Objective To compare healthcare utilisation outcomes among older hospitalised patients with and without cognitive impairment, and to compare the costs associated with these outcomes. Methods Retrospective cohort study of administrative data from a large teaching hospital in Melbourne, Australia from 1 July 2006 to 30 June 2012. People with cognitive impairment were defined as having dementia or delirium coded during the admission. Outcome measures included length of stay, unplanned readmissions within 28 days and costs associated with these outcomes. Regression analysis was used to compare differences between those with and without cognitive impairment. Results There were 93 300 hospital admissions included in the analysis. 6459 (6.9%) involved cognitively impaired patients. The adjusted median length of stay was significantly higher for the cognitively impaired group compared with the non‐cognitively impaired group (7.4 days 6.7–10.0 vs 6.6 days, interquartile range 5.7–8.3; p < 0.001). There were no differences in odds of 28‐day readmission. When only those discharged back to their usual residence were included in the analysis, the risk of 28‐day readmission was significantly higher for those with cognitive impairment compared with those without. The cost of admissions involving patients with cognitive impairment was 51% higher than the cost of those without cognitive impairment. Conclusions Hospitalised people with cognitive impairment experience significantly greater length of stay and when discharged to their usual residence are more likely to be readmitted to hospital within 28 days compared with those without cognitive impairment. The costs associated with hospital episodes and 28‐day readmissions are significantly higher for those with cognitive impairment. Copyright © 2016 John Wiley & Sons, Ltd.