comprehensive exam to the participant. An external ophthalmologist, a glaucoma specialist, will be appointed as a medical monitor. Any serious adverse events will be sent by the project manager to the medical monitor via encrypted email to be reviewed in real time. The medical monitor will determine whether or not the event was trial-related and will recommend next steps to ensure the trial participants’ safety. Two statisticians will have access to the trial data. The unmasked statistician will prepare the reports of all adverse events and protocol deviations for the medical monitor’s review and report during Data and Safety Monitoring meetings with the study team every 6 months. If the medical monitor identifies undue safety occurrences that are related to the trial’s exposure, he/she will bring this to the leadership of the trial with his/her recommendation(s), which may involve protocol modifications. This plan follows the National Eye Institute Guidelines for Data and Safety Monitoring of Clinical Trials.48 The University of Michigan IRBMED will be notified via Standard Continuing Review or Other Reporting Information (ORIO) in eResearch (University of Michigan regulatory web-based IRB system) of any serious or non-serious adverse event. Trial Outcomes All data collected during the ophthalmic technician screening and optometrist visits will be entered into the respective EHR (MiChart for University of Michigan (UM) for the technology-based screening and Athena for Hamilton Community Health Network (HCHN) for the optometrist visit. Images will be uploaded to the PACS (Continuum for UM and ModMed for HCHN) for review and patient care purposes and will be archived on PHI-compliant university storage for data archival. The primary outcome will be assessed during the first visit. The primary outcome is the presence or absence of these three leading causes of preventable blindness: (1) glaucoma/suspected glaucoma, (2) visually significant cataract (cataract visualized with best corrected visual acuity ≤ 20/40 in the absence of other pathology),49,50 and/or (3) diabetic retinopathy. The secondary outcome is vision-related quality of life (VRQOL) 6 weeks after initial visit, which will be assessed prior to helping control participants order low-cost glasses via online retailers. Exploratory outcomes assessed at 6 weeks include additional vision-related quality-of-life measures (AQoL-7D), whether glasses were purchased, whether participants were satisfied with their glasses, and participant satisfaction with the glaucoma screening program. At 18 months, recommended follow-up completion will be assessed. We will report accuracy (including sensitivity/specificity) of technology-based screening and standard optometrist-based screening as compared to glaucoma physician gold-standard assessment. Quantitative Analysis Plan We will perform an intent-to-treat analysis to evaluate our primary, secondary, and exploratory outcomes. Since participants, not eyes, need continued care, the primary analytic unit will be patients. All baseline characteristics and outcomes will be summarized with appropriate descriptive summary statistics and assessed for skewness, and extreme outlying values. Baseline participant characteristics will be compared between arms for imbalances. The primary outcomes will be summarized as the percentages screening positive for glaucoma/suspected glaucoma, visually significant cataract, and diabetic retinopathy, calculated separately among the total number of participants screened overall, and by initial treatment arm (first period, before cross-over). Percentages for each of the three eye diseases will be compared between arms using Chi-square tests (or Fisher’s exact tests as needed). Additionally, logistic regression models will be performed for each eye disease with an indicator for intervention arm as the primary predictor and adjusting for randomization strata and any non-balanced baseline characteristics. If the number screened positive for an eye disease is less than 5, we will use an exact logistic regression model. Because all data for the primary outcome will be obtained at the first study visit, missing data is expected to be minimal (2.6% of fundus photos and 2.9% of OCT images were ungradable and 1.4% of participants did not complete screening examinations in the initial MI-SIGHT study).17 Percentage missing data will be reported for the main outcomes and investigated for associations with treatment, participant demographics, and variables known to be associated with outcomes (eg, self-reported disease or family history of disease). If missingness is <10% and unrelated to investigated variables, complete case analysis will be performed. If missingness is ≥10% or related to investigated variables, sensitivity analysis that employs multiple imputation will be performed. Agreement between optometrist-based in-person assessment and technology-based assessment within each participant (cross-over) will be calculated overall, by initial randomization arm, and stratified by outcome, glaucoma/glaucoma suspect, diabetic retinopathy, and visually significant cataract. Imbalance in agreement will be assessed with McNemar’s tests. Sensitivity, specificity, and accuracy of (1) in-person optometric exam, and (2) technology-based optometric exam diagnoses (both obtained during the first period, before cross-over) will be calculated in comparison to the gold standard glaucoma specialist diagnosis. For the secondary outcome of vision-related quality of life measured using the NEI VFQ-9, the questionnaire composite scores will be calculated according to official documentation.42 Primary analysis will be restricted to complete cases, but sensitivity analyses will be performed using multiple imputation. The between‑arm difference in change in mean NEI VFQ-9 scores (exit visit – initial visit) will first be tested using a 2-sample t-tests (or 2-sample Wilcoxon tests as needed). Additionally, we will use a multiple regression model with an indicator for the intervention arm, adjusting for baseline NEI VFQ-9 score, and randomization strata. Exploratory outcomes will be compared between arms using Chi-square or Fisher’s exact tests for binary outcomes or using 2-sample t-test or 2-sample Wilcoxon tests for continuous outcomes. Additionally, logistic or linear regression models will be performed with an indicator for intervention arm as the primary predictor and following adjustments similar to those described for the primary outcome analysis. Sample Size and Power In 2021, chart review of usual care at Hamilton detected pathology at: 0.9% glaucoma/suspected glaucoma, 0.9% visually significant cataract likely needing surgery, and 0.9% diabetic retinopathy. In the first-year cohort of the initial MI-SIGHT, pathology was detected at: 26.5% glaucoma/suspected glaucoma, 4.4% visually significant cataract, and 9.3% diabetic retinopathy. The chart review denominator was the total number of patients seen in the eye clinic during the same period of time as the MI-SIGHT program, and we therefore assumed that chart review would reflect rates of disease detected by an optometrist, which is the control arm in our trial, and the technician or technology-based MI-SIGHT exam would have rates similar to the intervention arm in our trial as the protocol would be the same. From these estimates, a sample size of 450 participants per arm will provide 80% power to detect the cataract percentage difference, which is the smallest difference to detect between arms (Table 2), using a 2-sided Fisher’s exact test at alpha=0.01667 (adjusted for the three comparisons). For glaucoma/suspected glaucoma, if we were to compare the MI-SIGHT’s 26.5% to national (6%–9%)51 or state Medicaid (12.6%)52 percentages instead of Hamilton usual care percentages of 0.9%, we would need 72–106 or 180, respectively, per group to identify a significant difference between arms, instead of 35 per group. These are all within our currently planned sample size of 450 per group as dictated by the cataract comparison. For the secondary outcome, since the VRQOL data will be collected six weeks after the initial visit, we will assume a worse case of 50% missing data, decreasing the anticipated sample size for the secondary outcome to 225 per arm. Assuming a common standard deviation of 12.6 for composite NEI VFQ-9 score based on our initial MI-SIGHT study, the sample size of 225 per arm will yield 80% power to detect a difference between arms in mean composite NEI VFG-9 score as small as 3.4, using a 2-sided t-test, and assuming alpha = 0.05. This detectable difference is smaller than the 5.3-point difference observed before and after participation in the initial MI-SIGHT program.Table 2Sample Size for Percent Pathology Difference (α=0.01667)Eye DiseaseMI-SIGHTUsual CareORN per group∞Glaucoma26.5%0.9%39.738Cataract*4.4%0.9%5.1450DR9.3%0.9%11.3143Notes: *Cataract, defined as visually significant with best-corrected visual acuity <20/40; ∞Based on exact inference. MI-SIGHT prevalence rates were obtained from the first year cohort of participants at the Hamilton clinic site.Abbreviations: DR, diabetic retinopathy; MI-SIGHT, Michigan Screening and Intervention for Glaucoma and Eye Health through Telemedicine; OR, Odds Ratio. Table 2Sample Size for Percent Pathology Difference (α=0.01667)Eye DiseaseMI-SIGHTUsual CareORN per group∞Glaucoma26.5%0.9%39.738Cataract*4.4%0.9%5.1450DR9.3%0.9%11.3143Notes: *Cataract, defined as visually significant with best-corrected visual acuity <20/40; ∞Based on exact inference. MI-SIGHT prevalence rates were obtained from the first year cohort of participants at the Hamilton clinic site.Abbreviations: DR, diabetic retinopathy; MI-SIGHT, Michigan Screening and Intervention for Glaucoma and Eye Health through Telemedicine; OR, Odds Ratio. Sample Size for Percent Pathology Difference (α=0.01667) Notes: *Cataract, defined as visually significant with best-corrected visual acuity <20/40; ∞Based on exact inference. MI-SIGHT prevalence rates were obtained from the first year cohort of participants at the Hamilton clinic site. Abbreviations: DR, diabetic retinopathy; MI-SIGHT, Michigan Screening and Intervention for Glaucoma and Eye Health through Telemedicine; OR, Odds Ratio. Potential Sources of Bias and Their Mitigation Several potential sources of bias were considered in the design of this study, including recall bias, incorporation bias, and learning effects. To minimize recall bias with respect to diagnosis in the crossover study design, a washout period was enforced. Specifically, for participants randomized to control, a period of 4 to 5 weeks is observed before the optometrist reviews the technology-based intervention screening examination; for participants randomized to the intervention group, a period of 1–2 weeks is observed before the optometrist then sees the participant for an in-person standard care control examination. The shorter washout period in participants randomized to the intervention arm has the potential for a more impactful recall bias, and therefore analysis of crossover agreement in diagnosis will be performed overall and by initial randomization arm. This will allow for investigation of differences in agreement by arm to understand the potential impact of a shorter washout period and if better agreement in diagnosis is observed when participants receive the technology-based intervention examination first. Further, diagnoses for each arm are recorded in different electronic health records (control diagnoses recorded in the Hamilton EHR and treatment diagnoses recorded in the University of Michigan EHR) to further avoid any leakage. As for incorporation bias, the technology-based intervention screening results are used by the glaucoma specialists to obtain the gold-standard diagnoses, which will bias the intervention arm, that uses these same results toward better agreement in diagnosis with gold-standard. This is purposeful as these screening tools are the current standard of care at ophthalmology practices but are not standard at FQHC. Showing whether there is an increased prevalence of disease by using technology-based screening is a primary outcome of this study, but further showing the accuracy of those diagnoses is also an important outcome. If accurate, this would give confidence that clinicians working at FQHCs can provide disease diagnoses on par with specialists. If not accurate, refinement of the training protocol can be undertaken. Lastly, several safeguards were implemented to mitigate learning effects. First, the optometrist was provided with a protocol to assess diagnoses via the technology-based intervention examination. They also trained on up to 100 patients under a specialist ophthalmologist until diagnoses were consistently applied according to the protocol, thus attenuating learning effects before enrollment of study participants. Potential learning effects due to leakage of the gold-standard ophthalmologist diagnosis to the optometrist, or vice-versa, were avoided by masking ophthalmologist and optometrist to each other’s diagnoses. Implementation Analysis We will use the Practical, Robust, Implementation and Sustainability Model/Reach, Effectiveness, Adoption, Implementation and Maintenance (PRISM/RE-AIM) Model (Figure 2), to assess implementation outcomes.53–57 RE-AIM focuses on evaluating quantitative implementation indicators and metrics (extent of reach, adoption, implementation and maintenance at the patient, provider, and clinic level) and these variables are defined for our study in Table 3. PRISM complements the RE-AIM framework by adding a focus on understanding context.23,24 PRISM identifies multilevel contextual factors such as the external and internal environments and infrastructure that influence implementation. We will ask how the external environment (reimbursement, quality metrics) and the internal environment (infrastructure, staff, equipment, and technology) impact implementation and maintenance of glaucoma and eye disease screening at Hamilton in semi-structured interviews with key stakeholders, including FQHC leaders, managers, providers, and staff. We will interview patients from both arms after their optometrist visit. All interviews will be audio recorded, and transcripts will be purposively analyzed from those who screened positive and negative for disease to understand participant experiences. These data will inform refinement of the MI-SIGHT 2 program at Hamilton and identify challenges to address to scale delivery of glaucoma screening at FQHCs nationally.Table 3MI-SIGHT Trial PRISM/RE-AIM Metric Analysis PlanRE-AIM Metric DefinitionReach - The proportion and representativeness of individuals who are willing to participate in the intervention.1. Compare sociodemographic individual- and neighborhood-level characteristics between trial participants and non-participants. Assess uptake of those approached.2. Collect and trend data on number of participants enrolled over time from different referral sources (eg, word-of-mouth, Hamilton staff referral).Effectiveness - The impact of an intervention on important individual outcomes, quality of life, and economic outcomes and assessing whether there is variability across subgroups.1. Compare between trial groups: Primary outcome, proportion of eye disease detected; Secondary outcome, vision related quality of life; Exploratory outcomes, eyeglasses purchased, participant satisfaction, completion of recommended follow-up care.2. Sub-group analyses to measure robustness of effectiveness across income, race, and sex.3. Mixed methods comparison of participant experience between participants who screened positive and negative for disease, and between participants from different races/ethnicities using a joint display embedded in the PRISM framework.Adoption - The proportion and representativeness of clinicians who are willing to initiate the program and why.1. Trend proportion of MI-SIGHT participants referred from Hamilton physicians and staff over time to assess program integration into larger clinical structure.2. Assess institutional readiness for change among Hamilton leadership and eye clinic staff and providers 6 months before and after the trial using the Organizational Readiness for Implementing Change58 scale (measures change commitment and change valence with 10 items).3. Assess Hamilton, clinic staff, technicians, ophthalmologists, optometrists, administrators, and leaders’ attitudes toward MI-SIGHT Program adoption into clinical care through qualitative interviews using the PRISM framework.Implementation - Fidelity to the various elements of an intervention’s key components. Assess adaptations made to interventions and their rationale.1. Assess Hamilton optometrist fidelity to MI-SIGHT Program grading protocols by having UM specialists re-grade a random 10% of the participants.2. Assess adaptations through qualitative interviews with Hamilton clinic staff using the PRISM framework.Maintenance - Extent to which a program becomes part of routine organizational practice.1. Assess necessary prerequisites for maintaining the program after the grant period through interviews with Hamilton staff using the PRISM framework.2. Assess local implementation costs.Abbreviations: PRISM, Practical, Robust, Implementation and Sustainability Model; RE-AIM, Reach, Effectiveness, Adoption, Implementation and Maintenance; MI-SIGHT, Michigan Screening and Intervention for Glaucoma and Eye Health through Telemedicine. Figure 2PRISM/RE-AIM Model taking in perspectives of patients, the organization and the external environment.A flowchart illustrating the intervention and implementation processes in healthcare. It begins with ′Intervention′ divided into ′Organizational Perspective′ and ′Patient Perspective′, with ′Leaders′, ′Managers′ and ′Staff′ under the organizational side. Arrows lead to ′Implementation and Sustainability Infrastructure′, then to ′Recipients′, which is split into ′Organizational Characteristics′ and ′Patient Characteristics′, also showing ′Leaders′, ′Managers′ and ′Staff′. This connects to ′External Environment′. Arrows from ′External Environment′ lead to ′Adoption′, ′Implementation′ and ′Maintenance′, which then connect to ′Reach and Effectiveness′.A flowchart of intervention and implementation processes in healthcare settings. Table 3MI-SIGHT Trial PRISM/RE-AIM Metric Analysis PlanRE-AIM Metric DefinitionReach - The proportion and representativeness of individuals who are willing to participate in the intervention.1. Compare sociodemographic individual- and neighborhood-level characteristics between trial participants and non-participants. Assess uptake of those approached.2. Collect and trend data on number of participants enrolled over time from different referral sources (eg, word-of-mouth, Hamilton staff referral).Effectiveness - The impact of an intervention on important individual outcomes, quality of life, and economic outcomes and assessing whether there is variability across subgroups.1. Compare between trial groups: Primary outcome, proportion of eye disease detected; Secondary outcome, vision related quality of life; Exploratory outcomes, eyeglasses purchased, participant satisfaction, completion of recommended follow-up care.2. Sub-group analyses to measure robustness of effectiveness across income, race, and sex.3. Mixed methods comparison of participant experience between participants who screened positive and negative for disease, and between participants from different races/ethnicities using a joint display embedded in the PRISM framework.Adoption - The proportion and representativeness of clinicians who are willing to initiate the program and why.1. Trend proportion of MI-SIGHT participants referred from Hamilton physicians and staff over time to assess program integration into larger clinical structure.2. Assess institutional readiness for change among Hamilton leadership and eye clinic staff and providers 6 months before and after the trial using the Organizational Readiness for Implementing Change58 scale (measures change commitment and change valence with 10 items).3. Assess Hamilton, clinic staff, technicians, ophthalmologists, optometrists, administrators, and leaders’ attitudes toward MI-SIGHT Program adoption into clinical care through qualitative interviews using the PRISM framework.Implementation - Fidelity to the various elements of an intervention’s key components. Assess adaptations made to interventions and their rationale.1. Assess Hamilton optometrist fidelity to MI-SIGHT Program grading protocols by having UM specialists re-grade a random 10% of the participants.2. Assess adaptations through qualitative interviews with Hamilton clinic staff using the PRISM framework.Maintenance - Extent to which a program becomes part of routine organizational practice.1. Assess necessary prerequisites for maintaining the program after the grant period through interviews with Hamilton staff using the PRISM framework.2. Assess local implementation costs.Abbreviations: PRISM, Practical, Robust, Implementation and Sustainability Model; RE-AIM, Reach, Effectiveness, Adoption, Implementation and Maintenance; MI-SIGHT, Michigan Screening and Intervention for Glaucoma and Eye Health through Telemedicine. MI-SIGHT Trial PRISM/RE-AIM Metric Analysis Plan Abbreviations: PRISM, Practical, Robust, Implementation and Sustainability Model; RE-AIM, Reach, Effectiveness, Adoption, Implementation and Maintenance; MI-SIGHT, Michigan Screening and Intervention for Glaucoma and Eye Health through Telemedicine. PRISM/RE-AIM Model taking in perspectives of patients, the organization and the external environment. Qualitative Data Collection and Analysis We will create an interview guide using the PRISM/RE-AIM framework to understand Hamilton stakeholders’ view on how the external environment (reimbursement, quality metrics, costs) and the internal organizational environment (infrastructure, staff, capacity, equipment, and technology) will impact adoption, implementation and maintenance of evidence-based glaucoma screening after the trial. If the MI-SIGHT 2 program is effective, we will also ask how to optimize the MI-SIGHT 2 program and how to best integrate it into current care pathways. We will ask advisory board members to recommend staff, managers and leaders from the eye clinic, the primary care clinic, and clinic administration to participate in these semi-structured interviews. We will use popcorn-style recruitment, asking each participant for recommendations for additional key stakeholders to interview. Interviews will be audio-recorded and transcribed. Spanish transcripts will be translated and back-translated to ensure original meaning is captured.59 Data will be coded using PRISM constructs.57 A consensus process will be used to develop a codebook (Dedoose, Dedoose Version 9.0.17, Hermosa Beach, CA, USA). Quantitative Data Analysis Reach will be assessed by calculating the proportion of participants in the trial divided by the total number of eligible people who were contacted or inquired about study participation. We will compare individual characteristics abstracted from the EHR between trial participants and eligible non-participants. We will geocode (ArcGIS) individual addresses to obtain the Michigan Area Deprivation Index60,61 to compare neighborhood deprivation between participants and eligible non-participants using 2-sample Wilcoxon or t-tests. We will assess participants enrolled over time by different referral sources. The ten-item Organizational Readiness for Implementing Change scale from Hamilton leadership and eye care providers will be compared before and after the trial using paired Wilcoxon or t-tests (SAS, SAS version 9.4, Cary, NC, USA). Cost will also be assessed. Total costs are a combination of labor and equipment costs. We will collect process times for optometrists to (1) interpret and document the screening exams, and (2) meet face-to-face with participants when providing direct care. We will collect process times for the ophthalmic technicians to (1) conduct and document screening tests; (2) choose, order, and fit glasses; (3) provide education; and (4) provide care navigation for participants (eg, arranging transportation or medical appointments). We will multiply process times by provider salaries and benefits to estimate program labor costs, and we will calculate labor costs for standard care in a similar manner. Equipment costs will be amortized over a useful time period (10 years). We will calculate cost per case of glaucoma detected, per case of eye disease detected, and per participant served in the intervention and control arms. The cost analysis will focus on the incremental cost of the eye disease screening program protocol over and above a standard eye exam and will include both equipment and labor costs. This data will be used to inform a future cost-utility analysis to understand whether the costs of this program meet standardized willingness to pay thresholds. Implementation Initiatives We will continue to convene the advisory board quarterly before, during, and after the trial; the group will offer insight into implementation and maintenance of evidence-based glaucoma and eye disease screening. As part of the implementation strategy after the trial, we will convene a group of program champions, including eye care providers, key primary care providers, and staff involved in the eye care referral pathway, who will meet monthly for six months and will also be invited to join the advisory board. We will share the qualitative and quantitative implementation analysis results with the program champions and advisory board. We will work with the program champions to create a local implementation blueprint that includes a plan describing implementation strategies, goals, timelines, and key personnel necessary for eye disease screening to be successful. Potential implementation strategies include: (1) accessing new funding from increased in-person visits to defray program costs; (2) Hamilton leadership mandating change based on trial results; (3) having program champions strategize to address potential challenges; (4) having program champions build a coalition of providers to maintain program adoption.62 We will follow University of California Berkely guidelines63 to develop an online toolkit of materials and protocols that will be available at https://seeglaucoma.org/mi-sight and linked to a data repository so that the materials can be assigned a permanent DOI. The toolkit will contain overview documents, process documents, training curricula, and protocols for intervention implementation. The toolkit will also contain tools for engagement and communication, assessment, and descriptions of the funding model. These materials will help inform future implementation of whichever eye disease screening program was found to be most effective at the FQHC.64 Sample Size and Power, Implementation Analyses For the qualitative analysis, we will select as many interviews as needed to reach thematic saturation, defined as approximately 13 interviews per sub-group.65 For participant interviews, we will purposively select from each arm from the following categories to assess experiential differences in program effectiveness (n = approximately 26): (1) screened positive for eye disease; (2) screened negative for eye disease. For Hamilton key stakeholder interviews on adoption, implementation and maintenance of evidence-based glaucoma screening, we will interview all relevant stakeholders in an attempt to reach thematic saturation. Discussion This proposal begins to address the National Academy of Medicine’s call to “eliminate correctable and avoidable vision impairment”.66 Currently, screening for eye disease is only recommended for those with diabetes to screen for retinopathy. There are no recommendations to screen holistically for vision impairment, glaucoma, or other eye diseases in the primary care setting.67 In terms of glaucoma screening, the United States Preventive Services Task Force (USPSTF) determined that while the evidence to recommend glaucoma screening in the general population is insufficient, testing can successfully identify glaucoma. They identified a gap in the utilization of glaucoma testing in the primary care setting.67 Current testing strategies to screen for glaucoma have excellent (96%) specificity and good (73%) sensitivity.68 However, because of the low prevalence of glaucoma (1.4%) in the general US adult population, screening all adults would lead to a false discovery rate of 79%. This high false discovery rate from universal screening would place undue burden on patients and undue costs on the healthcare system.68 However, screening populations with higher rates of glaucoma such as populations with higher rates of poverty who are served by FQHCs14 – would lower the false discovery rate. For example, in the initial MI-SIGHT program, the prevalence of glaucoma/suspected glaucoma was 26.5%, a prevalence that decreases the false discovery rate from 79% to 13%, an acceptable rate given the potential mitigation of vision loss from case identification. However, though FQHCs are the medical home for low-income people, they are not equipped for eye care. FQHCs provide medical care to 1 in 11 Americans, serving the 27.5 million people who lack health insurance (8% of the US population)69 and 90.5 million people with Medicaid (27% of the population). The population cared for by FQHCs has multiple important risk factors for eye disease, including diabetes.14,70 Specialty care is limited for the uninsured and those on Medicaid. People with Medicaid, compared to those with private insurance, had 60% lower odds of obtaining an appointment with an eye clinician when they called to schedule.71 Resources at FQHCs are robust. The 1385 FQHCs nationwide employ 5086 full-time social workers to help with transportation and social services.40 FQHCs are uniquely poised to provide direct eye disease screening and help people attend recommended follow-up care. Yet in 202140, only 2.6% of people receiving medical care at FQHCs received eye care.14 The four leading causes of preventable vision loss in the US are uncorrected refractive error due to lack of glasses, cataract, glaucoma, and diabetic retinopathy, and with the advent of new treatments, vision loss from macular degeneration is becoming preventable.72–77 Uncorrected refractive error, cataract, and later stage macular degeneration all present with blurred vision often leading people to seek treatment, while neither glaucoma nor diabetic retinopathy present with vision symptoms until the diseases are very late in their courses and treatment is less effective.78 Diabetic retinopathy screening currently has objective guidelines and primary care-based screening strategies.79 However, glaucoma screening currently lacks this structure. Therefore, though blindness from glaucoma is preventable, only half of the three million Americans with glaucoma are ever diagnosed.75,80 In order to both mitigate preventable blindness and improve vision-related quality-of-life, we need structured protocols to holistically screen for eye disease and assist people in obtaining affordable eyeglasses. This manuscript details the protocol for the MI-SIGHT 2 pragmatic trial. This trial was designed to function within the constraints of a real-world setting, an FQHC. There are some limitations to this trial design. There is the potential for recall bias where the optometrist interpreting the technology-based protocol, after seeing the participant for an in-person exam 4–5 weeks earlier, remembers the exam and this influences their technology-based assessment. Though we attempted to mitigate this bias by having the optometrist review the technology-based exam 4–5 weeks after seeing the participant in-person, the attenuated risk remains. The potential for a learning bias, where the optometrist learns from the ophthalmologists’ gold standard diagnoses or the ophthalmologists learn from the optometrist’s in-person diagnoses, is mitigated by masking each grader to the others’ assessments. Though the potential for missing data is very low given that the outcomes are ascertained during the first visit, there remains the possibility of missing outcomes data due to missing or ungradable imaging, and our statistical analysis plan will address this issue. Conclusion This pragmatic effectiveness-implementation trial of MI-SIGHT 2 vs usual care is designed to generate evidence to inform effective and sustainable protocols for eye disease screening in FQHCs. It will also establish strategies for local implementation that could inform future nation-wide scaling. The work aligns with the American Academy of Ophthalmology and the American Optometric Association who both support expanding eye care services at FQHCs as a national strategy to decrease preventable blindness.81