Aims To evaluate agreement between ocular findings of a telemedicine eye screening (visit 1) with diagnoses of a comprehensive eye examination (visit 2).
Methods A primary care practice (PCP)–based telemedicine screening programme incorporating fundus photography, intraocular pressure (IOP) and clinical information was conducted. Eligible individuals were African American, Hispanic/Latino or Asian over the age of 40; Caucasian individuals over age 65; and adults of any ethnicity over age 40 with a family history of glaucoma or diabetes. Participants with abnormal images or elevated IOP were invited back for a complete eye examination. Both visit 1 and visit 2 were conducted at participants’ local PCP. Ocular findings at visit 1 and eye examination diagnoses at visit 2 are presented, including a cost analysis.
Results Of 906 participants who attended visit 1, 536 were invited to visit 2 due to ocular findings or unreadable images. Among the 347 (64.9%) who attended visit 2, 280 (80.7%) were diagnosed with at least one ocular condition. Participants were predominately women (59.9%) and African American (65.6%), with a mean age (±SD) of 60.6±11.0 years. A high diagnostic confirmation rate (86.0%) was found between visit 1 and visit 2 for any ocular finding. Of 183 with suspicious nerves at visit 1, 143 (78.1%) were diagnosed as glaucoma or glaucoma suspects at visit 2.
Conclusions This screening model may be adapted and scaled nationally and internationally. Referral to an ophthalmologist is warranted if abnormal or unreadable fundus images are detected or IOP is >21 mm Hg.
Trial registration number NCT02390245.
- Diagnostic tests/Investigation
- Intraocular pressure
- Public health
Statistics from Altmetric.com
Vision loss is a major public health issue in the USA with U$35.4 billion annually in direct costs and loss of productivity.1 2 Studies show that many Americans, especially those in underserved populations, do not obtain regular, preventive eye examinations.3 Glaucoma is the second leading cause of irreversible blindness in the USA with an annual economic burden of US$2.9 billion, yet studies consistently estimate that more than half of affected individuals are unaware of their diagnosis.4 5 While African Americans and Hispanics have the highest rates of open-angle glaucoma in the USA, these figures likely under-represent the true burden of disease, as many are undiagnosed and not currently accessing care.5 Early detection, long-term management and ongoing follow-up with an eye care provider are essential to delay glaucoma progression and consequent vision loss.6–8 Previous studies have shown that telemedicine improved glaucoma detection and access to care, reduced time from diagnosis to treatment, and reduced costs for both patients and healthcare systems.9–12 The use of new portable tonometers (eg, iCare Rebound Tonometry) that do not require corneal anaesthesia, and non-mydriatic, autofocus, hand-held fundus cameras, increases the safety and ease of community telemedicine glaucoma screenings.13 Researchers are also developing cloud-based, data-secured visual screening techniques that may aid testing outside of ophthalmic offices.14–16
The Philadelphia Telemedicine Glaucoma Detection and Follow-up Study was designed to assess the effectiveness of a primary care practice (PCP)–based telemedicine screening model for glaucoma and other eye diseases in targeted populations, with goals of improving access, detection and follow-up eye care for medically underserved, high-risk individuals.
This paper examines the rate of confirmation of ocular findings at the telemedicine eye screening (visit 1) with the eye examination diagnoses determined by an ophthalmologist (visit 2) in order to determine whether this eye screening model can reliably detect glaucoma and other eye diseases in urban PCPs and federally qualified health centres (FQHCs). The cost analysis data for both visit 1 and visit 2 are also described.
The detailed methods of this study have been previously published.17
Participants, recruitment and inclusion/exclusion criteria
This study recruited African Americans, Hispanics/Latinos or Asians over the age of 40; Caucasians over age 65; and adults of any ethnicity over age 40 with a family history of glaucoma and/or diabetes. We excluded individuals seen within the past year by an ophthalmologist or those currently following an ophthalmologist for a previous ocular diagnosis. Participants were recruited with the help of community partners which included the Public Health Management Corporation, Philadelphia Department of Public Health, Health Federation of Philadelphia and Temple Physicians Inc. They selected PCP and FQHCs based on Zip Code Tabulation Areas in Philadelphia with federal Medically Underserved Area/Population and Health Professional Shortage Area designations. Both visits were free of charge. Transportation expenses to the PCP and FQHCs were not covered.
Visit 1: telemedicine screening and detection at PCP and FQHC offices
From April 2015 to February 2017, an ocular technician and two health educators travelled to screening locations with the telemedicine screening equipment. Once consent was obtained, the trained health educators measured presenting visual acuity (with correction if available) using the digital acuity system Clear Chart 2 (Reichert Technologies, Depew, New York, USA). The certified, trained ocular technician took two monoscopic fundus photographs and one anterior segment photograph per eye using a non-mydriatic, autofocus, hand-held fundus camera (Volk Pictor, Mentor, Ohio, USA). The camera allowed for a 45-degree field to view the retina. One fundus photograph was centred on the macula and the other on the optic nerve.
Intraocular pressure (IOP) was measured by the ocular technician without topical anaesthesia, using a rebound tonometer TA01I (iCare, Vantaa, Finland). One IOP measurement was taken for each eye. A second measurement was taken if IOP was >21 mm Hg. If the final recorded IOP was ≥30 mm Hg, the participant bypassed visit 2 and was assigned to one of three ‘Fast Track’ schedules and immediately referred to a community ophthalmologist.
Blood pressure measurements were obtained from the PCP electronic medical record for all participants and were categorised according the American College of Cardiology Guidelines.18
All data and fundus/optic disc images were uploaded within 24 hours to the Wills Eye Telemedicine Department password-protected, encrypted, Health Insurance Portability and Accountability Act (HIPAA)–compliant server and were read by a glaucoma fellowship-trained ophthalmologist and a certified retina image reader within 5 days. Images were classified as either abnormal, normal or unreadable. Unreadable images were classified as having a macular image with less than three-fourths of the macula clearly visible (ie, blurry vessels and insufficient focus precluding detection of microaneurysms) or disc images with the entire optic nerve not clearly visible. Both those with unreadable images in one eye and normal in the other eye and unreadable in both eyes were considered unreadable.19 Participants who had a normal image but had IOP >21 mm Hg in at least one eye were diagnosed with ocular hypertension (OHTN). A suspicious nerve diagnosis was made if the fundus image had any of the following features:
Vertical cup:disc (C:D) ratio >0.65 in average and large discs or >0.5 in small discs.
Rim width <0.2 in any area (including optic disc notches).
Vertical C:D ratio asymmetry of >0.2 between eyes, any disc haemorrhage, nerve fibre layer defect or beta zone peripapillary atrophy in association with suspicious rim-thinning.
All participants’ ocular images were evaluated for diabetic retinopathy (DR), and levels of retinopathy were graded according to the National Health Service Diabetic Eye Screening Programme revised grading definitions.20
All participants with abnormal or unreadable images, or IOP 21–29 mm Hg, were called, up to six times, by the research staff and scheduled for a complete eye examination at visit 2 at the same location to confirm their ocular diagnosis within 6 months of the screening. Those with normal images were advised to follow-up in 1 year with an eye care provider.
Visit 2: confirmatory eye examination at PCP and FQHC offices
Participants underwent a full ophthalmic examination, including best-corrected visual acuity (BCVA), using Clear Chart 2 (Reichert Technologies) and G-Top 30-2 (G standard white/white Top) visual field test using the Octopus visual field analyser (Haag Streit Diagnostics, Bern, Switzerland). A glaucoma fellowship-trained ophthalmologist conducted the eye examination which included anterior segment slit-lamp biomicroscopy, central corneal thickness (Ipac (Reichert Technologies) or Pachmate (DGH Technology, Exton, Pennsylvania, USA)), IOP using Goldmann applanation tonometer (GAT), gonioscopy and funduscopic evaluation. All necessary diagnostic equipment was transported to each PCP location.
Fundus photographs, IOP and eye screening results from visit 1 were available for the ophthalmologist during the eye examination. All participants with diabetes or DR on screening received a dilated examination at visit 2. Other participants were dilated at the physician’s discretion. The ophthalmologist determined a final diagnosis, treatment plan and recommended follow-up for each participant based on the American Academy of Ophthalmology (AAO) Practice Pattern Guidelines.21 22
Patient satisfaction surveys
Satisfaction surveys (figure 1), created by the Wills Eye research team, with assistance from Westat, Inc. (Bethesda, Maryland, USA), were administered to all participants at the conclusion of visit 1 and visit 2.
Data management and statistical analysis
Study data were collected and managed using Research Electronic Data Capture and a HIPAA-compliant server hosted at Wills Eye Hospital. Rate of confirmation of findings from visit 1 and visit 2 was determined by calculating the proportion of participants diagnosed with suspicious nerve and other eye diseases whose diagnosis was confirmed at visit 2 along with an exact two-sided binomial CI. The analysis was repeated to determine the proportion of participants with abnormal fundus photography and OHTN on the eye screening who received any ocular diagnosis at the confirmatory eye examination (whether or not the exact abnormality was correct). All analyses were performed using R packages (V.3.5) VGAM and Ordinal.23
Staff time required for visit 1 and visit 2 was calculated using a tracking log. Staff time costs per visit component were calculated in US$ 2015 as mean time per visit component×(staff member’s wage rate+institutional fringe benefit costs). Staff wage rate assumptions were informed by the US Bureau of Labor Statistics (BLS) occupational wage rates for the Philadelphia metropolitan area, and the ophthalmologists’ wage rate was derived from the National Institutes of Health physician salary cap.24 25
The wage rate of a project manager was difficult to map to a BLS occupational listing and thus was estimated as being 25% above a medical assistant’s wage rate. For those diagnosed with either glaucoma suspect or confirmed glaucoma, the cost per case detected was calculated by dividing the total cost for visits 1 and 2 by the number of participants with ocular pathology detected.
A total of 906 participants completed visit 1, of whom 334 (36.9%) were found to have an abnormal image, 155 (17.1%) were unreadable and 62 (6.8%) had OHTN (figure 2). Therefore, 536 (59.2%) were invited to attend visit 2. Reasons for not attending visit 2 included ‘not interested in participating’, ‘no show/cancelled’ and ‘unable to be reached’ (figure 2).
Fifteen participants were fast-tracked immediately to a community ophthalmologist because of high IOP. Of the 15, 14 consented to the follow-up phase of the study. All 14 received a glaucoma-related diagnosis at the first ophthalmologist visit: glaucoma (n=5), glaucoma suspect (n=6) and OHTN (n=3). The 347 participants (64.7% of those invited) who ultimately attended visit 2 were predominately women (59.9%) and African American (65.6%), with a mean age (±SD) of 60.6±11.0 years (range, 40–99.4) (table 1). Mean IOP was 16.3 in the lower eye and 18.9 in the higher eye, and almost 18% of participants who completed the eye examination had an IOP >21 mm Hg in at least one eye. The self-reported prevalence of diabetes was 57.1%, hypertension was 69.5% and family history of glaucoma was 25.1% (table 1).
Of the 347 participants who attended visit 2, 280 (80.7%) were diagnosed with at least one ocular condition (table 2).
Of the 183 noted to have suspicious nerves on screening, 143 (78.1%) were diagnosed as glaucoma or glaucoma suspects at visit 2 (table 3). Of the 39 participants found to have DR on screening, 23 (59.0%) were confirmed with DR after or during visit 2. Overall, there was a high rate of pathology at visit 2 (86.0%) in those invited for follow-up due to abnormal fundus photography or high IOP at visit 1 (table 3).
Participant satisfaction was very high; 100% were satisfied with their visit 1 screening and 99% were satisfied with its convenience (figure 1A). All but two participants (99.8%) reported they were likely to return for follow-up at the same location (figure 1A). Satisfaction survey results after visit 2 indicated that 99% of participants were satisfied with the eye examination and 82% found it very convenient (figure 1B).
The mean duration of visit 1 was 22.7 min, corresponding to total personnel cost of US$7049, or a per-participant cost of US$7.78. Staffing for visit 1 included a health educator, an ocular technician and a translator when needed. The highest cost and time-consuming component of visit 1 was imaging (n=718), requiring a mean of 12.2 min per participant, corresponding to a per-participant cost of US$4.96. The mean duration of visit 2 was 52.3 min, costing US$14 362 in total, or $41.39 per participant. Visit 2 required the same staffing as visit 1, plus an ophthalmologist. The majority of time and cost for visit 2 was the ophthalmologist examination (n=239) and post-dilation examination (n=121) requiring 11.6 and 11.0 min, respectively, corresponding to per-participant costs of US$21.79 and US$20.64, respectively. Based on the cost per visit, the mean cost per case detected was US$13.15 at visit 1 and US$51.29 at visit 2. Thus, the total cost for each participant who completed both visits and were diagnosed with an ocular condition was US$64.44.
The AAO recommends glaucoma screening and eye examinations in high-risk populations.22 However, attending eye examinations can be challenging due to barriers including limited knowledge of the permanency of glaucoma-induced vision loss, perception of non-importance of follow-up visits, cost of eye examinations, poor access, scheduling conflicts and lack of insurance.3 26 Poor adherence may increase the risk of developing irreversible blindness from undetected glaucoma and other eye diseases.
The Philadelphia Telemedicine Glaucoma Detection and Follow-up Study intentionally targeted a high-risk population and diagnosed 245 cases of glaucoma-related eye disease (glaucoma (n=38), glaucoma suspects (n=159), OHTN (n=25), anatomically narrow angle (n=23)) totalling approximately 29.2% of the 906 participants who were screened. Over 80% of those with abnormalities or unreadable images on screening were found to have an ocular diagnosis at visit 2 requiring ongoing monitoring or treatment. For the 155 participants found to have unreadable images at visit 1, the most frequent diagnosis at visit 2 was cataracts (n=71). A complete analysis of unreadable images is discussed in a separate publication.27 28
This study demonstrates that the eye screening has adequate pathology detection in this targeted, underserved population to warrant direct referral to community ophthalmologists if abnormal or unreadable images are detected, or if the individual has an IOP >21 mm Hg. By using non-mydriatic fundus photographs and non-anaesthetised rebound tonometry, this study tested a model that can be performed with modest training by technicians in PCP and FQHCs in populations at high risk for glaucoma.
The use of fundus photography and telemedicine to screen and increase follow-up for DR has been extensively studied.29–34 Previous studies have shown that it is possible to detect glaucomatous damage in telemedicine screenings as well.32 35 However, detecting glaucoma with high accuracy has proven challenging because of the diagnostic testing required for correct diagnosis.22 Previous studies using only fundus photography to screen for glaucoma-related disease have had a sensitivity of 58% and kappa coefficient of 0.52 compared with a comprehensive eye examination.35
Our study used a more comprehensive eye screening model in order to maximise the accuracy of glaucoma screening.17 The results of our study reveal that non-mydriatic fundus photography, when evaluated by expert readers, along with demographics and clinical information in this targeted population, resulted in enhanced screening accuracy for glaucoma. Additionally, our study displayed high rates of satisfaction among patients with 100% being satisfied or very satisfied with visit 1 and 99% being satisfied after visit 2.
Compared with a previous glaucoma screening study that found only 56.4% agreement using only dilated fundus photography, our study’s confirmation of screening diagnosis for glaucoma or glaucoma suspects was higher (78.1%).35 There was also a high diagnostic confirmation (86.0 %) of having an abnormal image or OHTN at screening by the ocular diagnosis at visit 2. This study targeted a high-risk population with higher prevalence of disease, leading to greater disease detection and measured diagnostic accuracy. However, only one study, by Conlin et al, evaluated the effectiveness of medical, ocular and family history of ocular disease, combined with IOP for evaluation of DR and concomitant ocular diseases.27 28 One telemedicine programme directed by Maa et al that included refraction, corneal pachymetry, pupil size, mydriatic fundus photos, IOP, BCVA, and medical, ocular, social and family history in its screening protocol had a 93.7% agreement for glaucoma or glaucoma suspects.36 However, the additional cost of equipment and training may limit the scalability of this approach as a routine screening model.32
Based on the findings presented here, the cost of visit 2 was approximately five times the cost of visit 1, as was the cost per case of ocular pathology detected. Therefore, using the telemedicine screening approach may provide substantial cost benefits.
Our results showed a lower confirmation for the diagnosis of OHTN (37.8%). These results can be attributed to the low threshold for IOP (>21 mm Hg) of our screening model and the use of rebound tonometry at visit 1, which may not exactly correlate with GAT measurements at visit 2. Furthermore, IOP fluctuates daily, as was found in the Ocular Hypertension Treatment Study when subjects returned for IOP checks.7 In regards to the DR confirmation (59.0%), the quality of fundus images allowed some small blood vessels to appear as haemorrhages. All these participants were screened as suspicious for DR and invited to return to visit 2 to clarify these findings. This may have led to an overestimation of DR at visit 1.
Despite the comparatively lower diagnostic confirmation rates for DR and OHTN, of the 347 participants who attended visit 2, 80.7% (280/347) were found to have previously undiagnosed ocular pathology, with many having treatable diseases, thus validating the need to increase access to eye care by targeting underserved, high-risk populations. The consort diagram (figure 2) summarises the number of participants at different times for the eye screening and eye examination.
Limitations of our study include targeting optic nerves with moderate cupping values and reading monoscopic optic nerve photographs with a single site grader. Sensitivity and specificity were not calculated because participants with fundus images interpreted as normal were not invited to return for a confirming diagnosis at visit 2, as it was not financially feasible or practical considering the purpose of this public health intervention. Additionally, the ophthalmologists who performed the complete eye examination were not blinded to the eye screening results in order to mirror real-world clinical practice and maximise disease detection. However, this may have biased their diagnosis and overestimated the prevalence of ocular pathology.
In conclusion, a PCP-based telemedicine eye screening yielded substantial ocular disease detection. This targeted screening protocol is technician-based and of limited cost per screening, and thus potentially can improve access, detection and follow-up eye care for individuals at risk for glaucoma and other eye diseases. This targeted eye screening model may be adapted and scaled nationally and internationally as new telemedicine technology emerges, and a cloud-based, HIPAA-compliant, software platforms become available for transmitting encrypted health information.
The authors thank the CDC for funding the development of the Philadelphia Telemedicine Glaucoma Detection and Follow-up Study and for reviewing the manuscript. We thank the Wills Eye Glaucoma Research Center research team and the Wills Eye Telemedicine Department. We thank Saloni Sapru, PhD at Westat, Inc. for developing and analysing the satisfaction survey data. We thank our community partners: Temple Physicians Inc., Public Health Management Corporation, Philadelphia Department of Public Health–Philadelphia District Five Health Center, Health Federation of Philadelphia and Spectrum Health Center. We also thank our Scientific Advisory Board: George Spaeth, MD; Louis Schwartz, MD; Dennis Slochower, MD; Alan Forman, MD; and Joseph Markoff, MD for advising us on the study design, clinical decision-making and evaluation of outcomes, as well as Stella Stempel, BS, LSW; Elizabeth Murdakhayev, BS; and Cecile Truong, BS for assistance with the manuscript submission.
Contributors LAH had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. This manuscript publication is approved by all authors and by the responsible authorities where the work was carried out. LAH has obtained written permission to include the names of all individuals included in the AcknowledgEment section and confirms that such permission has been obtained in the Authorship Form. Concept and design: LAH, JSM, LJK, JAH, LTP, BEL. Acquisition, analysis or interpretation of data: LAH, JM, CB, MD, AVM, AI, AJ, KR, SJF, LTP, AVM, MW. Drafting of manuscript: LAH, JSM, LJK, JAH, BEL, SH, TZ, JH, MW, AI, AJ, KR, LRP, LTP, JS. Critical revision of the manuscript for important intellectual content: LAH, JSM, LJK, LRP, JAH, LTP, BEL, JS. Statistical analysis: BEL, TZ, SH. Obtained funding: LAH, JSM, LJK, JAH. Administrative, technical or material support: JM, CB, MD.
Funding This study was supported by the United States Centers for Disease Control and Prevention (CDC) (Cooperative Agreement: U01 DP005127). The United States Centers for Disease Control and Prevention provided funding for this Cooperative Agreement and ongoing advice about the study implementation. This manuscript was internally reviewed and cleared for submission by the CDC to disseminate the findings.
Competing interests JSM: Grant/Research Support: Allergan (Madison, NJ), Aerie Pharmaceuticals (Bedminster Township, NJ), Diopsys (Montville, NJ), Haag-Streit (Bern, Switzerland), Heidelberg Engineering (Heidelberg, Germany), Alcon/Novartis (Sinking Spring, PA), Glaukos (San Clemente, CA); Consultant/Advisory Board: Allergan (Madison, NJ), Alcon (Sinking Spring, PA), Aerie Pharmaceuticals (Bedminster Township, NJ), Glaukos, Inotek (Lexington, MA), MicroOptx (Minneapolis, MN); Speakers List: Aerie, Allergan (Madison, NJ), Alcon (Sinking Spring, PA). SJF: Consulting and Speaking: Allergan (Madison, NJ) and Novartis (Sinking Spring, PA); Consultant: Aerie Pharmaceuticals (Bedminster Township, NJ). AVM: Consulting and Speaking: Glaukos (San Clemente, CA); Consultant: Allergan (Madison, NJ) and Gore Medical (Flagstaff, Arizona). LRP: Grant/Research Support: NEI, Bethesda, MD; Advisory Board for Eyenovia, NY, NY; Consultant for Bausch+Lomb, Inc. (Bridgewater, NJ). JAH: Grant/Research Support: ThromboGenics (Iselin, NJ); Consultant: Janssen (Raritan, NJ), Merck (Kenilworth, NJ), Novartis (East Hanover, NJ), KalVista (Cambridge, MA), Spark Therapeutics (Philadelphia, PA), Lowy Medical Research Institute (La Jolla, CA); Board Member: Celgene Corporation (Summit, NJ). LJK: Grant/Research Support: Allergan (Madison, NJ), Diopsys (Montville, NJ), Heidelberg Engineering (Heidelberg, Germany), Zeiss (Oberkochen, Germany); Consultant/Advisory Board: Allergan (Madison, NJ), Alcon (Sinking Spring, PA), Glaukos (San Clemente, CA), Aerie Pharmaceuticals (Bedminster Township, NJ), Diopsys (Montville, NJ), Mati Therapeutics (Austin, Texas), Aerpio Therapeutics (Blue Ash, OH); Speakers List: Allergan (Madison, NJ), Alcon (Sinking Spring, PA), Glaukos (San Clemente, CA), Bausch+Lomb (Rochester, NY), Aerie Pharmaceuticals (Bedminster Township, NJ); Stock Shareholder: Glaukos (San Clemente, CA), Mati Therapeutics (Austin, TX), Aerie Pharmaceuticals (Bedminster Township, NJ); Chief Medical Officer: Glaukos (San Clemente, CA).
Patient consent for publication Obtained.
Ethics approval The study had approval of the Wills Eye Hospital Institutional Review Board (#14-441) and was conducted in accordance with the Declaration of Helsinki.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement Data will be made available upon request.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.