Purpose: To test in a 'real world' diabetic eye-screening programme, a computer-based personal risk evaluation for progression to sight-threatening diabetic retinopathy. Screening intervals were individualized, and clinical outcomes, safety and cost-effectiveness documented. Methods: The RETINARISK algorithm was used in an ophthalmology clinic in Norway. The diabetes cohort was divided on voluntary basis into two groups: one with variable screening intervals based on their personal risk profile and the other group with conventional fixed interval diabetic eye screening. Compliance, clinical outcomes, safety and health economics were evaluated. A total of 843 diabetic patients participated in the program 2014-2019. A total of 63 had type 1 and 780 type 2 diabetes. A total of 671 patients had no diabetic retinopathy at baseline and 171 had retinopathy. Results: A total of 444 (53%) diabetic patients were included in the personal risk profile program and 399 in the fixed interval group. The RETINARISK algorithm calculated 563 screening intervals for the variable interval group, which was 23 AE 16 months (mean AE SD), compared to 14 AE 5 months for the group with fixed screening intervals. Due to selection bias, the two groups could not be directly compared. We did not experience any delay in detecting diabetic retinal changes when using the personal risk profile program. Conclusion: The RETINARISK algorithm was safe and effective in a diabetic screening program in an ophthalmology clinic over 5 years. The use of the program reduces the mean frequency of screening visits and liberates valuable time in ophthalmic practice to be used on high-risk diabetic patients or other patient groups.