Fig. 11.5 Composite image of patient with large drusen with standard color fundus image (top image) and DigiScope® image (lower image)

With DigiScope® technology, the incidence of unreadable images was found to increase rapidly with age, resulting in a higher total rate of referral for older patients. Others have also reported that age is the strongest predictor of unreadable images for persons with diabetes, utilizing either mydriatic or nonmydriatic fundus imaging systems [8]. As expected, small pupil size and media opacity occurred more frequently in the older age group as reasons for unreadable DigiScope® images [12]. However, referral of these patients is not a drawback in a screening program since older individuals are at higher risk for ocular pathology and, ideally, should be seen by an ophthalmologist even if they do not have diabetic retinopathy. Cavallerano and collaborators found that the majority of their patients who were referred due to unreadable images actually had ocular disease which would have resulted in referral if adequate images had been obtained [14].

11.7Unreadable Images

One of the concerns when using fundus photography outside of the traditional eye care arena has been the potential for a large proportion of unreadable images. The relatively low rate of unreadable images with our system indicates that the DigiScope® and its operation by nonophthalmic staff can successfully produce useful fundus images. The percentage of unreadable images with the DigiScope® is similar to that reported by other diabetic retinopathy assessment systems, in particular those utilizing nonophthalmic personnel to capture images [13]. The explanation for poor images is not always clearly delineated, and there may be several causes for each unreadable image. In many cases, the reason for an unreadable image may be inferred from the video image of the pupil included with each DigiScope® fundus image. Reasons for unreadable images include poor patient fixation, poor pupil centering, small pupil size (dilating drops not used or not given ample time to take effect), media opacity, and instrument and operator problems [12]. A specific cause for the unreadable image cannot always be determined. The influence of age on the rate of unreadable images has also been investigated.

11.7.1Impact on Overall Diabetic Retinopathy Assessment Rates

In spite of the plethora of telemedicine diabetic retinopathy programs, research reporting on the overall relative value and impact of telemedicine digital retinal assessment is generally lacking. On a small scale, we have been able to show a significant impact on rates of diabetic retinopathy assessment in individual practice settings. In a retrospective 4-year study, all patients with diabetes in a multispecialty primary care practice were tracked for a documented evaluation for diabetic retinopathy with either remote telemedicine imaging or a comprehensive eye examination [19]. In the first year of the study, which served as a baseline prior to implementation of the DigiScope® system, the practice had 1,257 patients with diabetes, and only 15% (190 patients) had a documented assessment for diabetic retinopathy. The DigiScope® program was initiated in the second year of the study. Documented rates of diabetic retinopathy assessment increased steadily to 51% (698 patients) in year 2 and 71% (994 patients) in year 3. By year 4, when the practice had a total of

1,709 patients with diabetes, the rate of diabetic retinopathy assessment had increased to 85% (1,449 patients). Of the observed increase in number of evaluations, interestingly, 78% was due to increased examinations by ophthalmologists, and only 22% was attributable to remote telemedicine evaluation. The total number of patients with diabetes and the total number evaluated by an ophthalmologist increased significantly over 4 years, while the number of remote retinal imaging assessments increased initially and then stabilized by year 4. These results certainly suggest that implementation of a remote telemedicine evaluation system in the primary care setting can significantly improve compliance with rates of documented evaluations for diabetic retinopathy.

11.7.2Compliance with Recommendations

Clearly, a diabetic retinopathy surveillance system is effective only if a majority of the patients identified with vision-threatening disease seeks further evaluation by an ophthalmologist for possible treatment. However, limited information is available to evaluate if digital diabetic retinopathy surveillance systems in clinical practice actually result in an increase in the delivery of appropriate diagnosis and treatment of retinopathy in the diabetic population. In a retrospective observational study, we have assessed compliance with recommendations for further evaluation by an ophthalmologist based on recommendations made after review of DigiScope® images [20]. Over a 14-month study period, 1,453 consecutive patients with diabetes were evaluated with remote images obtained in their primary care physician’s office. Follow-up data were collected for patients with sight-threatening disease – namely, for the 52 (4%) patients for whom urgent referral was recommended. The majority of urgent referrals, 67%, were for proliferative diabetic retinopathy. Two patients were deceased at the time of fol- low-up. Of the remaining 50 patients, verification that an ophthalmic examination occurred was documented in 45 (90%) of the cases. Four of the five patients who did not follow through

with recommendations refused to see an ophthalmologist or did not keep their appointment, and one patient was lost to follow-up. Time to evaluation by an ophthalmologist varied from 1 to 319 days after imaging (mean 61 days). While it is recommended that urgent referrals be seen as soon as practical, and ideally within 1 week, only five patients were seen within this time frame after imaging. The time to evaluation includes an average of 24–48 h turnaround time between imaging and return of a report to the primary care physician. At least seven patients had documentation of laser treatment which was performed as a result of the referral recommendation, but information on treatment was not available for all patients. Since these patients were previously noncompliant with recommended diabetic eye care guidelines, it is most likely that their visionthreatening disease would not otherwise have been identified in a timely manner.

There are few studies specifically assessing the level of adherence with referral recommendations made by diabetic retinopathy assessment systems in the primary care setting. A recent study in an American Indian population found that, after implementation of telemedicine diabetic retinopathy surveillance technology, a 51% increase in the rate of laser treatment for diabetic retinopathy occurred over a 5-year period [21]. While not directly comparable to the DigiScope® platform described in this report, the United Kingdom national diabetic retinopathy screening program provides additional insight into compliance with consultation recommendations in “screen-positive” patients. At one program, 84% of 352 patients referred for diabetic retinopathy were evaluated by an ophthalmologist as recommended. However, only 33% with proliferative retinopathy were seen within 2 weeks as recommended [22]. A national survey of screening programs in the United Kingdom reported that almost half had waiting lists for patients who were identified as needing further assessment and treatment [23]. These examples demonstrate that implementation of diabetic retinopathy assessment programs may improve overall compliance with recommendations for further eye evaluations in patients with diabetes, but evaluations may not