A mixed methods validation of technology enhanced macula services

Background Ophthalmology's lead as NHS England's busiest outpatient specialty is growing, with 9 million 2018/19 appointments (8.2%). The increasing prevalence of exudative age-related macular degeneration (exAMD) has fed this growth. People with exAMD visit clinic every 4-16 weeks for intravitreal injections (IVI).

If IVIs are too infrequent patients irreversibly lose vision. Already, observed IVI intervals are often longer than intended, due to inadequate ophthalmologist availability. Consequently, in 2018 the Royal College of Ophthalmologists requested 22% more consultant ophthalmologists.

A pre-existent artificial intelligence (AI) package could remove this need through its potential to independently direct treatment.

However, the absence of data informing NHS adoption of this AI perpetuates its pre-clinical status and deprives patients from its benefits.

Aim To validate pre-existent AI for exAMD treatment planning and establish how it can be adopted to improve NHS service capacity and quality.

Objectives Synthesise qualitative evidence for AI adoption in all healthcare settings and inform the design of primary qualitative research targeting specified healthcare niches including exAMD services. Inform the placement of AI within the exAMD clinical pathway and disseminate barriers and facilitators to its adoption.

Validate a pre-existent AI package's treatment planning performance relative to real-world exAMD care.

Establish the consequences to patient care of using alternative strategies to address the demand-capacity mismatch in exAMD services; telemedicine and clinician diversification.

Methods Qualitative evidence synthesis of research examining the adoption of automated healthcare decision aids from clinician, patient and carer perspectives.

Semi-structured interviews with patients, carers, hospital ophthalmic professionals, community optometrists and NHS commissioners and managers will be conducted to explore perceptions of how AI should fit into the exAMD clinical pathway and the barriers and facilitators to its adoption. Imaging and clinical data archived from 384 exAMD patient visits will be identified in the electronic medical record.

A single AI package will analyse imaging to recommend a treatment interval for each visit.

Medical retina consultants will independently review imaging and clinical data to recommend a gold standard IVI interval for the same visits.

The disagreement of both the real-world and hypothetical AI IVI intervals with these gold standards will be calculated, then AI performance will be tested against the real-world for non-inferiority.

Consultants will undertake blinded telemedical review of their own archived face-to-face consultations, creating paired decisions for comparison.

Seniority and profession of clinicians making real-world clinic decisions will be analysed using Bland-Altman plots to define any impact on patient treatment.

Impact Personal: I will develop the ability to deliver patient benefit from AI advancements, through elevating my career trajectory towards translational clinical data science.

Patient: Hundreds of thousands of people depending on injections to keep their sight will be able to access timely and convenient treatment sooner.

Society: Outside of exAMD services, the public will benefit, as previously committed human and financial NHS resources are freed for other unmet healthcare needs and a route to AI adoption is set to lead innovation in other NHS services.