Development and Evaluation of the Electronic Frailty Index+ (eFI+) Tool: Integrated Prognostic-decision Modelling to Target Interventions for Older People With Moderate or Severe Frailty

Research questions i) How should electronic frailty index (eFI) components be combined with additional routine primary care data to develop prognostic models for predicting key outcomes of requirement for home care, falls/fractures, nursing home admission and mortality in older people with moderate or severe frailty? ii) Can model predictive performance be improved through addition of data from measures that are practical for primary care use, but not available in routine data? iii) How should risk predictions from the prognostic models be translated into a decision analytic model (DAM) to guide clinical management? iv) What is the potential cost-effectiveness of implementing interventions targeted at subgroups of older people with frailty in routine NHS care? Background Lead applicant Clegg led the eFI development, validation and national implementation. This has been translated into major UK health policy change through inclusion in the 2017/18 GP contract, which supports frailty stratification using the eFI, and UK National Health Service Long Term Plan. Aim To develop and evaluate the eFI+, a prognostic tool supplementing the original eFI including 4 integrated prognostic-decision models. The eFI+ will stratify older people with moderate or severe frailty into subgroups most likely to benefit from key interventions (community rehabilitation; falls prevention; comprehensive geriatric assessment; advance care planning). Methods Design Prognostic model development, internal validation and external validation using large datasets (ResearchOne, SAIL databank, Leeds Data Model) and cohort study data (CARE75+), with linked DAM and health economic analysis. Population Patients ≥65 with moderate or severe frailty, defined by the existing eFI. Key outcomes 12-month outcomes for prognostic models: * New/increased home care package * Emergency Department (ED) attendance/hospitalisation with fall/fracture * Nursing home admission * All-cause mortality Statistical methods i) Prognostic modelling The investigators will build 4 separate prognostic models for our 4 key outcomes by combining the eFI with additional individual-level routine data, informed by reviews to identify prognostic factors. Each model will be developed and internally validated in one large dataset, to adjust for potential overfitting, with subsequent external validation of predictive performance in a second large dataset. Separately, the investigators will use CARE75+ (n≈1,200) to investigate additional predictive value of clinical measures practical for primary care (e.g. gait speed, activities of daily living, loneliness). ii) Decision analytic model (DAM) The investigators will translate the prognostic models into a framework to support clinical decision-making, in co-production with stakeholders/PPI. The investigators will integrate prognostic models with effect size estimates from systematic reviews/meta-analyses to identify relevant thresholds of predicted risk, above which implementation of our key interventions would be warranted. iii) Health economic evaluation 12-month and long-term cost effectiveness models will be developed, informed by the DAM.