Skip to contentUnited States Department of Transportation - Federal Highway Administration FHWA Home
Research Home
Public Roads
Featuring developments in Federal highway policies, programs, and research and technology.
This magazine is an archived publication and may contain dated technical, contact, and link information.
Federal Highway Administration > Publications > Public Roads > Vol. 60· No. 3 > The Highway Safety Information System: Transforming Data Into Knowledge

Winter 1997
Vol. 60· No. 3

The Highway Safety Information System: Transforming Data Into Knowledge

by Jeffrey F. Paniati and Forrest M. Council

My dictionary defines information as "knowledge obtained from investigation, study, or instruction." This definition concisely describes the purpose of the Federal Highway Administration's (FHWA's) Highway Safety Information System (HSIS). HSIS exists to provide knowledge on the safety performance of the highway system and, more specifically, the effects that changes in highway design and operations have on safety.

Highway engineers and administrators are continually faced with decisions concerning safety on the highway. These decisions can be very broad, such as the safety impacts of proposed programs or policies, or very specific, such as the design of a particular intersection or section of highway. They may require the engineer to look into the future to foresee the safety implications of a particular decision or to defend designs and decisions after the fact. In any case, effective transportation decision-making requires knowledge of the safety impacts of the decision so that they might be considered along with traffic flow, the environment, cost, and other factors.

For example, the design of the width of highway medians can have important safety, environmental, and traffic flow impacts. In urban areas, when redesigning highways, medians are frequently being used to provide additional capacity; however, this additional capacity may be gained at the expense of safety. The questions the designer/administrator must answer (whether implicitly or explicitly) are whether or not the trade-off between mobility and safety is acceptable and/or how to mitigate any negative impacts. Similarly, in rural areas, the decision-maker may be trading off environmental concerns, such as the taking of additional wetlands, to provide improved safety by having a wider median. To effectively address the safety concerns when making trade-offs between these different factors, the decision-maker must have knowledge about the relationship between the width of the median and crashes. HSIS is designed to provide this type of knowledge.

The impact of changes in median width is just one example of the type of safety issues that are brought to the attention of FHWA. Resolving these issues requires determining whether or not a significant safety problem exists, conducting research to develop a detailed understanding of the problem, identifying and evaluating potential countermeasures, and developing programs or policies to address the problem. HSIS provides the information necessary to support each phase of this process.

HSIS is different from other "crash-based" safety systems, such as the National Fatal Accident Reporting System (FARS) and the National Accident Sampling System (NASS). By their inherent design, these databases are "outcome" databases only. Even if they contain specifics of the roadway at the crash site (e.g., whether divided or not, paved or unpaved shoulder, curve or tangent section), the records that are on file relate only to "failures" of the system -- instances in which the roadway design may have contributed to the vehicle being involved in a reportable crash.

However, to answer questions concerning the effects of roadway design or operations on safety, one must not be able to study only the "failures" that occur but also the "successes" -- the miles of highway with certain design or operational features where the crash rate is either zero or very low. And since the "safeness" of a section in terms of crashes is defined not only by the roadway design but also by the number of vehicles exposed to the design, some measure of traffic flow is necessary in the analysis. Thus, what is really needed to develop a relationship between design or operational features and crash rates is not only crash data but also roadway inventory and traffic volume data. HSIS contains these types of data.

Prior to the development of HSIS, obtaining crash, roadway inventory, and traffic data that could be linked together and studied required each researcher to acquire the data directly from a given state or states or conduct a special data-collection effort. This process was repeated each time a new research question was raised. This approach was time-consuming and expensive for both FHWA and the states providing data. It left FHWA without ready access to information for "quick-turnaround" questions and provided no way for researchers to document and transfer knowledge about the quality and applicability of the data from one study to another. HSIS solves these problems by providing information that is in a common computer format, documented, and prepared for analysis.

HSIS is a multistate database that contains crash data; roadway inventory information; traffic volume data; and special inventory data-related items such as intersections, interchanges, and roadside hardware. By design, HSIS uses data already collected by the state highway and safety agencies for their own purposes, thus maximizing the cost-effectiveness of both the system itself and the state data. In addition to providing a storehouse for data collected by the states, the system also provides storage capabilities for new data collected either in the HSIS laboratory or as part of other FHWA research efforts. HSIS is designed to complement, but not duplicate, other Department of Transportation databases maintained by FHWA and the National Highway Traffic Safety Administration. For example, bridge and roadway grade crossing inventory systems that have long been retained by FHWA can now be linked with additional crash and roadway inventory data within the HSIS system.

The Evolution of HSIS

In the late 1980s, FHWA recognized the need for a roadway-based safety data system to allow the agency to better address current and emerging highway safety issues. In response to this need, FHWA initiated the development of HSIS. A planning study was conducted to examine various approaches that could provide the necessary information. Options that were explored included the collection of additional crash data for Highway Performance Monitoring System (HPMS) sample sections nationwide, the expansion of FARS or NASS, and the acquisition of state data. Based on the need for large samples of detailed roadway and traffic inventory information, the need to upgrade the system annually, and the need to keep the long-term data-acquisition costs within budgetary constraints, it was determined that the most effective and cost-efficient way to proceed was to develop the HSIS database from data that were already being collected by state highway agencies for their own use.

In HSIS, state data are acquired annually by FHWA. The data then undergo a series of quality-control checks and are prepared in a format that allows the data from the various files -- accident, roadway, traffic, etc. -- to be merged. The merged data are then used in analyses. Instead of attempting to combine data from different states into a common format, which would result in the loss of much unique information, HSIS was developed to maintain the integrity of the individual mergeable files within each state. This type of system provides several advantages (in addition to the cost savings) for problem analysis. First, each state data set can be examined to determine which ones possess the most appropriate data variables, categories, sample sizes, and linkages. Second, state-by-state analyses can be done using the appropriate data for each state. Finally, the results can be compared across states to check for consistency and/or differences, leading to greater faith in generalization to other parts of the nation.

The initial version of HSIS was brought on-line in 1990 and included five states -- Illinois, Maine, Michigan, Minnesota, and Utah. The participating states were selected based on a detailed review of all 50 states. The review examined the variety, quality, and quantity of data and the ability to merge various files. Between 1990 and 1994, HSIS was used successfully in more than 70 safety analysis efforts. These studies ranged from problem-identification to preliminary analysis and scoping of potential research studies to large-scale research efforts. Clients included not only internal staff in various FHWA offices but also independent researchers conducting research for FHWA, the National Cooperative Highway Research Program, state departments of transportation, universities, and the private sector.

While the original system has proven to be quite valuable, the use over the past five years has indicated that the system is somewhat limited in terms of such factors as geographic coverage, climate, and road type. In addition, the research conducted to date and the questions that were being asked by the HSIS "customers" led FHWA to recognize the need for additional special files -- files related to intersections, interchanges, the roadside, and horizontal and vertical alignment.

The roadway-based safety data from the eight shaded states shaded is merged into the HSIS database for analysis.

To help overcome these limitations, FHWA decided to expand HSIS. In this expansion, three additional states were added to the system -- California, North Carolina, and Washington. These states were chosen because they possess very good accident, inventory, and traffic information that can be linked together and because they increase the geographic coverage, the quantity of data, and the variety of files available to the user. In particular, these new states bring in much-needed additional information on alignment, intersections, and interchanges.

At the same time the system was expanded in size, a complete overhaul of the hardware and software used for processing the data was undertaken. This overhaul was driven by the facts that between the late 1980s and mid-1990s, dramatic improvements and changes had taken place in the computer industry and that the new HSIS system would include more than eight gigabytes of data at its inception and could be expected to grow by one gigabyte per year. The focuses of the overhaul were: (1) to move from a hardware system where high-end stand-alone personal computers were used to a distributed system over a client-server network, and (2) to restructure the databases from flat files to a relational structure. The result of this overhaul is a much faster, more efficient system. The experience gained through the HSIS conversion can provide valuable guidance to other organizations, including state departments of transportation that are considering a similar conversion from older mainframe-based data systems.

The HSIS Laboratory

The HSIS laboratory is located at the Turner-Fairbank Highway Research Center in McLean, Va. It is staffed by two full-time engineer/analysts as well as several graduate research assistants. The laboratory provides a working environment where FHWA staff and visiting researchers can use a variety of analytical tools in the study of highway safety. These include:

  • Personal computers and workstations where users can access the HSIS files, extract pertinent data, and conduct analyses using the latest computer hardware and statistical analysis software.
  • Videodisc photologs to access roadway images for visual verification of existing data and/or collection of supplemental data. Currently, the HSIS laboratory contains images for the entire Michigan, Minnesota, and Utah state-maintained highway systems. These images are taken at intervals of approximately 16 meters (59 feet) and can be randomly accessed using the route number and milepost for the location of interest.
  • Geographic Information System (GIS) software where users can manipulate, display, and analyze HSIS data. The HSIS laboratory is also used to support the development and testing of new GIS-based safety analysis tools for use at the state and local levels.

The HSIS laboratory also serves as a one-stop center for information about HSIS. Copies of HSIS guidebooks and the results from HSIS research can be obtained from the laboratory staff. Those who wish to be added to the HSIS mailing list can contact the laboratory at (703) 285-2090 or mail a business card to HSIS Mailing List, c/o HSIS Laboratory, Turner-Fairbank Highway Research Center, T-212, 6300 Georgetown Pike, McLean, VA 22101-2296.

HSIS Specifics -- What's Available?

HSIS currently includes data on more than 5 million crashes on more than 252,000 kilometers of roadway. Table 1 shows the HSIS road distances inventoried by the states and the annual number of crashes that can be linked to that road distance. In most cases, the kilometers inventoried are those on the "state system," and include interstate, U.S., and state routes in both rural and urban locations. In some states, additional distances on county/township roadways are included in the inventory. Crash files include all crashes that meet the state's reporting threshold. In most cases, the threshold now ranges between $500 and $1,000 in property damage or personal injury in a crash.

Table 1 - Accidents Per Year on State-Maintained Roadways

First Year Available Average Accidents/Year State-Maintained Roadway (Kilometers)
California 1991 45,000 11,250
Illinois 1985 150,000 25,750
Maine 1985 39,000 35,400
Michigan 1985 140,000 15,450
Minnesota 1985 85,000 79,800
North Carolina 1991 118,000 49,900
Utah 1985 50,000 20,750
Washington 1991 34,000 13,850

The types of files that are included for use in HSIS are shown in table 2. As can be seen, all eight states include basic information on crashes, roadway inventory, and traffic volumes. The crash data are divided into three subfiles: (1) the specifics of the particular crash, (2) the specifics of each of the vehicles in the crash, and (3) details related to the occupants within each of the vehicles. In addition to these basic files, certain states provide information on horizontal curvature, grade, and/or vertical curvature; vehicle identification number (VIN); detailed intersection geometric and traffic control information; information on interchanges and ramps; and inventory information concerning guardrails and median barriers.

Table 2 - Types of Files Provided for Use in HSIS

CA IL ME MI MN NC UT WA
Accident X X X X X X X X
Roadway Cross Section X X X X X X X X
Traffic Volume X X X X X X X X
Curve and Grade X X X
VIN X X X X
Intersection X X
Interchange/Ramp X X X X
Guardrail/Barrier X X

Finally, to facilitate the use of this system, a multisection "guidebook" has been developed for each of the states. The guidebook is designed to provide sufficient information for both the analyst and the programmer to effectively use the system. It also documents data-quality issues uncovered through annual quality-control checks or reported by system users. The guidebook for each state consists of two volumes. Volume I contains a basic description of the state data system, an alphabetized listing (by file type) of all available variables, detailed definitions of each category present within each variable, and notes on the quality of the variable. Volume II contains single-variable tabulations for a large number of key variables within each of the files. The tables include data for the previous five years. The HSIS guidebooks are updated on a two-year cycle.

Transforming Data Into Knowledge

Types of HSIS Data Files

Accident -- Contains basic accident, vehicle, and occupant information on a case-by-case basis. Typical data include type of accident, types of vehicles, sex and age of occupants, fixed object struck, accident severity, and weather conditions.

Roadway Inventory -- Contains information on roadway cross section and the type of roadway. Data include the number of lanes, lane width, shoulder width and type, median width, rural/urban designation, and functional classification.

Traffic Volume -- Contains annual average daily traffic (AADT) data. Additional data on hourly volumes and percentage of trucks are also available for selected states and/or locations.

Roadway Geometrics -- Contains horizontal curve and vertical grade information. Data include degree of curve, length of curve, and percent of grade.

Intersection -- Contains information on highway intersections. Data include traffic-control type, intersection type, signal phasing, and turn lanes.

Interchange -- Contains information on highway interchanges. Data include interchange type and ramp characteristics.

Vehicle Identification Number (VIN) -- Contains VIN data decoded using the VINDICATOR program. Data include make and model, body style, body type, curb weight, and wheelbase.

Guardrail/Barrier -- Contains an inventory of guardrails. Data include guardrail type, post type, rail height, and terminal type.

Transforming the data contained within the HSIS databases into knowledge about the safety performance of the highway system requires access to individuals with expertise in the analysis of highway safety problems. That expertise is provided by staff in the FHWA Safety Design Division and the University of North Carolina Highway Safety Research Center (HSRC). HSIS is operated by HSRC under contract with FHWA. The FHWA and HSRC staffs include highway and traffic engineers, human factors researchers, statisticians, and computer programmers with expertise in the study of highway safety problems. Together they conduct research with HSIS and guide other users on the application of HSIS.

The HSIS data and the expertise of FHWA and HSRC are available for any effort whose purpose is to develop increased knowledge of the relationships between highway design and operations and safety. To explore the potential of HSIS for addressing a particular problem, a user can prepare a one-page description of the issue to be addressed and submit it to the Safety Design Division of FHWA. The request is then reviewed by both the FHWA staff and the HSIS research team at HSRC who will work with the user to either answer the question directly or to select the appropriate data and prepare extract files for additional analysis by the user. The staff is also available to consult with the user throughout the project to answer questions concerning the variables within the files or how the data might have been collected within the states. At the end of the project, the user is asked to send a copy of any final report produced to FHWA and to provide input on data-quality issues that were uncovered so that they can be incorporated by the HSIS staff into the revised guidebooks. HSRC also operates the HSIS laboratory at FHWA's Turner-Fairbank Highway Research Center in McLean, Va.

Since its inception, the FHWA and HSRC staffs have conducted or provided HSIS data for use in more than 90 different safety analysis efforts. These efforts have ranged from small efforts, including preliminary "problem-size" analyses and development of extract files for use by other researchers, to larger, complete research studies that resulted in technical papers or journal articles. These analysis efforts have resulted in better understanding of emerging problems, helped scope future research efforts, provided input to policies and program decisions, and guided changes in design policies and practices. In this work, HSIS has been used as a data resource by research consultants, national laboratories, and universities nationwide.

For those efforts where a significant finding was achieved, a two- to four-page HSIS summary report has been prepared to document the results of the analysis. The purpose of the HSIS summary reports is to transfer the knowledge gained in the major HSIS research efforts to a large audience of users; therefore, these reports focus on the results of the research and its practical implications.

The following HSIS summary reports are available by calling the FHWA Research and Technology Report Center at (301) 577-0906 or by downloading directly from the Turner-Fairbank Highway Research Center's Internet web site at www.tfhrc.gov:

  • Analysis of Commercial Bus Crashes (FHWA-RD-93-018).
  • Association of Median Width and Highway Accident Rates (FHWA-RD-93-046).
  • Accident Analysis of Older Drivers at Intersections (FHWA-RD-94-021).
  • Truck Accident Models (FHWA-RD-94-022).
  • Roadway Widths for Low-Traffic-Volume Roads (FHWA-RD-94-023).
  • The Magnitude and Severity of Passing Accidents on Two-Lane Rural Roads (FHWA-RD-94-068).
  • The Application of an Improved Accident Analysis Method for Highway Safety Evaluations (FHWA-RD-94-082).
  • Comparison of the Safety of Lighting Options on Urban Freeways (FHWA-RD-94-188).
  • Investigation of Crashes With Animals (FHWA-RD-94-196).
  • Analysis of Older Drivers on Freeways (FHWA-RD-96-035).
  • Investigation of Highway Work Zone Crashes (FHWA-RD-96-100).
  • Safety Effects of Cross-Section Design on Rural Multilane Highways (publication pending).

The development of HSIS has proven to be extremely valuable both to FHWA and the safety community. It provides the flexibility to analyze a wide range of safety issues from quick-turnaround problem-identification efforts to long-term, in-depth research studies. It serves as an example of the benefits that can accrue through cooperation between federal and state departments of transportation. FHWA and the safety community have benefitted through the willingness of the participating states to provide information and technical assistance. The states have also benefitted through a thorough review and quality control of their data by the HSIS staff, access to advancements in computer software, feedback from users of the data, and interaction with their peers in other HSIS states.

With the expansion of the number of states and diversity of data and the upgrading of the hardware and software tools, HSIS is now well-positioned for the future. The focus of additional development efforts will be on enhancing the data files available in the existing HSIS states. This will enable HSIS researchers to address a broader range of safety problems, increase the depth of analysis, continue to upgrade the computer hardware to take advantage of advances in technology, and improve the accessibility of the system.

HSIS Overhaul

The Old System

Prior to the overhaul, the sytem consisted of what were at the time considered to be high-end personal computers (PCs) that operated independently. In terms of data storage, all four-plus gigabytes of the original HSIS databases resided on the optical disks and were retrieved on an as-needed basis. Statistical Analysis System (SAS) served as both the database management system and the analytical and statistical package. For example, when a study was to be conducted, the analysts would copy the needed files from the optical disk and then write a SAS program that would sort, merge, and screen the files such that the needed subset of data would be created. These methods worked and met the analysts' needs; however, they were very time intensive. The size of the files to be merged (typically) containing 100,000 to 300,000 observations) also resulted in processing times of 30 minutes or more.

The New System

The system now consists of a typical client/server configuration, which allows remote access from the PC clients (the existing PCs from the old system) to the data residing on the server. Additions include all the server hardware and software, as well as upgrades to the client's central processing unit (CPU). The analyses can be initiated -- given the multiprocessing capabilities of the server -- it is possible to have the several analyses running at the same time from several clients. The features of this system include:

SERVER CLIENT (same a PC's above)
Hardware
IBM RS/6000 PowerServer 370
64 MB memory
11+ GB hard disk drive
5 GB external tape drive
Hardware
486/DX2, 66 MHz CPU
16 MB RAM
1.6+ GB hard disk drive
360 MB read/write optical disk drive
Software
Sybase SQL server
with PC open-client DB
and net library
Software (most recent version)
DOS
Windows 3.1
SAS

The Results

Under this new configuration, an analyst can use the best features of SAS and Sybase to access, query, and process the data. The table below provides a comparison of processing times between the old and new system. This shows how significant this overhaul has been in terms of reducing processing time, which, of course, causes less downtime for the analysts.

Database Procedure Processing Time
(in minutes)
Percent Reduction in
Processing Time
Old System New System
Merging the Accident and Roadway Files

(Accident file has 14 variables, 153,796 obs.; roadway file has 29 variables, 74,729 obs.)

39.2 13.5 66%
Merging the Accident and Vehicle Files

(Accident file has 14 variables, 153,796 obs.; vehicle file has 13 variables, 297,813 obs.)

48.6 13.8 72%

Jeffrey F. Paniati is chief of FHWA's Safety Design Division in the Office of Safety and Traffic Operations Research and Development.

Forrest M. Council is director of the University of North Carolina Highway Safety Research Center.

ResearchFHWA
FHWA
United States Department of Transportation - Federal Highway Administration