This study was to clarify the prevalence of do-not-resuscitate (DNR) and identify the factors related to critically ill patients who have DNR directives or not in a medical Intensive Care Unit (ICU) in Taiwan. Materials and Methods: A retrospective chart review of 100 critically ill patients expired between January and December 2012 were included. The outcome was DNR or not when patient expired. Other variables regarding patient’s demographics, disease-and DNR-related information were recorded. Logistic regression model was used to assess the related factor about DNR. A P < 0.05 was considered statistically signifi cant. Results: DNR rates were 87%, and the mean interval from DNR signature to death was 3.9 days. Compared with the patients without DNR signature, the patients with DNR signature had no statistical signifi cance of cancer diagnosis (odds ratio [OR] = 3.41, 95% confi dence interval [CI] = 0.88-13.25, P = 0.076), and frequency of ICU admission (OR = 4.17, 95% CI = 0.92-18.86, P = 0.063). In addition, there were 4.22-fold (95% CI = 0.90-19.89) but no statistical signifi cance (P = 0.068) of the frequency of ICU admission by patients with DNR directives compared to those without DNR directives after adjusting the variables of age, gender, economic status, primary diagnosis, and level of consciousness. Conclusion: Although this study indicated no statistical signifi cance, we found that a patient with a cancer diagnosis and more frequency of ICU admissions tended to infl uence on family members concerning DNR directives in clinical setting. Early initiation of palliative care and DNR discussion may enhance the quality of care for dying patients.