(SLA-163) Multidomain frailty index from routine health data and medication use in older patients with and without dementia in Australian hospitals

Background: Frailty is a key determinant of health outcomes in older adults, particularly in hospital settings. However, the relationships between medication use, dementia and frailty in older inpatients remain unclear.

Objectives: To determine the prevalence of a multidomain frailty index derived from routinely collected health data (electronic frailty index-acute hospital; eFI-AH) in older patients with and without dementia; to examine the association between dementia diagnosis, medication use and eFI-AH; and to evaluate the predictive ability of eFI-AH for in-hospital mortality and long hospital stays (>10 days).

Methods: This retrospective study included 30,492 patients aged ≥65 years on their first admission to six hospitals in NSW, Australia, between June 2022 and July 2023. Dementia diagnosis was identified using ICD-10-AM codes. The eFI-AH, derived from 54 health deficits across multiple domains, was assessed for patients with ≥30 items recorded. The associations of medication use and dementia diagnosis with the eFI-AH were evaluated using Gamma regression, adjusting for age, gender and the number of different medications prescribed during admission. The predictive ability of the eFI-AH for in-hospital mortality and long hospital stays was evaluated using ROC-AUC, PR-AUC, and calibration plots.

Results: The eFI-AH was calculated for 23,060 of 30,492 patients (75.6%), including 2,549 (10.1%) with dementia. The proportion of patients who were robust (eFI < 0.1), prefrail (0.1≤eFI < 0.25) and frail (eFI≥0.25) were 1.2%, 21.7% and 77.1% respectively in patients with dementia and 20.5%, 50.2% and 29.4% in patients without dementia. The median eFI-AH (0.34, IQR 0.26–0.42) in patients with dementia was higher than that in patients without dementia (0.17, IQR 0.11–0.27). During hospitalisation, 600/2,549 patients with dementia received at least one specific dementia medication, and 28 were prescribed two (donepezil 412, rivastigmine 97, memantine 91, galantamine 28). Dementia diagnosis and the number of medications prescribed during admission were significantly associated with eFI-AH (both p < 0.001). Prevalence of mortality among older patients was 4.1% (dementia 9.5%, non-dementia 3.4%), and long hospital stay was 24.9% (dementia 42.7%, non-dementia 22.7%). The ROC-AUC was 0.87 for mortality and 0.78 for long hospital stay. However, the PR-AUC for mortality was low (0.24) due to its low prevalence, whereas the PR-AUC for long hospital stay was 0.51 . Calibration analysis demonstrated slopes of 1.04 for mortality and 0.99 for long hospital stay.

Conclusions: Dementia diagnosis and total medication count were strongly associated with eFI-AH.