Aim: The performance of common mortality prediction models in the intensive care units (ICU) are extensively validated, predominantly in high-income countries. Simple and fast models with region specific features are needed.
Study design: Retrospective case-control study
Methods: We reviewed the medical records of 1057 ICU-admitted patients within three years. Patient survival was defined as discharge before 28 days. Multivariate logistic regression modeling was applied, basic parameters were selected, and a simple model was tried using four of them (age, albumin, platelet, C-reactive protein); as Quick Prediction of Mortality (Qpm) score, and then tested. The Qpm score predictions were compared to calculated APACHE II predicted mortality (APM) score predictions. Both scores were then weighted by calculated standardized mortality ratios (SMR).
Results: 933 patients were included into the analyses. The patients’ overall observed mortality rate was 47%. APACHEII score prediction was 49% (p< 0.001, AUC= 0.810, r: 0.518). Qpm score prediction was 57% (p< 0.001, AUC= 0.699, r: 0.338). The SMR for Qpm was 0.82 in comparison to APM score SMR = 0.96.
Conclusion: This simple prediction model has showed an acceptable performance in our ICU sample and needs to be prospectively evaluated for feasibility. In addition, further studies could be planned for external evaluations and validations in different settings.