Independent Predictors of Mortality in Polytrauma Patients

The study protocol was approved by the Institutional Medical Ethics Committee (CAPPesq 1081/09) and received financial support from the São Paulo State Research Foundation (Fundação de Amparo à Pesquisa do Estado de São Paulo - FAPESP) under no. 2010/03315-4.

At the end of data prospecting, 334 patients with severe trauma (Injury Severity Score, ISS, >15), assisted by the rescue system and taken to HCFMUSP-Hospital das Clínicas da Faculdade de Medicina da Universidade de São Paulo- (by land and helicopter) were included in the study. This screening strategy included trauma patients with severe confirmed bleeding, patients with severe traumatic brain injury (TBI) and patients facing high-energy trauma (a marked characteristic of this trauma center). Patients under 18 years of age were excluded ( Figure 1).

In total, 78 patients were excluded because they did not fit the above criteria, leaving 256 patients for analysis. Of this total, 34 patients died before arriving at the emergency room, and 22 were removed from the analysis due to incomplete data, not signing the informed consent form or due to any situation in which the data collection could compromise victim care(care Figure(Figure 1).

All patients had data recorded at the following times: 1, at the trauma scene; 2, in the emergency room; 3, at 3 hours after hospital admission; and 4, at 24 hours after hospitalization.

Data regarding gender, age, trauma and trauma mechanism, medical procedures performed at all stages, time until arrival at the hospital and comorbidities were recorded. The clinical data collected were as follows: systolic blood pressure (SBP); diastolic blood pressure (DBP); heart rate (HR); respiratory rate (RR); arterial hemoglobin oxygen saturation (SAT); Glasgow Coma Scale (GCS) score; and pupil pattern. The following laboratory data were collected through peripheral venipuncture at the hospital stages: pH(hydrogenic potential); base excess (BE); bicarbonate (BIC); partial oxygen pressure (pO2); partial carbon dioxide pressure (pCO2); arterial hemoglobin oxygen saturation; lactate; sodium (Na+); potassium (K+); ionized calcium (Ca2+); glucose; hemoglobin (Hb); hematocrit (Ht); serum creatine phosphokinase (CPK); platelets and leukocytes. Data regarding volume expansion [crystalloids (CRYSTAL), colloids (COLO)], diuresis and water balance were also collected. The use of blood products [packed red blood cells (PRBC), fresh frozen plasma (FFP), platelet concentrates (PLAT), cryoprecipitate concentrates (concentrates (CRYO)] was computed, and the use of vasoactive drugs (VAD) was scored at the respective times. Patient follow-up was conducted at the intensive care unit (ICU), and time of ICU stay (days in the ICU), days under mechanic ventilation (days under MV), the presence or absence of sepsis, coagulopathy and acute renal failure (ARF) were recorded.

Blood coagulation data were collected as follows: time point 1, International Normalized Ratio (INR) and prothrombin time (PT); and time points 2, 3 and 4, INR, PT, activated partial thromboplastin time (aPTT), the respective ratio between aPTT of patients and controls (R) and thrombin time (TT).

The following severity indexes were calculated: ISS, revised trauma score (RTS); trauma and injury severity score (TRISS); and simplified acute physiology score (SAPS) 3 (the latter when the patient was in the ICU).

At time point 1, blood tests were performed using the i-STAT® device (Abbott, USA; PT/INR and CG4+ and CG8+ kits). At the hospital stages, the analyses used the clinical analysis methodology of the HCFMUSP.

Clinical follow-up was conducted for 30 days after initial trauma.

Statistical analyses The sample power analysis was conducted with a significance level of 0.05, power of 0.80, moderate correlation of 0.5 between time periods and assumption that the variability is equal within each factor (non-sphericity). Due to the effect size between 0.1 and 0.5, there was no need for samples larger than 140 patients. A total of 200 patients waswere defined conservatively, with a margin for possible deaths. The software G*Power 3.1.7 was used for sample size calculation.

The data analysis was divided into three interconnected parts. The first part refers to the descriptive data analysis and tests of association with death. The second part addressed the profiles of the time-dependent measures and their relationship with death through analysis of nonparametric variance for repeated measures. The third part includes the set of all previous analyses that resulted in a Generalized Estimating Equation (GEE).

For the first part, the analyses were performed with the overall sample (n=200) subdivided into two groups: deaths (n=52) and non-deaths (n=148). For qualitative variables, the two-tailed Fisher's exact test was used, and for quantitative variables, the two-tailed t test or the two-tailed Mann-Whitney test was used according to the normality of the variable, which was verified by the Anderson-Darling test.

In the second part, the profiles of each measure over time were analyzed for groups of survivors and non-survivors. Nonparametric analyses of variance (NPar ANOVA) were conducted.

Finally, in the third part, a GEE model was developed considering the family of binomial distributions (dichotomous response variable) with logit link function. Only the main effects of each measure were considered. The NPar ANOVA (p>0.1) was the test used for inclusion in the model. The variables for the final models were selected using the backward method, with output alpha equal to 0.05. All part 3 results were interpreted by estimating odds ratios (OR), their corresponding 95% confidence intervals and significance tests (p-value).

The significance level was set at 0.05, and the free software R 3.0.2 was used for performing the statistical analyses.