Background Older people with dementia are high users of hospitals and suboptimal care processes are often associated with adverse outcomes. This study aims to measure the association between variability in care processes and healthcare outcomes. Methods Observational study using data from electronic medical records at a tertiary hospital in Melbourne, Australia. Patient cohort: Older people with dementia admitted to hospital for falls with injury, delirium, and hip fracture in 2022-2024.  Care processes included: restraint use, referrals to specialists/aged care/allied health, use of medications and nerve blocks, goals of care, and discharge orders. Patients' care was represented as sequences of events, then used to create patient trajectories. These were then compared using edit distance sequence analysis to calculate a variation score for each patient episode. The relationship between variation scores and outcomes (e.g. length of stay (LOS), readmissions) was analysed using Gamma regression. Results Overall, 766 admissions among 685 older people with dementia (53% female, median age 86 years) were included in the analysis. There were 3873 allied health and 753 aged care/delirium consults, 1208 goals of care, 206 restraint orders, and 165 discharge referrals. Patient trajectories with greater variability were associated with longer LOS. This became significant at 5 days. Further findings will be presented and implications for practice discussed. Discussion This project studied the problem of excessive variation in care and its association with poorer outcomes using novel process mining tools. This approach has the potential to be the starting point of understanding sources of unwarranted variability in care and could be scaled-up to other hospitals and datasets.