UK NSC Opens Consultation on AI Grading for Diabetic Eye Screening, Deadline Aug 2026

Context Diabetic retinopathy, damage to the retina caused by high blood sugar, can lead to vision loss if not detected early. In the UK, anyone aged 12 or older with diabetes undergoes eye screening every one to two years. Trained staff—primary graders—review retinal photographs, escalating ambiguous cases to second and third graders. Rising diabetes prevalence strains the NHS screening programme, prompting interest in automated retinal image analysis systems (ARIAS) that mimic a primary grader’s decisions.

Key Facts - The UK NSC has launched a formal consultation on using AI for automated grading in the diabetic eye screening (DES) programme. Stakeholders must submit feedback by 11:59 pm BST on 3 August 2026. - A 2025 evidence summary, updating a 2021 review, examined recent studies: two from the UK and five from comparable regions (Europe, USA). Most ARIAS matched human graders in identifying severe disease but lagged in correctly ruling out disease‑free eyes. Performance varied by vendor, setting, and patient demographics such as ethnicity and age. - Four additional studies explored ARIAS impacts on patients, staff, and programme logistics, but none confirmed that UK outcomes would mirror these findings. No new UK cost‑effectiveness analyses were identified. - The summary concludes that certain ARIAS achieve parity with the first human grader and could be trialled within NHS services to collect real‑world data on patient outcomes, staff workload, and costs.

What It Means If the consultation leads to a recommendation, AI grading could streamline the DES workflow by reducing the number of images needing human review. Faster triage may free staff for more complex cases and potentially lower operational costs, though the evidence does not yet prove cost savings. Clinicians should note that AI systems currently perform less reliably in confirming healthy eyes, which could increase false‑positive referrals if deployed without safeguards.

Practical takeaways for patients: participation in screening will likely continue unchanged in the short term. For providers, preparing for pilot projects may involve training staff to interpret AI outputs and monitoring performance across diverse patient groups.

Looking ahead, the UK NSC will assess consultation responses before deciding whether to endorse AI grading trials. Watch for the committee’s updated recommendation later in 2026, which could shape the next phase of diabetic eye screening in the NHS.