Objective: To evaluate and compare the outcome predictive power of Acute Physiologic and Chronic Health Evaluation (APACHE) Ⅳ、Ⅲ、Ⅱ scoring systems for intensive ill patients in an independent medical intensive care unit (MICU), and explore the best time point when they can most accurately predict outcome.

Design: Retrospective analysis of medical records. Methods and Materials: Collect data of patients admitted between July 2010 and July 2013 to the medical intensive care unit of the First Affiliated Hospital of Sun Yat-sen University in Guangzhou, China. Patients with an ICU stay less than 4 hours and age less than 16 years were excluded. Calculate the APACHE Ⅳ、Ⅲ、Ⅱ scores and corresponding predictive mortality risks in the first 24 hours of ICU admission and the 3th, 5th, 7th, 14th, 21 th, 28th day or the day of transferred out of ICU or death. The predictive power of each model was assessed through the ratio of observed death rates and predictive death rates (Standardized mortality ratios, SMR), the calibration of observed and predictive death rates and the discriminative ability between survivors and non-survivors. Hosmer-Lemeshow test was employed for assessing the calibration and the discriminative ability was assessed by the area under the receiver operating curve. Compare the predictive power of the three models at different time points and explore the the best time point when they can most accurately predict outcome.

Inclusion Criteria:

• Patients with an ICU stay more than 4 hours and age older than 16 years.

Exclusion Criteria:

• Patients with an ICU stay less than 4 hours and age less than 16 years.

Patients admitted between July 2010 and July 2013 to the medical intensive care unit of the First Affiliated Hospital of Sun Yat-sen University in Guangzhou, China.