Study Explores Clinician’s Accuracy at Diagnosing Age-Related Macular Degeneration Despite Potential Difficulty

Published on: 

Of 1023 color fundus photograph images, the 5 optometrists in the study were able to accuracy diagnosis 226 images with AMD.

Despite potential difficulty of diagnosing AMD, clinicians in a new study demonstrated excellent accuracy in diagnosing age-related macular degeneration (AMD).1

AMD, a highly prevalent retinal disease, leads to irreversible vision loss. The disease impacts quality of life and poses a financial burden to the individual, families, and society. In Australia, AMD costs AU $750 million overall. Clinicians typically detect AMD during a regular eye exam, but if undetected, they might do a visual acuity test, ophthalmoscopy, an amsler grid test, or an eye angiogram and optical coherence tomography if they suspect a patient might have wet AMD.2

AMD can be left undiagnosed—and about 25% of people with AMD do not even know they have the eye condition. Investigators of the new study suggest this could either be because of the difficulty of detecting AMD or because patients do not routinely get their eyes checked. If it is because of the difficulty of detecting AMD, the investigators explain detecting AMD can be tricky to detect due to its heterogenous appearance. AMD detections can overlap with other macular conditions like central serous chorioretinopathy, adult vitelliform, dystrophy, and malattia leventinese. People may not be receiving AMD diagnoses because of lack of clinician knowledge surrounding the diagnosis. Also, in the early stages of AMD, the individual may not experience any symptoms, so they might not think about getting their eyes checked.3

Led by Sharon Ho, of the centre for eye health at The University of New South Wales in Sydney, Australia, the investigators conducted a prospective diagnostic accuracy study of community optometrists practicing in Australia.1

“In the current study, the diagnostic accuracy of community optometrists for AMD was estimated as ‘excellent’ with an [receiver operating characteristics curve] of 0.86, sensitivity of 84.5% and specificity of 88.0%,” the investigators wrote.

The study included 5 practicing optometrists, with 60% being female (n = 3), and the mean age being 30.2 years old. Over half (60%) worked in an independently owned practice, and 40% in a corporate practice. Clinical practices were located either in a major city (60%) or a regional centre (40%) in New South Wales.

The 5 optometrists examined 1023 images obtained from color fundus photographs to identify AMD. The team unidentified images so they could not be linked to a patient’s age, gender, history, or other clinical characteristics. They also created 5 image datasets by random assigning.

Of the images, 226 featured AMD and 797 featured other ocular conditions or no abnormal findings. The investigators found 27% of the images with AMD had other significant ocular condition like other posterior diseases, anterior eye diseases, and glaucoma. For images without AMD, 62.2% had other significant ocular conditions like other posterior eye disease, glaucoma, and anterior eye disease. About half (52%) of the images were from females (n = 536), and the mean age of patients in the images was 61.2 years old.

The investigators assessed diagnosis accuracy by the area under the receiver operating characteristics curve (aROC). They also calculated specificity, as well as positive and negative predictive values. For each image, they had to say if AMD was present with a yes or no. If they answered “yes,” they had to identify the stage of AMD in the image as either early, intermediate, late geographic atrophy, or late neovascular AMD). The optometrists had 4 weeks to go through the images with an option of a 2-week extension.

Using the receiver operating characteristics curve, diagnostic performance was characterized as: no discrimination (aROC = 0.5), poor (0.5 < aROC < 0.7), acceptable (0.7 < aROC < 0.8) excellent (0.8 < aROC < 0.9), and outstanding (aROC > 0.9).

The receiver operating characteristics curve for community optometrists’ diagnosis of AMD was 0.86 (95% CI, 0.83 – 0.89), making it an excellent score. When testing for diagnosis sensitivity and specificity, the investigators received 84.5% for sensitivity (95% CI, 79.1 – 89.99) and 88% for specificity (95% CI, 5.5 – 90.1).

The team also tested the diagnosis accuracy for each stage of AMD, receiving 0.82 (95% CI, 0.73 – 0.91) for early AMD, 0.76 (95% CI, 0.72–0.81) for intermediate AMD, 0.69 (95% CI, 0.49 – 0.90) for late geographic atrophy AMD, and 0.55 (95% CI, 0.34 – 0.75) for late neovascular AMD.

“Wide 95% CIs indicated that performance ranged from ‘no discrimination’ to ‘excellent’ for diagnosing late geographic atrophy AMD and ‘no discrimination’ to ‘acceptable’ for diagnosing late neovascular AMD,” the investigators wrote. “This is likely due to the small participant sample size (n = 5) and small number of late AMD images in the study dataset (n = 10) for late geographic atrophy AMD, n = 9 for late neovascular AMD).”

The investigators noted optometrists often underestimate AMD severity, which can slow down patient care. Clinical guidelines recommend referral within 1 week of a treatable neovascular lesion—but if patients are undiagnosed, AMD vision loss will continue to worsen. While the study had a small sample size, the results still demonstrated clinician’s accuracy in AMD diagnosis.

“Therefore, future studies involving more participants and larger balanced test datasets are required to explore this issue,” the investigators concluded.


  1. Ho S, Doig GS, Ly A. Diagnostic accuracy of community optometrists for age-related macular degeneration using colour fundus photographs: A pilot evaluation. Ophthalmic Physiol Opt. Published online November 3, 2023. doi:10.1111/opo.13242
  2. Macular Degeneration Diagnosis. Standard Medicine. Accessed November 28, 2023.
  3. Age-Related Macular Degeneration (AMD). Hopkins Medicine. Accessed November 28, 2023.