Abstract
Objective:
to analyze the occurrence of severe injuries and deaths among crash victims transported to hospitals in relation to occupant and scene characteristics, including on-scene patient mobility, and their potential use in triaging patients to the appropriate level of care.
Methods:
the occurrence of death and ISS>15 were studied in relation to occupant, crash and mobility data readily available to EMS at the scene, using weighted NASS-CDS data. Data set was randomly split in two for model development and evaluation. Characteristics were combined to develop new triage schemes. Overtriage and undertriage rates were calculated for the NASS-CDS case trauma center allocation and for the newly developed triage schemes.
Results:
Compared to the NASS-CDS distribution, a scheme using patient mobility alone showed lower overtriage of those with ISS≤15 (38.8% vs. 55.5%) and lower undertriage of victims who died from their crash-related injuries (2.34% vs. 21.47%). Undertriage of injuries with ISS> 15 was similar (16.0 vs. 16.9). A scheme based on the presence of one of many scene risk factors (age>55, GCS<14, intrusion ≥18”, near lateral impact, far lateral impact with intrusion ≥12”, rollover or lack of restraint use) resulted in an undertriage of 0.86% (death) and 10.5% (ISS>15) and an overtriage of 63.4%. The combination of at least one of the scene risk factors and mobility status greatly decreased overtriage of those with ISS<15 (24.4%) with an increase in death undertriage (3.19%). Further combination of mobility and scene factors allowed for maintenance of a low undertriage (0.86%) as well as an acceptable overtriage (48%).
Conclusion:
Patient mobility data easily obtained at the scene of a crash allows triaging of injured patients to the appropriate facility with a high sensitivity and specificity. The addition of crash scene data to scene mobility allows further reductions on undertriaging or overtriaging.
BACKGROUND
The objective of civilian trauma triaging has been summarized as “getting the right patient to the right place in the right time” (COT ACS, 2006).
Existing triage criteria have evolved over time (Mckersie, 2006; Sasser, 2009). Criteria include physiologic, anatomic and mechanistic information ranked in a hierarchical fashion to match cases with appropriate care settings (i.e. Level I, II, III, IV trauma centers and non-specialized acute care facilities) (COT ACS, 2006; Mckersie, 2006; Sasser, 2009).
An ideal triage scheme would discriminate cases with a high likelihood of severe injury or mortality from those with low likelihood of those outcomes. Undertriage and overtriage represent the failure of the scheme to properly allocate high risk and low risk cases to the appropriate setting (COT ACS, 2006; Mckersie, 2006; Sasser, 2009). An inappropriate matching of patient needs to the appropriate level of care (i.e. trauma versus non-trauma center) worsens clinical outcomes, wastes costly resources, and decreases the surge capacity of trauma centers (COT ACS, 2006; Mckersie, 2006; Sasser, 2009).
While mechanistic and injury data obtained at the scene of car crashes are included as part of the triage criteria, little attention has been given to patient mobility status at the scene as a potential triage factor in non mass casualty civilian trauma. Since patient scene mobility information (i.e. “ejection”, “removed due to decreased mental status”, “self exited”, “exited with assistance”, “removed due to perceived serious injury”) is easily available and frequently reported by emergency medical services (EMS), its inclusion in triaging schemes would not require major changes in education of pre-hospital providers.
The objectives of this study are: 1) to analyze the occurrence of severe injuries and deaths among vehicular crash victims transported to hospital facilities in relation to occupant and scene characteristics, including on-scene patient mobility, which are readily available to EMS personnel, 2) to combine characteristics of interest into triage schemes and 3) to evaluate the proposed triage system performance in comparison to the actual allocation of cases (i.e. Trauma Center vs. undesignated hospital facility) occurring in the NASS-CDS cohort.
METHODS
The study used the National Automotive Sampling System Crashworthiness Data System (NASS-CDS) and included cases from 2003 through 2008 with crash occupants aged 15 and older who were transferred to a hospital facility. Study was limited to occupants of cars, light trucks, sport utility vehicles, and vans. Deaths at the scene or due to illness were excluded.
NASS-CDS is a probability sample of all police reported crashes in the United States and contains detailed data on thousands of minor, serious, and fatal crashes. NASS-CDS cases are crashes that involve a harmful event (property damage and/or personal injury) and at least one towed passenger car, light truck or van in transport on a traffic-way. After statistical weighting methods are applied, NASS-CDS becomes a representative, random sample of crashes in the US.
NASS-CDS enters approximately 5,000 cases per year over 24 regions (sampling units) and includes data that describe scene, crash, vehicle, occupant and injury characteristics. Information is collected by trained crash investigators, including data from crash sites, vehicle inspections, interviews with crash victims, and review of medical records to determine the nature and severity of injuries (NHTSA, 1998).
Analytic Strategy
The dataset was randomized into 2 groups: development set and evaluation set. The occurrence of death or ISS>15 within the training set was studied in relation to data readily available to EMS at the scene: age, gender, vehicles’ main general area of damage (GAD), degree of intrusion, rollover, patient mobility (self exited and exited with minimal assistance), vehicle type, restraint use, and patient mental status. Characteristics were analyzed to select the ones that could be used in triage schemes using χ2 and Mantel-Haenszel tests. P <0.001 was used for statistical significance.
Overtriage and undertriage rates were calculated for the natural allocation of cases to trauma centers in the weighted NASS-CDS cohort and for the newly developed triage schemes. Performance of each scheme (i.e. overtriage and undertriage) was compared to the observed triage in NASS-CDS. Triaging performance was calculated for the development set, evaluation set and for the entire cohort.
On-scene patient mobility was based on the following options given within NASS-CDC (NHTSA, 1998): ejection (i.e. occupant fully ejected), removed due to decreased mental status (i.e. removed from vehicle while unconscious or not oriented to time or place), self exited (i.e. exited from vehicle under own power), exited with minimal assistance (i.e. exited from vehicle with some assistance), removed due to perceived serious injury, removed from vehicle for other reasons, and unknown.
Undertriage was defined as the number of deaths or cases with ISS>15 who were not properly triaged by the scheme (1-sensitivity). Overtriage was defined as the number of cases without ISS>15 who were not properly triaged (1-specificity) (Mckersie, 2006; Sasser, 2009).
Research was compliant with the University of Maryland Institutional Review Board policy.
RESULTS
After exclusions were applied, 23,815 cases representing 4,726,252 weighted cases were included in the analysis. Characteristics of the cases in the entire cohort, and in the development and evaluation sets are presented in Table 1.
Mortality and ISS>15 occurred in 0.60%, and 4.07% of all the cases respectively; 56.7% of cases were transported to a trauma center, representing 78.5% of deaths and 84.1% of those with ISS>15.
Analysis of the development set revealed mortality and ISS>15 to be positively associated (p<0.001) with age>55, near lateral, top and under GAD, intrusion ≥ 30.5 cm, rollover, ejection and GCS<14, and negatively associated (p<0.001) with seatbelt use and with mobility at the scene (i.e. “self exited” or “exited with minimal assistance”) (Tables 2 and 3). Similar findings were present when the entire cohort was used (Tables 2 and 3).
The NASS-CDS cohort allocation of cases in the development set resulted in undertriage of 20.3% (death) and 16.2% (ISS>15) and an overtriage of 55.5% (ISS>15) (Table 4).
Four triage schemes were built: 1) a model using only mobility at the scene (Mobility) (i.e. those who were not “self exited” or “exited with minimal assistance”), 2) a model including any of the other scene factors (Scene) (i.e. age>55, GCS<14, intrusion ≥45.7 cm, near lateral impact, far lateral impact with intrusion ≥30.5cm, rollover or lack of restraint use) but not mobility, 3) a model that combines “Mobility” and “Scene” (i.e those that fulfilled both the “mobility” and “scene” model criteria) (Mobility+Scene), and 4) a model combining the Mobility+Scene model or GCS<14 or age > 55 (i.e. cases older than 55 or GCS<14, or fulfilling Mobility+Scene criteria) (Mobility+Scene/GCS/Age).
Performances of the first 3 models were calculated in the development set. The fourth model (Mobility+Scene/GCS/Age) was developed by selecting the combination of factors that would preserve the low mortality undertriage rate of the “Scene” scheme while reducing the overtriage rate. Performance of the NASS-CDS cohort allocation and the new four schemes in the development and evaluation sets are presented in tables 4 and 5, respectively.
While the NASS-CDS allocation undertriage and overtriage performance were similar in both the development and evaluation sets, mortality undertriage was markedly lower in the evaluation set than in the development set for the four new triage schemes (Tables 4 and 5). Furthermore, while exhibiting a relatively high mortality undertriage in the development set, the “mobility” scheme outperformed all other schemes in this regard in the evaluation set (Tables 4 and 5). Overtriage performance of all models did not differ greatly in the development and evaluation sets.
Table 6 reports the triaging performance of the NASS-CDS cohort allocation and the four new schemes for the entire cohort. The NASS-CDS cohort natural allocation of cases resulted in undertriage of 21.5% (death) and 16.0% (ISS>15) and an overtriage of 55.5% (ISS>15) (Table 4).
The “Mobility” scheme allocated 40.6% of the cases to trauma centers and showed undertriage levels of 2.34% for death and 16.9 for ISS>15, and overtriage levels of 38.8 (ISS>15) (Table 6). The “Scene” model allocated 64.4% of cases to a trauma center and showed an undertriage rate of 0.86% for death and 10.5% for ISS>15, and an overtriage of 63.4% (ISS>15) (Table 6). With the “Mobility+Scene” model only 26.5% of the cases were allocated to a trauma center with a resulting undertriage of 3.19% (death) and 24.3% (ISS>15), and overtriage of 24.4% (ISS>15) (Table 6). Finally the last model “Mobility+Scene/GCS/Age) allocated 49.5% of the cases to a trauma center, resulting in an undertriaging 0.86% (death) and 17.4 (ISS>15), and overtriage (ISS>15) of 48.0%.
DISCUSSION
The ideal field triage criteria should consist of a set of decision rules allowing pre-hospital providers to transport injured victims to a specialized acute care facility (i.e. trauma center) versus an undesignated, non-specialized acute care facility (Mackersie, 2006). The interest in developing rules that effectively allocate the most severely injured patients has grown over the last three decades with the recognition that outcomes of seriously injured victims are improved by their treatment at specialized trauma centers (Mackersie, 2006; Sasser, 2009).
Regionalized trauma systems rely on the ability of decision-making schemes to not only allocate those severely injured patients to trauma centers, but also to limit the number of minimally injured individuals that are brought to this specialized care setting (i.e. overtriage) (Mackersie, 2006; Sasser, 2009). The negative consequences of overtriage are not limited to the cost of the wasted resources but extend to the overburdening of the trauma system that becomes unable to properly care for those severely injured victims that could benefit the most from specialized care. These outcomes could be of particular importance during periods of time where there is a surge in demand for care (i.e. disasters, mass casualties, etc.) (Sasser, 2009; Frykeberg, 2002).
Vehicular crashes represent one of the main causes of injury requiring hospital admission (WISQARS, 2011) as well as deaths due to trauma, particularly among those in the 15–54 age group (NTDB). The importance of this mechanism is reflected by the presence of multiple triage criteria specific to vehicular crashes within the field triage criteria (ACS COT, 2006; Sasser, 2009).
While ability to walk at the scene is the first step in multiple algorithms for the triaging of mass casualties (Garner, 2001), it has been minimally studied as a criterion in non-mass casualty civilian trauma (Scheetz 2007).
We utilized a population representative of the universe of crashes resulting in property damage and/or personal injury in the USA to test the performance of triaging schemes that include basic crash and occupant scene data and mobility at the scene.
This analysis showed that using mobility at the scene (i.e. those who did not “self exit” or “exited with minimal assistance”) as the only triage criterion more accurately allocated fatally injured victims to a trauma center than the allocation actually experienced by the studied NASS-CDS cohort (undertriage of 0.4% vs. 22.6%). In regard to the allocation of ISS>15, scene mobility performed similarly to the NASS-CDS cohort in regard to undertriage (16.1% vs. 15.7%) and considerably better in regard to overtriaging (38.7% vs 55.5%).
The very high undertriage of fatalities experienced by the NASS/CDS cohort is worrisome and needs to be explained. Possible explanations of this high undertriage rate include poor access to trauma centers (absence of trauma systems, rural location, etcetera) or inadequate implementation of current triage guidelines. Alternatively, this high undertriage may be due to the transport of cases “in-extremis” to the closest facility.
Subgroup analysis (not shown) for NASS/CDS urban regions (i.e. large cities - one of the 60 largest standard metropolitan statistical areas), large suburban areas, and others revealed a lower death undertriage rate in the large suburban regions than in the remaining NASS/CDS population. This suggests that triage implementation is greatly influenced by geography and urbanization characteristics.
The combination of multiple scene criteria (excluding scene mobility) yielded a similar low undertriage of deaths (0.4%) as well as a reduction of severe injury undertriage (12.8%). However, overtriage worsened to 62.8%.
Rearrangement of mobility and scene criteria allowed a further improvement of overtriage to 48.0% (Mobility+Scene/GCS/Age) and 23.9% (Mobility+Scene). While the Mobility+Scene/GCS/Age model preserved the very low undertriage rate for death, the Mobility+Scene model showed an increase in the undertriage rate to 0.8%.
All models utilizing scene mobility resulted in lower overtriage than the one experienced by the NASS-CDS cohort and within the accepted 25–50% overtriage rate considered acceptable by the American College of Surgeons Committee on Trauma (COT ACS, 2006). While all the new models resulted in death undertriage rates within the ACS recommended range (i.e. <1%) in the evaluation set, only the Scene and Mobility+Scene/GCS/Age models showed that recommended performance in the entire cohort. The severe injury undertriage rates experienced by the NASS-CDS cohort and resulting from the application of all the new triage models exceeded the 5% recommended by the ACS in both the evaluation set as well as the entire cohort.
The study design used to avoid overoptimistic results due to “overfitting” (i.e. the use of “development” and “evaluation” sets) paradoxically resulted in an unexpected better performance in regard to mortality undertriage (p<0.0001) in the “evaluation” set than in the “development” set.
This unlikely scenario may be related to the nature of NASS-CDS sampling and weighting methodology. It is possible that a (random) sampling error during the randomization of the un-weighted sample may have been amplified to the weighted population in such a way that it resulted in significantly different populations.
Given that the weighted NASS-CDS cohort represents, not a sample, but the universe of crashes resulting in property damage and/or personal injury in the USA, we believe that the findings for the entire cohort are less prone to error than those for the “evaluation” cohort.
Furthermore, as our models are not very complex, with a relatively small number of parameters relative to the number of observations (i.e deaths), and since we did not use multiple logistic regression models with selection procedures, “overfitting” should not be a major concern. Nevertheless, the validity and actual performance of our models will be better established in a prospective fashion in a truly random sample.
Another limitation of this study is the lack of physiologic data (i.e. blood pressure and respiratory rate) and information regarding co-morbidities that would have allowed comparisons to the current CDC/NHTSA triaging scheme instead to the naturally occurring triage within NASS-CDS.
A model (not included in the results section) using the CDC/NHTSA scheme without blood pressure and respiratory rate when applied to the entire cohort yielded a fatality undertriage rate of 2.71%, and severe injury under- and overtriage rates of 16.3% and 49.6%, respectively. Anatomical criteria in this model were based on the discharge diagnosis reported within NASS-CDS.
The inclusion of blood pressure and respiratory rate (not available within NASS-CDS) might have decreased the undertriage in the CDC/NHTSA model with probably little effect on overtriage rates. Since “physiologic” factors have been described to reveal the highest yield among triage criteria (Esposito TJ, 1995), it should be investigated whether a model combining blood pressure and respiratory rate with mobility and scene factors could show some improvement of undertriage and/or overtriage rates.
Another factor not included in the analysis was the presence of alcohol intoxication as a potential confounder of “poor” mobility. Those reported to “consume alcohol” by the police, were less likely to exit with minimal assistance or self exit from their vehicle than those reported not to consume alcohol (44% vs. 62%). We chose not to include alcohol consumption as a variable because the “subjective” assessment of alcohol consumption or intoxication of the injured is known to be subjected to a great degree of bias and a definite objective assessment of its presence is unlikely to be available at the scene of a crash in most of cases (NASS-CDS variable for alcohol involvement includes data extracted from the police report based in information obtained both at the scene and after the fact).
Finally, the lack of documentation of on-scene patient mobility in close to one-fifth of the cases adds a degree of uncertainty to the data. Of interest, those classified as “unknown” scene mobility within NASS-CDS experienced higher mortality and injury severity than those that self exited or exited with minimal assistance. While they experienced more injuries in all body regions than those self exiting, this difference was more pronounced for lower extremity and head injuries.
The process followed by NASS-CDS investigators to adjudicate “unknown” in those cases is unclear. It is uncertain whether “unknown” represents poor documentation in the EMS record or a limitation in NASS-CDS classification of scene mobility that is unable to capture EMS data. This limitation calls for the evaluation of these criteria prospectively in a population where scene mobility is better characterized and clearly documented.
Our findings for the entire USA population would need further validation before their application to particular trauma systems or geographical locations.
Well-designed algorithms overcome deficiencies in human judgment by incorporating principles of statistics, decision theory, and epidemiology in clinically useful formats (Elstein, 2009). Algorithms based on a small number of simple data points can greatly assist in decision making, particularly in emergency situations.
The simplicity of using scene mobility as the first step in triage also represents a great advantage, allowing responders with even the most basic knowledge to assist in decision-making.
In conclusion, easily obtained scene mobility data allow triaging of injured MVC patients to the appropriate facility with better sensitivity and specificity than the naturally occurring triage currently experienced by a population representative of USA crashes. The addition of occupant and crash scene data to scene mobility, allows further reductions in overtriaging or undertriaging depending on the combinations of factors used.
Acknowledgments
The authors wish to acknowledge the valuable contribution of Cindy Burch, who assisted with the editing and formatting of the manuscript.
APPENDIX
TABLES:
Table 1.
Occupant, Crash, Scene and Outcome Characteristics (N=4,726,252)
| All Cases (%) | Development Set* (%) | Evaluation Set** | |
|---|---|---|---|
| Male | 46 | 47.5 | 44.5 |
| Age>55 | 17.9 | 18.8 | 17.1 |
| Restrained1 | 67 | 67.2 | 66.9 |
| Driver position | 80.4 | 79.9 | 80.8 |
| Rollover | 13.4 | 13.8 | 13.0 |
| GAD | |||
| Frontal | 45.6 | 45.4 | 45.7 |
| Near side (lateral) impact | 12.5 | 12.6 | 12.5 |
| Far side (lateral) impact | 9.4 | 8.6 | 10.1 |
| Rear | 7.7 | 7.3 | 8.1 |
| Top | 6.8 | 6.5 | 7.2 |
| Under | 0.3 | 0.3 | 0.3 |
| Unknown | 1.4 | 1.3 | 1.4 |
| Missing | 16.3 | 18.0 | 14.6 |
| Delta V (km/h) | |||
| 0–29 | 41.4 | 39.7 | 43.0 |
| ≥ 30 | 12.3 | 11.8 | 12.8 |
| Unknown | 46.3 | 48.5 | 44.0 |
| Intrusion (cm)2 | |||
| <2.54 | 64.6 | 66.0 | 63.6 |
| 2.54–27.9 | 27.8 | 26.4 | 29.0 |
| 30.5–43.2 | 5.0 | 5.4 | 4.7 |
| ≥ 45.7 | 2.6 | 2.2 | 3.0 |
| Scene Mobility | |||
| Ejected | 1.4 | 1.4 | 1.3 |
| Removed with ↓ mental status | 4.1 | 3.5 | 4.7 |
| Removed due to perceived injury | 13.9 | 14.3 | 13.4 |
| Exited with assistance | 18.2 | 17.5 | 18.9 |
| Self exited | 41.3 | 42.0 | 40.6 |
| Unknown | 21.2 | 21.3 | 21.1 |
| GCS <14 | 1.11 | 1.14 | 1.08 |
| Death | 0.60 | 0.59 | 0.61 |
| ISS>15 | 4.1 | 4.0 | 4.1 |
Development set N=2,310,063, random sample (48.88%) of weighted cohort.
Evaluation set N=2,416,189, random sample (51.22%) of weighted cohort.
Restraint status unknown in 11.6% of cases.
<1, 1–11, 12–17, and ≥18 inches.
Table 2.
Mortality and Injury Severity by Occupant and Crash Characteristics.
| Development Set* | All cases | |||
|---|---|---|---|---|
| Mortality (%) | ISS>15 (%) | Mortality (%) | ISS>15 (%) | |
| Gender | ||||
| Female | 0.42 | 3.4 | 0.44 | 3.2 |
| Male | 0.78 | 4.9 | 0.80 | 5.2 |
| Age | ||||
| ≤55 | 0.49 | 3.4 | 0.50 | 3.6 |
| >55 | 1.02 | 6.6 | 1.04 | 6.0 |
| Restraint | ||||
| Use | 0.31 | 2.6 | 0.28 | 2.6 |
| Non Use | 1.23 | 8.3 | 1.52 | 9.0 |
| Position | ||||
| Passenger | 0.56 | 3.6 | 0.69 | 3.9 |
| Driver | 0.59 | 4.1 | 0.57 | 4.1 |
| Rollover | ||||
| No | 0.53 | 3.5 | 0.48 | 3.4 |
| Yes | 0.86 | 6.7 | 1.12 | 6.9 |
| GAD | ||||
| Frontal | 0.59 | 3.3 | 0.53 | 3.4 |
| Near side lateral | 0.99 | 7.8 | 0.94 | 6.8 |
| Far side lateral | 0.67 | 5.7 | 0.53 | 4.8 |
| Rear | 0.51 | 1.6 | 0.28 | 1.4 |
| Top | 0.74 | 7.0 | 1.22 | 7.5 |
| Under | 0 | 0.8 | 2.34 | 4.0 |
| Unknown | 0 | 1.4 | 0.0 | 1.1 |
| Missing | 0.29 | 2.6 | 0.47 | 3.6 |
| Delta V (km/h) | ||||
| 0–29 | 0.19 | 2.0 | 0.15 | 1.7 |
| ≥ 30 | 1.59 | 11.2 | 1.57 | 10.1 |
| Unknown | 0.67 | 3.9 | 0.74 | 4.6 |
| Intrusion (cm)1 | ||||
| <2.54 | 0.26 | 1.8 | 0.24 | 1.9 |
| 2.54–27.9 | 0.67 | 5.2 | 0.64 | 5.4 |
| 30.5–43.2 | 1.96 | 17.4 | 2.52 | 15.7 |
| ≥ 45.7 | 5.85 | 23.9 | 5.08 | 21.8 |
| GCS | ||||
| 15–14 | 0.40 | 3.3 | 0.40 | 3.4 |
| < 14 | 16.25 | 71.3 | 17.83 | 68.2 |
Development set N=2,310,063, random (48.88%) sample of the entire cohort.
<1, 1–11, 12–17, and ≥18 inches.
Table 3.
Mortality and Injury Severity by Scene Mobility Characteristics.
| Development Set* | All cases | |||
|---|---|---|---|---|
| Mortality (%) | ISS>15 (%) | Mortality (%) | ISS>15 (%) | |
| Ejected | 6.86 | 33.4 | 9.75 | 36.3 |
| Removed with ↓ mental status | 5.14 | 24.0 | 4.91 | 21.6 |
| Removed due to perceived injury | 1.22 | 8.7 | 1.07 | 8.9 |
| Exited with minimal assistance | 0.01 | 1.3 | 0.01 | 1.3 |
| Self exited | 0.06 | 1.2 | 0.03 | 1.1 |
| Unknown | 0.51 | 3.5 | 0.47 | 3.6 |
Development set N=2,310,063, random (48.88%) sample of the entire cohort.
Table 4.
Development set*: Undertriage and Overtriage of Different Schemes
| NASS-CDS Cohort | Mobility1 | Scene2 | Mobility and Scene3 | Age or GCS<14 or (Mobilty and Scene) 4 | |
|---|---|---|---|---|---|
| Trauma triage (%) | 56.6 | 40.6 | 65.1 | 26.9 | 49.5 |
| Death | |||||
| Undertriage (%) | 20.27 | 4.46 | 1.32 | 5.78 | 1.32 |
| ISS>15 | |||||
| Undertriage (%) | 16.2 | 17.9 | 8.0 | 24.2 | 14.7 |
| Overtriage (%) | 55.5 | 38.8 | 64.0 | 28.9 | 48.0 |
Development set N=2,310,063, random sample (48.88%) of the entire cohort.
Those that did not self exited or exited with minimal assistance triaged to trauma centers.
Cases with intrusion ≥ 45.7cm, ejection, rollover, near GAD, lack of restraint use, GCS< 14 or age >55 triaged to trauma center.
Those following Mobility and one Scene criteria triaged to trauma centers.
Those following Mobility and Scene criteria (3 above) or age >55 or GCS <14 triaged to trauma centers.
Table 5.
Evaluation Set*: Undertriage and Overtriage of Different Schemes
| NASS-CDS Cohort | Mobility1 | Scene2 | Mobility and Scene3 | Age or GCS<14 or (Mobilty and Scene) 4 | |
|---|---|---|---|---|---|
| Trauma triage (%) | 56.7 | 40.7 | 63.8 | 26.1 | 49.4 |
| Death | |||||
| Undertriage (%) | 22.58 | 0.37 | 0.43 | 0.80 | 0.43 |
| ISS>15 | |||||
| Undertriage (%) | 15.7 | 16.1 | 12.81 | 24.36 | 19.8 |
| Overtriage (%) | 55.5 | 38.7 | 62.8 | 23.94 | 48.0 |
Development set N=2,310,063, random sample (48.88%) of the entire cohort.
Those that did not self exited or exited with minimal assistance triaged to trauma centers.
Cases with intrusion ≥ 45.7cm, ejection, rollover, near GAD, lack of restraint use, GCS< 14 or age >55 triaged to trauma center.
Those following Mobility and one Scene criteria triaged to trauma centers.
Those following Mobility and Scene criteria (3 above) or age >55 or GCS <14 triaged to trauma centers.
Table 6.
All cases*: Undertriage and Overtriage of Different Schemes
| NASS-CDS Cohort | Mobility1 | Scene2 | Mobility and Scene3 | Age or GCS<14 or (Mobilty and Scene) 4 | |
|---|---|---|---|---|---|
| Trauma triage (%) | 56.7 | 40.6 | 64.4 | 26.5 | 49.45 |
| Death | |||||
| Undertriage (%) | 21.47 | 2.34 | 0.86 | 3.19 | 0.86 |
| ISS>15 | |||||
| Undertriage (%) | 16.0 | 16.9 | 10.5 | 24.3 | 17.4 |
| Overtriage (%) | 55.5 | 38.8 | 63.4 | 24.4 | 48.0 |
Entire cohort n=4,726,252.
Those that did not self exited or exited with minimal assistance triaged to trauma centers.
Cases with intrusion >18”, ejection, rollover, near GAD, lack of restraint use, GCS< 14 or age >55 triaged to trauma center.
Those following Mobility and one Scene criteria triaged to trauma centers.
Those following Mobility and Scene criteria (3 above) or age >55 or GCS <14 triaged to trauma centers.
REFERENCES
- COT ACS ACS, Chicago IL (2006) Resources for Optimal Care of the injured patient. 2006.
- Elstein AS. Thinking about diagnostic thinking: a 30-year perspective. Adv in Health Sci Educ. 2009;14:7–18. doi: 10.1007/s10459-009-9184-0. [DOI] [PubMed] [Google Scholar]
- Esposito TJ, Offner PJ, Jurkovich GJ, et al. Do prehospital trauma center triage criteria identify major trauma victims? Arch Surg. 1995;130:171–176. doi: 10.1001/archsurg.1995.01430020061010. [DOI] [PubMed] [Google Scholar]
- Frykberg ER. Medical management of disasters and mass casualties from terrorist bombings: how can we cope? J Trauma. 2002;53:201–212. doi: 10.1097/00005373-200208000-00001. [DOI] [PubMed] [Google Scholar]
- Garner A, Lee A, Harrison K, et al. Comparative analysis of multiple-casualty incident triage algorithms. Ann Emerg Med. 2001;38:541–548. doi: 10.1067/mem.2001.119053. [DOI] [PubMed] [Google Scholar]
- Mackersie RC. History of trauma field triage development and the American College of Surgeons criteria. Prehosp Emerg Care. 2006 Jul-Sep;10(3):287–94. doi: 10.1080/10903120600721636. [DOI] [PubMed] [Google Scholar]
- National Automotive Sampling System, Crashworthiness Data System (NASS-CDS) http://www.nhtsa.gov/people/ncsa/nass_cds.html Accessed 3/13/2011. [DOI] [PubMed]
- National Highway Traffic Safety Administration . US Department of Transportation; Washington, DC: 1998. 1988 - 1996 NASS CDS Variable-Attribute Structure Manual - National Automotive Sampling System (NASS) Crashworthiness Data System. Available at http://www-nrd.nhtsa.dot.gov/Pubs/CDS8896.pdf Accessed 3/13/2011. [Google Scholar]
- National Trauma Data Bank (NTDB) 2010 Annual report. http://www.facs.org/trauma/ntdb/pdf/ntdbannualreport2010.pdf Accessed 2/14/2011.
- Sasser SM, Hunt RC, Sullivent EE, Wald MM, Mitchko J, Jurkovich GJ, Henry MC, Salomone JP, Wang SC, Galli RL, Cooper A, Brown LH, Sattin RW, National Expert Panel on Field Triage, Centers for Disease Control and Prevention (CDC) Guidelines for field triage of injured patients. Recommendations of the National Expert Panel on Field Triage. MMWR Recomm Rep. 2009;58(RR-1):1–35. [PubMed] [Google Scholar]
- Scheetz LJ, Zhang J, Kolassa JE. Using crash scene variables to predict the need for trauma center care in older persons. Research in Nursing and Health. 2007;30:399–412. doi: 10.1002/nur.20203. [DOI] [PubMed] [Google Scholar]
- WISQARS http://www.cdc.gov/injury/wisqars/index.html, accessed 2/4/2011.