Belted driver fatalities: Time of death and risk by injury severity

ABSTRACT

Purpose: This is a descriptive study of the fatality risk by injury severity and time of death for lap–shoulder-belted drivers without ejection in modern vehicles. It also determined the body region for severe injuries experienced by belted drivers using the most recent federal crash data.

Methods: 1997–2015 NASS-CDS data were evaluated for fatally injured lap–shoulder-belted drivers without ejection in light vehicles of 1997+ model year (MY). The severity of injuries sustained by belted drivers was assessed by the Maximum Abbreviated Injury Scale (MAIS) and individual injuries by Abbreviated Injury Scale (AIS) and body region. The change in fatality risk with MAIS was fit with a Logist function. Time of death was determined using the variable DEATH, which is reported hourly in unequal intervals up to 24 h and then daily up to 30 days after the crash. The fraction (f) and cumulative fraction (F) of the deaths are reported for each time period up to 30 days. A power or logarithmic curve was fit to the data using the trendline functions in Excel.

Results: The NASS-CDS sample included 20,610,000 belted drivers with 37,974 fatalities from 1997 to 2015. The fraction of driver deaths increased with maximum injury severity (MAIS). For example, 17.4% of drivers died within 30 days with MAIS 4 injury. Virtually all drivers (99.7%) died with MAIS 6 injury. The change in fatality risk with injury severity was r = [1 + exp^{(10.159 − 2.088MAIS)}]⁻¹, R² = 0.950. Overall, there were 19,772 driver deaths with MAIS 4–6 injury and 13,059 with MAIS 0–3 injury. In addition, 44.7% of driver deaths occurred within 1.5 h of the crash, 56.7% within 2.5 h, and 64.6% within 4.5 h after the crash. The cumulative fraction of the deaths (F) up to 30 days was fit with a logarithmic function. It was F = 0.0739ln(t) + 0.5302, R² = 0.976, for deaths after 3.5 h. There were 19,772 driver deaths with 52,130 AIS 4+ injuries. On average, the driver experienced 2.64 AIS 4+ injuries most commonly to the head (44.5%) and thorax (38.1%).

Conclusions: The risk for belted driver deaths exponentially increased with MAIS. A majority of deaths occurred within 2.5 h of the crash. On average, fatally injured drivers experienced 2.64 AIS 4+ injuries, primarily to the head and thorax.

KEYWORDS:

• Car crashes
• fatalities
• time of death
• fatal injuries
• probability of death

Introduction

There have been studies on the fatality risk by injury severity in motor vehicle crashes. Malliaris et al. (1982) were one of the first to report on fatality risk by injury severity in terms of the Maximum Abbreviated Injury Scale (MAIS). They used 1979–1980 NASS-CDS data on all occupants irrespective of belt use and ejection status in their hallmark study of the HARM from motor vehicle crashes. Human HARM from motor vehicle crashes was defined as a cost-weighted sum of the number of injuries sustained by the cost to society for each injury by The Abbreviated Injury Scale (AIS). Later, the Association for the Advancement of Automotive Medicine (AAAM 1990) reported on fatality risks in their update to the AIS. This sample included all occupants irrespective of belt use or ejection status. AAAM (2005), in a further update of the AIS, reported on trauma center data from the National Trauma Data Bank. They evaluated 181,707 patients and determined the fatality risk by injury severity for those with a single injury.

There have been studies on the time of death after traumatic injury. Meislin et al. (1997) studied 710 traumatic deaths and found that approximately half (52%) were pronounced dead at the scene. Neurologic dysfunction was the most common cause of death. Two distinct peaks of time were found with 23% dying within 60 min and 35% dying 24–48 h after injury.

Bansal et al. (2009) determined cause of death by autopsy in hourly intervals for those who died within 24 h. Of 167 deaths, 73 (43.7%) occurred within the first hour. Brain injury was the most common cause of death in all hourly intervals, but hemorrhage was as or more important than brain injury as the cause of death during the first 3–6 h. No deaths were attributable to hemorrhage after 12 h. Kauvar et al. (2006) found that hemorrhage was responsible for 30–40% of the deaths, 33–56% of which occurred during the prehospital period. Among those who reached care, early mortality was caused by continued hemorrhage, coagulopathy, and incomplete resuscitation.

Gunst et al. (2010) found that trauma deaths have a predominantly bimodal distribution. Near elimination of the late peak likely represented advancements in resuscitation and critical care that have reduced organ failure. Ghorbani et al. (2014) examined the proportion of trauma patients dying within 30 days and found that 10.5% of the late deaths were not directly related to injury. There have been many improvements in emergency medical services (EMS), advanced trauma life support (ATLS), and trauma centers as well as advances in the occupant restraints and crashworthiness of motor vehicles since the earlier research.

This study reports on the fatality risk by injury severity and time of death for lap–shoulder-belted drivers in motor vehicle crashes. The most recent NASS-CDS data from 1997 to 2015 were used with relatively modern vehicles (1997+ model year [MY]). The data provide an understanding of the current fatality risk by MAIS. There are no earlier studies with lap–shoulder-belted occupants to compare to, so this study is the first reporting on belted occupants who remain in the vehicle. It provides information on the current trend. The study also determined the body regions severely injured in belted driver deaths.

Methods

NASS-CDS

NASS-CDS (NHTSA 2017) is a stratified multiphase, unequal selection probability sample of motor vehicle crashes that are prospectively selected for in-depth investigation. Most of the vehicles were towed from the scene because of damage. The data include information based on crash investigation teams that gather information from the crash site, vehicle, medical records, police accident reports, and personal interviews. NASS-CDS data for calendar years 1997–2015 were evaluated for light vehicles with 1997+ MY. The data for calendar years 2009–2015 are representative of model year vehicles up to 9 years younger than the calendar year of the crash because older vehicles are not part of a detailed investigation and do not include belt use and ejection information.

Drivers (SEATPOS = 11)

Drivers who were lap–shoulder belted (MANUSE = 4), aged 15 years old or older (14 < AGE < 105), and nonejected (EJECTION = 0) were included in the study.

Injury severity

Injury severity was assessed using the MAIS and the “TREATMNT” and “INJSEV” variables. MAIS represents the assessment of the injury at the time of first medical evaluation and not long-term consequences. It ranges from MAIS 0 to 9, where MAIS 9 is an injury with unknown severity. Fatality was also used to determine whether the occupant died within 30 days of the crash. Fatality (F) was defined by the following:

•	TREATMNT = 1, which means that the occupant was fatally injured and not transported to the hospital.
•	Police injury severity: INJSEV = 4 represents a fatality from police rating.

Exposed occupants were defined as those with known MAIS (MAIS 0–6) or with a fatality. The shorthand notation is MAIS 0+F. Severely injured occupants were defined as those with MAIS 4–6 or fatality, because fatalities can occur at any MAIS level. The shorthand notation is MAIS 4+F. The risk for death by injury severity (MAIS) was fit with a Logist function, excluding the data at MAIS = 0. This followed the work of Huang and Marsh (1978) based on the AIS relationship to fatality risk by injury severity (Gennarelli and Wodzin 2006; Petrucelli et al. 1981; States 1969).

Injury by body region and severity

Injury was assessed using the AIS and body region (region90). Injuries were classified using AIS 98 coding (AAAM 1998). The number of severe-to-maximum (AIS 4+) injuries was determined for fatally injured drivers. The rate of injury was determined by dividing the number of AIS 4+ injuries by the number of fatally injured drivers with MAIS 4–6. The fatality risk was also determined by Glasgow Coma Scale (GCS) groupings of 3–8, 9–12, and 13–15.

Time of death

Time of death was assessed using the variable DEATH, which varied from 1 to 99 and represents a time period. Table A1 (see online supplement) provides the coding information for the variable DEATH. The information was taken from the 1997 NASS-CDS coding manual. The source for the time of death is a police report, hospital or medical record, autopsy report, or other official record for actual time of death for a fatally injured driver.

•	DEATH = 1–24 represents the number of hours after the crash. For example, DEATH = 1 represents <1.5 h after the crash. DEATH = 2 means it occurred between 1.5 and 2.5 h after the crash. The interval between the DEATH codes is not equal. In the analysis, the time of death is reported as the upper time limit.
•	DEATH = 31–60 represents the number of days after the crash minus 30. For example, DEATH = 32 is a death occurring between 1.5 and 2.5 days after the crash. The maximum number of days for a fatality was 30 (DEATH = 60). In the analysis, the number of days was converted to hours by subtracting 30 from the DEATH number, multiplying by 24 h, and adding 12 h to report deaths to the nearest day. For example, DEATH = 32 was equal to 60 h.
•	DEATH 96 represents a non-crash-related fatality when a medical examiner (or other official vested by the state to verify the cause of death) or an official medical report verifies that the death resulted from either a disease condition or non-crash-related injuries.
•	DEATH 99 represents unknown time of death when the length of time between the crash and the person was pronounced dead by a qualifying person (coroner, state medical examiner, etc.) is unknown.

Weighted data

National^{Table 1} estimates for the number of driver deaths were made using the ratio weight (RATWGT) variable in NASS-CDS. All calculations were based on weighted values. Cases with a RATWGT equal to 0 or with a negative RATWGT were excluded from the analysis.

Analysis

The fraction (f) of deaths and the cumulative fraction (F) of deaths are reported for each time period up to 30 days. Because the intervals between the coded times of death are not equal, the data for each interval are reported on a per hour basis and are identified at the maximum time for that interval. For example, the number of drivers who died within 1.5 h of the crash was normalized by 1.5 and is identified at 1.5 h after the crash. A power or logarithmic function was used to fit the fraction and cumulative fraction of deaths using the trendline functions in Excel.

Results

Table 1 shows the maximum injury severity for lap–shoulder-belted drivers who died or survived in the NASS-CDS sample. The sample included 20,610,000 belted drivers with 37,974 fatalities in 1997–2015 NASS-CDS. A fatality can occur at any injury severity. For example, 17.4% of drivers with MAIS 4 injury died within 30 days of the crash, whereas virtually all drivers (99.7%) with MAIS 6 injury died. There were 19,772 driver deaths with MAIS 4–6 injury and 13,059 with MAIS 0–3. For completeness, Table A2 (see online supplement) provides the same information on unbelted drivers without ejection. For example, the risk of death with MAIS 4 injury was 9.7%.

Table 1. Lap–shoulder-belted drivers, 15+ years old and not ejected in 1997+ MY vehicle crashes (NASS-CDS 1997–2015 weighted data).

			Fatal			% Fatal
Injury severity	MAIS	Not fatal	#	SE	Total	#	SE
No injury	0	11,860,000	1,569	1,153	11,860,000	0.013%	0.0097%
Minor injury	1	6,971,155	1,718	154	6,972,873	0.025%	0.0022%
Moderate injury	2	803,509	3,207	456	806,716	0.398%	0.057%
Serious injury	3	212,779	6,565	1,087	219,344	2.99%	0.50%
Severe injury	4	33,660	7,107	1,392	40,768	17.4%	3.4%
Critical injury	5	11,800	7,730	1,853	19,530	39.6%	9.5%
Maximum injury	6	15	4,935	809	4,950	99.7%	16.3%
Injured, unknown severity		449,193	5,101	2,135	454,294
Unknown if injured		227,385	42	38	227,427
Total		20,570,000	37,974	4,873	20,610,000	0.184%	0.024%
MAIS 0+F		19,892,918	32,831	4,873	19,924,180	0.165%	0.024%
MAIS 4–6		45,475	19,772	3,638	65,247	30.3%	5.58%
MAIS 4+F		45,475	37,974	3,489	83,449	45.5%	5.35%

Figure 1 shows the risk of death by injury severity (MAIS) for lap–shoulder-belted drivers. The risk is plotted with a logarithmic scale. MAIS is an ordinal scale. A Logist function nicely fit the change in risk with increasing MAIS. The fatality risk was r = [1 + exp^{(10.159 − 2.088MAIS)}]⁻¹, R² = 0.950. The actual fatality risk was greater than the function for MAIS 2–4 and 6 but was lower for MAIS 1 and 5 injuries.

Figure 1. Risk of death by injury severity with one standard error and Logist fit.

Figure 2 shows the fraction of driver deaths per hour up to 30 days after the crash. The time is in hours and is plotted on a logarithmic scale, as is the fraction of deaths per hour. The first data point on time of death is 1.5 h after the crash. It accounted for 44.7% of all driver deaths within 30 days. This is a fraction of 29.8%/h. The fraction decreased with time after the crash and is plotted using a logarithmic scale. The early trend in the data up to 12 h was fit with a power function. The fraction (f) was f = 0.756t^−2.138, R² = 0.830, where f is the fraction of deaths per hour and t is time in hours with t < 12 h, based on 14,762 deaths. The data after 12 h were fit with a power function with a declining rate with increased time of 12 h after the crash based on 5,472 deaths. The fraction (f) was f = 0.0891t^−1.083, R² = 0.730, with t > 12 h. Complete data by time of death code and calendar year of the NASS-CDS data collection are provided in Table A3 (see online supplement). Figure A1 (see online supplement) shows the fraction of deaths to belted drivers per hour up to 12 h after the crash.

Figure 2. Fraction of nonejected belted driver deaths per hour up to 30 days after the crash (NASS-CDS 1997–2015 weighted data).

Figure 3 shows the^{Table 2, Table 3} cumulative fraction of deaths to drivers up to 30 days (720 h) after the crash. The trend was fit with a logarithm function that closely follows the data after the first 2 data points at 1.5 and 2.5 h after the crash. The percentage of deaths within 4.5 h and 73.0% within 12 h after the crash was 64.6%. The cumulative fraction of deaths was F = 0.0739ln(t) + 0.5302, R² = 0.976, t > 3.5 h. Figure A2 (see online supplement) shows the cumulative fraction of deaths to drivers up to 12 h after the crash. The trend was fit with a logarithm function. The cumulative fraction of deaths was F = 0.1299ln(t) + 0.4334, R² = 0.946, t < 12 h.

Figure 3. Cumulative fraction of nonejected belted driver deaths up to 30 days after the crash (NASS-CDS 1997–2015 weighted data).

Table 2 shows the AIS 4+ injuries experienced by fatally injured drivers with MAIS 4–6 injury. There were 19,772 driver deaths with 52,130 AIS 4+ injuries. This is 2.64 AIS 4+ injuries per belted driver death, on average. Severe injury was most common to the head (44.5%) and thorax (38.1%). Table A4 (see online supplement) shows the risk of death by GCS for lap–shoulder-belted drivers 15+ years old and not ejected. Drivers with GCS 3–8 had a fatality risk of 43.5%. Those with GCS 9–12 had a 3.3% risk and those with GCS 13–15 had a risk of 0.093%.

Table 2. Severe-to-maximum (AIS 4+) injury in driver fatalities with lap–shoulder belts, 15+ years old, and not ejected in 1997+MY vehicles (NASS-CDS 1997–2015 weighted data).

		AIS 4+ injury by body region
Injury severity	MAIS	Head	Neck	Thorax	Abdomen	Spine	Lower extremity	Unspecified	Total
MAIS 0+F									19,924,180
Severe injury	4	4,383	0	4,894	864	318	240	0	10,700
Critical injury	5	12,863	0	7,282	1,569	800	263	54	22,831
Maximum injury	6	5,956	27	7,661	2,278	1,813	42	823	18,599
Total		23,203	27	19,837	4,712	2,931	545	877	52,130
Frequency
Severe injury	4	8.4%	0.0%	9.4%	1.7%	0.6%	0.5%	0.0%	20.5%
Critical injury	5	24.7%	0.0%	14.0%	3.0%	1.5%	0.5%	0.1%	43.8%
Maximum injury	6	11.4%	0.1%	14.7%	4.4%	3.5%	0.1%	1.6%	35.7%
Total		44.5%	0.1%	38.1%	9.0%	5.6%	1.0%	1.7%	100%
Rate/MAIS 4–6 fatals
Severe injury	4	0.22	0.000	0.25	0.04	0.02	0.01	0.00	0.54
Critical injury	5	0.65	0.000	0.37	0.08	0.04	0.01	0.00	1.15
Maximum injury	6	0.30	0.001	0.39	0.12	0.09	0.00	0.04	0.94
Total		1.17	0.001	1.00	0.24	0.15	0.03	0.04	2.64
MAIS 4–6 (all)									65,247
MAIS 4–6 (fatal)									19,772

There were 17 cases with driver death and MAIS = 0. The NASS-CDS electronic cases were reviewed. Eight of the cases involved the medical examiner ruling death by disease, 4 involved the police report listing death with no injury, one involved asphyxiation with roof crush from a 9.2-m (30-ft.) fall, and the remainder lacked information because the vehicle was older than 9 years in an NASS-CDS case investigated after 2009 or other reasons.

Discussion

Table 3 summarizes studies on fatality risk by injury severity (MAIS). Malliaris et al. (1982) used 1979–1980 NASS-CDS data on all occupants irrespective of belt use to determine the fatality risk by MAIS. They also included pedestrians and cyclists. AAAM (1990), cited in Nordhoff (2005), reported on fatality risk by MAIS in an update to the AIS scale. AAAM (2005), in a further update of the AIS scale, reported on trauma center data on fatality risk by injury severity for those with a single injury. These data were also reported by Gennarelli and Wodzin (2006). They concluded that the “AIS severity score performs well as a measure of mortality, but mortality is not the sole determinant of AIS severity.” AIS is the basis for reporting injury in NASS-CDS. It is an anatomically based, consensus-derived severity scoring system that classifies each injury by body region according to its severity using an ordinal 6-level scale. Due to different populations in the samples, the earlier studies cannot be directly compared to the current data on belted drivers.

Table 3. Summary of studies on fatality risk by injury severity (Note: There are different populations in the samples reported).

		Risk for death
					Current study of drivers
					Lap-shoulder belted		Unbelted
MAIS	Injury severity	Malliaris et al. (1982)	AAAM (1990)	AAAM (2005)	Risk	SE	Risk	SE
0	None				0.013%	0.0097%	0.004%	0.0025%
1	Mild	0.0%	0.0%	0.70%	0.025%	0.0022%	0.167%	0.0343%
2	Moderate	0.1%	0.1%–0.4%	0.80%	0.398%	0.057%	0.740%	0.188%
3	Serious	0.8%	0.8–2.3%	3.5%	2.99%	0.50%	5.53%	1.57%
4	Severe	7.9%	10%	14.6%	17.4%	3.4%	9.7%	1.8%
5	Critical	58.4%	50%	39.6%	39.6%	9.5%	44.0%	11.6%
6	Maximum	100%	Virtually unsurvivable	78.8%	99.7%	16.3%	98.1%	20.2%

The fatality data from the current study involve NASS-CDS cases that are 19–37 years newer than the original study by Malliaris et al. (1982). To our knowledge, it is the first that focuses on lap–shoulder-belted drivers in relatively modern vehicles with 1997+ MY. Though AAAM (2005) fit their National Trauma Data Bank data with a second-order polynomial, a Logist function nicely fits the results. The Logist function is preferred because it has very low risk at low MAIS and has an S shape, which tails off to 100% risk at the highest MAIS. The Logist function for the AAAM (2005) data is r = [1 + exp^{(7.374 − 1.378MAIS)}]⁻¹, R² = 0.999. Overall, the most recent data have similar fatality risks for MAIS 3–5 and a high risk with MAIS 6 compared to the AAAM (2005) data.

There have been many improvements in the capabilities of EMS and ATLS that have saved lives after severe injury because of rapid medical attention. Mears et al. (2012) reviewed EMS in the United States and provided guidelines for improved evaluation and management of patients with acute injury outside the hospital. van Olden et al. (2004) found that ATLS improved outcomes in the first hour after admission. Kerr et al. (1999) studied differences in mortality rates among trauma patients transported by helicopter and ambulance and found that rapid air transport had a positive effect on the outcome of severely injured patients.

The time of arrival of EMS to the severely injured is different in urban and rural settings and makes a difference for survival. Grossman et al. (1997) compared differences in response times, scene times, and transport times by advanced life support–trained paramedics to trauma incidents in urban and rural locations. They found that the mean response time for urban locations was 7.0 min compared to 13.6 min for rural locations (P <.0001). The mean transport times from the scene to the hospital were also significantly longer for rural incidents (17.2 vs. 8.2 min, P <.0001). Rural victims were over 7 times more likely to die before arrival (relative risk = 7.4, 95% confidence interval [CI], 2.4–22.8) if EMS response time was >30 min.

Gonzalez et al. (2006) studied fatality rates in rural and urban vehicular fatalities. The mean EMS response time for urban survivors was 11.2 min versus 13.9 min for rural survivors (P <.0002). In a rural setting, distance to the scene when patients were alive was 7.7 miles versus 10.5 miles when patients were dead (P <.001). They concluded that increased EMS response time in rural crashes, time on scene, and distance to the scene are associated with higher fatality rates. These studies followed earlier work of Maio et al. (1992), who conducted a case–control study of the risk of drivers dying in rural and urban areas. The relative risk for rural driver deaths was 1.96 greater than their nonrural counterparts.

Fatovich et al. (2011) found that mortality from trauma in rural areas increased compared to the urban environment. After adjusting for age, injury severity, and the effect of time with the initial rural deaths, there was a significantly increased risk of death (odds ratio = 2.60, 95% CI, 1.05–6.53, P =.039) in the rural group. For those rural patients, the adjusted odds ratio for death was 1.10 (95% CI, 0.66–1.84, P =.708). There is more than double the risk of major trauma death in rural areas.

Automatic crash notification is available in modern vehicles and offers another means of shortening the time to provide definitive care for those severely injured (Lerner et al. 2003). The data reported in this study involved relatively modern vehicles that were fit with advanced safety systems, including airbags, vehicle structures, and occupant restraint systems. The fatality rates occurring in modern vehicles often reflect many factors, such as the severity of motor vehicle crashes, lower tolerance of older occupants, and heavy truck involvement, among other factors.

NHTSA recently addressed the key factors in frontal crash deaths in modern vehicles. Bean et al. (2009) studied 122 fatal frontal crashes and identified extremely severe impacts, corner and oblique deformation of the vehicle, underriding heavy vehicles, and vulnerable occupants as key factors. Obviously, lowering of crash speeds, continuing development of occupant restraint systems and vehicle crashworthiness, and improving EMS and trauma center delivery of care are all part of a strategy to further lower fatality rates in motor vehicle crashes.

Limitations

The level of documentation in NASS-CDS cases varies, and an individual can sustain multiple injuries at a given AIS level and a range of injury severities. For those with fatal injuries, the level of information on injuries depends on the hospital stay and details of an autopsy. These variations should be considered in viewing the AIS 4+ injuries reported in Table 2. For MAIS = 0 and death, there are limitations based on a lack of more details in the NASS-CDS report. In addition, we wanted to make a comparison of time of death for rural and urban crashes, but NASS-CDS does not have a code for this variable. It would be helpful if future NASS-CDS coding included this variable.