Background: The Intensive Care Unit (ICU) is a unit in the hospital where seriously ill patients are cared for by specially trained staff.  The objectives of this study were to assess the survival, the incidence density and its determinants of ICU stay.

Methods and materials: Prospective cohort study design was conducted in the referral hospitals of Amhara region, Ethiopia. The sample size was calculated using Epi-info software. Initially baseline data were recorded from the patient chart at the time of admission to the ICU.  Patients' status was followed every day until they discharge with good prognosis. The Kaplan Meier curve was used to estimate time to discharge from the ICU. Cox proportional hazard model (Weibull) was used to identify the predictors of ICU stay.  Adjusted Hazard risks with their 95% CI were used to identify the predictors of ICU stay.

Results: A total of 2789 patients were included, the incidence density for discharge with prognosis was 1962/12448 person days and 50 % of patients discharge on 4^th day. The mean time of ICU stay for none post operative patients was 2.43 days and the mean time of ICU stay for post operative days was 1.31 days. ICU stay was determined by patient to nurse ratio AHR 0.59[95% CI: 0.56 - 0.64], chronic illness AHR 0.93 [95% CI: 0.88-0 .98],hemoglobin concentration AHR 1.09[ 95% CI: 1.05 - 1.14], ICU area to bed ratio AHR1.14 [95% CI: 1.06- 1.22], Nosocomial infection AHR  0.47 [95% CI: 0.37-0.59], Tracheotomy AHR 1.12 [95 % CI: 1.01- .24], time of admission AHR 0.83 [95 % CI: 0.75 - 0.93], formal education AHR  0.72 [95% CI: 0.64- 0.80].

Conclusion: Decision makers should give high priority for ICU infrastructures and increasing the number of nurses in the ICU wards.

Key words: ICU, survival, predictors, referral hospitals.