Strategic Highway Research: Saving Lives, Reducing Congestion, Improving Quality of Life -- Special Report 260 (2001)

Each year approximately 42,000 people are killed on the nation’s highways, and 3 million are injured. The cost of these crashes approached $182 billion in 1999 (NSC 2000). Highway fatalities account for approximately 95 percent of transportation-related deaths (BTS 1999, Table 3-1). Indeed, the annual highway death toll is equivalent to a jet airliner crashing and killing everyone on board every day of the year. Highway crashes represent the single largest category of accidental deaths and the most frequent cause of death among children and young adults (Insurance Institute for Highway Safety 2001). (Box 5-1 shows some additional statistical comparisons.)

Significant improvements have been made in highway safety during the last several decades. From 1988 to 1999, for example, the number of highway fatalities dropped from 47,087 to 41,611, an 11.6 percent decline, and the fatality rate in deaths per 100 million vehicle-miles traveled (VMT) dropped from 2.3 to 1.5, a 34.8 percent decline. During the same period, the injury rate dropped from 169 to 120 per 100 million VMT, a 29.0 percent decline. There have also been reductions in alcohol-related deaths and pedestrian and bicyclist fatalities and increases in the use of safety belts. These advances are due to improvements in vehicles (safety belts, airbags, crash-worthiness), infrastructure design (roundabouts, shoulder rumble strips), roadside hardware (guardrail end treatments, breakaway signposts and light

posts), traffic engineering (raised pavement markings, new reflective sign sheeting or special applications), and enforcement and public awareness efforts (safety belt and alcohol laws and programs). These improvements in turn are the result of federal and state legislation and applied safety research carried out by federal and state agencies, universities, and industry. However, decreases in fatality and injury rates have leveled off since the early 1990s, as shown in Figure 5-1. In addition, as noted earlier, low fatality and injury rates still mean large numbers of deaths and serious injuries because of the significant increase in VMT (see Figure 5-2). Between 1988 and 1999, VMT rose 35 percent from 2.0 trillion to 2.7 trillion.¹ This means improvements in safety, as reflected in declining fatality and injury rates, are not keeping up with increases in VMT. If these trends continue, more aggressive highway safety efforts will be needed just to keep the absolute numbers of fatalities and injuries from rising and certainly to reduce these numbers significantly.

Several other factors add to the challenge of achieving safer highways:

• Changes in vehicle design and in the popularity of particular vehicles (such as sport utility vehicles) make it difficult for highway agencies to keep roads and roadside hardware compatible with the vehicle fleet.

• Demographic changes raise a number of issues with regard to driver performance and crash survivability. First, there continues to be an increase in the number of older drivers, who are more susceptible to injury in a crash and less likely to survive their injuries. The oldest drivers may exhibit a decline in vision, quickness of reaction, and cognitive skills associated with the performance of complex tasks. In addition, an increase can be expected in the number of teenage drivers, who have little experience or mature judgment.

• New vehicle technologies, such as antilock brakes, onboard navigation systems, and adaptive cruise control, have certain safety benefits but are accompanied by changes in driving behavior that may reduce (or in some cases negate) those benefits.

• Even without the new technologies, highway designers are finding, driving behavior has changed in important safety-related ways. For instance, comparison of the new version of the Highway Capacity Manual (TRB 2000) (used to design highways) and the earlier version (TRB 1994) reveals that drivers are willing to accept shorter distances between vehicles for a given speed.

As is evident from the above considerations, the highway safety problem is both serious and extremely complex. The committee received many suggestions for safety-related research during the outreach process. After consultation with safety experts, the committee identified two knowledge gaps whose remediation through appropriate research would help the highway safety community use many existing countermeasures more effectively and potentially develop improved countermeasures as well.

First is the need for more precise understanding of how various factors contribute to crashes. Many factors contributing to the occurrence or severity of crashes have been identified. These factors include roadway geometry, speed, alcohol use, driver fatigue, poor visibility, unforgiving roadside objects, and many others. However, most of the research that has addressed these factors has been constrained by limitations on the availability or accuracy of the data or on the ability to draw conclusions about the relative contributions of individual factors. In particular, even the best studies have lacked adequate exposure data for the noncrash population and accurate precrash data. A variety of

new technologies offer unprecedented opportunities to collect data on crash factors and exposure that have never before been available.

The second major knowledge gap relates to the cost-effectiveness of the countermeasures already in use or under consideration. The highway community is spending many millions of dollars on a wide variety of counter-measures—flashing lights, driver education, enforcement strategies, new sign colors and sizes, different signal timing—without having adequate knowledge of their cost-effectiveness, whether singly or in combination, in preventing crashes or reducing crash severity.

These two major knowledge gaps are closely related. Without more precise knowledge of the contributions of various crash factors, it is difficult to determine how much effort should be focused on any one of them. Even if the most significant crash factors are identified, it is not clear which countermeasures address those factors most cost-effectively. In addition, some factors may be particularly resistant to countermeasures. This is the case with many human factors because it is more difficult to change human behavior than to change vehicle or roadway design. Moreover, even when a countermeasure is technically effective, it may be incompatible with social, political, or legal requirements or may be prohibitively expensive. All of these issues must be taken into account if optimal investments in highway safety are to be made. And the absence of accurate data on each aspect of the problem is a major barrier to achieving a substantial leap forward in highway safety.

The research proposed in this chapter is intended to overcome this barrier and provide the data and analysis needed to make the highway system as safe as possible. This research will require the collection and analysis of a larger quantity and wider variety of data than has previously been available for crash analysis studies, including both noncrash and precrash data. Moreover, the research will focus specifically on the effectiveness of selected crash counter-measures. To carry out such an ambitious study, the committee proposes that researchers use available advanced technologies to the extent possible. Additional description of the proposed research and potentially useful technologies is provided later in this chapter.

How highway safety meets the first of the criteria set forth in Chapter 1 for selecting the strategic focus areas for F-SHRP was described in the preceding section: it is an issue that bears on national transportation goals and is of con-

tinuing concern to highway agencies. The other two F-SHRP criteria—appropriateness for a SHRP-style program and the effectiveness or expected impact of the research—and how the proposed study of crash factors and cost-effectiveness of countermeasures meets these criteria are addressed in this section.

The study envisioned by the committee will require resources well beyond the capability of existing highway safety research programs. For any of these programs to attempt such a study would mean abandoning other critical contributions to highway safety. A study of crash factors will also require several years of continuous, predictable funding, not subject to the vicissitudes of the annual budget decisions that affect all existing programs. In addition, an integrated approach is required in the research design to encompass the various factors—driver, vehicle, and roadway—that must be studied. The conduct of the study will require coordination and cooperation among several players, including a number of federal agencies [FHWA, the National Highway Traffic Safety Administration (NHTSA), the Federal Motor Carrier Safety Administration (FMCSA)], state and local agencies, vehicle manufacturers, the insurance industry, the legal community, and others. A focused research program, independent of many of the institutional constraints appropriate to existing research programs, will be in a position to bring these players together more effectively and forge new alliances while building on the institutional strengths and historical alliances that characterize existing programs.

A study of the scope, scale, and rigorous research design envisioned by the committee is not likely to be carried out by an entity other than F-SHRP because of funding and institutional constraints. At the same time, both the ultimate goal—a significant improvement in highway safety—and the specific path chosen toward that goal—a comprehensive study of crash factors and the cost-effectiveness of countermeasures—are appropriate focuses of public-sector concern and resources. This is true for several reasons: the vast majority of highways are publicly owned; safety on the highways is a significant public health issue;² and longer-term, higher-risk research such as this is a traditional responsibility of the public sector.

In selecting focus areas for F-SHRP research, the committee sought to identify research opportunities with the potential to yield results that would surpass qualitatively or quantitatively what existing highway research programs are likely to provide on their own. While the outcomes of a study of crash factors and effectiveness of countermeasures are less certain than those of other, lower-risk types of research, this study by its very nature is oriented toward fundamental knowledge and therefore fundamental advancement. In addition to providing a basis for wiser investments in existing countermeasures, results of this study can potentially be applied to produce new countermeasures in a wide variety of safety programs, from intelligent transportation systems (ITS) strategies, to highway and vehicle design, to enforcement and driver training approaches. Application of more fundamental knowledge of crash factors and effectiveness of countermeasures could lead to sizable reductions in deaths and injuries, making it possible to outstrip the anticipated growth in VMT. Every 1 percent improvement in highway safety resulting from application of the results of this study would mean more than 400 lives saved, 30,000 injuries averted, and $1.8 billion in economic costs avoided annually.

The study is expected to take approximately 7 years. Some interim results could be available earlier, but it is expected that analysis of the full study results along with recommendations on cost-effective implementation of countermeasures could begin to be available within 10 years of the start of the study (assuming that the pilot study discussed below could be conducted in advance of the next highway reauthorization so that the full study could begin at the start of the next authorization period).

Barriers to the conduct of the research program are largely legal and institutional (as discussed in more detail below), but initial indications are that they can be overcome. Barriers to implementation are likely to be financial and institutional. However, these barriers, too, should be surmountable given the importance of the problem; the results obtained on the cost-effectiveness of specific countermeasures; and the existence of potential implementation mechanisms, such as the AASHTO Strategic Highway Safety Plan³ and the highway safety manual currently being initiated by NCHRP.

The highway safety research community has never conducted a study of this type. However, experience gained through NHTSA’s crash analysis pro-

gram, the truck crash causation study currently under way at FMCSA, the pilot study proposed below, and other related efforts should adequately prepare the research community for the undertaking. The ability to implement the study results will vary among the responsible groups—federal, state, and local agencies and industry. Moreover, where the burden of implementation lies will depend on what is learned from the study. For example, highway design issues will fall within the purview of government agencies, while vehicle design issues will have to be addressed by the automobile industry. One of the challenges involved in the conduct of the study will be to keep the various stakeholders apprised of its progress and to find appropriate ways of involving them, building on the strengths and interests of each.

The proposed research has two major objectives:

• To identify more accurately the contributions of various factors to highway crashes, fatalities, and injuries; and

• To determine the cost-effectiveness of selected countermeasures in preventing or reducing the severity of highway crashes.

What is entailed in achieving these objectives is examined in this section. In addition, since the development of certain data and communications technologies offers unprecedented opportunities to meet these objectives, these opportunities are explored as well.

The issue of what constitutes a crash factor is complicated. While such factors are sometimes referred to as causal factors, limitations in the research design of traditional crash analysis studies (as discussed in this section) make it difficult if not impossible to infer causality. Indeed, the committee has chosen to avoid using the term “causation” in recognition of the complexity of the problem and the different perspectives on what constitutes the cause of a crash. To illustrate the point, Box 5-2 describes a simple crash scenario, as well as the multiple judgments about causality and potential solutions that might result from analysis of the incident by various experts. The proposed study will not resolve these

differences in perspective or reduce the complexity of highway safety to a simple algorithm. Nonetheless, a comprehensive study of crash factors performed by a multidisciplinary team (representing human factors, traffic engineering, vehicle design, roadway design, enforcement, and other disciplines), using the most accurate data and most scientifically rigorous research design currently available, can greatly improve decision making with regard to highway safety. Most highway safety decisions are based on a small amount of data and analysis and a large amount of professional judgment and experience, tempered by political and financial constraints. A multidisciplinary team of experts can provide the fullest possible understanding of the interplay among various crash factors and the relative effectiveness of different countermeasures. Enhancing the quality of crash data and the robustness of safety analysis will lead to better decisions that will save lives and reduce injuries and property damage (the exact nature and extent of this improvement will depend on the results of the proposed research).

The main challenge of the proposed study and the key to its ultimate contribution is the scientific rigor of the research design. All other crash analysis studies⁴—from the seminal Indiana work of the 1970s, to the regular NHTSA investigations on which the industry depends, to the current FMCSA truck crash causation study—have had limited research designs as a result of con-

straints on funding, time, or technological capability, thus constraining the ability to draw causal inferences from the data produced. Researchers too often have resorted to gathering the data that are available instead of those needed to address the salient questions. To remedy this situation and provide the strongest possible intellectual basis for designing and implementing effective safety strategies, the proposed study of crash factors must have the following characteristics:

• A statistically significant, representative sample of all vehicles;

• Analysis of fatal, injury, and property damage–only crashes;

• A control of noncrash circumstances to measure exposure in the driving population; and

• New and more precise data about precrash circumstances and actions.

These characteristics are discussed in turn below.

Statistically Significant, Representative Sample of All Vehicles To achieve such a sample, (a) a large enough number of vehicles must be studied that the number of crashes expected to occur among these vehicles will itself be sufficient to meet the requirements of statistical significance; and (b) the vehicles must be chosen randomly so they are representative of all vehicles, thereby avoiding bias toward or against any significant crash factor (such as driver characteristics, geography, or type of vehicle). Although this characteristic implies a vehicle-based sampling design, the researchers should also be sure that the design will yield crashes on a representative distribution of roadway types and classifications (for example, rural and urban roads, Interstates and arterials). This characteristic of the research program is fundamental to the next two characteristics.

Analysis of Fatal, Injury, and Property Damage–Only Crashes Existing crash studies are based on a pool of crashes that have been reported to police. Therefore, crashes involving minor injuries or property damage only, which may never be reported, are systematically omitted from those studies. Detailed investigations of randomly selected crashes nationwide are performed under the NHTSA Crashworthiness Data System for crashes involving a towed passenger car, van, or truck that is less than or equal to 10,000 pounds gross vehicle weight. This level of investigation is not performed for other types of crashes. The proposed study will overcome this limitation and thus yield entirely new information about crash types not previously well researched.

Control of Noncrash Circumstances to Measure Exposure in the Driving Population One of the most critical elements in determining the significance of a given factor in crashes is to control for other factors. A perfect experimental control is not possible in a highway crash study because of the lack of information about exposure to risk in a large enough sample. However, statistical means can be used to control for various factors hypothesized to contribute to crashes if the study includes a control population of drivers and vehicles not involved in crashes. Since the unit of analysis in most crash analysis studies is the crash event, often triggered by a police report, these studies do not include analysis of drivers and vehicles that have not been involved in crashes.

An example may help illustrate the importance of this point with regard to designing and implementing effective safety measures. One may know from analyzing a number of crashes that 25 percent of the drivers were suffering from fatigue. It may appear obvious, then, that fatigue was a significant factor in the crashes. But if one were to discover that 25 percent of the driving population at large suffers from fatigue—including those who never experience a crash—it would clearly be necessary to look for other factors as significant contributors to the crash. Better knowledge of the role of various crash factors would then lead to much more effective design and implementation of safety policies and programs, helping highway safety professionals choose countermeasures that address the truly critical crash factors rather than factors that accompany crashes but may not be the most significant contributors to crash occurrence or severity.

New and More Precise Data About Precrash Circumstances and Actions Most existing information about crashes consists of the observable data after a crash, such as the type of crash, the nature and extent of damage to the vehicle, and the injuries sustained. It is sometimes possible to learn about precrash events by interviewing witnesses or by drawing inferences from postcrash evidence (such as skid marks), but for the most part the chain of events and circumstances leading up to a crash is not known with much detail or accuracy. As a result, the presence or absence of various factors can be difficult to determine, and investigators often resort to identification of the first harmful event of which they are aware (such as a collision with another vehicle), which may not in fact be the first or most harmful event. NHTSA has begun to use event data recorders (EDRs), or onboard recording devices, to gather some precrash data pertaining to vehicle performance and driver behavior. This approach offers much promise. The use of such devices is an important component of the pro-

posed program and is discussed further below, along with other technologies now available to facilitate the conduct of the study.

The conduct of this study will require a suite of data gathering and analysis methods, including on-site investigations, surveys, interviews, and use of police accident reports. The exact details of the research protocol will need to be developed during the interim work discussed later. However, several technologies currently available provide an opportunity to perform this study in a way that could only be imagined just a few years ago. These technologies include EDRs, video recording systems, cell phones, the Global Positioning System (GPS), and sensors. Those developing the research design should fully consider the possibilities offered by these technologies.

Event Data Recorders EDRs make it possible for researchers to gather objective data about certain precrash and crash conditions. Such devices have been available in one form or another in some motor vehicles since the 1970s when airbags were introduced. Early recorders were used to monitor the readiness of airbags, provide warning to the driver if a bag appeared likely to malfunction, and record the bags’ actual performance. Since that time, EDRs have increased in sophistication and have been included in more vehicle makes and models. They can now be designed to record such data as vehicle speed, acceleration, braking, and safety belt use, as well as airbag deployment. If made available to researchers, the data thus obtained would provide unprecedented insight into the circumstances and events leading up to a crash. The improvement in crash-related data potentially available from the use of EDRs is illustrated in Figure 5-3. The figure shows two examples of the Haddon matrix, which encompasses the human, vehicle, and environmental data available about the precrash, crash, and postcrash periods.⁵ The first matrix shows the data available without EDRs and thus has several empty cells for precrash and crash data. The second matrix shows those cells filled, plus additional data in other cells resulting from the EDR capability.

NHTSA has begun to take advantage of the data available from vehicles equipped with EDRs. The agency has also formed a working group with gov-

Figure 5-3 Improvement in availability of crash-related data with use of event data recorders [slightly modified from Chidester et al. (1999, Tables 4 and 5)].

ernment and industry members to encourage the use of EDRs, develop performance requirements and data definitions, and resolve legal and privacy issues.⁶ The National Transportation Safety Board encourages the use of on-board recording devices in all modes of transportation.⁷ Devices for downloading data from some EDRs are commercially available and relatively inexpensive.

The major issues associated with EDRs pertain to legal, privacy, and data consistency concerns. Clearly, the owners of vehicles included in the proposed research will need to be informed about the EDRs on their vehicles and consent to their use for the study purposes. At the same time, the vehicle owners

must be protected from third-party access to the data. Thus security measures will be required in the handling of the data, including masking the identity or source of specific data. In addition, the researchers and the data will have to be protected from subpoena in legal proceedings. Such protection is not unusual in research involving human subjects or data that individuals would be unlikely to reveal to researchers if they thought those data could be used against them. This legal protection also serves as a safeguard for the integrity of the research because individuals are less likely to provide inaccurate data to avoid a potential personal threat. In the case of the proposed study of crash factors, this means drivers will be less likely to modify their driving behavior to avoid getting into trouble, so researchers will be able to obtain a more accurate picture of their behavior.⁸ A possible mechanism for this purpose is to apply for a Certificate of Confidentiality that the Department of Health and Human Services is authorized to issue. This certificate acknowledges that certain types of research have special privacy requirements and that people need to be protected from use of the resulting data against them. The National Transportation Safety Board is similarly protected from releasing data it downloads from aircraft and motor vehicle recorders.

Video Recording Systems Video recording systems also offer unique advantages for this type of research. The Kentucky Transportation Cabinet has installed such a system at a high-incident intersection in Louisville (Urban Transportation Monitor 2001). The system uses continuous-loop cameras and microphones to monitor the intersection. When the microphone picks up the sounds of a collision, it saves several seconds of video and allows the cameras to capture several more seconds, providing visual data for before, during, and after the crash. If intersection crashes were chosen for particular analysis in this study, such a device could be installed at the sampled intersections. In addition to providing for crash footage, an appropriate research design could be developed for randomly sampling vehicles approaching and progressing through the intersections to gather visual data on driving behavior and intersection operational performance under noncrash circumstances.

In-vehicle cameras have been used to capture data about events both inside and immediately outside of a vehicle. FHWA is using in-vehicle cameras to

track driver eye and hand movements and other driving behaviors under experimental conditions. Some private and police vehicle fleets have cameras mounted on the rearview mirror to record events occurring in front of the vehicle should a crash occur.

As with the use of EDRs, privacy and legal issues must be addressed in the use of video recording systems. For example, cameras external to vehicles (such as intersection-mounted cameras) should not be aimed at private property and should not capture individuals in the vehicle. In-vehicle cameras would require owners’ permission. In practice, it may be possible to use such cameras only on vehicle fleets whose owners may have their own incentives for camera use. Video data (and audio data, if included) would also need to be protected from legal proceedings.

Cell Phones, GPS, and Sensors Cellular telephone technology could be used to alert researchers when a crash takes place (or at predetermined intervals to collect noncrash data) and to transmit recorded data efficiently and inexpensively. GPS could be installed on vehicles (as is already the case on some vehicles) to locate a crash for further, on-site investigation of highway geometry and roadside hardware. In the event of a crash, sensor technology could be used to gather data on weather, pavement condition (presence of ice or moisture), and traffic volumes and speeds. These data could be transmitted remotely as well. Technology for automated collection of site and crash geometry data could be used by researchers to improve the speed and accuracy of data collection.

In addition to sensors that can gather weather, roadway, and traffic data, recent advances have resulted in a number of technologies oriented toward driver performance that are being field tested by NHTSA, FMCSA, and the AAA Foundation for Traffic Safety. These technologies include passive alcohol sensing technology; eyelid monitoring (PERCLOS); and other devices that provide information about the driver’s state of alertness, sobriety, attention status, and use of cell phones or telematics devices. Such technologies could provide objective data about whether and to what degree these driver conditions and activities contribute to crashes and would be especially useful for gathering exposure data. By the time the proposed study is launched, these technologies will have seen several more generations of development. Similarly, lane tracking and night vision systems are already on the market and could be used to add environmental data to the vehicle data collected by EDRs.

Even if all of these technologies cannot be used throughout the study, it may be possible to test some of them in portions of the research as experimental data-gathering methods. These newer methods could be compared with more traditional data gathering techniques, such as interviews and use of police reports, which will also be employed in the research. As newer technologies emerge, they could be more fully employed in future studies.

In the course of the outreach conducted for this study, the committee encountered the argument that the money required for a comprehensive study of crash factors might be spent more effectively on implementing countermeasures already in existence or under development. This is a reasonable point of view, and certainly implementation of existing countermeasures must continue. However, as effective as certain countermeasures are, there are many whose effectiveness is not well established, and there is still little basis for determining which countermeasures are the most cost-effective. Improved knowledge of crash factors, together with knowledge about the cost-effectiveness of countermeasures, will help agencies prioritize the various countermeasures now available, make more rational investments in their implementation, and direct the development of new countermeasures. In addition, some countermeasures are controversial; one reason the political will to implement them is lacking is that there is no clear basis for weighing the safety benefits gained against social or economic costs. The proposed study can provide a basis for informed public discourse and policy development in this regard.

The specific countermeasures to be studied will be determined during the interim work stage. The kinds of countermeasures the committee has in mind are effectiveness of guardrails, impacts of roadway and shoulder width, aspects of intersection safety (signal type and phasing, geometrics), and enforcement strategies. In choosing countermeasures, researchers should examine the incidence of fatalities and injuries from various crash types and identify countermeasures intended to address these crash types. By using data from the study of crash factors, researchers should be able to identify countermeasures that are designed to address the most critical factors and then perform analyses of their relative cost-effectiveness. Researchers may also want to examine commonly used countermeasures to determine whether investments in those approaches are in fact likely to yield safety benefits comparable with their expense.

An example may be helpful to show how more accurate knowledge of crash factors, including better precrash and noncrash data, can lead to more effective use of crash countermeasures. Run-off-the-road crashes account for 33 percent of highway fatalities (FHWA, personal communication). Potential factors in these crashes include roadway geometry (a curved road, for example); excessive speed for that geometry; absence or inadequate visibility of lane markings or signs indicating the road’s curvature; the presence or absence of a guardrail; the presence or absence of objects (such as signposts, trees, and walls) that may be struck by an errant vehicle; vehicle malfunction, such as brake failure; driver behavior, such as safety belt use and braking; and driver distraction, fatigue, or substance abuse.

To know the importance of road geometry, it is necessary to have accurate data about crash location. Often these data are only approximate (to the nearest mile marker, for instance) or are not available at all (because they were inadvertently omitted from the police report or purposely excluded from the data set to protect privacy). As a result, it is difficult to pinpoint high-accident locations and establish reliable relationships between crash types or severity and particular roadway geometries. The use of GPS can help identify crash locations with much greater accuracy. More accurate crash location data can also allow researchers to see exactly what pavement marking, signage, guardrail, and other traffic control and safety appurtenances and roadside objects are at the scene.

Speed is usually estimated from physical evidence associated with the crash, such as skid marks (if the vehicle does not have antilock brakes) and damage to vehicles and other objects. In some cases, speed can be determined by interviewing witnesses or someone involved in the crash. These methods (especially interviews) are only approximate, however, and can be extremely inaccurate. EDRs can provide accurate speeds, as well as changes in speed (delta V) during the crash. In addition, if EDR and GPS technologies are used to gather information about typical speeds on similar geometries where crashes do not occur, researchers will have a much better idea of how significant a factor speed is in crashes and what ranges of speeds may be safer.

As noted, EDRs can provide highly accurate data on vehicle functions, as well as some aspects of driver performance or behavior. A driver may report that he or she applied the brakes or was wearing a safety belt at the time of a crash. But an EDR will record if and when the brakes were in fact applied and which, if any, of the safety belts in the vehicle were in use.

Measures of driver characteristics are more difficult to obtain. Police may administer tests related to alcohol use, but they do not always do so. Information about possible sources of driver distraction may be observed (a cell phone, several children in the vehicle), but it is difficult to know whether a given distraction was actually taking place just before the crash. Interviews are again the main source of this kind of information, as well as that concerning fatigue or possible mental distractions (such as whether the driver was worried about a problem at home or at work). Yet the technologies described earlier can help provide more accurate data on driver activities through use of video and audio recording, if appropriate permission is obtained and privacy issues are adequately addressed. Similar information (for example, on cell phone use, external distractions, eye movement, drowsiness) from noncrash circumstances would help determine the extent to which these factors differ between crash and noncrash situations.

Such improved data could inform more effective use of countermeasures in several ways. Changes in velocity during a crash, for instance, have implications for the design of vehicle safety features and roadside hardware because they determine the amount of energy that must be absorbed by the vehicle or hardware to reduce injury to the vehicle’s occupants. More accurate data on speed not only could be used to design better roadside hardware, but also could affect geometric design, signage, traffic control devices, and enforcement strategies. More accurate data on crash locations would allow researchers to compare crash severity at locations with various geometric designs and under use of different countermeasures, such as the design and placement of guardrails or crash attenuators.

There are many other considerations related to run-off-the-road crashes, and similar examples could be developed for intersection crashes, vehicle rollovers, and other types of crashes. Given the complexity of these crashes, the researchers will have to strategically select which factors and countermeasures should be the focus of the research design and analysis. This selection can be made on the basis of existing data from NHTSA and FHWA on the incidence of various crash types and countermeasures.

This proposed research program is intended to identify a strategic direction (critical knowledge requirements for substantially improved highway safety) and potential opportunities in this area afforded by research and the use of advanced technologies. The committee is aware that this is a very broad

and ambitious proposal. While all four proposed research programs will require additional definition during the interim period, the safety program depends most critically on this interim work. The committee recognizes that the research cannot address the entire universe of possible crash types, crash factors, and crash countermeasures. However, the human suffering and economic burden resulting from highway crashes warrants something significantly more than continued small-scale, incremental improvement. Given recent advances in computer data storage, processing, and communications technologies, what once would have been an overwhelming if not impossible task is now only challenging. The advanced technologies currently available make it feasible to gather more accurate and some previously unobtainable data to develop a robust and scientific understanding of highway crashes upon which decisions about cost-effective, life-saving safety investments can be based. While expert interpretation of the objective data will still be required, the increased objective data provided by the proposed study will reduce much of the need for conjecture and subjective opinion in the analysis of crashes.

A feasibility study should be conducted and a detailed research plan developed before the full-scale study is initiated. The feasibility study corresponds to the interim work (described in Chapter 8) that is required for all four F-SHRP programs and should include the following tasks:

• Define clearly the goals of the study, the questions to be addressed, and the data needed to answer these questions. Research questions will focus on issues such as the contribution of various factors to particular crash types and the relative cost-effectiveness of selected countermeasures in addressing these factors. The feasibility study should help identify crash types, factors, and countermeasures for which the full-scale study can be expected to yield beneficial results. The relative strengths and weaknesses of different options for the focus of the full-scale study should be identified. The study might also address such issues as the following: Are drivers compensating for the benefits of new safety measures by pushing the limits more? How can emerging advanced communications technologies (telematics) be used strategically? How can drivers be informed about the potential benefits and risks of these technologies?

• Assess the capability of EDRs, GPS, video technology, and other methods to capture the required data. Define critical shortcomings, and propose alternative means of acquiring data in these areas.

• Perform an analysis to assess the feasibility of carrying out the study of crash factors as described. This analysis should address such issues as the following:

  • Whether the EDRs and GPS devices already in vehicles can be used for the study, or the researchers will need to equip vehicles with specially designed devices.

  • How other in-vehicle devices, such as cameras and sensors, might be used. For example, can they be installed in a representative sample of vehicles, or should a separate sample of fleet vehicles be equipped with these technologies? What are the trade-offs in terms of the cost and value of the data?

  • Whether the legal and privacy issues concerning EDRs, cameras, and other technologies significantly compromise their use in the proposed study.

  • If any of these technologies are not feasible, whether other methods can be used to obtain equivalent data.

  • How data will be handled: who will collect them, where they will reside, how privacy will be protected, and how and which data (if any) will be available to other researchers.

  • Whether use of the Department of Health and Human Services Certificate of Confidentiality is appropriate and feasible for this study.

  • How coordination with other, related programs can be achieved to avoid duplication and leverage financial resources and expertise.

• Design a research protocol that will result in obtaining the appropriate data and performing the required analyses, taking into account the results of the feasibility study.

• Investigate possible incentives that could be offered to vehicle owners for participation in the study, such as provision of navigation technology (if this does not introduce a bias to the research) or automatic notification of 911 in the event of a serious crash.

• Develop a cost estimate and schedule for the full-scale research program.

If possible, a pilot study to test the research protocol should also be performed before embarking on a large-scale research program. If this is not possible, such a pilot study would be the first task under the formal research program. Other broad tasks within the study, pending development of a detailed plan, are as follows:

• Conduct of the study—This task will be accomplished using whatever suite of research methods is determined to be appropriate in the detailed research plan. These methods may include surveys, interviews, on-site crash investigation, use of police accident reports, and collection of data by means of

advanced technologies. The study should focus on a set of crash types, crash factors, and countermeasures, taking into account the strengths and weaknesses of the options identified in the feasibility study. The research team should include the following disciplines, and others as appropriate: human factors, traffic engineering, vehicle design, roadway design, enforcement, driver education, and public awareness.

• Analysis of data—The detailed research plan will include the specific questions to be addressed by the research (which will drive decisions about which data to collect). Appropriate analyses will be performed on the data to address these questions and develop conclusions about the relative contributions of various crash factors in certain crash types and the cost-effectiveness of selected countermeasures.

• Recommendations concerning countermeasures—On the basis of the knowledge gained from the study, the selected crash countermeasures (which may include roadway designs, vehicle designs, education, public awareness programs, and enforcement strategies) should be assessed and recommendations made about those most cost-effective to implement.

Other highway safety efforts can provide opportunities for dialogue with and involvement of stakeholders, complementary research results, articulation of safety goals, and venues for implementation. These other efforts include the following:

• NHTSA’s Fatality Analysis Reporting System uses state DOT, police, and other data to study fatal crashes. The agency’s National Automotive Sampling System examines all types of crashes using the General Estimates System, which looks at a nationally representative sample of crashes using information from police reports; and the Crashworthiness Data System, which performs detailed analyses of 5,000 crashes per year, supplementing police reports with on-site investigation and other data (including some use of EDRs).

• NHTSA is conducting a study in Georgia using EDRs on 1,100 vehicles to test the use of this technology in crash studies. The results of that effort should be considered in the feasibility study for this proposed research.

• As noted, FMCSA is conducting a congressionally mandated $15 million to $20 million truck crash causation study. The study will include in-depth

investigations of a representative sample of large-truck crashes involving fatalities or serious injuries. Trained investigators from NHTSA and FMCSA are involved in the study.

• FHWA performs limited crash analyses using the data from eight states in the agency’s Highway Safety Information System. FHWA has also developed the Interactive Highway Safety Design Model (IHSDM), which can be used to better integrate safety analysis into the planning and design phases of highway development. IHSDM design modules use estimates of the effectiveness of various countermeasures (such as wider lanes or larger curve radii) in reducing accidents. At present, this model focuses predominantly on two-lane rural highways, but it will be expanded to higher-volume and more complex road configurations in coordination with NCHRP’s development of a highway safety manual (discussed below). FHWA has also funded the development of the ALERT vehicle, a law enforcement vehicle equipped with advanced communications technology to aid police officers in collecting more accurate data at crash scenes.

• The Comprehensive Highway Safety Improvement Model is a 6-year, $2.5 million project to develop an expert system for use by each state, with its particular databases, to screen the road network, identify high- or higher-than-expected accident locations, diagnose the accident causes from the patterns seen in collision diagrams and site visits, select countermeasures related to the particular accident types, and compare the cost–benefits of these countermeasures to select those most appropriate.

• Under the Intelligent Vehicle Initiative (IVI), the U.S. Department of Transportation’s Joint Program Office for Intelligent Transportation Systems has created a partnership with industry aimed at developing in-vehicle technologies that can help in avoiding common types of crashes, such as rear-end collisions, single-vehicle run-off-the-road crashes, and intersection collisions. A promising area is the development of cooperative vehicle–infrastructure ITS to enhance safety. FHWA has formulated initial plans in this area, but little work has been done to date because of a lack of funding. A crash causation study of limited scope may be proposed as part of the IVI activities, as well as studies of naturalistic driving (to learn more about driver behavior).

• NCHRP has initiated a study on the use of EDR technology for the collection and analysis of highway crash data.

• The development of prototype countermeasures under the proposed program should be closely coordinated with related ongoing safety research being conducted by NCHRP, state DOTs, universities, and industry.

• The AASHTO Strategic Highway Safety Plan⁹ was developed in 1997 by AASHTO in cooperation with FHWA, NHTSA, and TRB and with input from a broad range of highway safety stakeholders and interest groups. The plan targets the goal of saving 5,000 to 7,000 lives annually and substantially reducing health care costs related to highway crashes over a period of 5 to 7 years. Emphasis is placed on making a concerted effort to implement proven highway safety strategies and on conducting some model development and demonstration of emerging strategies in 22 key emphasis areas. The plan does not identify long-term knowledge gaps or research needs; however, the implementation of proven strategies builds on historical safety research, and the model development and demonstration effort requires the support of existing safety research programs at FHWA, NCHRP, and state DOTs. NCHRP Project 17-18, with additional pooled funding from state DOTs, involves a series of tasks aimed at implementing the AASHTO plan. One of these tasks (Task 5) will involve developing an integrated management system for highway safety.

• The development of a highway safety manual through NCHRP is a new effort to evaluate, standardize, and disseminate best practices in highway safety.

• AASHTO is developing safety software that will, among other things, facilitate more accurate crash reporting for police.

Lessons learned from this study about the role of highway design and roadside features will be applicable to highway renewal projects, as well as to the construction of new highways that may be required to meet increased demand. Improved safety will also contribute to highway system reliability by reducing the number of crashes that take place within the system.

Several considerations should be kept in mind during the development and performance of the proposed study. For example, research should be done in close cooperation with FHWA, NHTSA, and FMCSA. Lessons from the ongoing truck crash causation study being sponsored by FMCSA and from the EDR studies being undertaken by NHTSA and NCHRP should be incor-

porated into the design and conduct of this research program. The use of EDR technology will require the active involvement of vehicle manufacturers, whose cooperation must be sought as early as possible in the study. In addition, the insurance industry may be interested in participating.

Additional considerations pertain more directly to implementation of the research results. For instance, the recently initiated NCHRP highway safety manual could be an ideal implementation vehicle. Assessments of the cost-effectiveness of selected countermeasures would be extremely useful information for users of the manual. Researchers engaged in improving existing countermeasures or in developing new ones could benefit from data about crash factors. In general, the research program staff should maintain regular contact with safety professionals within state and local agencies to keep them abreast of the program as it develops and to obtain their input in the development of prototype countermeasures.

BTS Bureau of Transportation Statistics

NHTSA National Highway Traffic Safety Administration

NSC National Safety Council

BTS. 1999. National Transportation Statistics 1999. BTS99-03, U.S. Department of Transportation, Washington, D.C.

Chidester, A., J. Hinch, T. C. Mercer, and K. S. Schultz. 1999. Recording Automotive Crash Event Data. Presented at the International Symposium on Transportation Recorders, May 3–5, Arlington, Va. www.ntsb.gov/events/symp_rec/symp_rec.htm.

Cirillo, J. A. 2001. Safety Agenda for the Nation. Presented at the Institute of Transportation Engineers Spring Conference, Monterey, Calif., March 26.

Insurance Institute for Highway Safety. 2001. www.iihs.org/safety_facts/fatality_facts/general.htm. Accessed May.

Lehmann, G., and T. Reynolds. 1999. The Contribution of Onboard Recording Systems to Road Safety and Accident Analysis. Presented at the International Symposium on Transportation Recorders, May 3–5, in Arlington, Va. www.ntsb.gov/events/symp_rec/proceedings/authors/lehmann.htm.

NHTSA. 2000. Traffic Safety Facts 1999. DOT HS 809 100. U.S. Department of Transportation, Washington, D.C., Dec. www.nhtsa.dot.gov/people/ncsa/tsf1999.pdf.

NHTSA. 2001. Event Data Recorders. Summary of Findings by the NHTSA EDR Working Group (draft final report), Aug. www-nrd.nhtsa.dot.gov/pdf/nrd-10/EDR/WKGrp0801.pdf.

NSC. 2000. Injury Facts (2000 ed.). Itasca, Ill.

Rice, D. P., E. J. MacKenzie, et al. 1989. Cost of Injury in the United States. Institute for Health and Aging, University of California, San Francisco, and Injury Prevention Center, The Johns Hopkins University.

TRB. 1994. Special Report 209: Highway Capacity Manual, 3rd ed. (1994 update). National Research Council, Washington, D.C.

TRB. 2000. Highway Capacity Manual. National Research Council, Washington, D.C.

Urban Transportation Monitor. 2001. Vol. 15, No. 15, Aug. 3.

AASHTO American Association of State Highway and Transportation Officials

DOT U.S. Department of Transportation

GAO General Accounting Office

TRB Transportation Research Board

AASHTO. Strategic Highway Safety Plan. safetyplan.tamu.edu.

Automotive News Insight. 2000. Heads Up. March 6, pp. 38i–42i.

Barbosa, H. M., M. R. Tight, and A. D. May. 2000. A Model of Speed Profiles for Traffic Calmed Roads. Transportation Research A, Vol. 34, No. 2, Feb., pp. 103–123.

Bolte, K., L. Jackson, V. Roberts, and S. McComb. 1999. Accident Reconstruction/Simulation with Event Recorders. Presented at the International Symposium on Transportation Recorders, May 3–5, Arlington, Va. www.ntsb.gov/events/symp_rec/symp_rec.htm.

Chidester, A. B., J. Hinch, and T. A. Roston. n.d. Real World Experience with Event Data Recorders. Paper 247 (on the basis of content, paper would have been written in 2000 or early 2001). National Highway Traffic Safety Administration.

Dole, L. 1999. On-Board Recorders: The “Black Boxes” of the Trucking Industry. Presented at the International Symposium on Transportation Recorders, May 3–5, Arlington, Va. www.ntsb.gov/events/symp_rec/symp_rec.htm.

DOT. 1999. Human-Centered Systems: The Next Challenge in Transportation. Washington, D.C., June.

DOT. 1999. Operator Performance-Enhancing Technologies to Improve Safety: A US DOT Safety Initiative for Meeting the Human-Centered Systems Challenge. Washington, D.C., Nov.

DOT. 2000. Intelligent Vehicle Initiative Business Plan. Intelligent Transportation Systems Joint Program Office, Washington, D.C., July.

English, L. P. 1999. Plain English on Data Quality: Seven Deadly Misconceptions. DM Review, June/July/Sept.

Fenwick, L. 1999. Security of Recorded Information. Presented at the International Symposium on Transportation Recorders, May 3–5, Arlington, Va. www.ntsb.gov/events/symp_rec/symp_rec.htm.

GAO. 1995. Highway Safety: Causes of Injury in Automobile Crashes. GAO/PEMD-95-4, Washington, D.C., May.

Kaysi, I., and G. Alam. 2000. Driver Behavior and Traffic Stream Interactions at Unsignalized Intersections. Journal of Transportation Engineering, American Society of Civil Engineers, Vol. 126, No. 6, Nov.–Dec., pp. 498–505.

McKnight, A. J. 2001. Too Old to Drive? Issues in Science and Technology, Vol. 17, No. 2, Winter 2000–2001.

Mening, P., and C. Coverdill. 1999. Transportation Recorders on Commercial Vehicles. Presented at the International Symposium on Transportation Recorders, May 3–5, Arlington, Va. www.ntsb.gov/events/symp_rec/symp_rec.htm.

Pfefer, R., and T. Neuman. 1999. Improved Safety Information to Support Highway Design. TR News, No. 204, Sept.–Oct., pp. 23–27, 42–43.

Smiley, A. 2001. Auto Safety and Human Adaptation. Issues in Science and Technology, Vol. 17, No. 2, Winter 2000–2001.

TRB. n.d. Highway Safety Research: A National Agenda. National Research and Technology Partnership Forum, Infrastructure Renewal Working Group. www.TRB.org, under “R&T Forum.”

TRB. 1990. Special Report 229: Safety Research for a Changing Highway Environment. National Research Council, Washington, D.C.

TRB. 1999. Motor Vehicle Crashes—Data Analyses and IVI Program Emphases. From handout at the Nov. 9–10 meeting of the TRB Committee for the Review of the Intelligent Vehicle Initiative.