Poster Presentation Australian and New Zealand Society for Geriatric Medicine Annual Scientific Meeting 2026

Measuring what matters: Psychotropic stewardship metrics for AI-enabled dementia care in residential aged care. (#255)

Bashar Adam 1 2 , Joby Alex 2 3 , Yenna Salamonson 2 , Caleb Ferguson 2 3
  1. Community Geriatric Outreach Services, Sutherland Hospital, South East Sydney Local Health District, Sydney, NSW, Australia
  2. School of Nursing, University of Wollongong, Wollongong, NSW, Australia
  3. Centre for Chronic & Complex Care Research, Blacktown Hospital, Western Sydney Local Health District, Blacktown, NSW, Australia

Aims: To propose a measurement-first evaluation framework using artificial intelligence (AI)-enabled decision support in Australian residential aged care facilities (RACFs), focusing on psychotropic stewardship for behavioural and psychological symptoms of dementia (BPSD).

 

Design & Methods: A horizon-thinking synthesis of AI functions in the gerontological literature was conducted to identify measurable, audit-ready endpoints and associated implementation risks.

 

Results: Dementia affects over 50% of Australian RACF residents, with BPSD occurring in up to 90% of cases.(1) Currently, 40-66% of residents with dementia are prescribed psychotropics, medications linked to significant side effects and around 1.4-fold increase in mortality within 180 days of initiation.(2-6)

 

Two primary AI functional domains were identified:

(1) Clinical monitoring and predictive analytics: Identifying risk states such as falls, delirium and dehydration

(2) Behaviour decision strategies: Prompting reversible-cause screening and prioritised non-pharmacological responses.(9)

 

A four-indicator stewardship core was specified for routine audit-and-feedback: PRN antipsychotics per 100 resident-days; benzodiazepine exposure, time to first documented non-drug action, and refusal-of-care episodes resolved without PRN psychotropics.

Seven governance requirements were mapped for AI-enabled decision support in RACFs: (1) the tool should be lawful, (2) privacy-preserving use, (3) transparency at the point of care, (4) clinical validation, (5) ability for human override, (6) bias and performance drift monitoring, and (7) accountable audit trails.(9-11)

 

Conclusions: As a literature-derived framework, AI implementation in RACFs should be evaluated by its capacity to drive measurable reductions in psychotropic reliance through timely, non-pharmacological care, rather than its technical novelty. This framework enables audit-ready psychotropic stewardship and consistent non-pharmacological-first dementia care.