Linking data from national trauma and rehabilitation registries.

OBJECTIVE: To evaluate te feasibility of retrospectively creating a data base useful in trauma systems evaluations. MATERIALS AND METHODS: Records for 375 patients in both the Major Trauma Outcome Study and the Uniform Data System for Medical Rehabilitation were linked to create an injury-through-rehabilitation data base, including patients from four impairment groups: traumatic brain injury (TBI); spinal cord injury --paraplegic complete (SCI-PARA) and quadriplegic complete (SCI-QUAD); and hip fracture (HIP-FX). MEASUREMENTS AND MAIN RESULTS: The average ages (25.1 years SCI-QUAD, 72.6 years HIP-FX); Injury Severity Score (10.2 HIP-FX, 31.7 SCI-PARA); Revised Trauma Score (5.9 TBI, 7.8 HIP-FX); and acute care lengths of stay (13.3 days HIP-FX, 24.2 days TBI) varied substantially over the four groups. On average, patients spent from approximately 20 days (HIP-FX) to nearly 100 days (SCI-QUAD) in rehabilitation. Functional gains during rehabilitation were primarily in motor skills, but TBI patients also made substantial cognitive gains. Nearly 90% of TBI and SCI patients were discharged to their homes; the percentage of HIP-FX patients discharged to their homes, however, was lower (74%). Across all impairment groups, more patients lived with their relatives after rather than before injury. The correlation between a summary Major Trauma Outcome Study-Functional Independence Measure assessed at acute care discharge and the complete Uniform Data System for Medical Rehabilitation-Functional Independence Measure assessed on admission to rehabilitation was significant for all study patients and for each impairment group except SCI_PARA. CONCLUSIONS: Linking records to create the study data base was arduous and could not be practically accomplished on a large scale or on a continuing basis. Because of the growing emphases on trauma system evaluations and outcomes beyond survival at acute care discharge, we recommend the routine inclusion of rehabilitation data in hospital-based trauma registries.

An evaluation of expert human and automated Abbreviated Injury Scale and ICD-9-CM injury coding.

Two hundred ninety-five injury descriptions from 135 consecutive patients treated at a level-I trauma center were coded by three human coders (H1, H2, H3) and by TRI-CODE (T), a PC-based artificial intelligence software program. Two study coders are nationally recognized experts who teach AIS coding for its developers (the Association for the Advancement of Automotive Medicine); the third has 5 years experience in ICD and AIS coding. A "correct coding" (CC) was established for the study injury descriptions. Coding results were obtained for each coder relative to the CC. The correct ICD codes were selected in 96% of cases for H2, 92% for H1, 91% for T, and 86% for H3. The three human coders agreed on 222 (75%) injuries. The correct 7 digit AIS codes (six identifying digits and the severity digit) were selected in 93% of cases for H2, 87% for T, 77% for H3, and 73% for H1. The correct AIS severity codes (seventh digit only) were selected in 98.3% of cases for H2, 96.3% for T, 93.9% for H3, and 90.8% for H1. On the basis of the weighted kappa statistic TRI-CODE had excellent agreement with the correct coding (CC) of AIS severities. Each human coder had excellent agreement with CC and with TRI-CODE. Coders H1 and H2 were in excellent agreement. Coder H3 was in good agreement with H1 and H2. However, errors among the human coders often occur for different codes, accentuating the variability.(ABSTRACT TRUNCATED AT 250 WORDS)

Results of a multi-institutional outcome assessment: results of a structured peer review of TRISS-designated unexpected outcomes.

The utility of TRISS as a component of trauma center quality assurance (QA) was evaluated. TRISS survival probabilities were estimated for a total of 2,023 consecutive trauma patients admitted to three level-I trauma centers during a 6-month period. A structured peer review was performed of the 50 patients (2.1%) having statistically unexpected outcomes. For 23 (18 survivors, five deaths) TRISS-designated outcomes were sustained in peer review. In 27 cases (one survivor, 26 deaths) TRISS-designated outcomes were not sustained by peer review and TRISS. Limitations were identified in each case. Peer review of unexpected outcomes identified by TRISS provided a consistent and objective QA methodology. An understanding of TRISS as an objective component of the trauma center QA process is essential in blending it with what is, at present, a largely subjective process in many hospitals. Use of TRISS standardizes the peer review process, resulting in a more reliable base for development and improvement of trauma center QA programs.

The Major Trauma Outcome Study: establishing national norms for trauma care.

The Major Trauma Outcome Study (MTOS) is a retrospective descriptive study of injury severity and outcome coordinated through the American College of Surgeons' Committee on Trauma. From 1982 through 1987, 139 North American hospitals submitted demographic, etiologic, injury severity, and outcome data for 80,544 trauma patients. Motor vehicle related injuries were most frequent (34.7%). Twenty-one per cent of patients had penetrating injuries. The overall mortality rate was 9.0%. The mortality rate for direct admissions was strongly related to the presence of serious head injury, 5.0% and 40.0%, when head injuries were less than or equal to AIS (Abbreviated Injury Scale) 3 or greater than or equal to AIS 4, respectively. Survival probability norms use the Revised Trauma Score, Injury Severity Score, patient age, and injury mechanism. Patients with unexpected outcomes were identified and statistical comparisons of actual and expected numbers of survivors made for each institution. Results provide a description of injury and outcome and support evaluation and quality assurance activities.

Progress in characterizing anatomic injury.

A three-valued description of anatomic injury is presented. Anatomic profile (AP) components A, B, and C summarize serious injuries (greater than AIS 2) to the head/brain or spinal cord; to the thorax or front of the neck; and all remaining serious injuries. Relationships between AP components and survival rate reaffirm the seriousness of head injury. Logistic function models relating AP components and the Injury Severity Score (ISS) to survival probability were based on 20,946 Major Trauma Outcome Study (MTOS) patients (9.2% mortality rate) submitted through 1986. Model performance comparisons were based on 5,939 MTOS patients (7.8% mortality rate) submitted during 1987. The AP better discriminated survivors from nonsurvivors and provided a 31% increase in sensitivity when compared with the ISS. Neither the ISS nor the AP alone reliably predict patient outcome. The predictive power of methods for estimating patient survival probability which include physiologic indices or profiles, patient age, and an anatomic profile should be compared with current methods. The AP, which is based on the severity and location of all serious injuries, provides a more rational basis for comparing patient samples than the ISS.

A new characterization of injury severity.

ASCOT (A Severity Characterization of Trauma) is a physiologic and anatomic characterization of injury severity which combines emergency department admission values of Glasgow Coma Scale, systolic blood pressure, respiratory rate, patient age, and AIS-85 anatomic injury scores in a way that obviates ISS shortcomings. ASCOT values are related to survival probability using the logistic function and regression weights reaffirm the importance of head injury and coma to the prediction of patient outcome. The ability of TRISS and ASCOT to discriminate survivors from non-survivors and the reliability of their predictions, as measured by the Hosmer-Lemeshow statistic, were compared using Major Trauma Outcome Study (MTOS) patient data. ASCOT performance matched or exceeded TRISS's for blunt-injured patients and for penetrating-injured patients. ASCOT performance gains were modest for blunt-injured patients. The Hosmer-Lemeshow statistics suggest that ASCOT reliably predicts patient outcome for penetrating-injured patients and nearly so for blunt-injured patients. Statistically reliable predictions were not achieved by TRISS for either set. ASCOT provides a more precise description of patient physiologic status and injury number, location, and severity than TRISS. The ASCOT patient description may be useful in relating to other important outcomes not highly correlated with TRISS or the Injury Severity Score (ISS) such as disability, length of stay, and resources required for treatment.

Progress toward a new injury severity characterization: severity profiles.

Presented is a new seven-dimensional injury severity profile. The profile includes three physiologic assessments and four variables which express the number, location, and severity of a patient's injuries in terms of 'Abbreviated injury scale' values. The physiologic assessments are coded values for the 'Glasgow coma scale', systolic blood pressure, and respiratory rate. Also presented are survival-death predictive values of a cluster model based on survival rates of clusters of profiles of 2569 blunt-injured and penetrating-injured patients. The cluster model has a relative information gain (R) of 0.90. R is a measure of predictive value relative to an infallible predictor. It varies from 0 to 1, the higher the value the better the predictive value. The model had 26 false negatives (deaths predicted to survive) and 35 false positives (survivors predicted to die) giving rise to a false negative rate of 9.3%, a false positive rate of 1.4% and a misclassification rate of 2.4%. The R value and false negative rate are particularly noteworthy, the R value being higher than, and the false negative rate much lower than typical values of 30-40% achieved by TRISS (a combination index based on trauma score, injury severity score and patient age). Also noteworthy is that the clustering was independent of survival/death outcome information and that the good results were achieved even though patient age has not yet been incorporated into the model.

A comparison of Abbreviated Injury Scale 1980 and 1985 versions.

The 1980 and 1985 versions of the Abbreviated Injury Scale (AIS) are quantitatively and qualitatively compared based on experience gained during the recent coding of nearly 115,000 injuries from more than 33,000 seriously injured patients using both AIS versions. Quantitative comparisons are based on differences in AIS scores and Injury Severity Score (ISS) values which result under the two schemes. Qualitative comparisons concern the completeness and clinical usability of the two scales in a trauma center setting.

The Trauma Score as applied to penetrating trauma.

The Trauma Score is a simple physiological measure of injury severity that has been shown to have a high correlation with mortality for blunt trauma patients. In this study, the Trauma Score was evaluated on two subsets of penetrating trauma patients. Results showed that the Trauma Score had a relative information gain of 0.83 and 0.87, an excellent rating in comparison with a possible perfect rating of 1.0. The Score had modest numbers of false negatives (13/64 for the design set, and 3/380 for the test set) and a low number of false positives (5/380 for the design set, and 3/380 for the test set). The Trauma Score is thus an accurate predictor of mortality for both blunt and penetrating trauma patients.