Cover Image

Not for Sale



View/Hide Left Panel
Click for next page ( 9


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 8
8 SECTION II Data Types Used in Preparing the Safety Plan There are various types of data that can be used in the database custodian differs from state to state, but is usually preparation of a safety plan. The procedures described in this either the State Police (or State Highway Patrol), the State guide are designed such that crash data are required as a min- Department of Revenue or Motor Vehicles, or the State imum. For the higher-order procedures, the crash data must Department of Transportation. In addition, the safety be "location-coded," and for some procedures, additional engineering staff within the state highways department (often roadway inventory and traffic volume data are required. Thus, the Traffic Engineering Branch) is a major user of the state as described in Section III, the choice of procedure will depend crash files and can often provide assistance and information on the data available in the user's jurisdiction. concerning how the local jurisdiction can obtain data. In There are other types of safety data that can also be used in the many cases, this staff is also responsible for assigning location development of safety plans. These would include information codes to the crashes and may maintain a separate crash data- ranging from driver citation/conviction data to observational base with more complete location coding. If the analyst is surveys of occupant restraint usage. The following provides a looking for crash data that have gone through a location vali- brief description of the major types of data that might be used, dation process at the state level, it is often necessary to contact and some information on where they might be found if not in these staff. Note that in most cases, the location-validation the user's own jurisdiction. Some of the following information process will only be conducted for state-system roads (i.e., was taken from NCHRP Report 501 (18). Interstate, U.S., and state routes) in local jurisdictions, and not for all local streets and roads. However, the continuing development and refinement of spatial data systems in states Crash Data and Related Files (e.g., GIS systems) is increasing the ability to locate crashes to all roads in the state. Local and Statewide Crash Data As noted in NCHRP Report 501 (18), although these Records on traffic crashes are derived from the police archives contain a wealth of information on the driver, the report form that is usually completed by investigating police vehicle, and the circumstances of the crash, some caution is officers at the crash sites. A typical crash report contains data warranted. It is critical for the analyst to understand the on about 100 different pieces of information that describe the boundaries and shortcomings in the crash database before crash, the location, and the people and vehicles involved. using these data to support decision-making. Every state has Crash reports may be used individually to explore the a reporting threshold usually a dollar amount of damage or circumstances and factors that contributed to a particular a specific level of injury sustained in the crash below which event, and they may be used in aggregate to develop a picture a crash report will not be entered into the statewide database. of the safety performance of a given location or jurisdiction. Local jurisdictions, however, may want to include all crashes At the local level, the analyst will generally use police- in their automated systems, not just those resulting in reported crash data from his/her own files. However, some damages or injuries above the state reporting threshold. Con- localities that investigate crashes may not retain their own files versely, some local jurisdictions may use a higher threshold or may not automate their paper files. In these cases, the ana- such that local police respond to fewer crashes. The statewide lyst should contact the state agency that serves as the custodian crash database may also contain additional crashes for a for the statewide crash database and request copies of the jurisdiction if more than one law enforcement agency reports computerized data for their jurisdiction. The statewide crash (e.g., the local police department, the county sheriff's office,

OCR for page 8
9 and the state highway patrol), with the possibility of each how supervisors review each crash report before it is finally agency using a different crash reporting threshold. submitted. When using data from more than one source, In several states, crash reports are collected from the such as when combining data from the state police, the local drivers involved in a crash (i.e., an operator report). These sheriff's office, and the municipal police department, or when operator reports may take the place of police-reported data comparing crash experience among several jurisdictions, the under certain conditions; for example, below-threshold differences in reporting practices among the various agencies crashes, crashes in which no one was seriously injured or can be critical. In these cases, the analyst should plan to con- killed, and weather- or animal-related crashes during peak duct validation tests as part of the overall analytic process. For seasons. In other states, operator reports and officer reports example, the analyst might compare the proportions of injury are blended in the final statewide crash database, thus affect- and property-damage-only crashes or the distributions of ing the overall consistency of data from one record to the property damage amounts in the data received from various next. Moreover, the NHTSA reports, ". . . various sources agencies. If that is not possible, the analyst should learn about suggest that about half of motor vehicle crashes in the coun- the thresholds and reporting practices of each agency as a try are not reported to police . . ." (19, p. 5). minimum. Crash data depend heavily on the subjective judgments of Note that these warnings are not to indicate that police the officers who attempt to describe the crash after the fact. data are of questionable value in safety planning, but only to These judgments cannot be error-free since the officer does alert the analyst of issues that can affect conclusions. Police not see the crash occur, and must rely on physical evidence data provide large samples of data even in local jurisdictions. and driver and witness statements to draw conclusions. Moreover, even with the known problems, these data have Moreover, officers are not experts in some areas that may be been used successfully for years in identifying safety prob- of crucial interest to a highway safety analyst, but which the lems, in choosing and targeting treatments, and in evaluating officer is called upon to record for the crash report. A com- the treatment effects. It is also worth noting that decisions parison of police-reported data to those collected by multi- made using the available crash data are going to be better than disciplinary crash-investigation teams on the same crashes decisions made without any data at all. indicate that the most reliable police data were those concerned with crash descriptors and the least reliable were Fatal Analysis Reporting System (FARS) Data driver/vehicle variables (20). The police were most accurate on six variables: location (note that most safety engineers in While most states maintain a database of all reportable highway departments question this finding), date, day of crashes that occur on all public roadways, there are frequently week, number of drivers, number of passengers, and number limitations to the crash data available for local roadways, of vehicles. The least reliable police data concerned vertical especially with respect to location coding and crashes involv- and horizontal road character, crash severity, road surface ing property damage only. In addition, some states and local composition, and speed limit. The accuracy of police judg- agencies may question the reliability of crash data for less ment of crash causes varies by the cause type, with more severe crashes, may not have a usable computerized file, or reliability in detecting human-direct causes than in detecting may wish to conduct at least some safety-related analyses vehicular, environmental, and human-indirect causes. How- using only fatal crashes. In these cases, the Fatality Analy- ever, the authors also noted that in the area of human-direct sis Reporting System (FARS) data system maintained by causes, the police performance was relatively good in identi- NHTSA is an excellent database for analysis. FARS contains fying "failure to yield" and "failure to stop" but was relatively annual data on a census of fatal traffic crashes within the poor with respect to "speeding," "driving left of center," and 50 states, the District of Columbia, and Puerto Rico. According "other improper turns." to NHTSA, ". . . to be included in FARS, a crash must involve Data accuracy and data completeness also vary from state a motor vehicle traveling on a trafficway customarily open to to state and among jurisdictions within a state. While it is the public, and must result in the death of an occupant of a generally agreed that state police and highway patrol officers vehicle or a non-motorist within 30 days of the crash." FARS provide more consistent and accurate crash data than their data are available annually back to 1975. Each new year's file local counterparts by virtue of their greater familiarity with is typically available about 6 months following the end of the crashes and, typically, more extensive training, few states year; however, the FARS analyst in each state has access to maintain the kind of quality-control measurement program his/her own state's fatal crashes at all times. FARS contains that would support interagency comparisons of accuracy or more than 100 data elements related to the driver, vehicle, in- completeness of crash report data. When using crash data volved persons, and the crash itself. FARS has proven to be a from a single source, such as a municipal police department, rich information source for research and program evaluation it is important to know that agency's reporting practices and focusing on fatal crashes. The quality of the FARS data is quite

OCR for page 8
10 high due to extensive training provided to the FARS analysts crash data are linked with detailed information on the med- and the attention paid to each case. ical consequences of the crash. Originally, seven states were To assist users in analyzing these data related to fatal funded by NHTSA to develop the CODES system (Hawaii, crashes, NHTSA provides a "FARS Web-Based Encyclope- Maine, Missouri, New York, Pennsylvania, Utah, and dia" at http://www-fars.nhtsa.dot.gov/Main/index.aspx that Wisconsin). Twenty-two other states have had CODES provides links to national reports and statistics. By clicking on demonstration grants or special projects (Alaska, Arizona, the "Create a Query" link, the user can choose her/his own Connecticut, Delaware, Georgia, Indiana, Iowa, Kentucky, state and run univariate tables and cross-tabulations for any Maryland, Massachusetts, Minnesota, Nebraska, Nevada, of the 100+data items. City and county codes make it possi- New Hampshire, New Mexico, North Dakota, Oklahoma, ble to isolate fatal crash data for local jurisdictions. (Note that Rhode Island, South Carolina, South Dakota, Tennessee, and the FARS data contain information on every vehicle and Texas). At a minimum, basic statewide police crash data are every occupant within each fatal crash, not just the fatally supplemented with hospital data and either Emergency injured occupant. Thus, care must be taken in choosing the Medical Services (EMS) data or emergency department data. screens to be used and in interpreting the results.) While Some states also add data for each driver and other occupants multi-year queries are no longer supported, the site contains concerning driver license status, vehicle registration, multi-year reports for individual states under the "States" and citation/conviction records, insurance claims, rehabilitation the "Trends" links, and analysts can run an analysis for and long-term care, and other items. The linkage of medical multiple years, one year at a time, to develop multi-year com- information to crash and driver data is done through parisons. FARS data can also be obtained from NHTSA (see "probabilistic linkage technology" since direct linkage is not "Request Data" link) on a CD or via download from an ftp often possible due to missing personal information and site (ftp://ftp.nhtsa.dot.gov/fars/). privacy concerns. It is noted that the reliance on fatal-only data has its CODES data are used in safety studies of specific injury- drawbacks in targeting treatments, with the main one being related issues such as seatbelt and motorcycle helmet sample size. In addition, targeting roadway treatments with fatal effectiveness any crash-related issue in which more detailed data is questionable since most factors that turn a crash into a injury data or injury cost data would be helpful. These data fatal crash are not roadway-related they are other factors such could also be used to study specific injuries occurring in road- as driver age or seatbelt use. It is often tempting to conduct side object crashes (e.g., head injury to right-front passengers analyses using only fatal crash data because the state and in guardrail impacts) and truck crashes (e.g., abdominal injury national targets all address the number and rate of deaths on for lap-belted truck drivers). Many states use their CODES data highways. It is strongly recommended, however, that crash fre- to develop a state-specific estimate of the economic impact of quency and rate at all levels of severity be used in safety analyses crashes that is based on the state's own data. to avoid the problems of small, unrepresentative samples. While CODES is funded by NHTSA, access is controlled by It is also recommended that if the analyst is not already famil- the states. More detail on CODES including links to some iar with FARS data, an effort be made to learn about the way the participating states may be obtained at http://www-nrd. records are created and the proper use of important data fields. nhtsa.dot.gov. Imputed Blood Alcohol Content (BAC) data is one example of a data field in FARS that contains data unlike that in most state Motor Carriers Management Information System or local crash databases. In FARS, BAC values are recorded from (MCMIS) the information supplied by police officers and an additional im- puted value is calculated for each driver for which a BAC value This database is a very comprehensive truck safety database was not supplied on the original crash report or cannot be that is the source of data for many of the Federal Motor Carrier obtained from follow-up contacts with local hospitals or med- Safety Administration's (FMCSA) other data files (e.g., ical examiners. Analyses of alcohol-involved crash frequency can SAFER) and analysis procedures (e.g., SafeStat). Data are obtain markedly different results depending on whether the entered into MCMIS via the SafetyNet system accessed by per- records or recorded-plus-imputed values are used. This is but sonnel in each state's Motor Carrier Safety Assistance Program one example of how use of FARS data may differ from the use of (MCSAP) agency. The database is maintained by the FMCSA more familiar local or statewide crash databases. to allow analysis of motor-carrier issues. MCMIS consists of five files, with input to each file from each state and from car- riers subject to federal truck safety regulations in all states: Crash Outcome Data Evaluation System (CODES) The Crash Outcome Data Evaluation System (CODES) is Registration ("Census") file Carrier information including an enhanced state-based crash data system in which police DOT numbers and descriptive information about a motor