Background: - Mortality prediction for ICU patients is critical and crucial for establishing mortality risk scores based on prognostic determinants. However, the available models currently are developed for developed countries in their contexts which make them inapplicable in low-income settings, and models in low-income settings are scarce and are not validated in more similar settings. In the Ethiopian setting in general and in the current study settings in particular, there is no clinical prediction model that can assist decision-making. Objective- The aim of this study was to develop a mortality risk score in the intensive care unit of referral hospitals in Bahir Dar city, northwest Ethiopia, 2022. Method-a retrospective cohort study was conducted among 852 ICU patients admitted from January 1, 2019-December 31, 2021. The data were extracted from March 1/2022-May 15/2022 using a checklist. Four emergency and critical care nurses (three MSc and one BSc) have extracted the data. The data were entered to EpiData version 3.1 and were exported to R-software for analysis. Multivariable logistic regression was employed to identify the independent prognostic determinants and then a mortality risk score model was fitted. After internal validation was carried out, a score for each prognostic determinant was calculated to show the relative weight of determinant variables. Results- The overall mortality rate among ICU patients was found to be 35.9%. Age, intensive care unit stay, respiratory rate, mean arterial pressure, and white blood cell count were some of the prognostic determinants identified in this study. The model developed in this study has a discrimination performance of area under the curve 0.90 (95% confidence interval 0.88 to 0.92) with a calibration p-value of 0.65 and calibration curve slope almost overlapping in 45 degrees. This does mean an excellent agreement between the predicted and the observed probabilities. Conclusion and recommendation- Our model is well-calibrated and perfectly discriminates between survived and non-survived patients. Variables used to develop the risk score are easily measurable which makes its usage simple. The overall ICU mortality rate was high. Thus, measures to reduce mortality should be strengthened. Keywords- Mortality, risk score, intensive care unit, referral hospitals, Bahir Dar city