CHAPTER 6 TRAUMA SCORING
Trauma was the first medical specialty to regionalize health care delivery to specialized centers and to systematically measure health care outcomes. The first trauma scores were designed for a specific purpose: to standardize injury descriptions and rank injury severity to effectively triage injured patients to the appropriate trauma center.1 Since then trauma scores have evolved to serve two new purposes: to allow risk adjustment for comparisons of outcomes for research and quality performance, and to predict the probability of survival.1–3 An additional purpose that trauma scores have only started to address is predicting functional impairment or disability.4 Currently, trauma scores play a major role in quality improvement processes and patient safety by identifying unexpected deaths for peer review audit.2,5,6 While existing scoring systems are reasonably predictive of survival, they are inadequate for measuring quality performance.7–10
Most trauma scoring is based on anatomical injury descriptors or physiological derangements. Current scoring systems have been modeled to address one principle outcome—mortality—while little attention has been paid to other quality performance outcomes such as functional impairment and quality of life issues.11,12 Only recently have efforts been made to incorporate into scoring systems the impacts of demographics, comorbidities, and mechanism of injury. Unlike other medical scoring systems involving more uniform populations of patients and conditions (i.e., ischemic heart disease), it has proven extremely difficult to design a satisfactory scoring system in the heterogeneous trauma population. For example, there are a handful of ICD-9-CM descriptors that fully describe ischemic heart disease versus approximately 2000 descriptors for traumatic injuries. Patients with ischemic heart disease tend toward a uniform set of comorbidities and demographics, whereas trauma patients span the entire spectrum. Thus, scoring of traumatic injuries in a way that reduces the variables to a single numeric score, results in loss of detail, and generates similar or identical numeric scores for patients whose conditions are not comparable.
ANATOMIC SCORING SYSTEMS
Anatomic scoring systems require a lexicon to describe the anatomy and severity of the large number of potential injuries that result from trauma. Traditionally, this was provided by the Abbreviated Injury Scale (AIS), but more recently descriptors from the ICD-9-CM (Clinical Modification of the 9th revision of International Classification of Disease) diagnosis codes have been used. The Injury Severity Scale (ISS), Anatomic Profile (AP), and New Injury Severity Scale (NISS) are based on AIS rubrics, whereas the ICD-9 Injury Severity Scale (ICISS) is based on ICD-9-CM injury codes. Despite an ever-increasing number of injury descriptors in both the AIS and ICD-9, there are still a number of injuries that are difficult to classify accurately. The soon-to-be-released ICD-10-CM has even a larger number of injury descriptors. A further limitation of anatomic scoring systems is the difficulty in identifying all of a patient’s significant injuries, particularly in patients who die at the scene or early in their hospitalization and do not undergo autopsy.13
Abbreviated Injury Scale
In 1971, the American Medical Association Committee on Medical Aspects of Automotive Safety, later to become the Association for the Advancement of Automotive Medicine (AAAM), published the AIS, the first widely recognized anatomic injury scale.14 The AIS rated the severity of tissue damage secondary to motor vehicle crashes, and provided standardized terminology to describe injuries. The AIS divides the body into nine regions: head, face, neck, thorax, spine, abdomen/pelvis, upper extremities, lower extremities, and unspecified. For each region a consensus-derived scale was developed for grading injuries from 1 (minor) to 6 (virtually unsurvivable). The AIS is not an interval scale; the increase in mortality from 4 to 5 is much higher than from 2 to 3. The first published AIS described 73 blunt injuries for five body regions. Since then the AIS has been updated six times, most recently in 2005 (AIS 2005), and now includes descriptors for more than 1300 injuries covering blunt, penetrating, and pediatric injuries.15 For the first time, AIS 2005 addressed the prediction of functional impairment or disability in its classification.16 The AIS remains the foundation of most anatomic trauma scoring systems used by trauma registries as well as the National Highway Traffic Safety Administration (NHTSA) and other injury research and education organizations. The Organ Injury Scale (OIS) is a similar injury scaling system developed by the American Association for the Surgery of Trauma (AAST).17 The OIS provides a common terminology and severity score to allow comparisons of equivalent injuries for clinical research. Unlike the AIS, the OIS is not used as part of any trauma scoring system.
Injury Severity Scale
The AIS failed to account for the cumulative effect of injury in different body regions, so in 1974 Baker proposed the ISS, an algorithm based on the AIS designed to improve the ability of the AIS to predict mortality.1,18 The ISS divides injuries into six body regions compared to nine in the AIS. The ISS is calculated by taking the sum of the squares of the highest AIS from each of the three most severely injured body regions to achieve a score that ranges from 3 (least) to 75 (most) injured. By definition, an unsurvivable injury with an AIS of 6 is automatically given an ISS of 75. An ISS of 1–8 is considered minor, 9–15 moderate, 16–24 severe, and 25 and higher very severe. The ISS reduces the great variability of injury patterns to a much smaller range of values that can be used in outcomes research. Although the ISS score correlates well with mortality, the relationship is not linear and ISS methodology was not designed to predict disability or other outcomes.19 The ISS is integral to most trauma registries, and is the basis for the anatomic component of TRISS (Trauma Injury and Severity Score) discussed later.
A significant limitation of the AIS and ISS is the cost, time, and training involved in capturing the data and calculating the scores (hand coding), particularly in hospitals that do not use a trauma registry.20 Determination of the AIS requires abstraction of the injuries from the medical record and appropriate training of the trauma registrar or coder, and is dependent on the methodology used to assign the AIS codes and the version of AIS or algorithm used by the registry software to calculate the ISS. There can be significant differences in the calculation of the AIS and ISS due to registry software or personnel. These factors limit the ability to compare outcomes with data derived from varying institutional practices.21 Commercial computerized applications (ICDMAP) are available that convert ICD-9-CM discharge diagnosis codes into AIS scores (ICD/AIS), which in turn can be used to calculate the ISS score.22 The level to which injuries can currently be mapped by ICD-9-CM is crude compared to the AIS, as the detail of the injury descriptors is inadequate. AIS and ICDMAP are proprietary software, and this limits their availability. Despite these limitations, there is good correlation between AIS and ICD/AIS.6,22 The most recent iteration of AIS (AIS 2005) was considered in developing the injury portion of the upcoming ICD-10-CM; thus, mapping between ICD-10-CM and AIS 2005 is likely to be even better once software becomes available.16
The ISS is statistically problematic because it is based on the sum of squares of triplets. As a result, it is nonlinear and nonmonotonic, which means that mortality does not necessarily increase with successive values of ISS. This characteristic is frequently not accounted for in outcomes research.19 Of the 75 potential values, only 44 are represented by ISS scores and 11 of these scores are generated by pairs of triplets. Eight of these triplet pairs have mortality rates that are statistically different.23 The reason that this difference exists is the variable maximal AIS scores within pairs of triplets. For example, an ISS score of 25 is generated both by the triplets 5,0,0 and 4,3,0. Intuitively, one would expect the ISS score based on the triplet containing the near lethal 5 AIS score to have a higher mortality. This was confirmed by Russell and colleagues,24 who retrospectively calculated a mortality rate of 20.6% associated with the triplet 5,0,0 compared with 0% for the triplet 4,3,0. Thus, a trauma center with a higher percentage of the lethal triplets among its patients will have worse outcomes than expected if assessment is based on the ISS alone. Even for a single-value ISS triplet, one would expect significant variability in mortality rates depending on the body region affected. For example, the mortality rate for the same AIS of an isolated injury of the head would likely be more lethal than an isolated injury to the extremity. This was shown to be true when the mortality rate for an AIS of 4 of an isolated head injury was compared with an extremity injury and found to be 17.2% versus 0%.19 And finally, at the highest ISS values of 75, there are unexpected survivors due to AIS 6 patients who do not die. The statistical problems of the ISS could potentially be improved by representing the numerical data as a categorical rather than a continuous variable in regression models; however, this does not correct the underlying problem with its methodology.
Anatomic Profile and New Injury Severity Score
Another problem with the ISS is that it underestimates mortality resulting from multiple injuries to a single body region or organ because only the single most severe injury in each region is considered.25,26 The AP and NISS were designed specifically to address this limitation of the ISS. The AP score is a modification of the AIS and ISS that uses only four regions: brain and spinal cord, thorax and neck, all other serious injuries, and all other nonserious injuries. The AP score is calculated by taking the square root of the sum of the squares of all of the AIS scores within each region to give a summation score for each region, which is then used to calculate the ISS.27 The AP performs better than the ISS in single-system injury.27 The modified AP (mAP) only considers AIS values greater than 3, and coefficients derived from logistic regression analysis are then used to calculate the Anatomic Profile Score (APS) to predict survival.27 The AP has found limited use as the anatomic component of ASCOT (A Severity Characterization of Trauma), detailed later in this chapter.28 The NISS sums the squares of the three highest AIS score regardless of body region.25 The NISS and APS predict mortality better than the ISS, especially in head injuries and higher injury-severity patients, but have not gained widespread use.25,26,28–31
The ideal number of injuries to include in trauma scoring is unknown. The ISS and NISS score up to three injuries per patient, while the AP includes all injuries in its score. Multiply injured patients are currently modeled as if the effects of their injuries are independent, not cumulative; some combinations of injuries are likely to be more lethal than predicted by individual models. However, including additional injuries in trauma scoring models has not improved performance. Indeed, it has been shown that regardless of scoring system, a patient’s worst injury predicts survival best.6,32 Accounting for multiple injuries may be more important when outcomes such as morbidity, length of stay, and disability rather than mortality are being evaluated.33
ICD-9 Injury Severity Score
The ICISS skirts all of the issues with the AIS and ISS by directly calculating the probability of survival (survival risk ratio [SRR]) from approximately 2000 individual trauma-related ICD-9-CM diagnoses.33 The coefficients for the SRR are calculated from logistic regression from large databases. SRRs are only estimates of true survival and are database specific; however, they have been shown to be robust in terms of their application to other sets of injured patients from comparable populations.33,34 In general, SRRs are not calculated independently of other injuries, and thus are not true representations of individual injury risk; however, independent SRRs based on single-injury cases are available.35
The original mortality tables for the ICISS SRRs are based on the non-trauma North Carolina Hospital Discharge Diagnosis (NCHDD) database.36 The NCHDD is criticized for not being comparable to most populations of trauma patients with its overall low mortality, low numbers of trauma patients, and atypical injury patterns. Recalculated ICISS SRRs based on the National Trauma Data Bank (NTDB) and other databases have confirmed this, underscoring the need for adequate comparisons of SRRs from various sources.7,36,37
The ICISS carries the advantage that ICD-9-CM codes are readily available from hospital discharge codes; thus, no additional costs are incurred or trained personnel needed for capturing the data. Furthermore ICD-9-CM is universally available, and most medical personnel are familiar with ICD-based diagnosis coding in contrast to AIS coding. Another advantage of ICD-9 scoring is that risk stratification can easily be expanded to include coded comorbidities.35 ICISS does not include physiologic data; however, it predicts mortality, costs, and length of stay as well as or better than risk adjustment models like TRISS and ASCOT that do.35–40
In ICD-9-CM, there are a limited number of rubrics for orthopedic, vascular, and solid organ injury descriptors, and severity of injury is not accounted for. Therefore, coding the best diagnosis with sufficient detail of the various potential injuries is problematic in ICD-9-CM. There has been an effort to correct these discrepancies in the ICD-10-CM, whose draft version is now available. ICD-10 is already in use in the United States for coding fatal injuries, but the clinical modification has not been finalized and approved yet. The number of injury descriptors in the ICD-10-CM is large and allows precise location of injuries, in particular of interest to researchers in transportation safety. A disadvantage that results when large numbers of descriptors are available is that the number of cases on which to base each SRR will be small, thus diminishing the accuracy of the SRRs. ICISS is rapidly becoming the trauma score of choice for mortality prediction and quality improvement processes and this trend will likely continue as ICD-10-CM becomes available.37
PHYSIOLOGIC SCORING SYSTEMS
Physiologic derangements including hypotension, tachypnea, and diminished mental status reflect the response of the patient to injury and have prognostic value. Physiologic scoring systems are hampered by the fact that physiologic parameters are constantly changing after injury and during resuscitation, and the timing and duration of these changes are not accounted for in existing systems. Typically, the ED admission or initial prehospital vital sign set is used for scoring, although there has always been a concern that prehospital vital signs may not be sufficiently accurate. Currently, there is no consensus on which data time point is the best predictor of outcome. Some patients with severe injury will not be identified by physiologic scores because they are able to compensate, or the field response is so rapid that physiologic compromise has not yet occurred. Physiologic scores overestimate injury severity when physiologic changes are the result of other factors such as drugs and alcohol rather than the consequences of trauma.41
Glasgow Coma Scale
The Glasgow Coma Scale (GCS) is a component of numerous trauma scoring systems since head injury and mental status carry significant prognostications. The GCS is the sum of three coded values: motor, verbal, and eye opening. However, the GCS may lead to overclassification of injury severity in patients with depression of the central nervous system secondary to drugs or alcohol or when the patient is intubated resulting in loss of the verbal score. It has been proposed that the best motor score of the GCS be used rather than the total GCS, as this tends to most accurately reflect true head injury and thus patient outcome.41,42
Revised Trauma Score
The Trauma Score (TS) and Revised Trauma Score (RTS) are physiologic trauma scores designed for field triage of patients who are significantly injured and require trauma center transfer. The TS is a simple sum of points based on the degree of derangement of the GCS, systolic blood pressure (SBP), respiratory rate (RR), respiratory expansion, and capillary refill time (CRT).43 The RTS is a simplification of the TS that includes only the GCS, BP, and RR.44 The RTS has been used as a tool for predicting survival by adding weighted coefficients based on logistic regression with values range from 0 (worst) to 7.84 (best). The RTS is heavily weighted toward the GCS to compensate for major head injury without significant physiological changes, and correlates well with survival.44
Acute Physiology and Chronic Health Evaluation
The Acute Physiology and Chronic Health Evaluation (APACHE II) is a widely used system to predict mortality in intensive care units, but has performed poorly in trauma patients most likely because it lacks an anatomical component.35 APACHE III corrected this deficiency by including trauma-specific injury descriptors and equations, and accounting for head injury. However, this scoring system has not gained wide acceptance in part due to its proprietary nature, and it has not been validated in trauma patients.45
Physiologic Reserve
Physiologic reserve reflects a patient’s ability to cope with injury, and is based on age, gender, comorbidities, and possibly genetic predisposition. Age has an effect on mortality in trauma patients gradually up to age 65 and increasing rapidly thereafter.45 In-patient length of stay and discharge to long-term care are affected by age older than 55 and by some comorbidities.45 The addition of an age factor improves the predictive ability for survival of most trauma scoring systems.20 Comorbidities have a profound effect on individual patient outcome, even after controlling for age, anatomic and physiologic severity, and mechanism of injury.46 Institutional outcomes may not be influenced by comorbidities, due to their low incidence in trauma patients.47
RISK-ADJUSTMENT SCORING SYSTEMS
Risk-adjustment scoring systems use regression analysis of large databases to determine probability of survival based on anatomical and physiological data and age. The addition of age or physiologic data to injury severity improves prediction of mortality in all trauma scoring models examined.5
Trauma and Injury Severity Score
The TRISS combines physiologic data from the RTS, anatomic data from the ISS, and age (less than or 55 years and older) and mechanism of injury to give a probability of survival or TRISS score.2 A “pre-chart” analysis of RTS plotted against ISS can be used to calculate a survival probability of 0.5 based on regression analysis to identify patients with unexpected outcomes. These “TRISS unexpected survivors” are a widely used audit tool in identifying patients for peer review to investigate prehospital and hospital factors that contribute to outcome.10 The usefulness of this practice was recently called into question when a chart review of TRISS unexpected survivors revealed only 10% to be “unexpected survivors” based on clinical findings.10
Software to calculate TRISS is available and includes NATIONAL TRACS based on model coefficients derived from the MTOS (Major Trauma Outcome Study).2 To determine the actual probability of survival, the calculated TRISS score is compared with the model data set using three statistics, W, Z, and M.2,3 A positive W-statistic indicates that the institution has more survivors than predicted. The Z-statistic is used to assess whether the W-statistic is significantly different from zero, and hence whether the institution’s performance is significantly different from that defined by the model data set. Z-statistics can be compared with a standard normal distribution. The M-statistic is used to examine the similarity in the case mix of the observed data, compared with the model data set. The value of M is between 0 and 1, with values close to 1 indicating a very similar mix of injury severities. A value of less than 0.88 has been deemed unacceptable for the purpose of comparison with the model database, and hence for interpretation of the W- and Z-statistics.48 A relative outcome score (ROS) can be used to compare W-statistics against a perfect outcome of 100% survival. The ROS can be used as a benchmark to monitor improvement in institutional trauma care over time. An alternative to the Z- and W-statistics is the standardized mortality ratio (SMR). The SMR is defined as the ratio of the observed mortality rate (OMR) to the expected mortality rate (EMR) to identify hospital quality outliers. The SMR is the standard measure of quality used in critical care medicine.9
The TRISS has the best predictive value when studying patients with multiple injuries from blunt trauma. TRISS has poor predictive ability in isolated severe head trauma and multiple severe injuries to a single body region, and at the extremes of age. TRISS also does not distinguish between types of penetrating injuries, that is, stabbing versus gunshot, which are known to have disparate outcomes.49 TRISS underestimates survival in the lowest predicted survival group because it is based on the ISS. TRISS methodology is currently advocated as the standard for benchmarking performance in the United States, and is widely accepted in many parts of the world.50 Existing TRISS coefficients are based on MTOS data from U.S. trauma centers with a high percentage of penetrating trauma that is nearly 20 years old, and thus may not be applicable to foreign trauma centers and too outdated for current trauma systems. TRISS coefficients can be updated to reflect local databases, which should improve its predictive properties.50
A Severity Characterization of Trauma
To overcome the outcome limitations of TRISS, Champion and the American College of Surgeons Committee on Trauma proposed ASCOT, which uses AIS descriptors, physiologic data, mechanism, and age.51 ASCOT incorporates all severe patient injuries in the prediction model via the AP, in contrast to TRISS, which considers only ISS injuries. ASCOT proved to be equivalent or better than TRISS in most studies, particularly penetrating trauma, but failed to be widely accepted, most likely because of the complex computations involved in deriving the score.51 Like TRISS, the coefficients for ASCOT are based on the MTOS, which is biased toward severely injured and penetrating trauma patients.3,52
Risk-adjustment models like TRISS and ASCOT allow outcomes from different institutions to be adjusted for differences in injury severity, making it possible to compare hospital quality. Inaccurate risk adjustment may lead to some hospitals being labeled as poor quality and vice versa. However, a study comparing TRISS and ASCOT for identifying high-quality hospitals disagreed on the status of 35 of 69 hospitals studied.9 A second study comparing trauma centers using TRISS found an unacceptably high misclassification rate in patients with severe trauma, further supporting the conclusion that currently these tools are unable to accurately provide benchmarking for quality improvement.8 The addition of comorbidities was recently shown to improve TRISS performance for prediction of survival.53
SCORING SYSTEMS EVALUATION
Data Collection
The survival probability model is the most popular tool for evaluating trauma care.54 Current models are based on linear logistic regression analysis of patient variables to identify those independently associated with mortality. Formulas are then derived to predict the probability of survival using weighted coefficients according to the effect of the variable on mortality. To be statistically sound, this multivariate analysis requires large databases of trauma patients. These databases must include data on a large number of variables, including patient demographics, comorbidities, injury type and severity, mechanism of injury, prehospital care, emergency department care, in-hospital care, and postdischarge follow-up. Complete and accurate data gathering into a database is dependent on operator input and data availability. Missing data are a particular problem with multivariate analysis, as often the entire patient record containing the missing piece of data must be discarded.
Databases
Trauma scores are derived from several types of databases: hospital administrative databases, trauma registries, and the NTDB. Administrative databases are derived from ICD-9-CM hospital discharge data that were collected for billing purposes.46 They reflect the coding conventions of the institutions from which they were derived, and may be affected by reimbursement considerations. Furthermore, only the most significant injuries may be coded.55 Administrative databases suffer from significant gaps in data, lacking such details as prehospital and emergency department care, physiologic data, and postdischarge follow-up. Trauma registries are designed to have no such gaps, capturing all phases of trauma care, but require dedicated personnel to administer. Trauma registries vary from hospital to hospital, mostly in the manner in which AIS and ISS are coded, which render comparisons between them difficult or even invalid.21,48,56
The NTDB functions as a national repository of trauma data to be used for epidemiology, injury prevention, clinical research, education, and resource allocation.57 The NTDB voluntarily collects data from 565 U.S. hospitals, including 70% of the Level I trauma centers and 50% of the Level II trauma centers. It has a standardized data entry format that can be hand entered or automatically derived from existing trauma registry data. The NTDB collects data on a large number of variables felt to potentially impact quality of care in addition to patient demographics, complications, diagnosis, TRISS/ISS scores, and outcomes. It also documents the methodology used to determine AIS, ISS, TRISS, and diagnosis. The NTDB was created in 1989 by the American College of Surgeons, and participation by trauma centers has increased substantially in the past few years. The NTDB is nonproprietary, and its reports are available at no charge with a benchmark report for quality improvement processes provided annually to each participating hospital.
Outcome Measures
The most common outcome measured by trauma scores is mortality.6 The timeframe for inclusion of mortality is not uniformly defined; thus, data on all fatalities may not be captured.58–60 Mortality after injury may be variously defined as prehospital, in-hospital, 30- or 60-day postinjury, or all injury-related mortality identified postdischarge regardless of time period. For example, elderly patients are less able to survive mild to moderate injuries, and more likely to die of complications several weeks or months after the incident.60 Such patients would not meet the mortality inclusion criteria of in-hospital or 30-day mortality definition. Postdischarge mortality is not captured by administrative databases, and is only sporadically captured by trauma registries.61,62 Estimates of injury mortality substantially increase when using multiple independent databases to capture postdischarge fatalities.63
Prehospital deaths are not captured by trauma registries or administrative databases, but may affect mortality predictions for many injuries. Due to improved EMS, patients suffering fatal injuries that previously would have died, now make it to the hospital only to die soon after arrival.58,64 In-hospital mortality is also affected by withdrawal of care practices. Hospitals with more liberal policies for withdrawal of care during the in-hospital period will report artificially higher in-hospital mortality. Lower in-hospital mortality rates will be reported by hospitals whose policy is to transfer early significantly disabled trauma patients to skilled nursing facilities. Withdrawal of care is usually documented in trauma registries but not in administrative databases.
Injury outcome is dependent on which outcome is measured, and may be impacted by factors not related to quality. Type of injury, age, and comorbidities affect various outcomes differently. For example, aortic injuries have a high mortality but low disability, compared with head injuries, which have moderate mortality and high disability. Young patients with head injuries have less disability and mortality than old patients.65 Trauma patients with significant comorbidities are more likely to have complications. For example; diabetics are more likely to develop infections, obese patients are more likely to develop organ failure and patients with significant aortic stenosis have increased risk of death after injury.66 The reported intensive care unit or hospital length of stay can be impacted by availability of ward beds or skilled nursing beds, and delay in discharge may be related to transportation and patient or family issues. Length of stay is increased in elderly patients and those with significant comorbidities. Length of stay is shorter when patients die early in their hospitalization, and these patients should be excluded from length-of-stay analysis. Trauma registries perform better than administrative databases for analyzing these situations.
Disability is a significant problem in trauma patients, and is an important outcome measure for quality improvement processes.12 The Functional Capacity Index (FCI), Glasgow Outcome Scale score, and modified Functional Independence Measure (FIM) are all measures of functional impairment used in trauma research.67,68 The predicted FCI (pFCI12) is matched to descriptors in AIS-90 and measures the impact of injuries on function at 1 year. The original pFCI12 did not discriminate well, and a consensus group was convened to address these issues. Those changes are currently being validated in the new version of AIS2005.69 Hopefully, the pFCI12 and other measures of functional impairment will prove useful in trauma research and quality improvement processes in the future.
1 Baker SP, O’Neill B, Haddon W, Long WB. The injury severity score: a method for describing patients with multiple injuries and evaluating emergency care. J Trauma. 1974;14:167-196.
2 Boyd CR, Tolson MA, Copes WS. Evaluating trauma care: the TRISS method. Trauma Score and the Injury Severity Score. J Trauma. 1987;27:370-378.
3 Champion HR, Copes WS, Sacco WJ, Lawnick MM, et al. The Major Trauma Outcome Study: establishing national norms for trauma care. J Trauma. 1990;30:1356-1365.
4 MacKenzie EJ, Damiano A, Miller T, Luchter S. The development of the Functional Capacity Index. J Trauma. 1996;41:799-807.
5 Hannan EL, Hicks Waller C, Szypulski Farrell L, Cayten GC. A comparison among the abilities of various injury severity measures to predict mortality with and without accompanying physiologic information. J Trauma. 2005;58:244-251.
6 Meredith JW, Evans G, Kilgo PD, MacKenzie E, et al. A comparison of the abilities of nine scoring algorithms in predicting mortality. J Trauma. 2002;53:621-628.
7 Meredith JW, Kilgo PD, Osler TM. Independently derived survival risk ratios yield better estimates of survival than traditional survival risk ratios when using the ICISS. J Trauma. 2003;55:933-938.
8 Demetriades D, Chan L, Velmanos GV, Sava J, et al. TRISS methodology: an inappropriate tool for comparing outcomes between trauma centers. J Am Coll Surg. 2001;193:250-254.
9 Glance LG, Osler TM, Dick AW. Evaluating trauma center quality: does the choice of the severity-adjustment model make a difference? J Trauma. 2005;56:1265-1271.
10 Norris R, Woods R, Harbrecht B, Fabian T, et al. TRISS unexpected survivors: an outdated standard? J Trauma. 2002;52:229-234.
11 Jones JM. An approach to the analysis of trauma data having a response variable of death or survival. J Trauma. 1995;38:123-128.
12 Glance LG, Dick A, Osler TM, Mukamel D. Judging trauma center quality: does it depend on the choice of outcomes? J Trauma. 2004;56:165-172.
13 Harviel JD, Landsman I, Greenberg A, Copes WS, et al. The effect of autopsy on injury severity and survival probability calculations. J Trauma. 1989;29:766-772.
14 Committee on Medical Aspects of Automotive Safety. Rating the severity of tissue damage—1. The Abbreviated Injury Scale. JAMA. 1971;215:277-280.
15 Copes WS, Lawnick M, Champion HR, Sacco WJ. A comparison of abbreviated injury scale 1980 and 1985 versions. J Trauma. 1988;28:78-86.
16 Gennarelli T, Wodzin E. The Abbreviated Injury Scale–2005. Des Plaines, IL: Association for the Advancement of Automotive Medicine, 2005.
17 Moore EE, Cogbill TH, Malangoni MA, Jurkovich GJ, et al. Organ injury scaling. Surg Clin North Am. 1995;75:293-303.
18 Baker SP, O’Neill B. The injury severity score: an update. J Trauma. 1976;16:882-885.
19 Copes WS, Champion HR, Sacco WJ, Lawnick MM, et al. The injury severity score revisited. J Trauma. 1988;28:69-77.
20 Stephenson SCR, Langley JD, Civil ID. Comparing measures of injury severity for use with large databases. J Trauma. 2002;53:326-332.
21 Garthe E, State JD, Mango NK. Abbreviated Injury Scale Unifi cation: the case for unified injury system for global use. J Trauma. 1999;47:309-323.
22 MacKenzie EJ, Steinwachs DM, Shankar B. Classifying trauma severity based on hospital discharge diagnoses. Validation of an ICD-9CM to AIS-85 conversion table. Med Care. 1989;27:412-422.
23 Kilgo PD, Meredith JW, Hensberry R, Osler TM. A note on the disjointed nature of the injury severity score. J Trauma. 2004;57:479-485.
24 Russell RM, Halcomb BN, Caldwell BA, Sugrue M. Differences in mortality predictions between injury severity score triplets: a significant flaw. J Trauma. 2004;56:1321-1324.
25 Osler TM, Baker SP, Long W. A modification of the injury severity score that both improves accuracy and simplifies scoring. J Trauma. 1997;43:922-926.
26 Brenneman FD, Boulanger BR, McLellan BA, Redelmeier DA. Measuring injury severity: time for a change? J Trauma. 1998;44:580-582.
27 Copes WS, Champion HR, Sacco WJ, Lawnick MM, et al. Progress in characterizing anatomic injury. J Trauma. 1990;30:1200-1207.
28 Champion HR, Copes WS, Sacco WJ, Lawnick MM, et al. A new characterization of injury severity. J Trauma. 1990;30:539-545.
29 Tay SY, Sloan EP, Zun L, Zaret P. Comparison of the New Injury Severity Score and the Injury Severity Score. J Trauma. 2004;56:162-164.
30 Frankema SPG, Steryerberg EW, Edwards MJR, vanVugt AB. Comparison of current injury scales for survival chance estimation: an evaluation comparing the predictive performance of the ISS, NISS, and AP scores in a Dutch local trauma registration. J Trauma. 2005;58:596-604.
31 Lavoie A, Moore L, LeSage N, Liberman M, Sampalis JS. The New Injury Severity Score: a more accurate predictor of in-hospital mortality than the injury severity score. J Trauma. 2004;56:1312-1320.
32 Kilgo PD, Osler TM, Meredith W. The worst injury predicts mortality outcome the best: rethinking the role of multiple injuries in trauma outcome scoring. J Trauma. 2003;55:599-607.
33 Osler T, Rutledge R, Deis J, Bedrick E. ICISS: an international classification of disease-9 based injury severity score. J Trauma. 1996;41:380-386.
34 Meredith JW, Kilgo PD, Osler T. A fresh set of survival risk ratios derived from incidents in the National Trauma Data Bank from which the ICISS may be calculated. J Trauma. 2003;55:924-932.
35 Osler TM, Rogers FB, Glance LG, Cohen M, et al. Predicting survival, length of stay, and cost in the surgical intensive care unit: APACHE II versus ICISS. J Trauma. 1998;45:234-237.
36 Rutledge R, Osler T, Kromhout-Schiro S. Illness severity adjustment for outcomes analysis: validation of the ICISS methodology in all 821,455 patients hospitalized in North Carolina in 1996. Surgery. 1998;124:187-194.
37 Clarke JR, Ragone AV, Greenwald L. Comparisons of survival predictions using survival risk ratios based on International classification of Diseases, Ninth Revision and Abbreviated Injury Scale trauma diagnosis codes. J Trauma. 2005;59:563-569.
38 Osler TM, Cohen M, Rogers FB, Camp L, et al. Trauma registry injury coding is superfluous: a comparison of outcome prediction based on trauma registry International classification of Diseases-Ninth Revision (ICD-9) and hospital information system ICD-9 codes. J Trauma. 1997;43(2):253-256.
39 Rutledge R, Osler T, Emery S, Kromhout-Schiro S. The end of the Injury Severity Score (ISS) and the Trauma and Injury Severity Score (TRISS): ICISS, an International classification of Diseases, ninth revision–based prediction tool, outperforms both ISS and TRISS as predictors of trauma patient survival, hospital charges, and hospital length of stay. J Trauma. 1998;44:41-49.
40 Rutledge R, Osler T. The ICD-9-based illness severity score: a new model that outperforms both DRG and APR-DRG as predictors of survival and resource utilization. J Trauma. 1998;45:791-799.
41 Healey C, Osler TM, Rogers FB, Healey MA, et al. Improving the Glasgow coma scale: motor score alone is a better predictor. J Trauma. 2003;54:671-680.
42 Offner PJ, Jurkovich GJ, Gurney J, Rivara FP. Revision of TRISS for intu-bated patients. J Trauma. 1992;32:32-35.
43 Champion HR, Sacco WJ, Carnazzo AJ, Copes W, Fouty WJ. Trauma Score. Crit Care Med. 1981;9:672-676.
44 Champion HR, Sacco WJ, Copes WS, Gann DS, et al. A revision of the Trauma Score. J Trauma. 1989;29:623-629.
45 Vassar MJ, Lewis FRJr, Chambers JA, Mullins RJ, et al. Prediction of outcome in intensive care unit trauma patients: a multicenter study of Acute Physiology and Chronic Health Evaluation (APACHE), Trauma and Injury Severity Score (TRISS), and a 24-hour intensive care unit (ICU) point system. J Trauma. 1999;47(2):324-329.
46 Clark DE, Winchell RJ. Risk adjustment for injured patients using administrative data. J Trauma. 2004;57:130-140.
47 Sacco WJ, Copes WS, Bain LWJr, MacKenzie EJ, et al. Effect of preinjury illness on trauma patient survival outcome. J Trauma. 1993;35:538-542.
48 Joosse P, Goslings JC, Luitse JSK, Ponsen KJ. M-study: arguments for regional trauma databases. J Trauma. 2005;58:1272-1276.
49 Cayton CG, Stahl WM, Murphy JG, Agarwal N, Byrne DW. Limitations of the TRISS method for interhospital comparisons: a multihospital study. J Trauma. 1991;31:471-482.
50 Clark DE. Comparing institutional trauma survival to a standard: current limitations and suggested alternatives. J Trauma. 1999;47:S92-S98.
51 Champion HR, Copes WS, Sacco WJ, Frey CF, et al. Improved predictions from a severity characterization of trauma (ASCOT) over Trauma and Injury Severity Score (TRISS): results of an independent evaluation. J Trauma. 1996;40:42-48.
52 Hannan EL, Medeloff J, Farrell LS, Cayten CG, Murphy JG. Validation of TRISS and ASCOT using a non-MTOS trauma registry. J Trauma. 1995;38:94-95.
53 Bergeron E, Rossignol M, Osler T, Clas D, Lavoie A. Improving the TRISS methodology by restructuring age categories and adding comorbidities. J Trauma. 2004;56:760-767.
54 Jones JM, Redmond AD, Templeton J. Uses and abuses of statistical models for evaluating trauma care. J Trauma. 1995;38:89-93.
55 Hunt JP, Cherr GS, Hunter C, Wright MJ, et al. Accuracy of administrative data in trauma: splenic injuries as an example. J Trauma. 2000;49:679-686.
56 Jurkovich GJ, Mock C. Systematic review of trauma system effectiveness based on registry comparisons. J Trauma. 1999;47:S46-S55.
57 American College of Surgeons: National Trauma Data Bank Report 2005. Available at www.facs.org/trauma/ntdb.html
58 Demetriades D, Murray J, Charalambides K, Alo K, et al. Trauma fatalities: time and location of hospital deaths. J Am Coll Surg. 2004;198:20-26.
59 Lucas CE, Buechter KJ, Coscia RL, Hurst JM, et al. The effect of trauma program registry on reported mortality rates. J Trauma. 2001;51:1122-1126.
60 Olson CJ, Brand D, Mullins RJ, Harrahill M, Trunkey DD. Time to death of hospitalized injured patients as a measure of quality of care. J Trauma. 2003;55:45-52.
61 Mullins RJ, Mann NC, Hedges JR, Worrall W, et al. Adequacy of hospital discharge status as a measure of outcome among injured patients. JAMA. 1998;279:1727-1731.
62 Mullins RJ, Mann NC, Brand DM, Lenfesty BS. Specifications for calculation of risk-adjusted odds of death using trauma registry data. Am J Surg. 1997;173:422-425.
63 Mann NC, Knight S, Olson LM, Cook LJ. Underestimating injury mortality using statewide databases. J Trauma. 2005;58:162-167.
64 Riddick L, Long WB, Copes WS, Dove DM, Sacco W. Automated coding of injuries from autopsy reports. Am J Forensic Med Pathol. 1998;19:269-274.
65 Demetriades D, Kuncir E, Murray J, Velmahos GC, et al. Mortality prediction of head Abbreviated Injury Score and Glasgow Coma Scale: analysis of 7,764 head injuries. J Am Coll Surg. 2004;199(2):216-222.
66 Neville AL, Brown CV, Weng J, Demetriades D, Velmahos GC. Obesity is an independent risk factor of mortality in severely injured blunt trauma patients. Arch Surg. 2004;139:983-987.
67 Livingston DH, Lavery RF, Mosenthal AC, Knudson MM, et al. Recovery at one year following isolated traumatic brain injury: a Western Trauma Association prospective multicenter trial. J Trauma. 2005;59:1298-1304.
68 MacKenzie EJ, Sacco WJ, Luchter S, Ditunno JF, et al. Validating the Functional Capacity Index as a measure of outcome following blunt multiple Trauma. Qual Life Res. 2002;11:797-808.
69 Gotschall CS. The Functional Capacity Index, second revision: morbidity in the first year post injury. Int J Inj Contr Saf Promot. 2005;12:254-256.
