Skip to main content
. 2023 Feb 23;9:e41624. doi: 10.2196/41624

Table 1.

Characteristics of the simulated cohort and comparisons between community health centers using artificial intelligence and those using manual grading.


All CHCsa CHCs using artificial intelligence CHCs using manual grading P value
Community health center characteristicsb, mean (SD)

Number of full-time or part-time ophthalmologists 1.1 (0.7) 1.1 (0.5) 1.2 (0.7) .62

Annual numbers of ophthalmology outpatients 6248.6 (6019.0) 6774.8 (6238.2) 6016.2 (5946.4) .54
Screened residents’ characteristicsc

Age (years), mean (SD) 69.3 (7.2) 69.8 (6.9) 68.8 (7.4) <.001

Sex (male), n (%) 15,032 (46.0) 7225 (46.1) 7807 (45.8) .60

Duration of diabetes (years), mean (SD) 10.3 (7.7) 9.8 (8.0) 10.7 (7.3) <.001

Visual acuity of right eye (logMAR), mean (SD) 0.4 (0.4) 0.4 (0.4) 0.3 (0.4) <.001

Visual acuity of left eye (logMAR), mean (SD) 0.4 (0.4) 0.4 (0.4) 0.3 (0.4) <.001

aCHC: community health center.

bIn 2019, we conducted an investigation of the ophthalmic resources in Shanghai. Out of 250 community health centers, 111 were randomly selected. After matching with the Shanghai Digital Eye Disease Screening program data, 34 community health centers using artificial intelligence and 77 community health centers using manual grading–based telemedicine screening were investigated. We compared both these groups to see whether the community health centers’ characteristics were different between those choosing new technology and those who did not.

cA total of 32,695 residents with diabetes were screened in the Shanghai Digital Eye Disease Screening program. Among them, 15,663 residents were in the community health centers with artificial intelligence, and the rest 17,032 were in the community health centers with manual grading. We compared the residents’ characteristics in both groups of community health centers. Although some residents’ characteristics were significantly different owing to the large sample size in the 2 groups, the differences were not clinically meaningful. Therefore, it can be assumed that there is no practical difference between the 2 groups.