Introducing personalised risk based intervals in screening for diabetic retinopathy: development, implementation and assessment of safety, cost-effectiveness and patient experience.

This five year research programme aims to develop a major enhancement to the existing screening process for sight threatening diabetic retinopathy (DR, STDR) for implementation in the NHS. Important evidence gaps in screening for DR will be tackled comprising collection of prospective patient centred data in a whole population cohort study and new knowledge produced on outcomes relevant to patients such as visual impairment (VI). An individual risk-based approach will be developed and evaluated based on evidence collected from a well established diabetes eye care pathway in Liverpool. Safety, cost-effectiveness and acceptability to patients and staff will be assessed. Our mixed quantitative and qualitative approach has been piloted in a NIHR funded programme development grant (PDG) (report available at Appendix 2) and barriers to success identified and overcome. Inter-linked workstreams (WS) will run in parallel, each informing the progress of others. Systematic review In year 1 a systematic review of evidence of variable compared with annual screening intervals will be undertaken using standard methodology. We will estimate from published data sets the rates of visual impairment (VI) and treatment. This information will inform current national screening policies and our related workstreams. Establishment of a study data warehouse In a continuation of our PDG work, demographic, systemic and ocular data will be collected retrospectively and prospectively from primary care and hospital eye service (HES) systems on people with diabetes in Liverpool, linked electronically for individual patients using NHS number. For the first time, data from those already screened positive and attending the HES will be added (retinopathy grading, visual acuity, appointment, treatment) allowing comprehensive analyses. The data warehouse will provide the routinely collected demographic and clinical data needed for the prospective observational analysis, risk engine development, economic modelling, case detection and randomised controlled trial (RCT). Prospective observational study We will use standard statistical techniques to describe the population-based frequency of development of STDR and VI, and the progression to treatment, and to model risk factors for each outcome. The effect of change in screening interval will be simulated for likelihood of progression to first laser treatment and VI. This information will inform a health economics WS and will be used to cross-validate a risk engine. Risk calculation development and testing Known risk factors will be used to develop a risk engine for progression to STDR. A generalised model using a time-varying Cox model will be used. Towards the end of year one the tool will be used to set screening intervals for a RCT based on structured assessment of risk levels and their acceptability to PPI and health professional teams. Data will be added throughout the programme to add further robustness to the model. The risk engine will be extended in later years to model risk of progression from STDR to need for treatment for use in the HES. RCT standard and test screening interval protocols During years 2 to 5 we will evaluate usual care (annual photographic screening) versus individualised variable interval screening (intervals assigned by risk engine) in people with diabetes (PWD) registered with participating Liverpool GPs and eligible for photographic screening. The primary outcome will be non-attendance rate. Secondary outcomes will include STDR detection rate, retinopathy level, false positives, cost of screening and treatment, visual acuity, VI due to DR, missed appointments, patient acceptability measures and health related quality of life (EQ-5D). Health economics modelling An economic model will be constructed for usual care versus variable interval screening. The model will be populated with data from other work streams, such as the risk engine, in addit