Cover Image

Not for Sale



View/Hide Left Panel
Click for next page ( 100


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 99
Potential Measures for Improving the Identification of Root Causes for Hazardous Materials Crashes 99 5.3 Improving the Effectiveness of All Databases Required to Identify Root Causes 5.3.1 Ensure Data Completeness and Accuracy Data completeness includes accidents that are not reported as well as accident reports with incomplete data. If the system reports only one-half of the accidents that occurred, then it may take twice as long to obtain a good understanding of how the system responds to a set of changing conditions. Even then, this may not be possible because those accidents that were unreported may not be representative of those that are reported. For example, accident sever- ity, driver, and/or carrier characteristics might differ. Typically, less severe accidents are underreported. This is true whether the cases are self-reported or not. In addition, carriers with poor records are less likely to self-report. This means less safe carriers are less likely to be identified in a timely fashion. Incomplete reports can also have a negative effect on accident analysis since a complete record of cause and effect is not captured. Moreover, when one parameter is not filled out for one accident record and a different parameter is missing in another accident record, the sit- uation is worse than if the same parameter value is not filled in for each accident. When the negative consequences of underreported accidents and incomplete reporting are combined, the ability of an analyst to draw conclusions from the data is significantly compromised. For example, for semitrailer truck accidents with hazmat releases, if only 70% of the accidents are reported and, for those that are reported, only 75% of the truck configurations are known and speed at the time of a crash is populated only 50% of the time, it is virtually impossible to have any confidence in an estimate of the annual number of semitrailer truck hazmat releases in which excess speed is a contributing cause. This is because there would be com- plete data for only about 25% of the accidents that occurred (0.70 0.75 0.50 = 0.2625). Moreover, there is no assurance that every accident with complete information is represen- tative of the three that are missing. 5.3.2 Complete Values for All Parameters No credible information system can operate using records in which many fields are blank. Parameters are specified in a database because they can be collected and are important to collect. Seemingly unimportant parameters are essential for specifying the causal chain of events or char- acterizing the circumstances of the event. For any database incorporated into the proposed information system, required fields should not be left blank. In some instances, it may be necessary to provide an estimate of the appro- priate data entry. Some databases allow for this, but maintain a code to distinguish between measured and estimated values. This type of information is valuable for parameters that change as a result of the accident sequence, such as speed. Although the code for "unknown" will always be an option, it should be used only if it is truly impossible to estimate a value for one of the required parameters. Often, an estimate is good enough, since categories are gen- erally aggregated for analysis. Techniques would need to be implemented for ensuring that information required for the data fields is both collected and correctly entered into the database. For example, the overwhelm- ing majority of truck carriers surveyed believe that PHMSA should contact carriers to collect data for fields that are incomplete. Some techniques that could be implemented by database man- agers include the following: Computerize data collection to enforce and validate data coding. Create required fields that will not accept the record until those fields are complete.

OCR for page 99
100 Hazardous Materials Transportation Incident Data for Root Cause Analysis Develop incentives that will reward those that provide complete and accurate data into a data- base. For example, for states, a system could be developed to provide feedback in the form of crash data for the state. Or for carriers, data could be provided that defines how a company compares to its competitors. 5.3.2.1 Add Latitude and Longitude Add latitude and longitude to all databases to provide the exact location of a hazmat crash and enable the identification of the same accident when it appears in more than one database. This would facilitate linkages between databases where information on the same crash is com- pared and also would aid in obtaining site-specific information associated with a particular location. 5.3.2.2 Add a Specific Description of the Hazmat A specific hazmat description could be added to the database and would include the type or hazmat class identified on the placard and/or shipping papers, name of the chemical being trans- ported, quantity being transported, and--if there is a spill--the quantity spilled. Adding this information will provide significantly more detail on the hazmat being shipped and make the database more valuable for identifying root causes of hazmat crashes. 5.3.2.3 Electronic Submission of All Crash Reports to the Major Databases Using the Web to Facilitate Accuracy All crash reports could be submitted electronically using the web to facilitate accuracy. Cur- rently, information is submitted from the states using the SafetyNet System. Although this approach is undoubtedly fixed for the near future, the electronic submission of the complete police accident reports (PARs), including truck and hazmat supplements, to the states could reduce the opportunity for error when the information is entered into SafetyNet. Similarly, if HMIRS reports were submitted using a web interface, an added source of errors would be removed by eliminating one data transfer step. In addition, an automated system could be devel- oped to check the forms for completeness and identify incorrect coding. 5.3.2.4 Add "Error Trapping" "Error trapping" could be applied to all databases in order to eliminate errors by applying a program that would test whether certain logical connections have been made within the same accident record. One example of an error that could be eliminated is a situation where a travel speed of 40 mph is listed for a crash associated with a truck backing into a loading dock. Another example of a "logical inconsistency" could involve an accident where a single-truck cargo tank is reported to have lost 20,000 gallons of a hazmat liquid. The error-trapping pro- gram would flag this spill size as erroneous because it is inconsistent with realistic truck cargo tank capacities. 5.3.2.5 Increase State Checks of the Quality of Hazmat Crash Submittals to FMCSA As part of their reporting obligation, each state could quality-check reports for all hazmat crashes in MCMIS. States could read the reports, complete missing fields, and make any needed corrections before submittal to FMCSA. This step would increase the reliability of the MCMIS Crash file. As the use of electronic submissions becomes the norm at all levels, states could be encouraged to develop translational programs that automatically populate the fields in the MCMIS Crash file from the electronic entries in the PARs, thereby eliminat- ing translation errors and the need to check for such translation errors between the PAR and the MCMIS Crash file. States with demonstrated higher quality submittals might be rewarded.

OCR for page 99
Potential Measures for Improving the Identification of Root Causes for Hazardous Materials Crashes 101 5.3.2.6 Ensure that the Databases' MCMIS Report Number for Serious Hazmat Crashes Can Be Linked to the PAR Agencies checking the quality of MCMIS crash reports should be able to link the PARs with the MCMIS report. Including the PAR number in all MCMIS hazmat crash reports would allow the PAR associated with each crash to be easily identified and quickly retrieved. Currently, in many states this is a time-consuming manual process that could be easily remedied if the PAR number was referenced in the MCMIS Crash file and the PAR was electronically retrievable. Linking the MCMIS report number to the PAR could be a suggestion for police reports and require the following: States could be encouraged to automate collecting standard administrative data as much as possible since MCMIS data are extracted from police reports. Include non-spill hazmat crashes in HMIRS for all "serious crashes" involving placarded shipments. Include digital photos of the accident scene in all of the major databases. For truck carriers in MCMIS and HMIRS, include carrier out-of-service (OOS) rate per mil- lion miles. MCMIS inspection data could be able to determine the count of OOS vehicles for a carrier. Could then be normalized by the VMT estimate in the Census file. For HMIRS and MCMIS, include violations for drivers during the crash. Involve other parties such as emergency responders and insurance companies in data collec- tion and reporting. In HMIRS, the IEVENT Table could be used to capture remarks made by multiple individuals. 5.3.2.7 Include All Applicable Hazmat Accidents Each database should include a process for ensuring that all hazmat accidents that meet the specifications for inclusion are recorded. In order to accomplish this objective, agencies manag- ing these databases could consider the following: Conduct training for carriers and police in how to complete accident reports. For truck acci- dents reported to MCMIS, because there are tens of thousands of different police officers fill- ing out PARs each year and individuals from more than 50 states and territories filing crash reports, Web-based training is likely to be the most cost-effective method of delivering such training. Although the situation is simpler for other modes (e.g., rail and barge) because fewer individuals are responsible for compiling and reporting crash data, Web-based training might be cost effective for these individuals as well. Develop more efficient systems for entering data at the accident scene, such as greater depen- dence on electronic data entry and use of bar codes. Review news reports of hazmat accidents to identify those that should have been included in the database. Investigate the techniques used by financial news reporting services to automatically flag newsworthy statements made by company officials and reported in news reports. Check PARs to identify those accidents that should be included. For truck transport, this would be greatly facilitated if PARs were available electronically, thereby making the identifi- cation of reportable crashes more accurate and uniform. Provide incentives to states to verify that all hazmat truck crashes that should be reported have been included in crash databases. One suggested project would be to establish a multi-agency task force to identify all hazmat accidents that meet the reporting threshold, considering all modes of transport. The project would perform the searches listed above and compare the results with HMIRS, FARS, RAIRS, and MCMIS, as appropriate for the mode. The outcome would come close to being a census of hazmat accidents. This census could then be used by HMIRS, FARS, MCMIS, and RAIRS as a

OCR for page 99
102 Hazardous Materials Transportation Incident Data for Root Cause Analysis measuring tool to determine their respective completeness. It could also provide the sampling frame for in-depth investigations as suggested in this subsection. 5.3.2.8 Ensure Accuracy and Consistency of All Data Entered into the Databases Reporting incorrect data can make it impossible to identify causal relationships, even with an infinite amount of data. Thousands of vehicles might experience the same initiating event and correct for it without incident. The next vehicle might experience the same initiating event and, through a series of unique conditions, be unable to correct for the initiating event. If these unique conditions are not accurately recorded, then the true cause(s) will never be identified. Potential measures for ensuring data accuracy fall into the following categories: 1. Include parameters that are directly measurable such as how far from the roadway a vehi- cle was after a crash or parameters that are directly observable, such as whether it was dark or light. Such parameters should be recorded with high accuracy. 2. Include event data recorders on hazmat transporters. Hazmat vehicles are a good test bed for these devices. Knowing vehicle parameters for the 5 seconds prior to a crash would be invaluable. Just knowing the vehicle speed, when the brakes were applied, and with what force the brakes were applied would be insightful. 3. Record parameters that occurred before or during an accident and could be objectively described but are no longer apparent to the accident reporter. An example of this type of parameter could be an accident that occurred as a result of a white-out condition that is no longer present at the accident scene. 4. Include parameters that are based on highly subjective witness statements. Police officers are highly trained to recognize unusual driver behavior and these subjective observations should be recorded. The identification of contributing causes falls into this class of recorded information. 5. Record parameters that become available some time after the accident. Investigating offi- cers will frequently require a driver submit to a drug test when impaired performance is sus- pected. Even though this information is not available for several weeks, such information can be obtained and is routinely captured in the TIFA database. Other accident databases could adopt the TIFA procedures for capturing this information. 5.3.2.9 Place All Components of the Proposed Information System Under a Quality Assurance Program A process should be implemented that provides periodic reports stating the accuracy of the data entry process. The program could include the following measures: 1. A consistent data dictionary that provides accurate definitions of the parameters such that two individuals, when assessing the same accident, would fill in the same cause codes (where multiple interpretations are possible, the question/answer format adopted in RAIRS provides an effective means to ensure more consistent reporting); 2. Consistent PAR forms that contain the minimum amount of data required to identify root causes; 3. Training for individuals responsible for completing PARs or reports for HMIRS and MCMIS. As stated previously, since an individual might only need to fill out a report once every few years, Web-based training that could be accessed to assist the reporting official might be a very cost-effective training tool; and 4. A data checking procedure to ensure quality control of information using such techniques as: Checking a sample of all data submitted and entered into the database, Making telephone calls to participants for a sample of accidents, Comparing reports with press accounts of hazmat crashes, and Carefully reviewing the database to identify inaccuracies and inconsistencies.