Estimated Injury Risk for Specific Injuries and Body Regions in Frontal Motor Vehicle Crashes

Abstract

Objective: Injury risk curves estimate motor vehicle crash (MVC) occupant injury risk from vehicle, crash, and/or occupant factors. Many vehicles are equipped with event data recorders (EDRs) that collect data including the crash speed and restraint status during a MVC. This study's goal was to use regulation-required data elements for EDRs to compute occupant injury risk for (1) specific injuries and (2) specific body regions in frontal MVCs from weighted NASS-CDS data.

Methods: Logistic regression analysis of NASS-CDS single-impact frontal MVCs involving front seat occupants with frontal airbag deployment was used to produce 23 risk curves for specific injuries and 17 risk curves for Abbreviated Injury Scale (AIS) 2+ to 5+ body region injuries. Risk curves were produced for the following body regions: head and thorax (AIS 2+, 3+, 4+, 5+), face (AIS 2+), abdomen, spine, upper extremity, and lower extremity (AIS 2+, 3+). Injury risk with 95% confidence intervals was estimated for 15–105 km/h longitudinal delta-Vs and belt status was adjusted for as a covariate.

Results: Overall, belted occupants had lower estimated risks compared to unbelted occupants and the risk of injury increased as longitudinal delta-V increased. Belt status was a significant predictor for 13 specific injuries and all body region injuries with the exception of AIS 2+ and 3+ spine injuries. Specific injuries and body region injuries that occurred more frequently in NASS-CDS also tended to carry higher risks when evaluated at a 56 km/h longitudinal delta-V. In the belted population, injury risks that ranked in the top 33% included 4 upper extremity fractures (ulna, radius, clavicle, carpus/metacarpus), 2 lower extremity fractures (fibula, metatarsal/tarsal), and a knee sprain (2.4–4.6% risk). Unbelted injury risks ranked in the top 33% included 4 lower extremity fractures (femur, fibula, metatarsal/tarsal, patella), 2 head injuries with less than one hour or unspecified prior unconsciousness, and a lung contusion (4.6–9.9% risk). The 6 body region curves with the highest risks were for AIS 2+ lower extremity, upper extremity, thorax, and head injury and AIS 3+ lower extremity and thorax injury (15.9–43.8% risk).

Conclusions: These injury risk curves can be implemented into advanced automatic crash notification (AACN) algorithms that utilize vehicle EDR measurements to predict occupant injury immediately following a MVC. Through integration with AACN, these injury risk curves can provide emergency medical services (EMS) and other patient care providers with information on suspected occupant injuries to improve injury detection and patient triage.

Keywords:

• Abbreviated Injury Scale
• advanced automatic crash notification
• event data recorder
• injury risk curve
• logistic regression

Introduction

Over 5.3 million police-reported motor vehicle crashes (MVCs) occur annually in the United States, resulting in over 32,000 fatalities and over 2,200,000 injured victims. Though vehicle safety improvements (i.e., seat belts, airbags) have lowered the incidence of morbidity and mortality in MVCs, MVC injuries are still a major cause of death and hospitalization. The U.S. annual cost of reported and unreported MVCs is estimated to be over $230 billion (NHTSA 2011).

Injury risk curves estimate MVC occupant injury risk from vehicle, crash, and/or occupant factors. Logistic regression is a statistical method used commonly in injury biomechanics to estimate injury risk (Hosmer et al. 2013; Petitjean et al. 2009). Utilizing data from several sources including the NASS-CDS and event data recorders (EDRs), previous studies identified important parameters for estimating occupant injury such as crash configuration, delta-V, age, ejection, restraint use, intrusion, and seating position (Augenstein et al. 2003; Austin and Faigin 2003; Clark 2003; Clark and Cushing 2002, 2004; Kullgren et al. 2003; O’Donnell and Connor 1996). Injury risk curves for specific body regions or injuries have been developed previously using a variety of statistical models (Kent and Funk 2004; Mertz et al. 1997; Prasad et al. 2010; Rupp et al. 2010a, 2010b).

Table 1 List of the injury risk curves produced for specific body regions and the injury risk model coefficients (α, β₁, β₂). The unweighted (N_U) and weighted (N_W) sample sizes, area under the receiver operating characteristic curve (AUROC), P value for belt status, and estimated risk at a 56 km/h longitudinal delta-V is provided for belted versus unbelted cases. For each risk curve, the higher injury risk at 56 km/h longitudinal delta-V between belted and unbelted occupants is bolded

	Model coefficients			Belted			Unbelted
Body region injury											P value for
risk curve	α, Intercept	β1, Belt	β2, DV	NU	NW	Risk, 56 km/h (%)	NU	NW	Risk, 56 km/h (%)	AUROC	belt status
Head
AIS 2+	−4.763	−1.428	0.066	358	38,501	7.56	245	26,906	25.44	0.76	<.0001*
AIS 3+	−6.440	−1.223	0.066	133	9,267	1.84	92	5,908	5.99	0.81	.0002*
AIS 4+	−6.729	−1.340	0.062	89	5,550	1.02	59	3,966	3.80	0.81	.0008*
AIS 5+	−7.046	−1.279	0.059	55	3,835	0.65	34	2,556	2.31	0.80	.0091*
Face
AIS 2+	−6.708	−1.803	0.060	52	3,355	0.59	62	3,755	3.49	0.82	<.0001*
Thorax
AIS 2+	−6.263	−0.439	0.085	409	42,873	12.36	203	13,482	17.94	0.81	.0294*
AIS 3+	−6.868	−0.857	0.093	269	22,334	7.40	169	10,724	15.85	0.85	.0004*
AIS 4+	−7.789	−1.251	0.095	129	7,422	2.35	84	5,358	7.75	0.88	.0006*
AIS 5+	−8.204	−1.635	0.091	45	2,978	0.84	41	3,166	4.17	0.88	.0012*
Abdomen
AIS 2+	−6.167	−1.508	0.077	175	13,482	3.36	111	11,310	13.58	0.85	<.0001*
AIS 3+	−7.086	−1.155	0.072	69	6,418	1.43	46	4,012	4.39	0.85	.0032*
Spine
AIS 2+	−7.317	0.063	0.075	185	18,839	4.55	84	3,745	4.29	0.77	.8593
AIS 3+	−8.019	−1.071	0.073	52	3,001	0.68	37	1,802	1.96	0.79	.0639
Upper extremity
AIS 2+	−4.901	−0.317	0.064	598	90,353	15.99	190	22,259	20.71	0.72	.0348*
AIS 3+	−6.262	−0.627	0.068	212	21,361	4.49	79	7,650	8.09	0.73	.0164*
Lower extremity
AIS 2+	−4.163	−0.870	0.070	948	12,7180	24.62	459	4,9439	43.82	0.78	<.0001*
AIS 3+	−6.240	−1.495	0.100	399	27,776	10.31	258	21,938	33.90	0.84	<.0001*

* indicates statistical significance (p < 0.05)

Automatic crash notification (ACN) reduces response times and reduces fatalities by an estimated 1.5–20% (Bachman and Preziotti 2001; Clark and Cushing 2002; Lahausse et al. 2008; Pieske et al. 2002; Wu et al. 2013). An estimated 17–29% of fatalities are preventable with improved emergency medical services (EMS) and treatment (Rauscher et al. 2009). Advanced ACN (AACN) provides 9-1-1 telecommunicators, EMS, or hospitals with additional information on occupant injury severity using data transmitted from EDRs. EMS can use AACN data in conjunction with field triage guidelines to assess occupant injury risk and make more informed decisions on whether to transport occupants to a trauma center or non-trauma center.

Existing AACN systems include OnStar AACN and URGENCY. OnStar AACN uses multivariate logistic regression with the following variables to predict serious injury, defined as Injury Severity Score (ISS) 15+ for any occupant in the vehicle: crash direction, delta-V, multi-events, belt use, vehicle type, age, and sex (Kononen et al. 2011). Over 10 varieties of URGENCY exist with different variables used in multivariate logistic regression to predict Maximum Abbreviated Injury Scale (MAIS) 3+ injury risk for individual occupants (Augenstein et al. 2001, 2002, 2003, 2005; Bahouth et al. 2004, 2012; Champion et al. 2005; Malliaris et al. 1997; Rauscher et al. 2009). Many URGENCY variables are not obtainable from current EDRs (i.e., age, sex, height, ejection). Demographics (i.e., age, sex) improve the predictive power of OnStar AACN and URGENCY but are not required EDR data elements specified in regulation.

Tables I and II of the Part 563 regulation specify data elements required for all EDR-equipped vehicles and required under specified conditions (NHTSA 2006). Regulatory EDR data elements include longitudinal delta-V, multi-events, and driver and right front passenger belt status and frontal airbag deployment. Risk functions previously developed from EDR data use logistic regression to predict MAIS 2+ or 3+ injury (Gabauer and Gabler 2006, 2008; Stigson et al. 2012). This study's goal was to use NASS-CDS variables corresponding to required EDR data elements to compute occupant injury risk for (1) specific injuries and (2) specific body regions in frontal MVCs.

Materials and Methods

Injury risk curves were created using NASS-CDS 2000–2011, a representative, random sample of thousands of minor, serious, and fatal tow-away U.S. MVCs (NHTSA 2011). NASS-CDS includes nearly 1,000 variables detailing the crash, vehicle, occupant, and injuries coded with Abbreviated Injury Scale (AIS) 98 (Association for the Advancement of Automotive Medicine 2001). By applying NASS-CDS weights, a U.S. representative population can be analyzed. NASS-CDS 2009–2011 cases with greater than 10-year-old vehicles were excluded because 2009–2011 investigations did not collect injury information from these older vehicles.

Weighted NASS-CDS 2000–2011 data was used to predict risk for (1) a specific injury, or (2) injury to a specific body region of a given AIS severity (i.e., AIS 2+ head, AIS 3+ head) for specified frontal crash conditions. NASS-CDS variables (Table A1, see online supplement) were used to select frontal, single-impact crashes with deployed frontal airbags involving drivers and front right passengers ages 16 and older. Cases with a frontal general area of deformation and principal direction of force between 0–60° or 300–360° or any general area of deformation and principal direction of force of 0–30° or 330–360° were defined as frontal based on criteria from other studies and regulation (Bahouth et al. 2012; NHTSA Office of Regulatory Analysis & Evaluation Plans and Policy 2000). Occupants with missing data for longitudinal delta-V, sampling weight, or belt status were excluded. There were 10,930 unweighted (9,003 belted; 1,927 unbelted) and 3,851,978 weighted (3,378,714 belted; 473,264 unbelted) occupants included. Analyses focused on AIS 2+ injuries because AACN risk predictions that improve occupant triage can potentially benefit occupants sustaining these moderate to maximum severity injuries.

Specific Injury Risk

The specific injury analysis focused on the top 50% most frequently occurring AIS 2+ injuries in NASS-CDS 2000–2011, which includes 33 unique AIS 2+ injuries (Figure 1). Risk curves controlling for belt status and longitudinal delta-V were produced for 23 individual AIS 2+ injuries selected from the 33 injuries because they contained sufficient belted and unbelted sample sizes (Table A2, see online supplement). The 10 injuries excluded either did not converge in the risk model or were not informative due to inadequate injury variability.

Figure 1 Weighted injury count and cumulative percent for the top 50% of AIS 2+ injuries in NASS-CDS 2000–2011. On the x-axis, there are 33 injuries that make up the top 50% of AIS 2+ injuries in NASS-CDS 2000–2011 and these are ranked in descending order in terms of weighted frequency.

Body Region Injury Risk

Specific body region injury risk curves of a given AIS severity that controlled for belt status and longitudinal delta-V were produced when the sample sizes produced adequate injury variability and model convergence (Table 1). Risk curves were produced for the following body regions: head and thorax (AIS 2+, 3+, 4+, 5+), face (AIS 2+), abdomen, spine, upper extremity, and lower extremity (AIS 2+, 3+).

Injury Risk Model

Logistic regression models estimated risk for (1) specific injuries and (2) injury to specific body regions in relation to longitudinal delta-V using weighted NASS-CDS injury data and adjusting for belt status as a covariate. Each NASS-CDS weighting factor was normalized by dividing by the mean of all weighting factors, which did not change the relative values of the weights but normalized them to have a mean of one where the weighted and unweighted number of cases was equivalent. Each injury risk curve with 95% confidence intervals was calculated for longitudinal delta-Vs ranging from 15 to 105 km/h using Eq. (1)(1) . To account for random variation, 95% Wald confidence intervals were computed, representing the notion that if the study is repeated 100 times, the confidence interval would be expected to include true risk on 95 occasions. Logistic regression was performed in SAS 9.3 (SAS Institute Inc, Cary, NC). (1) with Belt = 1/0 (belted/unbelted); DV is the longitudinal delta-V; and α, β₁, β₂ are model parameters.

Results

Specific Injury Risk Curves

Table A2 includes the risk for each of the 23 specific injuries at a 56 km/h longitudinal delta-V analogous to the FMVSS 208 crash test speed. Risks are relatively low (0.4–9.9%) because they are estimated for very specific injuries.

Eighteen (78%) of the injuries have a higher risk for unbelted compared to belted occupants and belt status was a significant model predictor for 13 of the 18 injuries (Table A2). For these 18 injuries, unbelted risk was 7.8–412.5% higher and was on average 198.0% higher compared to belted risk. Unbelted occupants had a higher risk for subarachnoid hemorrhage, unconsciousness, rib fractures, lung contusions, spleen laceration, radius and ulnar fractures, knee sprain, pelvic fractures, and other lower extremity fractures. Cerebrum subarachnoid hemorrhage (AIS 3), unilateral lung contusion (AIS 3), and patella fracture (AIS 2) risk curves illustrate an elevated unbelted risk for 15–105 km/h longitudinal delta-Vs (Figure 2).

Figure 2 Injury risk curves with 95% confidence intervals (dashed lines) for select injuries: cerebrum subarachnoid hemorrhage, unilateral lung contusion, clavicle fracture, and patella fracture. Injury risk is plotted versus longitudinal delta-V (km/h) and stratified by belt status.

Belted risk was higher for sternum, rib, lumbar transverse process, carpus/metacarpus, and clavicle fractures. For these 5 injuries, belted risk was 8.5–122.3% higher and was on average 49.7% higher compared to unbelted risk. Belt status was also not a significant model predictor for these 5 injuries (Table A2). The clavicle fracture risk curve illustrates an elevated belted risk, though not significant, for 15–105 km/h longitudinal delta-Vs (Figure 2).

Risk was highest for the most frequently occurring injuries (Figure 3a). Belted and unbelted AIS 2 injury risks ranged from 0.4–4.6% and 0.6–7.4%, respectively. AIS 2 risk was highest for upper and lower extremity fractures for belted occupants and head injuries and lower extremity fractures for unbelted occupants. For both belted and unbelted occupants, AIS 2 head injuries with less than one hour or unspecified prior unconsciousness, knee sprains, and carpus/metacarpus, clavicle, patella, fibula, and metatarsal/tarsal fractures occur at high frequency and carry the highest risks (1.4–7.4%). Belted and unbelted AIS 3 risks ranged from 0.6 to 3.2% and 1.2 to 9.9%, respectively. AIS 3 lung contusions and radius, ulna, and femur fractures occur at high frequency and carry the highest risks (1.2–9.9%). Focusing on belted occupants, injuries with risks ranked in the upper third include 4 upper extremity fractures, 2 lower extremity fractures, and a knee sprain (2.4–4.6% risk). For unbelted occupants, injuries with risks ranked in the upper third include 4 lower extremity fractures, 2 head injuries, and a lung contusion (4.6–9.9% risk).

Figure 3 Risk at 56 km/h longitudinal delta-V for (a) specific injuries and (b) specific body regions plotted versus the weighted injury count in NASS-CDS and stratified by belt status and AIS severity.

Body Region Injury Risk Curves

Risk at 56 km/h longitudinal delta-V ranged from 0.6 to 43.8% for AIS 2+, 0.7 to 33.9% for AIS 3+, 1.0 to 7.8% for AIS 4+, and 0.7 to 4.2% for AIS 5+ body region injuries (Table 1). Within each body region, injury risk decreased as AIS severity increased.

AIS 2+ curves with 95% confidence intervals for the head, face, thorax, abdomen, spine, upper extremity, and lower extremity are illustrated as a function of longitudinal delta-V and belt status (Figure 4). Belt status was a significant predictor in all body region risk models except for the AIS 2+ and 3+ spine curves (Table 1). At a 56 km/h longitudinal delta-V, unbelted risk is higher than belted risk for all body region injuries with the exception of AIS 2+ spine injury (Table 1). This is consistent with the lumbar transverse process fracture curve, which had a higher risk for belted compared to unbelted occupants (Table A2). Unbelted risk across all body region risk curves was on average 3.0 times higher than belted risk.

Figure 4 AIS 2+ injury risk curves with 95% confidence intervals (dashed lines) for the head, face, thorax, abdomen, spine, upper extremity, and lower extremity. Injury risk is plotted versus longitudinal delta-V (km/h) and stratified by belt status.

Risk for injury of a given AIS severity varied depending on body region, delta-V, and belt status. Body region injuries that occurred at higher frequency also carried higher risks at 56 km/h (Figure 3b). The 6 curves with the highest risks were the AIS 2+ lower extremity, upper extremity, thorax, and head injuries and AIS 3+ lower extremity and thorax injuries (15.9–43.8% risk). For unbelted occupants, the lower extremity had the highest risk of AIS 2+ injury (43.8%), followed by the head (25.4%), upper extremity (20.7%), thorax (17.9%), abdomen (13.6%), spine (4.3%), and face (3.5%). For belted occupants, the lower extremity had the highest risk of AIS 2+ injury (24.6%), followed by the upper extremity (16.0%), thorax (12.4%), head (7.6%), spine (4.6%), abdomen (3.4%), and face (0.6%). AIS 3+ injury in both unbelted and belted occupants carried the highest risk for the lower extremity, followed by the thorax, upper extremity, head, abdomen, and spine. The thorax had a higher risk compared to the head for AIS 4+ and 5+ injury in both belted and unbelted occupants.

Discussion

Logistic regression of weighted NASS-CDS data was used to estimate injury risk for front seat adult occupants with frontal airbag deployment that were involved in frontal, single-impact MVCs. A major strength of NASS-CDS is that it is a population sample for deriving U.S.-representative MVC injury risks. Longitudinal delta-V and belt status were used in logistic regression models to produce 23 risk curves for specific injuries and 17 risk curves for body region injuries of a given AIS severity. Longitudinal delta-V, belt status, as well as airbag deployment and multi-impact events are data elements recorded by all EDRs meeting current regulatory requirements. Risk estimation for specific injuries and specific body region injuries from required EDR data elements has important implications for AACN. AACN provides earlier crash notification and information on estimated occupant injury risk, which has the potential to improve trauma triage decisions and reduce the time from injury to definitive treatment. The OnStar and URGENCY AACN algorithms report overall injury risk using MAIS 3+ or ISS 15+. Risk curves developed in this study are supplemental and more specific injury predictions that could be included in AACN algorithms that also estimate overall occupant injury risk. These risk curves, when implemented into an AACN algorithm, can provide EMS and other patient care providers with additional information to improve detection and subsequent treatment of injuries that are highly severe, time sensitive, or occult (Schoell et al. 2013, 2015a, 2015b; Weaver et al. 2013).

Similar trends are observed when comparing this study's results against head, thorax, spine, and knee–thigh–hip (KTH) injury risk functions from other studies (Prasad et al. 2010). Prasad et al. (2010) estimated injury risk to specific body regions at a frontal crash test speed of 56 km/h for NASS-CDS belted frontal MVCs and found that AIS 2+ KTH risk was greatest (14.0%), followed by AIS 3+ thorax risk (10.6%), AIS 3+ head/face risk (3.2%), and AIS 3+ neck/spine risk (0.7%). Although differences exist in the NASS-CDS sample analyzed and the risk curve methodology, these risks follow the same ordering by body region as belted risks from the current study for AIS 2+ lower extremity (24.6%), AIS 3+ thorax (7.4%), AIS 3+ head (1.8%), and AIS 3+ spine (0.7%) injury. AIS 2+ lower extremity risk varies the most (14.0% versus 24.6%) and may be higher because AIS 2+ non-KTH injuries including foot and ankle fractures are included in this study. AIS 3+ thorax, head, and spine risks are remarkably similar and fall within the range of NASS-CDS risks reported by other studies for 56 km/h frontal crashes (Laituri et al. 2009; Prasad et al. 2010).

A multitude of injury risk curves have been produced from experimental testing, but direct comparison is difficult because risk is often reported in relation to accelerations, forces, or deflections measured on postmortem human subjects or anthropomorphic test devices. Even when comparing this study's results to other NASS-derived risks, there may be differences in the sample of cases analyzed, statistical models, and variables included when calculating risk. Although a variety of statistical methods exist, logistic regression is commonly used in injury biomechanics and is a parametric model that allows for better representation of injury risk at the low and high end regions of the risk curve where less actual injury data exists in the sample (Hertz 1993; Hosmer et al. 2011, 2013; Mertz et al. 1996; Mertz and Weber 1982; Nuscholtz and Mosier 1999; Petitjean et al. 2009; Prasad et al. 2010).

As expected, belted occupants generally had reduced risks for body region injury and for specific injuries. At a 56 km/h longitudinal delta-V, unbelted risk for facial injury was 5.9 times that of belted occupants. This is likely due to the seat belt limiting forward excursion of the occupant and contact of the face with vehicle interior components. Likewise, the average unbelted risk across all AIS severities of abdomen, head, thorax, and lower extremity injury was 3.6, 3.5, 3.0, and 2.5 times that of belted occupants, respectively. These results illustrate the relative effect of seat belts on mitigation of injury to particular body regions for crash conditions similar to FMVSS 208.

Individual injury risk was relatively low compared to body region risks, which is to be expected because the prediction is for a specific injury. Belt use significantly contributed to a reduction in risk for 13 of the 23 specific injuries yet was a nonsignificant predictor for rib, sternum, lumbar spine transverse process, carpus/metacarpus, and clavicle fractures associated with higher risk for belted occupants. Sternum and clavicle fracture incidence is known to be increased for belted compared to unbelted occupants due to belt loading (Hill et al. 1994; Kemper et al. 2009; Porter and Zhao 1998). Thoracolumbar spine fracture incidence has also increased over time despite increasing rates of belt use (Doud et al. 2015), and belt use has been linked to an increased odds of lumbar compression fractures (Kaufman et al. 2013).

Limitations

Due to sample sizes, risk curves were only produced for select injuries and select body regions. Inadequate sample sizes limited the creation of risk curves for the neck or for higher severity injuries in some body regions. Though a larger number of injuries included in the risk curves were AIS 2–3 severity, risk curves were produced for higher severity AIS 4 injuries when sample sizes were adequate. Body region analyses are likely more robust because individual injuries are aggregated. Higher severity (AIS 4+, 5+) body region risk curves were produced for the head and thorax. Because the face, spine, upper extremity, and lower extremity rarely sustain AIS 4+ injuries, sample size was a limiting factor only for the abdomen AIS 4+ and 5+ risk curves. Exclusion of NASS-CDS 2009–2011 cases with missing data presents a limitation but allows for inclusion of newer NASS-CDS cases. NASS-CDS longitudinal delta-Vs are estimated from crash reconstruction measurements input into WinSMASH software. Frontal crash studies have shown that WinSMASH-obtained delta-Vs are underestimated by 2–27% compared to delta-Vs measured with EDR or vehicle accelerometers (Hampton and Gabler 2010; Niehoff and Gabler 2006; Sharma et al. 2007). Delta-Vs are also 6–7% lower when measured with EDR compared to vehicle accelerometers (Niehoff et al. 2005; Tsoi et al. 2013). Furthermore, driver belt status in NASS-CDS and EDR reports disagrees 21–30% of the time (daSilva 2008; Gabler et al. 2003; Gabler, Hampton, and Hinch et al. 2004). This study used NASS-CDS to extract crash characteristics including delta-V, belt status, and airbag deployment without adjusting for potential inaccuracies, which is a limitation. Inaccuracies may need to be adjusted for when using EDR-collected data to calculate risk. Additionally, EDR capabilities such as duration of precrash, crash, and postcrash recordings, number of events recorded, and the definition of time zero can vary widely between manufacturers and model years, which may present a challenge to implementing these risk curves into different AACN systems (Gabler, Gabauer, et al. 2004).

The models could be improved by including additional crash or occupant variables that influence risk. Crash variables such as peak acceleration and compartment intrusion can influence risk, but delta-V was selected because it is (1) widely used as indicator of crash severity, (2) correlated to occupant injury, and (3) a required EDR data element (Gabauer and Gabler 2008; NHTSA 2006). Occupant variables (i.e., age, sex) that improve the predictive power of risk models were not controlled for. However, the study's focus was to predict injury risk using only required EDR data elements and future studies may develop models with optional EDR data elements or additional crash or occupant variables not measured with EDR.

Future Work and Applications

Future work involves incorporating these risk curves into an AACN algorithm that also estimates overall risk of serious occupant injury based on criteria such as MAIS 3+ (Bahouth et al. 2004); MAIS 3+ through fatal injury, termed “MAIS 3+F” (Digges and Eigen 2006); or ISS > 15 (American College of Surgeons Committee on Trauma 2006). Because airbag deployment is a common AACN trigger, these risk curves developed for frontal crashes with deployed frontal airbags are particularly relevant for AACN. Our methodology could be extended to create risk curves for side impacts, rear impacts, or frontal impacts with nondeployed airbags. When implemented into an AACN algorithm, these risk curves will provide EMS and patient care providers with information on suspected occupant injuries. This additional information could lead to improved detection of specific injuries, better triage decisions on where to transport an occupant (trauma center versus nontrauma center), and a reduction in the time that elapses before an occupant receives definitive treatment.

Logistic regression of NASS-CDS single-impact frontal MVCs involving front seat occupants with frontal airbag deployment was used to produce 23 risk curves for specific injuries and 17 risk curves for AIS 2+ to 5+ body region injuries. Risk was estimated for 15–105 km/h longitudinal delta-Vs while adjusting for belt status. Overall, belted occupants had lower estimated risks compared to unbelted occupants and risks increased with longitudinal delta-V. Belt status was a significant predictor for 13 specific injuries and all body region injuries with the exception of AIS 2+ and 3+ spine injuries. Specific injuries and body region injuries that occurred more frequently in NASS-CDS also tended to carry higher risks at a 56 km/h longitudinal delta-V. Belted risks ranked in the top 33% included 4 upper extremity fractures, 2 lower extremity fractures, and a knee sprain (2.4–4.6% risk). Unbelted risks ranked in the top 33% included 4 lower extremity fractures, 2 head injuries with less than one hour or unspecified prior unconsciousness, and a lung contusion (4.6–9.9% risk). The 6 body region curves with the highest risks were for AIS 2+ lower extremity, upper extremity, thorax, and head injury and AIS 3+ lower extremity and thorax injury (15.9–43.8% risk). These risk curves, which use EDR data elements required by regulation, can be readily integrated into AACN systems to provide EMS and other patient care providers with information on suspected occupant injuries to improve injury detection and patient triage.

Acknowledgment

The authors thank Nathan Croswell for his assistance with literature review and preparation of this article.

Funding

Funding has been provided by Toyota Motor Corporation and Toyota's Collaborative Safety Research Center. Views expressed are those of the authors and do not represent the views of any of the sponsors.

Supplemental Materials

Supplemental data for this article can be accessed on the publisher's website