Background: Established general risk score models for intensive care patients incorporate several clinical and laboratory data. However, the collection, documentation and classification of clinical data are time-consuming, incur labor-related costs, and are dependent on the experience of the examiner. Therefore, in the present study a general score for medical intensive care patients based solely on routine laboratory parameters is presented.
Methods: Parameter selection was performed using stepwise logistic regression analysis. The maximum likelihood estimate of variable influence on mortality provided a relative weighting for each variable. The new score was compared to two established risk models (Acute Physiology And Chronic Health Evaluation II, APACHE II; and Simplified Acute Physiology Score II, SAPS II).
Results: The study included 528 medical intensive care patients with a mean age of 65.4+/-0.7 years. The in-hospital mortality was 16.5% (87/528). Multiple logistic regression analysis revealed eight parameters with significant prognostic power: alanine aminotransferase, cholesterol, creatinine, leukocytes, sodium, thrombocytes, urea, and age. These parameters were used to build a new laboratory score called Critical Risk Evaluation by Early Keys (CREEK). The area under the receiver operating characteristics curve was 0.857 (0.814-0.900). Pearson correlation analysis showed significant correlation between CREEK and APACHE II (r=0.550) and SAPS II (r=0.516; p<0.001; n=387). The areas under curve of the APACHE II and the SAPS II were 0.869 and 0.874, respectively.
Conclusions: We show that a general risk score for medical intensive care patients on admission based solely on routine laboratory parameters is feasible. The quality of risk estimation using CREEK is comparable to established risk models. Furthermore, this new score is based on quality controlled low-cost laboratory parameters that are routinely measured on admission to the intensive care unit. Therefore, no additional costs are involved.