Purpose of the study: To compare and evaluate preoperative assessment in ASA scale to predicted death rate (PDR) numbers, computed in commonly used severity-of-illness and prognostic scoring systems (Portsmouth-POSSUM, SAPS 2, MPM 2, MPM for cancer patients, LODS, ODIN i TRIOS) on the first and on the third postoperative days. Evaluation of the mean PDR calculated from the scales.
Material and methods: There were analyzed 187 cases of non-survivors and 100 cases of survivors of 187 patients treated in surgical intensive care unit at University Hospital in Kraków. In each case PDR was calculated in seven severity-of illness and prognostic scoring systems on the first and on the third postoperative day and compared to the ASA group and mean PDR computed from seven PDR numbers. Discrimination and calibration characteristics of the scoring systems was analyzed as area under receiver operating characteristic curves (AUROC) and predictive values.
Results: Length of hospital stay was shorter in survivors (16.6 days) as compared to nonsurvivors (25.3 days); similarly the time period between the hospital admittance and surgery was shorter in survivors (1.6 days vs 7.4 days). There were almost twice more frequent repeated surgical procedures in nonsurvivors (45.4% vs 26%). The mean ASA scale in non-survivors was 3.74 and 3.20 in survivors (p < 0.001). The mean PDR computed from seven scoring systems on the first postoperative day was 55.2 in non-survivors vs 21.2 in survivors (p < 0.001) and on the third postoperative day was 64.1 vs 32.3 (p < 0.001). The best discriminative properties, calculated as AUROC, showed: mean PDR computed from the used scoring systems on the first postoperative day (0.859), then ODIN (0.847), MPM2 (0.833), Portsmouth-POSSUM (0.83) and mean PDR computed on the third postoperative day (0827).
Conclusions: There is none severity-of-illness nor prognostic scoring system that could be commonly used in intensive care unit patients. There are discrepancies in predicted death rate (PDR) cal. culated in each of available risk models in population of intensive care unit patients. The mean PDR value computed from available scales could be a reasonable descriptive and prognostic alternative.