Introduction: Dementia increases the risk of falls and fall-related injuries, which may be caused by inappropriate medication use. To date, there is little evidence on which medications are more likely to cause falls. We therefore investigated the effects of medication use and medication changes 48 h before falls in hospitalised patients with dementia. Methods: This matched case-control study included 74 patients with a mean age of 83 years (38% women) who had been hospitalised for at least 7 days. Information on medications, diagnoses, disease severity, use of walking aids, falls, and demographics was collected from electronic medical records. The effects of the number of medications and psychotropics, equivalent daily doses of antidepressants, antipsychotics and benzodiazepines, anticholinergic burden, medication initiation, dose change, medication discontinuation, as-needed medications, opioid use and the presence of fall-increasing diseases were examined separately for the periods 0 h–24 h and 24 h–48 h before the falls using binomial logistic regression analyses. Results: Falls increased significantly with higher daily antipsychotic doses 24 h before the fall. In addition, the rate of falls increased with higher anticholinergic burden and prevalence of medication discontinuation 24–48 h before the fall. Notably, the total number of medications and psychotropic medications had no effect on the incidence of falls. Conclusion: With regard to the short-term effects of medication on fall risk, particular attention should be paid to the daily dose of antipsychotics, anticholinergic burden and medication discontinuation. Further studies with larger samples are needed to confirm the results of this study.