Background Best practice guidelines recommend regular evaluation of antipsychotics in managing behaviours for dementia patients with a view to de-prescribing, given its significant mortality and adverse outcomes (Reus et al. in Am J Psychiatry 173(5):543-546, 2016, Deprescribing Guidelines and Algorithms in https://deprescribing.org/resources/deprescribing-guidelines-algorithms/ , 2019). The relationship between the dose of antipsychotic and the probability of discontinuation remains unknown in hospitalized dementia patients. Objectives This study aims to examine the relationship between high dose antipsychotic (greater than 62mg chlorpromazine equivalent daily dose) and antipsychotics discontinuation in hospitalized dementia patients. Setting Specialized Dementia Behavioral Health Program in Hamilton, Ontario, Canada. Method A retrospective chart review was completed from August to December of 2019. A univariate logistic regression model was applied to antipsychotic doss (in chlorpromazine equivalent) and antipsychotic discontinuation outcome at 60days (Narayan and Nishtala in Eur J Clin Pharmacol 73(12):1665-1672, 2017). A multivariant model was used to assess potential confounders, including other psychiatric medication exposure and Medicines Comorbidity Index (Luthra in J Gerontol Geriatr Res 4(260):2, 2015). Regression and dose-response models were utilized to identify the threshold dose (maximum daily dose). Main outcome measures Antipsychotic discontinuation at 60days after the last dose. Results A total of 42 patients were eligible for outcome analysis. High dose antipsychotic was associated with worse discontinuation outcomes in both unadjusted (odds ratio, 0.09; 95% confidence interval, 0.02-0.37; p < 0.01) and adjusted generalized estimation equation models (odds ratio 0.65; 95% confidence interval, 0.59-0.72; p = 0.01). There were no statistically significant associations between baseline comorbidities (Medicines Comorbidity Index) (p = 0.68), mood stabilizer (p = 0.14), benzodiazepines (p = 0.93) and antidepressant exposure (p = 0.68) with antipsychotic discontinuation. The logistic regression model identified 40.7mg of quetiapine, 1.7mg of olanzapine and 0.51mg of risperidone as the threshold dose, balancing sensitivity and specificity. The dose-response model also identified similar doses of 42mg of quetiapine, 1.76mg of olanzapine and 0.53mg of risperidone. Conclusion The use of high dose antipsychotics is associated with worse discontinuation outcomes in hospitalized dementia patients. Therefore, our results suggest not exceeding a daily dose of 50mg of quetiapine, 1.75mg of olanzapine and 0.5mg of risperidone when used for responsive behaviours and reassess the benefits and risks for each patient regularly.