The use of the electronic frailty index to optimise HbA1c targets in older individuals with type 2 diabetes

Abstract Introduction Current guidelines advocate individualised HbA1c targets among frail individuals, but the evidence base remains limited. We therefore evaluated whether electronic frailty index (eFI) could improve model performance for mortality and hospital admission, and examined the association between HbA1c levels and clinical outcomes across frailty strata in large UK primary care population. Methods We included patients with type 2 diabetes mellitus aged ≥65 years within the UK based Clinical Practice Research Datalink (CPRD GOLD) and Hospital Episode Statistics from 2014–19. Step-wise multivariate Cox regression models were developed for all-cause mortality and hospital admission. Model performance before and after the addition of eFI using Akaike’s information criteria (AIC) and Harrell’s C statistics was analysed and the risk for both outcomes associated with eFI and HbA1c groups were assessed. Results N = 179 723 individuals (mean age 74.8 ± 7.6 years; diabetes duration 7.3 ± 6.1 years) were included and followed over 1.7 years. Incorporating eFI may improve model performance for both mortality and admission outcomes, reducing AIC (ΔAIC -1423.1 and ΔAIC -308.4 for admission and mortality, respectively) and improving Harrell’s C statistics (standard model vs. EFI; 0.616 vs. 0.631 and 0.776 vs. 0.778). All-cause mortality risk was lowest at HbA1c 48–57.9 mmol/mol and increased at both lower and higher HbA1c levels across all frailty groups. The highest risk of hospital admission was consistently observed at HbA1c 48 mmol/mol, particularly in individuals with severe frailty (HR 4.33; 95% CI: 3.64–5.15). Conclusions eFI may enhance model performance for mortality and hospital admissions and may support clinical decision-making by identifying frail individuals for whom intensive glycaemic control is less appropriate.