Design and Evaluation of a Telemedicine Program for Diagnosis of Urgent Ophthalmic Complaints

Abstract

Purpose : The need for cost-effective, accessible ophthalmic care has led to a growing interest in telemedicine. Past studies have validated the use of teleophthalmology for screening, but not for diagnosis. This prospective, observational pilot study assesses the ability of a generalizable teleophthalmology program to accurately diagnose and triage urgent ophthalmic complaints.

Methods : 45 patients who presented to the Massachusetts Eye and Ear Infirmary Emergency Ward with urgent ophthalmic complaints (aged 24-83, mean age 54.8; 23 female) completed a questionnaire consisting of a brief medical history, chief complaint and history of present illness. Anterior segment photos of the eye were collected via smartphone camera and posterior segment photos by digital fundus camera. These materials, in addition to ophthalmic vital signs typically collected by referring physicians, were given to three staff physicians uninvolved in the patient’s in-clinic care for remote diagnosis and determination of urgency. The remote findings were then compared to in-person clinical findings. Primary outcomes included the concordance on triage status and diagnosis between virtual and live examination, as well as the degree of agreement among virtual examiners’ triage statuses and diagnoses. The confidence level of the virtual examiners for their exam was also measured.

Results : Of the 45 patients, 24 (53.3%) had urgent triage status and there were 23 distinct diagnoses. Sensitivity for the triage status was 79.2% and specificity was 95%. Five patients were incorrectly triaged as non-urgent by teleophthalmology and one patient was incorrectly triaged as urgent relative to the live reference standard. Of the 45 patients in the pilot, 33.3% did not have concordance of diagnosis; however 10 out of 15 (66.6%) of these patients did have a correct triage status despite an incorrect virtual diagnosis. The average confidence for all examinations was 5.85 (out of 10), with an average confidence of 6.03 for correct triage, 4.72 for incorrect triage, 6.38 for correct diagnoses, and 4.80 for incorrect diagnoses.

Conclusions : In this pilot, we found that a teleophthalmic approach may reliably triage patients with urgent complaints, but is less reliable for diagnosis. A larger study is needed to establish significance and to determine subgroups of chief complaint or presenting symptoms for which the protocol can be used with high sensitivity.

This is an abstract that was submitted for the 2017 ARVO Annual Meeting, held in Baltimore, MD, May 7-11, 2017.