INTRODUCTION: The integration of advance care planning (ACP) as part of the comprehensive geriatric assessment (CGA) of hospitalised frail elderly patients, together with the clinical and demographic factors that determine successful ACP discussion, has not been previously explored. METHODS: A cross-sectional study on patients and family caregivers admitted under the geriatric medicine department of a tertiary hospital was conducted from October 2015 to December 2016. RESULTS: Among 311 eligible patients, 116 (37.3%) patients completed ACP discussion while 166 (53.4%) patients declined, with 62 (37.3%) of the decliners providing reasons for refusal. Univariate logistic regression analysis showed that older age, higher Charlson Comorbidity Index, poorer functional status and cognitive impairment had statistically significant associations with agreeing to ACP discussion (p < 0.05). On multivariate logistic regression analysis, only poorer functional status was significantly associated (odds ratio 2.22 [95% confidence interval 1.27-3.87]; p = 0.005). Among those who completed ACP discussion, a majority declined cardiopulmonary resuscitation (79.3%), preferred limited medical intervention or comfort care (82.8%), and opted for blood transfusion (62.9%), antibiotics (73.3%) and intravenous fluid (74.1%) but declined haemodialysis (50.9%). Decision-making was divided for enteral feeding. Among decliners, the main reasons for refusal were 'not keen' (33.9%), 'deferring to doctors' decision' (11.3%) and 'lack of ACP awareness' (11.3%). CONCLUSION The feasibility and utility of integrating ACP as part of CGA has been demonstrated. Poorer functional status is significantly associated with successful ACP discussion. Greater public education on end-of-life care choices (besides cardiopulmonary resuscitation) and follow-up with decliners are recommended.