Purpose
To directly assess diagnostic and referral-time variability between an optometrist, general ophthalmologist, and retinal subspecialist in diabetic teleretinal evaluation in an urban, predominantly Hispanic population.
Methods
Three telemedicine readers evaluated masked images obtained using a non-mydriatic camera from 178 eyes of 90 patients with diabetes mellitus participating in a teleretinal imaging program. These readers consisted of a VA-affiliated optometrist (D.H.), a general ophthalmologist (C.K.), and a fellowship-trained retina specialist (L.K.). All readers completed the standard telemedicine reader training program instituted by the Veteran’s Affairs Hospital System. In addition to the images, each masked telemedicine case included a brief summary of the patient's demographics and medical history (including last hemoglobin A1c). In order to assess intra-reader reliability, 10 cases were duplicated. All images were acquired at a single urban primary care clinic located in Chelsea, MA from 2008 to 2010. Kappa statistics were used to assess inter- and intra-reader variability.
Results
The readers demonstrated moderate to substantial inter-observer agreement upon the presence or absence of retinopathy (κ=0.67; p≤0.0008), macular edema (κ=0.67; p≤0.0001), and the severity of retinopathy (κ=0.59; p ≤ 0.005). In addition, there was moderate agreement regarding recommended referral-time to clinic exam by an ophthalmologist (κ=0.55; p ≤ 0.01) and regarding which images were of adequate quality for diabetic teleretinal evaluation (κ=0.43; p ≥ 0.56), but the latter was not statistically-significant. In addition, intra-reader agreement was slight to almost perfect (κ=0.11-0.84; p<0.05) in each of the above categories.
Conclusions
Moderate to substantial intra- and inter-observer agreement was found amongst three telemedicine readers: an optometrist, general ophthalmologist, and retinal subspecialist. This is the first study directly comparing the variability in diabetic teleretinal evaluation between readers with these three training backgrounds, and suggests that the type of eye care provider training does not contribute to significant variability in diabetic teleretinal evaluation when a standardized training program is utilized.
Keywords: 499 diabetic retinopathy •
550 imaging/image analysis: clinical •
465 clinical (human) or epidemiologic studies: systems/equipment/techniques