SEVERITY OF TRAUMA VICTIMS ADMITTED IN INTENSIVE CARE UNITS : COMPARATIVE STUDY AMONG DIFFERENT INDEXES

This study compared the performance of the Injury Severity Score (ISS) with the New Injury Severity Score (NISS) and also the Simplified Acute Physiology Score II (SAPS II) with the Logistic Organ Dysfunction System (LODS) in trauma victims, in order to predict mortality and length of stay in Intensive Care Units (ICU), besides identifying which indexes have been the most effective to estimate these results. A retrospective analysis was done in the records of 185 victims admitted in ICU between June and December 2006. None of the four indexes properly discriminated the patients according to length of stay at the ICU. The ISS and the NISS did not show a good discriminating capacity in case of death, but the SAPS II and the LODS presented good performance to estimate mortality at the ICU. Results pointed towards the use of SAPS II and LODS when trauma victims are admitted in an ICU.


INTRODUCTION
Today, global advances in technology and violence are contributing to the rising number of deaths or disabilities due to traumas.To improve care delivery to these victims, trauma severity indices have been created which, through uniform language, permit assessing the severity of anatomic injuries and the trauma population's probability of survival.These scoring systems make it possible to assess care delivery, plan emergency care and document epidemiological characteristics.Indices include the Injury Severity Score (ISS), developed as a result of the acknowledged fragility of the Abbreviated Injury Scale (AIS) as a prognostic measure for patients with multiple injuries (1)   .
The AIS determines the individual severity of injuries in trauma victims, but does not assess the cumulative effect of multiple injuries in different body regions, which are common in severe trauma patients.
The ISS attempts to picture trauma victims' global severity and consists of the sum of highest squared score of three different body regions where the most severe traumas are located, according to the AIS code.The higher the score, which can range from 1 to 75 points, the greater the trauma severity and, hence, the greater the probability of death (1) .
Errors were identified when applying the ISS to multiple injury patients, located in the same body region, as this index only considers the most severe injury, ignoring the second and third most severe injuries, which are often located in the same body segment as the first.To correct this distortion, the New Injury Severity Score (NISS) was created, whose score is obtained by adding up the squared AIS scores of the three most severe injuries, independently of the body region (2) .
Due to their severity and high complexity, many trauma victims need to be admitted to intensive care units (ICU).These units increasingly use severity indices due to their importance to assess the unit's performance and the efficiency of the applied treatment. .
The Logistic Organ Dysfunction System (LODS) also permits the identification of hospital mortality probability, focusing on patients' organ dysfunction during their first day of hospitalization at the ICU.Physiological variables are used and, by quantifying the severity of the organ dysfunction, the probability of hospital mortality can be identified (4) .
Facility and similarity in the application of these indices, besides their international recognition, were decisive to choose the indices used in this research.
A historical analysis of scientific research published in this journal revealed that only two articles had comparatively analyzed trauma severity indices, one of which was a literature review (5) and the other original research (6) , which strengthens the importance and contribution of this study to the scientific community.
Moreover, until data, no research has been located in literature which compares the predictive capacity of the ISS, NISS, SAPS II and LODS for mortality and length of stay of trauma victims admitted to ICU.This research aimed to make this comparison and, thus, identify which of these severity indices performs better to estimate mortality and length of stay of trauma victims admitted at ICU.

MATERIAL AND METHODS
An exploratory, descriptive, retrospective field study with a quantitative approach was developed, involving patients hospitalized at the ICU of the Hospital das Clínicas at the University of São Paulo Medical School (HCFMUSP), between June 1st and December 31st 2006.
The inclusion criteria adopted to select the sample were as follows: being 18 years of age or older, being a victim of blunt or penetrating trauma, being hospitalized at the ICU for more than 24 hours and being admitted to hospital within 48 hours after the trauma occurred.

RESULTS
The cases of 185 victims were considered.

DISCUSSION
In total, 185 files of trauma victims hospitalized at ICU were analyzed, 39 of whom died during their stay at the ICU, representing a mortality rate of 21.08% at the unit.International research on trauma victims in ICU reported on mortality rates ranging between 13.8 and 23% (7-8)   .A study on victims of traumatic brain injury identified hospital mortality rates of 20%, and all victims who died had been admitted to the ICU at some time during hospitalization (9)   .Mean length of stay at the ICU (16.55 days) can be considered high in comparison with other studies.International research describing trauma victims at ICU indicated lower averages, between 4.9 and 10 days (7,10-11)   .
In this research, the mean score on the SAPS II (34.10) approximates that in other international studies on trauma victims at ICU: 32 (11)   , 36,6 (7) .No studies were found in literature that used the LODS to identify the severity of trauma victims at ICU.Only one research used that index in trauma victims at the emergency room, with a mean score of five and mortality risk of 30% (12) .
When applying the ISS to the group of trauma victims hospitalized at the ICU, the mean score was 18.34 and the median 17.More than half of the victims (61.62%) showed ISS scores ≥16.Research on trauma victims at ICU sometimes showed lower results, with a mean score of 6 (13) and a median of 9 (14) , and at other times higher scores, with medians of 24 (10) and 25 (7) .A Brazilian study revealed that 77.5% of 40 traumatic brain injury victims hospitalized at ICU scored ≥16 on the ISS, as opposed to a minority of patients hospitalized at nursing wards reaching this severity levels (7.50%) (9)   .
The NISS has been continuously tested in comparison with the ISS and other indices.A literature review on research using the NISS in comparison with the ISS concluded that results favor the new version of the instrument, as most of the analyses evidenced the superiority of the NISS and none showed better performance for the ISS than for the NISS (5) .
In to distinguish between survivors and non-survivors in the sample (8) .
The comparison of the four indices used in this research revealed that the SAPS II and LODS better predicted mortality at ICU than the ISS and NISS.In a study of 325 trauma victims admitted at ICU, patient survival was analyzed in the short and long terms.In that sample, the SAPS II also appeared as a better mortality predictor than the ISS in this group of victims (7) .
In the present study, however, none of the SAPS II and LODS are physiology-based, and that some studies have been proposing the combination of anatomic and physiological indices to improve accuracy levels when predicting the mortality of trauma victims (15)(16) .Perhaps this combination can also be a route to improve the accuracy of other outcomes, also related to these victims' length of stay at ICU.

Four instruments were elaborated
to guide data collection from the patients' files: the first permitted the recording of patients' characterization data, departure conditions from the ICU (discharge, death) and length of stay at the unit; the remaining instruments were used to compile the variables Severity of trauma victims admitted in intensive care units… Nogueira LS, Sousa RMC, Domingues CA. included in the calculation of the SAPS II and LODS, to list the descriptions of injuries resulting from the trauma and to calculate the ISS and NISS.After approval by the Ethics Committee for Research Project Analysis -CAPPesq (Protocol No 0168/07), the files of patients hospitalized at ICU were requested from the hospital's Medical Filing and Statistics Division and fully read, with a view to filling out the proposed data collection instruments.The ISS and NISS were calculated on a printed instrument that permitted distributing injuries according to body region, the coding of injuries according to the AIS 2005 manual and the identification of the most severe injuries, in general for the calculation of the NISS and per body region for the ISS.For the final calculation of the SAPS II and LODS scores, the worksheets available on-line at http://www.sfar.org/scores2/saps2.htmland http:/ /www.sfar.org/scores2/lods2.htmlwere used, on which the obtained data were compiled, thus permitting the electronic calculation of these indices and of the patient's mortality risk.NCSS for Windows was used to analyze and interpret results.With a view to assessing the predictive capacity of the ISS, NISS, SAPS II and LODS indices to foresee the dependent study variables (length of stay and mortality at ICU), Receiver Operating Characteristic (ROC) curves were constructed for each of the studied outcomes, obtaining accuracy, measure of the area under the curve (AUC) and confidence interval.The cut-off points obtained by the ROC curve simultaneously considered the best sensitivity and specificity with regard to the addressed variable.As SAPS II and LODS results indicate both a total score and mortality risk, it should be highlighted that, in the construction of the ROC curves, mortality risk estimates were used.The comparative test of the areas under the curve was based on the Z test.In all analyses, the significance level was set at 5% (p value d"5%).Hosmer-Lemeshow's Goodness of Fit test was used to analyze the model's calibration, considered satisfactory when p was >0.05.
under the curve; CI 95%: confidence interval at 95% Similar accuracy levels were observed for the ISS and NISS to predict length of stay at ICU.According to the p-value, no significant difference was found between the two areas under the curve.In the sample, the ISS and NISS did not show good discriminatory capacity to predict length of stay at ICU (AUC=0.64 and 0.67).Hosmer-Lemeshow's Goodness of Fit test indicated satisfactory calibration for the model.When comparing SAPS II and LODS, similar accuracy was observed, an equal area under the curve, besides little discriminatory capacity to predict length of stay at ICU. Hosmer-Lemeshow's Goodness of Fit test indicated unsatisfactory calibration for the model.

Figure 1 -
Figure 1-ROC curve of ISS, NISS, SAPS II and LODS indices for prediction of mortality at ICU. São Paulo, June to December 2006

Figure 2
Figure 2 reveals that the ISS, NISS, SAPS II and LODS curves are very close when analyzing the length of stay at ICU.Therefore, it cannot be affirmed that one index is better than another to predict length of stay at this unit, but only that results are similar.

Figure 2 -
Figure 2 -ROC curve of ISS, NISS, SAPS II and LODS indices for prediction of length of stay at ICU. São Paulo, June to December 2006 In general, these research results evidenced better capacity of the SAPS II and LODS to predict mortality in trauma victims admitted at ICU than ISS and NISS, indicating that severity indices at ICU (SAPS II and LODS) should be preferred, even in trauma victims, when the goal is to predict mortality or assess care results at this unit in view of observed deaths.Moreover, the simplicity and speed to apply the SAPS II and LODS strengthen the use of these indices in trauma victims at ICU, as professionals working at the unit have little time available and instantly need to obtain information from patients with a view to clinical decision-making.Finally, some limitations should be highlighted for the present research: it was carried out at a single institution that is a referral center for care delivery to trauma victims, not permitting the characterization of other populations.Besides, restrictions in terms of sample size, client type (adults only) and type of institution (university hospital) should be taken into account when applying or comparing the results.Hence, future research could broaden this study in terms of sample size and population diversity, and also carry out comparative analyses of different anatomy and physiology-based indices.

Table 1 -
The mortality rate at the ICU was 21.08% and, at the hospital, 21.62%.The mean length of stay at the ICU was 16.55 days and, at the hospital, 21.71 days.Comparison between predictive capacity of ISS and NISS and between SAPS II and LODS for mortality at ICU, according to cut-off point, accuracy, area under the curve, confidence interval and Hosmer-Lemeshow test.São Paulo, June to December 2006 The results showed higher frequencies of young persons (mean age 38.95 years), male (76.76%), who had been subject to surgery (57.84%).The predominant external causes were traffic accidents (63.79%), falls (15.13%) and aggressions (11.90%).The behavior of the LODS was similar to that of the SAPS II: patients with mortality risk levels below 25% were more frequent and the number of victims decreased as the risk of death increased.Out of the ten patients in this research with death risk levels above 75%, identified by the LODS, nine died and one was transferred to another institution.The mean mortality risk according to the LODS was 21.14%, with a standard deviation of 22.47% and median 15%.NISS scores.The mean NISS found was 23.60, the standard deviation was 8.89 and the median 24 points.Of all files analyzed in this research, 31.35% of victims obtained the same score on the ISS and NISS, while 68.65% scored higher on the latter than on the former.AUC: area under the curve; CI 95%: confidence interval at 95% Similar accuracy levels of the ISS and NISS were observed to predict mortality in ICU.Based on the p-value, no significant difference was found between the two areas under the curve.In the sample, the ISS and NISS did not present good discriminatory capacity for the occurrence of death at the ICU (AUC=0.63 and 0.58), although Hosmer-Lemeshow's Goodness of Fit test indicated satisfactory calibration (pe"0.05)for the model.
indices showed good capacity to predict hospitalization time at ICU.Hence, it should be reminded that the ISS and NISS are anatomy-based severity indices, while Rev