Survival prediction of trauma patients: a study on US National Trauma Data Bank

Background

Exceptional circumstances like major incidents or natural disasters may cause a huge number of victims that might not be immediately and simultaneously saved. In these cases it is important to define priorities avoiding to waste time and resources for not savable victims. Trauma and Injury Severity Score (TRISS) methodology is the well-known and standard system usually used by practitioners to predict the survival probability of trauma patients. However, practitioners have noted that the accuracy of TRISS predictions is unacceptable especially for severely injured patients. Thus, alternative methods should be proposed.

Methods

In this work we evaluate different approaches for predicting whether a patient will survive or not according to simple and easily measurable observations. We conducted a rigorous, comparative study based on the most important prediction techniques using real clinical data of the US National Trauma Data Bank.

Results

Empirical results show that well-known Machine Learning classifiers can outperform the TRISS methodology. Based on our findings, we can say that the best approach we evaluated is Random Forest: it has the best accuracy, the best area under the curve, and k-statistic, as well as the second-best sensitivity and specificity. It has also a good calibration curve. Furthermore, its performance monotonically increases as the dataset size grows, meaning that it can be very effective to exploit incoming knowledge. Considering the whole dataset, it is always better than TRISS. Finally, we implemented a new tool to compute the survival of victims. This will help medical practitioners to obtain a better accuracy than the TRISS tools.

Conclusion

Random Forests may be a good candidate solution for improving the predictions on survival upon the standard TRISS methodology.

     

Predictors of outcome following severe head trauma: Follow-up data from the Traumatic Coma Data Bank

Outcome as a function of employment status or return to school was evaluated in severely head-injured patients. A priori we selected the most salient demographic, physiological, neuropsychological and psychosocial outcome predictors with the aim of identifying which of there variables captured at baseline or 6 months would best predict employability at 6 or 12 months. Based on the patients evaluated at 6 months, 18% of former workers had returned to gainful employment and 62% of former students had returned to school. For those not back to work or school at 6 months, 31% of the former workers and 66% of the former students had returned by 12 months. Age, length of coma, speed for both attending and motor movements, spatial integration, and intact vocabulary were all significantly related to returning to work or school. The three most potent predictors for returning to work or school were intactness of the patient's verbal intellectual power, speed of information processing and age.     

Timing of mortality in pediatric trauma patients: A National Trauma Data Bank analysis

Background/purpose

The classic “trimodal” distribution of death has been described in adult patients, but the timing of mortality in injured children is not well understood. The purpose of this study was to define the temporal distribution of mortality in pediatric trauma patients.

Use of vena cava filters in pediatric trauma patients: data from the National Trauma Data Bank

BACKGROUND: Placement of vena cava filters (VCFs) in high-risk adult trauma patients is a well-described intervention for prophylaxis against pulmonary embolism (PE). Few data exist regarding the use of VCFs in pediatric trauma. METHODS: We performed a cross-sectional study using the National Trauma Data Bank of the American College of Surgeons. Patients 17 years old or younger were included. Data regarding demographics, injuries, hospitalization, survival, and treating institution were analyzed. The prevalence of deep vein thrombosis (DVT), PE, and VCF placement were calculated. Odds ratios (ORs) for predictors of VCF placement were determined using multivariate logistic regression. RESULTS: There were 116,357 pediatric patients in the National Trauma Data Bank. VCFs were placed in 214 (0.18%) patients. VCF patients had longer mean hospital (23.99 vs. 4.12 days) and intensive care unit stays (13.65 vs. 1.12 days) and more severe injuries (mean Injury Severity Score, 30.89 vs. 9.04) than those without VCFs. Sixty-five patients had DVT, and PE was diagnosed in 28 patients, representing 0.06% and 0.02% of the cohort, respectively. University-associated teaching hospitals placed 72.4% (95% confidence interval, 65.9-78.3%) of VCFs and Level I trauma centers placed 46.3% (95% confidence interval, 39.4-53.2%) of VCFs. In multivariate analysis, significant predictors of VCF use were DVT (OR, 33.13), spinal cord injury (OR, 15.28), probability of survival (OR, 10.52), severe femur fracture (OR, 3.39), increasing age (OR, 1.99), ISS (OR, 1.05), intensive care unit stay (OR, 1.04), and length of stay (OR, 1.02). Higher Glasgow Coma Scale score decreased the likelihood of VCF use (OR, 0.87). CONCLUSION: Placement of VCFs in pediatric trauma patients is uncommon and is associated with several characteristics of the patient, the injury, and the treating institution. Long-term VCF efficacy in pediatric trauma is not known, and application of VCFs in these patients requires further investigation.     

The impact of diabetes on outcome in traumatically injured patients: an analysis of the National Trauma Data Bank

BACKGROUND: Studies on stress hyperglycemia in trauma patients have largely ignored diabetes, a potential confounder. The purpose of this study was to assess the relationship between diabetes and outcome in trauma patients. METHODS: Data were obtained from the National Trauma Data Bank (version 4.0). The primary outcome measures were mortality and infections. Age, injury severity, and comorbidities were analyzed as independent variables using logistic regression. RESULTS: A total of 343,250 patients were analyzed, of whom 2.7% were diabetic. On multivariate analysis, insulin-dependent diabetes was an independent although weak predictor of infectious morbidity and intensive care unit length of stay. However, diabetes was not associated with mortality or hospital length of stay. Age and injury severity were the main predictors for all outcome measures. CONCLUSIONS: Diabetes was an independent, although weak, risk factor for infectious complications in trauma patients. Age and injury severity were the most important predictors of outcome.     

Admission base deficit in pediatric trauma: a study using the National Trauma Data Bank

BACKGROUND: The base deficit, an important indicator of physiologic derangement after severe injury in adults, has not been specifically examined in the pediatric trauma population. The purpose of this study was to assess the ability of the admission base deficit to predict injury severity and outcome in the pediatric trauma population. METHODS: The study group included all patients in the National Trauma Data Bank over a 2-year period aged 0 to 12 years with a base deficit (0 to -30 mEq/L) recorded from the emergency department. Age, presence of a severe closed head injury, and base deficit were analyzed with respect to mortality and other indicators of injury severity. RESULTS: A total of 515 patients constituted the study group. Base deficit less than -4 mEq/L (p < 0.001) and the presence of a closed head injury (odds ratio, 3.8; p < 0.05) were predictors of mortality. For the group, an admission base deficit of -8 mEq/L corresponded to a probability of mortality of 25%. Significant correlations were found between base deficit and emergency department systolic blood pressure, Injury Severity Score, and Revised Trauma Score. There was no relationship between age and mortality. CONCLUSION: In injured children, the admission base deficit reflects injury severity and predicts mortality. The probability of mortality increases precipitously in children with a base deficit less than -8 mEq/L, and should alert the clinician to the presence of potentially lethal injuries or uncompensated shock.     

Incidence of acute deep vein thrombosis and pulmonary embolism in foot and ankle trauma: analysis of the National Trauma Data Bank

The incidence of deep vein thrombosis (DVT) after foot and ankle surgery is generally believed to be low. However, little information is available regarding DVT as it specifically relates to foot and ankle trauma. The National Trauma Data Bank data set (2007 to 2009) was used to evaluate the incidence of thromboembolism in foot and ankle trauma. Also, the risk factors associated with the thromboembolic events were identified. Data regarding the demographics, comorbidities, procedures, trauma types, and complications, including DVT and pulmonary embolism (PE), were collected from the data set for analysis. The incidence of DVT and PE was 0.28% and 0.21%, respectively. The risk factors statistically significantly associated and clinically relevant for both DVT and PE in foot and ankle trauma were older age (DVT, odds ratio [OR] 1.02, 95% confidence interval [CI] 1.01 to 1.03; PE, OR 1.02, 95% CI 1.01 to 1.03), obesity (DVT, OR 2.35, 95% CI 1.33 to 4.14; PE, OR 3.06, 95% CI 1.68 to 5.59), and higher injury severity score (DVT, OR 1.22, 95% CI 1.16 to 1.28; PE, OR 1.21, 95% CI 1.14 to 1.29). Owing to the low incidence, routine pharmacologic thromboprophylaxis might be contraindicated in foot and ankle trauma. Instead, careful, individualized assessment of the risk factors associated with DVT/PE is important.     

Patterns of injury and functional outcome after hanging: analysis of the National Trauma Data Bank

PURPOSE: Hanging has become the second most common method of attempted suicide among adolescents, but there is little relevant epidemiologic or outcome data in the trauma literature. Additionally, there are no studies examining the degree of functional disability among survivors of hanging injury. METHODS: The National Trauma Data Bank was queried for all patients with an E-code diagnosis of hanging injury. Demographic and injury pattern data were analyzed. Disability at discharge was assessed using the functional independence measure (FIM) scores for feeding, locomotion, and expression (range 1 = full disability to 4 = no disability). Univariate and multivariate analysis was performed to identify independent predictors of mortality and degree of functional disability at discharge. RESULTS: There were 655 patients identified (84% male) with a mean age of 30.3 years and mean injury severity score (ISS) of 9. There were 92 (14%) deaths in the emergency department (ED) and 119 (18%) deaths after admission, for an overall mortality rate of 33%. Excluding ED deaths, survivors had significantly higher Glasgow coma scores (GCS) at the scene (8 vs. 4) and in the ED (9 vs. 3), a lower ED base deficit (4 vs. 9), and lower ISS (6 vs. 15, all P < .01) compared with nonsurvivors. The strongest independent predictor of hospital mortality was ED GCS <15 (odds ratio 16.1, P < .01); the mortality rate was 1.5% for patients with an ED GCS of 15 versus 29% for any GCS <15. Of patients who survived to discharge (n = 277), 84% were functionally independent (total FIM = 12), and 10% had severe functional disabilities in feeding, expression, or locomotion (FIM <3). Patients with severe disability had a higher incidence of intracranial (38% vs. 19%) and chest injury (19% vs. 5%) but surprisingly demonstrated equivalent rates of vascular (0% vs. 2.6%) and spinal injury (11% vs. 12%) compared with those without severe disability. Independent predictors of functional outcome were ISS and ED GCS (both P < .01). There was no severe functional disability at discharge among patients with an ED GCS of 15 compared with a 15% severe disability rate if the ED GCS was <15. CONCLUSIONS: Hanging injuries are associated with a high overall mortality rate, with the admission GCS being the best independent predictor of outcome. However, the majority of survivors have little to no functional disability. The presence of severe disability at discharge is mainly attributed to intracranial and thoracic injury.     

Factors associated with mortality in trauma: re-evaluation of the TRISS method using the National Trauma Data Bank

BACKGROUND: TRISS remains a standard method for predicting survival and correcting for severity in outcome analysis. The National Trauma Data Bank (NTDB) is emerging as a major source of trauma data that will be used for both primary research and outcome benchmarking. We used NTDB data, to determine whether TRISS is still an accurate predictor of survival coefficients and to determine whether the ability of TRISS to predict survival could be improved by updating the coefficients or by building predictive models that include information on co-morbidities. METHODS: To compare the utility of different methods of TRISS calculation we identified the records of 72,517 trauma patients (62,103 blunt trauma and 10,414 penetrating trauma) who had complete information for all of the covariates to be considered in the analysis. Multiple logistic regression was used to recalculate the TRISS coefficients in models using both the original TRISS covariates and in models which also included variables for co-morbidities that could potentially affect survival. Model discrimination was evaluated by calculating the area under the receiver operating characteristic curves (AUC), and model calibration was evaluated with the Hosmer-Lemeshow Goodness-of-Fit Statistic (H-L). RESULTS: For penetrating trauma the original TRISS equation had good discriminative ability (AUC=0.98), but was poorly calibrated (H-L=267.04). When logistic regression was used to generate revised coefficients, discrimination was unchanged, but calibration improved (H-L=38.66). The only co-morbid factor significantly associated with survival after penetrating trauma was acute alcohol consumption, which was associated with increased survival (p < 0.0001). However, its inclusion in a logistic model did not improve discrimination, but improved calibration somewhat (AUC =0.98; H-L=19.95). The original TRISS equation was a less accurate predictor of survival after blunt trauma (AUC = 0.84; H-L= 10,720.7). When logistic regression was used to generate revised coefficients for the original TRISS covariates, predictions after blunt trauma improved (AUC = 0.94; H-L=25.45). With blunt trauma, acute alcohol consumption and prior hypertension were associated with increased survival, and male gender, congestive failure, cirrhosis, and prior myocardial infarction were associated with decreased survival. However, inclusion of these covariates in a logistic model did not improve predictions of survival (AUC = 0.94; H-L= 34.83). CONCLUSIONS: In the NTDB the traditional TRISS had limited ability to predict survival after trauma. Accuracy of prediction was improved by recalculating the TRISS coefficients, but further improvements were not seen with models that included information about co-morbidities.     

Functional outcome after blunt and penetrating carotid artery injuries: analysis of the National Trauma Data Bank

BACKGROUND: We analyzed the functional outcome at discharge among a cohort of patients suffering traumatic carotid arterial injury (CAI) and compared them with matched controls. METHODS: CAI were identified by International Classification of Diseases-9 codes from the National Trauma Data Bank. Control groups were matched by age, mechanism, head Abbreviated Injury Scale score, Injury Severity Score, base deficit, and admission Glasgow Coma Scale score. Outcome was assessed using the functional independence measure (FIM) score for feeding, locomotion, and expression. FIM scores for each category ranged from 1 (full assistance required) to 4 (fully independent), with a maximum total FIM score of 12 representing full independence. RESULTS: There were 967 CAI identified among 474,024 patients for an overall incidence of 0.2%. The mechanism of injury was blunt in 570 (59%) and penetrating in 397 (41%) patients. Seventy-eight percent of penetrating CAI patients were fully independent for locomotion at discharge (FIM = 4) compared with only 37% of blunt patients (p < 0.01). Compared with the control group of similarly injured patients without carotid injury (n = 14,119), the blunt CAI group demonstrated more severe functional disability (FIM < or = 2) at discharge for feeding, expression, and locomotion. Fifty-five percent of blunt control patients were discharged fully independent (FIM total = 12) compared with only 33% of blunt CAI (odds ratio= 2.5; p < 0.01). The penetrating CAI group showed no significant difference in severe disability for feeding, expression, or locomotion compared with controls. However, 80% of penetrating control patients were discharged fully independent (FIM total = 12) compared with only 67% of patients with penetrating CAI (odds ratio = 2.0; p = 0.01). CONCLUSION: Blunt CAI is associated with more severe functional disability at discharge than penetrating CAI. Even when compared with a similarly injured control group, blunt CAI results in significant additional severe functional disabilities in feeding, expression, and locomotion.     

The effect of a trauma system on the outcome of patients with pancreatic trauma

Patients suffering pancreatic trauma before and after the creation of a regional trauma center were analyzed. Between 1975 and 1982, before the trauma system was established, 80 patients (10 per year) with pancreatic injury were treated. From 1984 to 1987 (after trauma system implementation), 58 patients (19 per year) with pancreatic injury were treated. There were no significant differences in demographics, mechanisms of injury, types of pancreatic wound, or associated injuries. Sixteen (20%) patients with pancreatic injury during the study period died before the trauma center was established. Of these, 13 (81%) died of hemorrhage. In contrast, five (9%) patients with pancreatic injury who were treated after the trauma center was operational died. Only one death was due to hemorrhage. By relative risk, a patient was 2.67 times more likely to die and 17 times more likely to die of hemorrhage before the presence of the trauma center than after (P less than .03). In this study of a patient population suffering severe intra-abdominal injuries, Organ Injury Outcome Analysis demonstrates that development of a trauma system significantly improves outcome.     

Thromboembolism after trauma: an analysis of 1602 episodes from the American College of Surgeons National Trauma Data Bank

OBJECTIVE: Venous thromboembolic events (VTE) are potentially preventable causes of morbidity and mortality after injury. We hypothesized that the current clinical incidence of VTE is relatively low and that VTE risk factors could be identified. METHODS: We queried the ACS National Trauma Data Bank for episodes of deep venous thrombosis (DVT) and/or pulmonary embolism (PE). We examined demographic data, VTE risk factors, outcomes, and VTE prophylaxis measures in patients admitted to the 131 contributing trauma centers. RESULTS: From a total of 450,375 patients, 1602 (0.36%) had a VTE (998 DVT, 522 PE, 82 both), for an incidence of 0.36%. Ninety percent of patients with VTE had 1 of the 9 risk factors commonly associated with VTE. Six risk factors found to be independently significant in multivariate logistic regression for VTE were age > or = 40 years (odds ratio [OR] 2.01; 95% confidence interval [CI] 1.74 to 2.32), lower extremity fracture with AIS > or = 3 (OR 1.92; 95% CI 1.64 to 2.26), head injury with AIS > or = 3 (OR 1.24; 95% CI 1.05 to 1.46), ventilator days >3 (OR 8.08; 95% CI 6.86 to 9.52), venous injury (OR 3.56; 95% CI 2.22 to 5.72), and a major operative procedure (OR 1.53; 95% CI 1.30 to 1.80). Vena cava filters were placed in 3,883 patients, 86% as PE prophylaxis, including in 410 patients without an identifiable risk factor for VTE. CONCLUSIONS: Patients who need VTE prophylaxis after trauma can be identified based on risk factors. The use of prophylactic vena cava filters should be re-examined.     

Excess mortality, length of stay, and costs associated with serious hemorrhage among trauma patients: findings from the National Trauma Data Bank

Trauma is a serious injury or shock to the body from violence or crash and is an important and growing global health risk. Using 2000 to 2004 data from a comprehensive trauma registry, we estimated the prevalence of serious blunt and penetrating trauma-related hemorrhage among patients admitted to U.S. trauma centers along with excess in-hospital mortality, length of hospital stay, and inpatient costs. There were 65,750 patients with blunt trauma and 12,992 patients with penetrating trauma included in our analyses. Of patients sustaining blunt trauma, 7.6 per cent had serious hemorrhage; 18.8 per cent of patients sustaining penetrating trauma had serious hemorrhage. In-hospital mortality rates were significantly (P < 0.05) higher for patients with serious hemorrhage than for patients without (24.9 per cent versus 8.4 per cent for blunt; 23.4 per cent versus 4.2 per cent for penetrating). Patients with serious hemorrhage had adjusted mean excess lengths of stay of 0.4 days for blunt trauma and 2.7 days for penetrating trauma (P < 0.05); adjusted excess costs were $296 per day for patients sustaining blunt trauma and $637 per day for patients sustaining penetrating trauma (P < 0.05). In both blunt and penetrating trauma cases, serious hemorrhage is significantly associated with excess mortality, longer hospital stays, and higher costs.     

The effect of ventilator-associated pneumonia on the prognosis of head trauma patients

BACKGROUND: To investigate the effect of ventilator-associated pneumonia (VAP) on the prognosis of head trauma patients. METHODS: We performed a retrospective case-control study in which 57 head trauma patients with VAP were matched to 57 head trauma patients without VAP. Matching criteria were age (+/-5 years), Glasgow Coma Scale score (+/-2), Injury Severity Score (+/-5), Simplified Acute Physiology Score II (+/-5), and duration of exposure to mechanical ventilation. RESULTS: The most causative organisms of VAP were Pseudomonas aeruginosa, and Acinetobactor baumannii (36.8% and 33.8% of isolated organisms, respectively). The duration of mechanical ventilation, intensive care unit stay, and hospital stay were significantly increased in case patients (13 +/- 8.4, 24.5 +/- 18, and 30.8 +/- 18.6 days, respectively) compared with control patients (8.3 +/- 4.3, 12.3 +/- 8, and 20.3 +/- 18.7 days, respectively). Mortality rate was also higher in case (29.8%) than in control (12.3%) patients (p = 0.02). CONCLUSION: We conclude that the occurrence of VAP caused by high-risk organisms in cranial trauma patients may increase the risk of death, the mechanical ventilation duration, the intensive care unit stay, and the hospital stay.     

Development and validation of a new pediatric resuscitation and trauma outcome (PRESTO) model using the U.S. National Trauma Data Bank

Background

There is a need for a pediatric trauma outcomes benchmarking model that is adapted for Low-and-Middle-Income Countries (LMICs). We used the National-Trauma-Data-Bank (NTDB) and applied constraints specific to resource-poor environments to develop and validate an LMIC-specific pediatric trauma score.