Efficiency and effectiveness of real-world diabetic retinopathy screening by artificial intelligent (AI) are limited. Investigators will implement AI for diabetic retinopathy screening in 13 health districts in Thailand and investigate the efficiency, effectiveness as well as patients and health care personnel's satisfaction by an implementation research.

Inclusion Criteria:

1. Type 1 or 2 diabetes mellitus patients whose name are in primary hospital record
2. No full-time ophthalmologists in those primary hospital
3. Age more than or equal to 18 years
4. Eligible for fundus photo imaging at least 1 eye

Exclusion Criteria:

1. Type 1 or 2 diabetes mellitus patients whose name are in primary hospital record that have full-time ophthalmologists
2. Patients who previously diagnosed with other causes of macular edema, for example, Age-related Macular Degeneration, Radiation Retinopathy, Retinal Vein Occlusion etc.
3. History of retinal laser or surgery
4. Other ocular diseases that require referral to ophthalmologists
5. Not eligible for fundus photo imaging for both eyes (any causes)

All diabetes mellitus patients who visit for diabetic retinopathy screening at selected primary care units and hospitals in 13 health districts in Thailand

Fear of inappropriate use of data