Introduction. Mortality rate prediction in severely ill patients plays a key role in establishing an adequate therapeutic conduct and allocation of necessary resources into the early identification of high risk mortality. The aim. Comparative evaluation of mortality prediction scores and identification of the areas of most efficient use of prognostic tools in ICU (SAPS, APACHE, SOFA, MODS). Material and methods. We have analysed multiple analytical and descriptive studies, identified via searches on the databases Medline and Pubmed, in the period between 2000 and 2022, which evaluates the performance of mortality prediction models for patients hospitalised in ICU (SAPS, APACHE, SOFA, MODS). Results. The most worthwhile scores in clinical practice are held by APACHE II > SAPS II > SOFA (Δ SOFA). The APACHE II score offers the highest accuracy and is the most efficient in the analysis of the evolution of intoxicated patients. The SOFA score provides a specificity for patients with sepsis, whereas, Δ SOFA characterises the reflection of therapeutic strategies in ICU. At the admission of critical patients with uncertain diagnosis, the calculation of SAPS II and APACHE II scores presents the excellent calibration of their evolution during hospitalization. SAPS III score has a superior research research value, thus being more used in retrospective cohort studies. MODS score estimates the mortality in ICU, hospital acquired mortality and the duration of hospitalization in ICUs. Conclusions. All mortality prediction scores or a combination between them are available for use in ICUs, but each of them displays limitations.