. 2022 Mar 17;140(4):383–391. doi: 10.1001/jamaophthalmol.2022.0244

Table 3. Diagnostic Accuracy of the Model Developed Using the Ocular Hypertension Treatment Study Endpoint Committee Determination of End Points Based on Optic Disc Changes in Independent Test Sets^a.

Source	Country	Ground truth determination	Test set size (images), No.		AUROC (95% CI)		Sensitivity
Source	Country	Ground truth determination	Healthy	Glaucoma	All eyes	Mild glaucoma (VF MD ≥−6 dB)	80% Specificity	85% Specificity	90% Specificity	95% Specificity
DIGS¹⁸	US	Agreement by 2 expert graders from the University of California, San Diego, reviewing photographs only	5184	4289	0.74 (0.69-0.79)	0.71 (0.65-0.76)	0.59	0.52	0.43	0.30
ACRIMA¹⁶	Spain	1 Expert grader reviewing photographs only	309	396	0.74 (0.70-0.77)	Not available	0.55	0.46	0.38	0.29
LAG¹⁷	China	Glaucoma specialists from Beijing Tongren Hospital reviewing photographs, IOP, and VF	3143	1711	0.79 (0.78-0.81)	Not available	0.66	0.59	0.53	0.42

Abbreviations: ADAGES, African Descent and Glaucoma Evaluation Study; AUROC, area under the receiver operating characteristic curve; DIGS, Diagnostic Innovations in Glaucoma Study; IOP, intraocular pressure; LAG, Large-scale Attention-Based Glaucoma; MD, mean deviation; VF, visual field.

^{^a}

The DIGS/ADAGES data set consisted of photographs of 227 women and 151 men with a mean age of 64.0 years (95% CI, 63.5-64.6). Demographic characteristics of the LAG and ACRIMA study population are not readily available.

Table 3. Diagnostic Accuracy of the Model Developed Using the Ocular Hypertension Treatment Study Endpoint Committee Determination of End Points Based on Optic Disc Changes in Independent Test Setsa.

Table 3. Diagnostic Accuracy of the Model Developed Using the Ocular Hypertension Treatment Study Endpoint Committee Determination of End Points Based on Optic Disc Changes in Independent Test Sets^a.