Estimate of mortality reduction with implementation of advanced automatic collision notification

ABSTRACT

Objective: Advanced Automatic Collision Notification (AACN) is a system on a motor vehicle that notifies a public safety answering point (PSAP), either directly or through a third party, that the vehicle has had a crash. AACN systems enable earlier notification of a motor vehicle crash and provide an injury prediction that can help dispatchers and first responders make better decisions about how and where to transport the patient, thus getting the patient to definitive care sooner. The purposes of the current research are to identify the target population that could benefit from AACN, and to develop a reasonable estimate range of potential lives saved with implementation of AACN within the vehicle fleet.

Methods: Data from the Fatality Analysis Reporting System (FARS) years 2009–2015 and National Automotive Sampling System–Crashworthiness Data System (NASS-CDS) years 2000–2015 were obtained. FARS data were used to determine absolute estimates of the target population who may receive benefit from AACN. These estimates accounted for a number of factors, such as whether a fatal occupant had nearby access to a trauma center and also was correctly identified by the injury severity prediction algorithm as having a “high probability of severe injury.” NASS-CDS data were used to provide relative comparisons among subsets of the population. Specifically, relative survival rate ratios between occupants treated at trauma centers versus at non-trauma centers were determined using the nonparametric Kaplan–Meier estimator. Finally, the fatality reduction rate associated with trauma center care was combined with the previously published fatality reduction rate for faster notification time to develop a range for possible lives saved.

Results: Two relevant target populations were identified. A larger subset of 6893 fatalities can benefit only from earlier notification associated with AACN. A smaller subgroup of between 1495 and 2330 fatalities can benefit from both earlier notification and change in treatment destination (i.e., non-trauma center to trauma center). A Kaplan–Meier life curve and a multiple proportional hazard model were used to predict the benefits associated with transport to a trauma center. The resulting range for potential lives saved annually was 360 to 721.

Conclusions: This analysis provides the estimates of lives that could potentially be saved with full implementation of AACN and universal cell coverage availability. This represents a fatality reduction of approximately 1.6% to 3.3% per year, and more than double the lives saved by earlier notification alone. In conclusion, AACN is a postcrash technology with a promising potential for safety benefit. AACN is therefore a key component of integrated safety systems that aim to protect occupants across the entire crash spectrum.

KEYWORDS:

• Telematics
• injury prediction
• crash notification
• postcrash
• triage
• prehospital care

Introduction

Advanced Automatic Collision Notification (AACN) is a system on a motor vehicle that notifies a public safety answering point (PSAP), either directly or through a third party, that the vehicle has had a motor vehicle crash that reached a minimum severity (e.g., airbag deployment). AACN systems contain cellular equipment (i.e., telematics module) that allows the vehicle to place a call and transmit data. The data transmitted via cellular signal include the current vehicle location, vehicle identification information (make, model, color), and a severe injury risk prediction. AACN systems enable earlier notification of a motor vehicle crash, improve location identification for first responders, and provide an injury prediction that can help dispatchers and first responders make better decisions about how and where to transport the patient, thus getting the patient to the right hospital faster.

Automatic collision notification (ACN) is a similar system that notifies a PSAP of a crash, but transmits only vehicle location and identification information. The benefits of ACN, limited to faster notification time, have previously been published. A study by Wu et al. (2015) used data from the Fatality Reporting Analysis System (FARS), years 2009 to 2012, to show that the mean notification and mean emergency medical services (EMS) arrival time post crash for fatalities were 6 min and 16 min, respectively. Survival analysis demonstrated that earlier notification within 1 to 2 min resulted in a 1–2% higher survival rate compared with later notification. Thus, ACN could significantly improve crash survivability and save approximately 177–244 lives annually in the United States. Note that this previous work did not consider the current market penetration of ACN technology (or how many U.S. vehicles already had ACN) when arriving at this estimate.

In addition to earlier notification, it is believed that AACN can produce additional benefits such as improved dispatch decision making (i.e., what resources to send to the scene) and improved transport decision making (i.e., whether to take a patient to the nearest community hospital or bypass the nearest hospital and go directly to a trauma center) (Champion et al. 2005). For severely injured patients, the benefits of treatment at trauma centers have been demonstrated. MacKenzie et al. (2006) found a 24% reduction in mortality (for deaths within 30 days) when comparing patients who were admitted to a Level I trauma center versus those admitted to a non-trauma center. In a meta-analysis of studies involving establishment of trauma systems, Celso et al. (2006) found a 15% reduction in mortality in favor of the presence of trauma systems. Others have documented improved survival rates for occupants treated immediately at a trauma center compared with those who were initially transported to a non-trauma center and later transferred (Garwe et al. 2011a; Garwe et al. 2011b).

The purpose of this research is to define the target population in the United States that may benefit from AACN. The target population is the entire group of fatalities that share key characteristics, such as being in a light vehicle (currently only light vehicles have available AACN) and having access to a trauma center within a reasonable amount of time. Those characteristics are described and are used to determine a reasonable estimate range of how many lives could be saved with full implementation of AACN throughout the entire U.S. vehicle fleet. For this effort, full implementation assumes 100% cooperation by emergency dispatch and prehospital care providers to act on the AACN severe injury risk prediction by transporting these patients to a trauma center (where available). This research combines both the benefits of trauma center care for severely injured patients and the benefits of faster crash notification. AACN injury prediction may also provide information on the most appropriate type of emergency response (e.g., advanced life support or air medical services rather than basic life support) required, which may result in highly trained emergency personnel on scene sooner; however, these additional benefits are not estimated in the current effort.

Methods

Data set

To determine the target population that could benefit from AACN, data from both the FARS and National Automotive Sampling System–Crashworthiness Data System (NASS-CDS) were obtained. Being a census of fatalities, FARS data can provide estimates of absolute numbers of fatalities. For both overall fatalities and light vehicle occupant fatalities, trends indicate a significant decrease in fatalities between 2000 and 2014, followed by an increase in 2015 (Figure A1, see online supplement). In order for the benefits estimates derived in this article to be consistent with this trend, only 2009–2015 FARS data were used in this analysis. There was an annual average of 21,934 light vehicle occupant fatalities for this period. Light vehicles were examined because AACN is currently only available in this sector of the fleet.

While absolute numbers were determined with FARS, NASS-CDS data were used to provide relative comparisons between different subsets of occupants, in order to identify the target population of interest. For this analysis, NASS-CDS data were compiled from case years 2000 to 2015. Only occupants within light passenger vehicles and cars were considered because the CDS data system focuses on light vehicles (weight ≤10,000 lb) and occupants only, and no data are collected from heavier trucks or buses.

Target population

Not all of the 21,934 annual average fatalities can benefit from AACN. To identify the subset of the population who can receive benefit, a number of “reduction factors” were applied to the overall population. These factors are described in detail in the following subsections and are summarized in Figure 1 and Table 1.

Figure 1. Flowchart of reduction factors applied to target population and demonstration of which occupants do not benefit from AACN (group C), which occupants benefit from earlier notification (group B, n = 6,893), and which occupants benefit from the AACN injury prediction via change in destination to trauma center (group A, n = 1,495 to 2,330). There is some overlap among occupants in groups A and B.

Table 1. Summary of reduction factors for AACN target population.

	Reduction factor	Overall	Pre-admission	Postadmission	Source of estimate
1	Instant deaths	0.72	—	—	FARS
2	AACN activation threshold	0.97	—	—	NASS-CDS
3	Vehicles equipped with ACN	0.45	—	—	NHTSA
4	Original transport decision	—	—	0.26	NASS-CDS
5	Access to trauma center	—	0.8	0.8	Published data
6	Injury prediction algorithm	—	0.9	0.9	ACS guidelines
7	Time to destination	—	0.43 to 0.56	—	FARS
8	Triage protocol	—	0.44 to 0.55	0.44 to 0.55	Published data
9	Vehicles equipped with AACN	0.8	—	—	NHTSA

Instant deaths

Previous research (Wu et al. 2015) demonstrated that when examining deaths within 6 h, 30% are instant deaths. Expanding the time window to deaths within 24 h (Figure A2, see online supplement) demonstrates that approximately 28% of occupant fatalities are instant deaths. Since these will not be helped by earlier notification or injury prediction associated with an AACN system, a reduction factor of 0.72 was applied to exclude instant deaths from the total target population.

AACN activation threshold

A severely injured occupant will not benefit from the AACN system if an automatic call/notification is not made. Airbag deployment and/or 15 mph delta-V have been cited as the threshold for AACN notification (Kononen et al. 2011). Thus, fatalities occurring below this threshold will not benefit and the target population needs to be reduced accordingly. For the NASS-CDS population examined in this study, approximately 97% of fatalities had either an airbag deployment or a delta-V greater than 15 mph (reduction factor = 0.97).

Vehicles equipped with ACN systems

It is acknowledged that not all new vehicles can be assumed to benefit from AACN, because some already have the technology. For this effort, the National Highway Traffic Safety Administration (NHTSA) estimated that approximately 35% and 20% of the model year 2016 fleet are predicted to be equipped with ACN and AACN, respectively. This accounts for vehicle technology but not necessarily active subscription status (due to lack of available information), which may reduce these percentages. Therefore, 45% of the fleet can be assumed to benefit from earlier notification (ACN reduction factor = 0.45).

Pre-admission versus postadmission deaths

After excluding instant deaths and those in which the AACN activation threshold (airbag deployment or delta-V ≥ 15 mph) is not met, the remaining population was divided into pre-admission and postadmission deaths (Figure 1). This was defined using the NASS-CDS variable “STAY,” where hospital stay greater than or equal to 1 indicated a postadmission death. Most fatalities (73.2%) occurred pre-admission and 26.8% occurred postadmission. This is not a reduction factor, but rather a split in the target population. This was done because there are additional factors to identify whether the occupant can benefit from an injury severity prediction that results in transport to a trauma center, which vary depending on whether the occupant died before or after reaching a hospital/trauma center.

Original transport decision

While all postadmission deaths may gain some benefit from earlier crash notification, it is necessary to consider the original transport decision to determine whether additional AACN benefits can be derived. If the occupant was already transferred to a trauma center, the injury prediction supplied by the AACN system is assumed to have no effect on transport decision. However, for occupants transported to hospitals, the injury risk prediction might result in a change in transport, from hospital to trauma center. Using the NASS-CDS data set, the percentage of fatal occupants who are admitted to either a trauma center or hospital is 73.8% and 26.2%, respectively. Thus, a reduction factor of 0.26 is applied to determine those occupants who could benefit from a change in destination.

Access to trauma center

Not all seriously injured or fatal occupants have access to a trauma center. Without access to a trauma center, patients will likely be transported to the same destination regardless of the injury severity prediction of the AACN system, and therefore may benefit only from earlier crash notification, but not from a destination decision informed by the injury prediction. Previous studies have found that the odds of fatality are significantly greater for occupants outside a 60-min coverage area, compared with a 45-min coverage area (NHTSA 2012a). Thus, for this analysis, a time window of 45 to 60 min for transport to a trauma center is assumed to be a reasonable time window for which benefits of change in destination may be realized. Geospatial analysis demonstrated that 80% of fatal crashes occur within a 20-min coverage area of helicopter emergency medical response (equivalent of 60-min response given 20-min flight time, out and back, and flight preparation time) (NHTSA 2012b). A reduction factor of 0.80 was applied.

Injury prediction algorithm

A severely injured occupant will not benefit from the AACN system (besides earlier notification) if the injury severity prediction algorithm does not identify the patient as having a high risk of severe injury. Thus, the sensitivity of the algorithm, or true positive rate (proportion of occupants with ISS 16+ who are correctly identified by the algorithm as having a high risk of injury), is an important consideration in target population development. Various injury prediction algorithms have published sensitivity rates ranging from 40% (Kononen et al. 2011) to 95% (Stitzel et al. 2016).

The 2014 edition of the American College of Surgeons (ACS) Resources for Optimal Care (American College of Surgeons 2014) defines undertriage as severely injured patients transported to lower level trauma centers or other facilities, and overtriage as minimally injured patients transported to higher level trauma centers. The ACS gives higher priority to reduction of undertriage, because undertriage may result in preventable mortality or morbidity from delays in definitive care. The recommended level for undertriage is 5%. Overtriage may result in higher costs and also increase the burden for higher level trauma centers because resources needed for more severely injured patients are unnecessarily being used for minimally injured patients. Acceptable rates for overtriage are in the range of 25% to 35%.

The sensitivity of an injury severity prediction algorithm is equal to 100% minus the undertriage rate (i.e., a sensitivity of 95% will result in 95% of ISS 16+ occupants being correctly triaged to a trauma center, and 5% being undertriaged to a hospital). Specificity, or the true negative rate (proportion of occupants with ISS <16 who are correctly identified by the algorithm as having a low risk of injury), is equal to 100% minus the overtriage rate (i.e., a specificity of 65% means that 65% of minimally injured occupants are correctly triaged to a hospital and 35% are overtriaged to a trauma center). Undertriage and overtriage generally have an inverse relationship: Lowering undertriage will almost always increase overtriage, which in turn affects resource availability of the whole system. For the purposes of the present study, we are not examining this relationship in detail, but rather selecting a single set of parameters that is both in line with ACS recommendations and achievable by an injury prediction algorithm.

Thus, an algorithm sensitivity of 90% (reduction factor = 0.9) is being used for the current analysis because this sensitivity is achievable by an injury prediction algorithm that also meets the specificity requirement of 65%, while achieving 95% sensitivity requires a reduction in specificity below 65% (Enriquez and Lee 2017). Note that because injury prediction algorithms are optimized to identify severely injured patients (ISS 16+), the sensitivity and specificity rates do not directly apply to fatal occupants. Nonetheless, it is assumed here that an algorithm that correctly identifies 90% of ISS 16+ patients will also correctly identify at least that proportion of fatal occupants.

Time to destination

Previous research reported that the median time for fatalities to reach a hospital or trauma center was between 45 and 60 min and the mean time to death was 67 min (Wu et al. 2015). Using the previously published survival curve, expanded for the time window of 24 h (Figure A2, see online supplement), the survival rates at 45 and 60 min post crash were 40% and 31%, respectively. Excluding the 28% instant deaths yields survival rates of 56% and 43%, for 45 and 60 min post crash, respectively. The assumption here is that the 44% to 57% of occupants who expired prior to this time frame have injuries that are simply untreatable, or could not have been saved even with trauma center treatment. Thus, a reduction factor range of 0.43 to 0.56 was used to represent the proportion of pre-admission occupants who can benefit from improved care. For the postadmission group, this factor is not relevant since these occupants did in fact, reach a destination alive.

Triage protocol

Since 1986, the American College of Surgeons Committee on Trauma (ACS-COT) has published a resource manual that provides guidance for the field triage process through a Field Triage Decision Scheme (Sasser et al. 2011). The Field Triage Decision Scheme protocol is based on sequential evaluation of different aspects of trauma patient presentation, with the outcome being a determination of appropriate transport decision for a patient (e.g., hospital or trauma center). Step 1 is defined by physiologic and level of consciousness indicators and Step 2 is defined by anatomic signs of injury. If a patient meets either Step 1 or Step 2 criteria, that patient will likely be transported to a trauma center, regardless of the AACN system severity prediction. Vehicle telematics “consistent with high risk for injury” is currently listed in Step 3 of the protocol, meaning that if a patient is negative for Step 1 and 2, the AACN injury prediction can be used to inform transport decision to either hospital or trauma center.

A multisite assessment of the Field Triage Decision Scheme determined that Steps 1 and 2 were cumulatively 45% sensitive for identifying patients with Injury Severity Score (ISS) 16 or greater (16+) (Newgard et al. 2011). An analysis using NASS-CDS data found that Steps 1 and 2 combined identified about 52% of ISS 16+ patients (Davidson et al. 2014). Similarly, an analysis using the National Trauma Databank (NTDB) found that Steps 1 and 2 combined were 56% sensitive for identifying patients with ISS 16+ (Brown et al. 2011). Because 45% to 56% of severely injured patients are being identified by the first two steps in the Decision Scheme, the 44% to 55% who are not could receive benefit from an AACN system that correctly identifies them as high risk of severe injury (reduction factor: 0.44 to 0.55). Note that fatalities may be Step 1 or 2 positive more often than severely injured nonfatal occupants, but this has not been specifically evaluated in any published studies.

Vehicles equipped with AACN systems

As noted earlier, approximately 20% of the model year 2016 fleet are predicted to be equipped with AACN. Thus, 80% of the fleet can benefit from AACN implementation (AACN reduction factor = 0.80). Although penetration of the use of AACN injury predictions by emergency medical personnel may be as or more important than the penetration of the technology within the vehicle fleet, such predictions are beyond the scope of the current effort. As such, this factor only accounts for penetration of the technology within the future vehicle fleet.

Estimate of lives saved (change in destination plus earlier notification)

Once the relevant reduction factors (summarized in Table 1) were identified, they were applied to the original target population of 21,934 fatalities in successive order as shown in Figure 1. Next, a fatality reduction estimate range was developed. This included both the effects of earlier notification and the effect of improved transport decision. The benefits of earlier notification were assumed to be 1–2% fatality reduction, based on previously published estimates (Wu et al. 2015). To determine the effect of transport decision, the NASS-CDS data set was used to develop relative survival rate ratios between different medical facilities, using two different statistical approaches. NASS-CDS documents whether a fatal crash victim was admitted to a Level I or II trauma center (hereafter just “trauma center”) versus a Level III center or lower (hereafter just “hospital”) using variable “MEDFACIL.” Fatal outcome was defined using the variable TREATMNT = 1,2 in the NASS-CDS database.

One tool to compare the survival probability over time is a nonparametric method proposed by Kaplan and Meier. The Kaplan–Meier method is commonly used for medical research and reliability engineering (Kaplan and Meier 1958; Hosmer and Lemeshow 1999). The Kaplan–Meier estimator, or life curve, at any time is described by Eq. (1(1) $$\widehat{S}\left(t\right)=\prod _{t_i<=t}\left(1-\frac{d_i}{n_i}\right)=\prod _{t_i<=t}\left(\frac{s_i}{n_i}\right)$$(1) ): (1) $$\widehat{S}\left(t\right)=\prod _{t_i<=t}\left(1-\frac{d_i}{n_i}\right)=\prod _{t_i<=t}\left(\frac{s_i}{n_i}\right)$$(1) where d_i is “deceased” subjects or fatalities, and s_i is the “survivor” subjects or alive (“censored”) drivers/passengers (but the survival status depends on the time window), and n_i is total subject number (total occupants in the related time window). In CDS data, the only available time variable is time to death, which is recorded in hours (up to 24 h) or days (1+ days) using the variable DEATH. The occupant sample for the Kaplan–Meier life curve was any occupant with an available DEATH of 1–60. A time window of 24 h after the crash was considered, such that the occupants within this 24-h time window had two survival statuses—“died” and “still alive” relative to the time window. The effect of relative survival over time was considered important in the context of AACN because the benefits of AACN (e.g., faster transport to trauma center and definitive care, vs. transport to hospital and possibly requiring a subsequent transfer to trauma center) are time dependent. It is assumed that the faster response associated with AACN may not be as critical for patients who are able to survive for many days before expiring.

Because the Kaplan–Meier survival method uses time to death (the only time variable in CDS) to estimate survival rate over time, the analysis does not include any surviving occupants. Therefore, another statistical approach used to include all occupants and to estimate survival ratio was the multiple proportional hazard model, or hazard model. The hazard function, h(t), was introduced by Cox; the hazard, h(t), and survival probability functions, S(t), are closely related to each other, described by h(t) = – S’(t) /S(t), where S’(t) is the derivative of S(t). Cox proposed that the hazard function can be further expressed in Eq. (2(2) $$h\left(t\right)=h_{0}exp(\beta 1Age+\beta 2Facility+\beta 3Belt+\beta 4DeltaV)$$(2) ), known as the Cox proportional hazard model (Hosmer and Lemeshow 1999). The goal of this model is to establish a relationship between the hazard function with multiple risk factors simultaneously, while the previously discussed Kaplan–Meier curves explore a single risk factor, Medical Facility, only. The model included four independent factors treated as categorical variables: delta-V (>35 mph or not), belt use (belted or not), age (>65 years or younger), medical facility (no treatment, hospital, or trauma center). The binary dependent outcome is “fatal or not” (fatal if “Treatment = 1, 2” in NASS-CDS data): (2) $$h\left(t\right)=h_{0}exp(\beta 1Age+\beta 2Facility+\beta 3Belt+\beta 4DeltaV)$$(2)

Results

Revised target population

Figure 1 shows the application of the reduction factors to the original target population. After excluding instant deaths, those in which the activation threshold (airbag deployment or delta-V ≥ 15 mph) is not met, and the portion of the current fleet that already has ACN technology, the resulting 6893 fatalities (group B from Figure 1) can all receive some benefit from implementation of either an ACN or an AACN system, specifically the benefit of earlier notification. Only a small subgroup (group A from Figure 1) meets all the AACN criteria (access to trauma center, survivable injuries, negative for Step 1 and 2 of the Field Triage Decision Scheme, and positive for being predicted by the injury severity prediction algorithm) and therefore benefits from both earlier notification and injury severity prediction via a change in destination. This group encompassed between 1495 and 2330 occupants, given the range of reduction factors used. There is also some overlap between occupants in groups A and B.

Benefit of injury prediction (change in destination)

For patients with known time to death, the Kaplan–Meier survival analysis demonstrated better relative survival rates at trauma centers compared with hospital and no treatment (Figure 2). At 24 h post crash, the survival rate at hospital was 0.79 that of trauma center (Table 2).

Figure 2. Occupant survival rate vs. time to death for various medical facility destinations (if “Treatment = 1,2” as Fatal), using Kaplan–Meier estimator.

Table 2. Survival rates over time for occupants who died within 30 days, based on Kaplan–Meier estimator.

Time since crash (h)	No treatment	Trauma center	Hospital	Survival rate ratio (hospital/trauma center)
3	0.0422	0.5046	0.3906	0.774
6	0.0258	0.4111	0.3131	0.762
12	0.0211	0.3447	0.2593	0.752
24	0.00468	0.2674	0.2121	0.793

When all dead and surviving patients were considered, the Cox proportional hazard model demonstrated that medical facility had a significant effect on survival (Table 3). Specifically, the hazard ratio (hospital vs. trauma center) was 1.335, meaning that occupants sent to hospital had 33.5% higher fatality probability (or relative survival rate of 1/1.335 = 0.75) than if they were sent to trauma center. Other variables considered (belt use, age, delta-V) also had significant effects on survival.

Table 3. Multiple hazard ratios for various independent variables, using Cox proportional hazard model (NASS-CDS 2000–2015).

Parameter	Parameter estimate	Standard error	Chi-squared	Pr > ChiSq	Hazard ratio
Hospital vs.  trauma center	0.28885	0.01287	503.3567	<.0001	1.335
No treatment vs. trauma center	1.06258	0.00913	13536.8767	<.0001	2.894
Not belted vs. belted	0.20967	0.00798	691.1399	<.0001	1.233
Older (>65 years) vs. younger	−0.50045	0.00945	2805.3909	<.0001	0.606
Delta-V > 35 vs. slower	0.26559	0.00921	832.2262	<.0001	1.304

Thus, for the fatalities identified in the target population that can potentially benefit from a change in destination (group A from Figure 1), 0.75 to 0.79 would still have died given a change in destination (based on the survival ratios from the two methods of statistical analysis). Therefore, 0.21 to 0.25 of the population, or 314 to 583 people, could be saved by change in destination alone.

Estimate of lives saved (change in destination plus earlier notification)

Once benefits from change in destination were established, these were combined with the benefits of earlier notification. For the lower bound of the estimate range, group A was assumed to exclusively derive benefits from injury prediction and not from earlier notification. Thus, the occupants in group A (n = 1495 to 2330) were removed from group B (n = 6893), to ensure no overlap, and the number of occupants benefitting from earlier notification was 4563 to 5398. For the upper bound, group A was assumed to derive additive benefits from earlier notification and injury prediction. Thus, the number of occupants benefitting from earlier notification was from the entire group B (n = 6893).

Using the 1–2% fatality reduction for earlier notification, based on previously published estimates (Wu et al. 2015), the portion of lives saved by earlier notification is approximately 46 (4563 × 0.01) to 138 (6893 × 0.02). Thus, a total of between 360 (314 + 46) and 721 (583 + 138) total lives could potentially be saved given AACN.

Discussion

The purpose of this research was to identify the U.S. target population of fatalities that could benefit from AACN. Seriously injured but not fatal occupants were not considered here. The factors affecting the target population were described and used to determine a reasonable estimate range of how many lives could be saved with full implementation of AACN in the vehicle fleet (assuming full utilization of information by prehospital care providers) and universal cell coverage availability. This research combined both the benefits of trauma center care for severely injured patients and the benefits of faster crash notification. The benefits of trauma center care determined from this study, which combined Level 1 and 2 trauma centers (which is all that is available from NASS-CDS) and compared those to non-trauma centers, were comparable to those from earlier estimates (MacKenzie et al. 2006; Celso et al. 2006).

The estimate of lives saved (360 to 721) represents a fatality reduction of approximately 1.6 to 3.3% per year, and more than double the potential lives saved by earlier notification alone. Others have previously estimated mortality reduction associated with earlier notification (i.e., ACN only), both in the United States and around the world. These prior estimates have ranged from 1% to 10% fatality reduction (Chauvel et al. 2011; Clark and Cushing 2002; Lahausse et al. 2008). Similarly, one study estimated the fatality reduction for AACN in Sweden to be around 8% (Jonsson et al. 2015). While the current estimate is near the low end of the published fatality reduction range, factors such as the triage protocol were included that have not previously been utilized in other estimates. Thus, the current study demonstrates that AACN is a postcrash technology with a promising potential for safety benefit.

In addition to the benefits associated with reducing fatalities, AACN can provide additional benefits for seriously injured nonfatal occupants. Many injuries, such as brain hemorrhage or aorta laceration, are time sensitive (Schoell et al. 2015). Time-sensitive injuries in particular will benefit both from earlier crash notification and from a prediction of severe injury that prompts immediate transport to a trauma center rather than a local hospital. Trauma center care can also result in better outcomes and reduced readmissions for severely injured patients (Staudenmayer et al. 2015). For injuries such as hemorrhaging, Abbreviated Injury Scale (AIS) severity levels are often based on blood volume. Since faster treatment may result in reduced blood volume, this can also reduce maximum AIS score, leading to better outcomes. While the benefits for severely injured nonfatal occupants were not estimated in the current research, it is clear that AACN can benefit this group of occupants by improving response time and triage/transport decisions. Since nonfatal injured occupants greatly outnumber fatal occupants, the actual societal benefit is likely much greater than the fatality reductions estimated in this paper.

Other benefits can be realized for non-occupants (e.g., pedestrians, pedal cyclists and motorcyclists) or occupants of the non-AACN-equipped vehicle in a crash. In these situations, the AACN system can be used (through either automatic call or manual call) to contact emergency services. Finally, the benefits of AACN injury prediction estimated here were limited to the change in destination (from hospital to trauma center) based on injury prediction outcome. However, AACN injury prediction could also provide information on the most appropriate type of emergency response (e.g., advanced life support or air medical services rather than basic life support) required, which may result in highly trained emergency personnel on scene sooner. Also, correctly identifying patients as being severely injured can shorten response time at the trauma center, since a trauma team may be activated in advance of the patient arrival. While the effect cannot be quantified at this time, it may be substantial, since unlike the benefits of change in destination, these factors can benefit patients who were already transported to a trauma center (73.8% of postadmission patients already went to trauma center).

Limitations of this research include that the benefits of AACN rely upon others, such as 911 dispatchers and EMS first responders, to “act differently.” Upon receiving an automatic collision notification with a high probability of severe injury, these end users need to send different resources to the scene or make the decision to transport a patient to a trauma center who might otherwise be sent to a local hospital, in order for the benefits to be attained. Although the ACS Field Triage Decision Scheme was cited here, many states, counties, and municipalities have their own triage protocols, and do not necessarily follow the ACS Decision Scheme. Few, if any, prehospital care systems currently have EMS protocols in place that would dictate what first responders should do when receiving an AACN message with a high probability of severe injury. The estimates presented in this paper assume a 100% cooperation or compliance with AACN in step 3 of the triage protocol. It is hoped that the current research, along with prior research, will demonstrate the potential for benefits associated with AACN, leading others to adopt and implement protocols that would allow these benefits to be realized. If it is considered that there is currently almost no implementation of AACN injury prediction in EMS protocols, the potential benefits to be derived could be greater than predicted here, since any vehicle, regardless of whether it already has AACN technology, stands to potentially benefit from its implementation. Current efforts considered only new vehicles entering the fleet that do not already have an AACN system. Another limitation is that the current analysis also assumes universal cell coverage availability. Information is currently not available demonstrating the proportion of fatal crashes that occur outside of cell phone coverage areas and thus would not have access to an AACN call. Finally, a limitation to the approach used here is that the reduction factors were applied to the target population in a successive approach. This approach assumes that the relative proportions of occupants do not change as reduction factors are applied. Although considerable thought was given to the most logical order in which reduction factors should be applied, this nevertheless represents a limitation of the approach.

In conclusion, this analysis provides an estimate of lives that could potentially be saved with implementation of AACN. An estimated fatality reduction of approximately 1.6% to 3.3% per year, and more than double the lives saved by earlier notification (i.e., ACN) alone, demonstrates that AACN is a postcrash technology with a promising potential for safety benefit. AACN is therefore a key component of integrated safety systems that aim to protect occupants across the entire crash spectrum.