Percentage of individuals examined during the 6-month or less through greater than 42-month through 54-month time bin compared by original group assignment (telemedicine or traditional) with type of examination (telemedicine, traditional, or both) noted.
eTable 1. Prevalence and Severity of Diabetic Retinopathy (DR) and Macular Edema (ME) in the Eye With More Advanced Retinal Disease Using Any Exam, Tribal Vision Project 2014
eTable 2. Prevalence and Severity of Diabetic Retinopathy (DR) and Macular Edema (ME) in the Eye With More Advanced Retinal Disease After Telemedicine Was Available to Both Groups, Tribal Vision Project 2014
Mansberger SL, Sheppler C, Barker G, Gardiner SK, Demirel S, Wooten K, Becker TM. Long-term Comparative Effectiveness of Telemedicine in Providing Diabetic Retinopathy Screening ExaminationsA Randomized Clinical Trial. JAMA Ophthalmol. 2015;133(5):518-525. doi:10.1001/jamaophthalmol.2015.1
Minimal information exists regarding the long-term comparative effectiveness of telemedicine to provide diabetic retinopathy screening examinations.
To compare telemedicine to traditional eye examinations in their ability to provide diabetic retinopathy screening examinations.
Design, Setting, and Participants
From August 1, 2006, through September 31, 2009, 567 participants with diabetes were randomized and followed up to 5 years of follow-up (last date of patient follow-up occurred on August 6, 2012) as part of a multicenter randomized clinical trial with an intent to treat analysis. We assigned participants to telemedicine with a nonmydriatic camera in a primary care medical clinic (n = 296) or traditional surveillance with an eye care professional (n = 271). Two years after enrollment, we offered telemedicine to all participants.
Main Outcomes and Measures
Percentage of participants receiving annual diabetic retinopathy screening examinations, percentage of eyes with worsening diabetic retinopathy during the follow-up period using a validated scale from stage 0 (none) to stage 4 (proliferative diabetic retinopathy), and percentage of telemedicine participants who would require referral to an eye care professional for follow-up care using a cutoff of moderate diabetic retinopathy or worse, the presence of macular edema, or an unable-to-determine result for retinopathy or macular edema.
The telemedicine group was more likely to receive a diabetic retinopathy screening examination when compared with the traditional surveillance group during the 6-month or less (94.6% [280/296] vs 43.9% [119/271]; 95% CI, 46.6%-54.8%; P < .001) and greater than 6-month through 18-month (53.0% [157/296] vs 33.2% [90/271]; 95% CI, 16.5%-23.1%; P < .001) time bins. After we offered telemedicine to both groups, we could not identify a difference between the groups in the percentage of diabetic retinopathy screening examinations. Diabetic retinopathy worsened by 2 stages or more in 35 (8.6%) of 409 participants (95% CI, 5.8%-11.2%) and improved by 2 stages or more in 5 (1.2%) of 409 participants (95% CI, 0.1%-2.3%) during the 4-year period. The percent of telemedicine participants requiring referral ranged from 19.2% (52/271) to 27.9% (58/208).
Conclusions and Relevance
Telemedicine increased the percentage of diabetic retinopathy screening examinations, most participants did not require referral to an eye care professional, and diabetic retinopathy levels were generally stable during the study period. This finding suggests that primary care clinics can use telemedicine to screen for diabetic retinopathy and monitor for disease worsening over a long period.
clinicaltrials.gov Identifier: NCT01364129
Research suggests that the prevalence of diabetes mellitus in the US adult population will increase from 14% in 2010 to approximately 33% by 2050.1 Diagnosis and treatment of diabetic retinopathy are key public health interventions because they can greatly reduce the likelihood of vision loss.2,3 However, less than 50% of those with diabetes receive annual diabetic retinopathy screening examinations.4- 6
Health clinics can use store-and-forward telemedicine with nonmydriatic cameras to acquire retinal images without dilation and send images for remote evaluation. Studies7- 9 have found excellent diagnostic precision for diabetic retinopathy when compared with examinations with dilated pupils in eye care professionals’ offices. However, studies10- 17 have only evaluated the ability of telemedicine to provide diabetic retinopathy screening examinations with short-term follow-up.
The Tribal Vision Project10 was designed to determine the comparative effectiveness of telemedicine vs traditional surveillance techniques (examinations with eye care professionals) for providing diabetic retinopathy screening examinations using a multicenter randomized clinical trial design. The project addresses 2 recommendations from the Institute of Medicine’s priority topics for comparative effectiveness research, including decreasing health disparities in diabetes and comparing the effectiveness of new remote monitoring technologies.18 We are unaware of any previous reports that have examined the long-term effectiveness of telemedicine in providing diabetic retinopathy screening. Researchers and primary care clinics can use this information to determine the long-term effect of using telemedicine to detect diabetic retinopathy.
We determined that a sample size of 194 participants (97 participants per group) was required to detect a 10% increase in the percentage of diabetic retinopathy screenings in the telemedicine group using an α level of .05 and a power of 0.80 in the 1-year and 2-year time bins. We consecutively enrolled more participants than required to allow for attrition during long-term follow-up (last date of patient follow-up occurred on August 6, 2012). From August 1, 2006, through September 31, 2009, we recruited adults from 2 primary care clinics that serve a large number of American Indian/Alaska Native patients with diabetes (Figure 1). Participants self-classified their race/ethnicity. We included people with diabetes aged 18 years or older who were scheduled to visit their primary care physician. Exclusion criteria were cognitive impairment that prevented the ability to give informed consent or a disability that would not allow camera imaging.10
The institutional review boards of Legacy Health, Oregon Health & Science University, and the Portland Area Indian Health Service reviewed and approved the study protocol (Supplement 1). All patients provided written informed consent, and the study was conducted in accordance with the tenets of the Declaration of Helsinki for human subjects research.
We used a random number generator to randomly assign participants to the telemedicine group or the traditional surveillance group. Tribal leadership wanted all participants to have access to telemedicine during the study, so we offered telemedicine screening to those in the traditional surveillance group after they had been enrolled in the study for 2 years. This staged-intervention design allowed us to examine the use of telemedicine when offered to participants who were originally assigned to traditional surveillance.
Participants in the telemedicine group received retinal imaging at their primary care clinic during a regular visit. Although telemedicine with nonmydriatic cameras can screen for diabetic retinopathy, telemedicine cannot replace a comprehensive eye examination.19 Therefore, the project staff encouraged all participants to see an eye care professional once per year for a comprehensive eye examination.
We describe our category 4 telemedicine program19 in detail in a previous manuscript.10 Briefly, we used a modified Diabetic Retinopathy Study protocol to capture 6 undilated, 1.5-megapixel, 45° fundus photographs of each eye, including a stereo pair of photographs centered on the optic disc, a stereo pair centered on the macula, a single image centered on the superior temporal retina, and a single image centered on the inferior temporal retina.20,21
Devers Eye Institute created a telemedicine system to encrypt, compress, and securely transfer retinal images and participant data to a Health Insurance Portability and Accountability Act–adherent database. The system automatically e-mailed a notification to the image reviewers when clinic staff uploaded new photographs for evaluation. The reviewers (S.D., S.L.M.) graded images according to standard scalable criteria (Table 1).22,23 The reviewers entered their findings into electronic forms within the telemedicine system, and the system automatically sent the evaluation reports to the clinics via e-mail or fax. The telemedicine imaging process, including a video demonstration of how the software works, can be viewed at https://www.youtube.com/watch?v=dpN1Sp-P074&feature=email.
Medical staff recommended annual eye examinations to all participants during their primary care physician visits. If a participant did not already have an eye care professional, the primary care physician referred the participant to an eye care professional in the community. A study investigator (S.L.M.) telephoned the community eye care professionals to introduce the project and request their participation in completing the data collection forms for the study. Research staff sent data entry forms to eye care professionals that contained the same diabetic retinopathy grading criteria as the telemedicine group (Table 1). The eye care professional’s office faxed or mailed the data entry form back to research staff for data entry. Research staff also reviewed participants’ clinic medical records at regular intervals to identify missing eye examinations and contacted eye care professionals for the results.
All analyses were performed using the R statistical program (R Foundation; www.R-project.org) or SPSS statistical software, version 19.0 (IBM Corp). We compared baseline characteristics (age, sex, primary race/ethnicity, blood pressure, hemoglobin A1c [HbA1c], and duration of diabetes) for the 2 groups using independent sample t tests for continuous variables and Fisher exact test for categorical variables. We calculated the percentage of patients who received diabetic retinopathy screening examinations, the percentage of telemedicine examinations requiring referral to an eye care professional, and the percentage of eyes that had higher, lower, or the same level of diabetic retinopathy.
We created time bins based on number of days since enrollment to determine the percentage of participants who received a diabetic retinopathy screening examination. We defined enrollment as the date participants signed the consent form. We defined the time bins based on their latest time point: the 6-month or less time bin was 6 months before through 6 months after enrollment (−182 through 181 days); the greater than 6-month through 18-month time bin was greater than 6 through 18 months (182 through 546 days) after enrollment; the greater than 18-month through 30-month time bin was greater than 18 through 30 months (547 through 911 days) after enrollment; the greater than 30-month through 42-month time bin was greater than 30 through 42 months (912 through 1276 days) after enrollment; and the greater than 42-month through 54-month time bin was greater than 42 through 54 months (1277 through 1641 days) after enrollment. We included 6 months before enrollment as part of the 6-month or less time bin to include examination results from participants who had recently completed an examination. We included participants in a time bin if they were active in the study for the entire length of time defined by the bin.
We defined a diabetic retinopathy screening examination as any type of examination (traditional or telemedicine) within a time bin but excluded examinations that did not evaluate the retina (eg, refractions or anterior chamber exams after cataract surgery). We compared the percentage of diabetic retinopathy screening examinations between the telemedicine and traditional surveillance groups for each time bin using a Fisher exact test.
We calculated the percentage of patients who would require referral to an eye care professional based on their telemedicine examination results. The criteria for referral were (1) a diabetic retinopathy grade of moderate nonproliferative diabetic retinopathy or worse, (2) the presence of clinically significant macular edema, or (3) an unable-to-determine result in either eye for diabetic retinopathy or macular edema.10,24,25 For this analysis, we identified the number of examinations per participant in each time bin and reported the highest (ie, worse) level of diabetic retinopathy or macular edema (Table 1) between eyes. An unable-to-determine result was coded as the highest category for both diabetic retinopathy and macular edema. However, we allowed additional testing for poor images. Therefore, the analysis used an unable-to-determine result if all eligible tests were unable to determine in this time bin.
We determined the percentage of eyes with a change in the diabetic retinopathy stage during study follow-up. For this analysis, we used both examination types (traditional and telemedicine) and excluded all unable-to-determine results. In this analysis, we evaluated each eye separately. We determined the baseline stage of diabetic retinopathy by selecting the first examination result within the 6-month or less time bin (−6 months through 6 months) that was not an unable-to-determine result. For the greater than 6-month through 18-month through greater than 42-month through 54-month time bins, we used the worst grade of retinopathy recorded within the bin. We then compared the stage of retinopathy in each time bin to the 6-month or less time bin to determine whether the stage of diabetic retinopathy had increased (worsened), remained stable, or decreased (improved).
We evaluated 646 people for eligibility; 567 (87.8%) were enrolled and 79 (12.2%) were not enrolled (78 declined participation and 1 was ineligible because he/she was not a health clinic patient). We did not find differences in age, duration of diabetes, or HbA1c level between those enrolled and those who were not. However, females were more likely to enroll than males (295 [52%] of 567 vs 272 [48%] of 567; P = .03).
Table 2 lists the demographic characteristics. A total of 411 participants (72.5%) reported a nonwhite primary, secondary, or tertiary race/ethnicity. The mean number of years since receiving a diagnosis of diabetes was 9.5 years, and participants had a mean HbA1c level of 8.3%. There were no differences in demographic and medical characteristics at enrollment between the telemedicine (n = 296) and traditional surveillance (n = 271) groups.
Figure 2 shows that the telemedicine group underwent a diabetic retinopathy screening examination more frequently than the traditional surveillance group in the 6-month or less time bin with a 50.7% difference (94.6% [280/296] vs 43.9% [119/271]; 95% CI, 46.6%-54.8%; P < .001) and the 6-month through 18-month time bin with a difference of 19.8% (53.0% [157/296] vs 33.2% [90/271]; 95% CI, 16.5%-23.1%; P < .001). After we offered telemedicine to both groups, the percentage of diabetic screening examinations was similar for subsequent years (>18-month through 30-month time bin: 131 [44.3%] of 296 vs 107 [39.5%] of 271; difference, 4.8%; 95% CI, 3.0%-6.6%; P = .27; >30-month through 42-month time bin: 127 [45.0] of 282 vs 121 [46.4%] of 261; difference, 1.4%; 95% CI, 0.4%-2.4%; P = .80; and >42-month through 54-month time bin: 115 [51.1%] of 225 vs 117 [56.0%] of 209; difference, 4.9%; 95% CI, 2.9%-6.9%; P = .34).
Figure 2 shows that the traditional surveillance group had an increase in the percentage of eye examinations starting at the greater than 18-month through 30-month time bin through the greater than 42-month through 54-month time bin, suggesting that the availability of telemedicine increased the percentage of participants receiving retinopathy screening examinations. A small percentage of participants in the traditional surveillance group continued to only use traditional examinations even after telemedicine was offered to them, with this percentage decreasing during the follow-up period. The percentage of patients receiving only telemedicine examinations vs patients receiving only traditional examinations in the traditional surveillance group was 40.5% vs 59.4% (P = .12), 61.0% vs 39.0% (P = .17), and 89.0% vs 11.0% (P < .01) in the greater than 18-month through 30-month, greater than 30-month through 42-month, and greater than 42-month through 54-month time bins, respectively. This finding suggests that, when onsite telemedicine and offsite traditional eye examinations were both an option, most participants eventually opted for telemedicine. In addition, when participants only received one type of examination, they were more likely to have a telemedicine examination.
For the 6-month or less through greater than 42-month through 54-month time bins, we could not identify a difference in the prevalence of any stage of diabetic retinopathy between those in the telemedicine group and those in the traditional surveillance group (eTable 1 and eTable 2 in Supplement 2). However, the telemedicine group had a higher percentage of unable-to-determine results for macular edema when compared with the traditional surveillance group during the 6-month or less and greater than 6-month through 18-month time bins (≤6-month: 15.4% [43/280] vs 0% [0/119]; P < .001; >6-month through 18-month: 17.2% [27/157] vs 0% [0/90]; P < .001).
Table 3 lists the percentages of patients who received a telemedicine examination that would require referral to an eye care professional based on telemedicine examination results. The data indicate that most participants did not need to be referred for follow-up. Table 3 also indicates that an unable-to-determine result for diabetic retinopathy or macular edema was a common reason for referral.
Table 4 lists the changes in diabetic retinopathy stage throughout the trial. During the study, more than 90% (range, 90.4%-94.1%) of eyes had their diabetic retinopathy stage within ±1 of their baseline diabetic retinopathy stage throughout the study. At the greater than 42-month through 54-month time bin, 35 (8.6%) of 409 participants (95% CI, 5.8%-11.2%) experienced worsening by 2 stages or more, and 5 (1.2%) of 409 (95% CI, 0.1%-2.3%) had an improvement in diabetic retinopathy by 2 stages or more. Overall, this finding suggests that levels of diabetic retinopathy were relatively stable during the study period.
This project addressed the Institute of Medicine’s recommendations for the escalating public health issue of diabetes and diabetic retinopathy18 using a first quartile priority topic of “Compare the effectiveness of interventions to reduce health disparities in diabetes…” and a second quartile priority topic of “Compare the effectiveness of new remote monitoring technologies (e.g., telemedicine) and usual care in managing chronic diseases, especially in rural settings.” We found that telemedicine increased the percentage of participants who obtained diabetic retinopathy screening examinations when compared with traditional surveillance. After all participants had access to telemedicine, the data indicate that telemedicine increased the percentage of participants receiving examinations in the long term. The severity of diabetic retinopathy remained relatively stable during the study period, and most telemedicine participants did not have levels of diabetic retinopathy that warranted referral to an eye care professional. Overall, these results suggest that primary care clinics could use telemedicine to triage and monitor patients for diabetic retinopathy over a long period.
Similar to previous studies,10,11,26,27 we found that telemedicine increased the percentage of participants who obtained diabetic retinopathy screening examinations at baseline when compared with traditional surveillance with eye care professionals. However, when our study offered telemedicine to both groups after 2 years of enrollment, the percentages receiving screening examinations became similar. When participants are offered both traditional and telemedicine diabetic screening examinations, approximately 30% of patients will use only telemedicine (Figure 2). Therefore, even when eye care professionals are available, telemedicine will increase the percentage of diabetic retinopathy screening examinations.
One advantage of screening for diabetic retinopathy with telemedicine is that it may decrease the societal burden of providing a full eye examination for every patient with diabetes. However, if most participants require subsequent referral, screening examinations would actually increase health care costs. Similar to previous studies,24,25 we used moderate diabetic retinopathy or worse, macular edema, or an unable-to-determine finding from a telemedicine examination as the cutoff for recommending further evaluation with an eye care professional. Using this cutoff, only a few participants would be referred to an eye care professional. We also found that most patients had stable levels of retinopathy during the study.
We found that the percentage of annual diabetic screening examinations from greater than 18 months through 54 months was between 40% and 55% despite the availability of telemedicine to all participants. This percentage is below the National Committee for Quality Assurance’s requirement of 60% for the Diabetes Recognition Program.28 This finding suggests that, although telemedicine has the potential to increase the percentage of patients receiving diabetic retinopathy screening examinations, other barriers to obtaining examinations exist. Sheppler and colleagues29 found decreased adherence when participants (1) believed their medical insurance did not sufficiently cover the costs of diabetic eye examinations, (2) had uncontrolled blood glucose levels, or (3) had a shorter duration of diabetes. Future studies should include multiple types of interventions, such as telemedicine with health education and promotion, to improve diabetic retinopathy screening percentages.
Poor-quality images are a common reason for referral in a telemedicine diabetic retinopathy screening program10 as indicated by the current study. Poor-quality nonmydriatic imaging may occur because of small pupil size or ocular media abnormalities (eg, cataract). Future studies are needed to determine the best imaging method to decrease the percentage of unreadable images.
Our study has several important findings for researchers and physicians proposing telemedicine as a tool to increase the percentage of patients screened for diabetic retinopathy. However, our study also has limitations. The study population included a high percentage of participants who had transient housing and moved in and out of the health care system. Consequently, communities that display more stable housing may actually observe higher percentages of patients receiving long-term follow-up. We developed a health belief questionnaire during the last year of the study and invited all active participants to complete the survey.29 This approach may have increased the percentage of follow-up in both groups during this time because we offered a small monetary incentive ($25) for completing the questionnaire.
Overall, our findings suggest that primary care clinics can effectively use telemedicine to triage and monitor patients for diabetic retinopathy over a long period. Although telemedicine with nonmydriatic cameras may detect many eye diseases, it may miss ocular hypertension or refractive error. Therefore, we encouraged all participants to see an eye care professional regardless of their group assignment. Future studies should evaluate whether patients require a comprehensive eye examination with an eye care professional if a telemedicine result does not meet referral criteria and participants have no symptoms of eye disease.
Submitted for Publication: October 10, 2014; final revision received December 12, 2014; accepted December 20, 2014.
Corresponding Author: Steven L. Mansberger, MD, MPH, Devers Eye Institute/Discoveries in Sight, Legacy Health, 1040 NW 22nd Ave, Ste 200, Portland, OR 97210 (firstname.lastname@example.org).
Published Online: March 5, 2015. doi:10.1001/jamaophthalmol.2015.1.
Author Contributions: Dr Mansberger 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.
Study concept and design: Mansberger, Wooten, Becker.
Acquisition, analysis, or interpretation of data: Mansberger, Sheppler, Barker, Gardiner, Demirel, Wooten.
Drafting of the manuscript: Mansberger, Sheppler, Barker, Wooten.
Critical revision of the manuscript for important intellectual content: Mansberger, Barker, Gardiner, Demirel, Becker.
Statistical analysis: Mansberger, Sheppler, Barker, Gardiner.
Obtained funding: Mansberger, Becker.
Administrative, technical, or material support: Sheppler, Demirel, Wooten.
Study supervision: Mansberger, Becker.
Conflict of Interest Disclosures: None reported.
Funding/Support: This research was supported by grant NEI 3 K23 EY0155501-01 from the National Eye Institute, grants CDC U48DP000024-01 and 1U48DP002673-01 from the Centers for Disease Control and Prevention, and the Good Samaritan Foundation at Legacy Health (Dr Mansberger).
Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and the decision to submit the manuscript for publication.
Disclaimer: The findings and conclusions in this journal article are those of the author(s) and do not necessarily represent the official position of the Centers for Disease Control and Prevention.
Previous Presentation: This study was presented at the Association for Research in Vision and Ophthalmology Annual Conference; May 6, 2013; Seattle, Washington.