Role of Human Factors in Preventing Cargo Tank Truck Rollovers (2012)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

7 The study of the causes of prior cargo tank truck rollovers established a framework to support the later activities in this project. Given the number of heavy truck rollover crashes expected for the 2007–2009 timeframe, the standard detailed- level root cause analysis is unfeasible. At a more practical level, typical crash data reports lack the level of detail nor- mally required to provide clear answers about the full range of causal relationships. For these reasons, our basic approach is to conduct a higher-level survey of rollover crashes with the objective of defining the broader problem space that can be used to identify and evaluate best practices in the cargo tank truck industry. For comparison purposes, non-cargo tank truck rollover crashes are also studied. The approach was segmented into four activities: 1. Identify crash data set, 2. Develop analysis framework, 3. Conduct root factor analysis, and 4. Summarize findings from root factor analysis. 2.1 The Crash Data Set The first activity in Task 1 is to identify crash data sets for driver-related root factor evaluation. The research team has reviewed U.S. cargo tank truck rollover crash data from a vari- ety of sources to identify potential driver-related root factors. The research includes a review of HMCRP Report 1: Hazard- ous Materials Transportation Incident Data for Root Cause Analysis, and a detailed review of the key databases identified in that report: TIFA, HMIRS, and MCMIS. Using those data- bases, the team identified 400 individual cargo tank truck roll- over crashes worthy of review and examined PARs for them. The MCMIS and HMIRS databases did not contain a level of detail sufficient to perform an effective analysis of driver- related factors. The TIFA database, while containing good data, did not have data on a sufficient number of cargo tank truck rollover incidents to stand alone as the project’s data source for root factors. To gain sufficient knowledge about driver-related factors, the research team adopted a revised approach to iden- tify additional sources of information. The combined Tasks 1 and 2 approach is shown in Figure 1 (see Chapter 1). The project team sought incident analyses from large tank truck carriers. In some cases, carriers perform a true root cause analysis, delving much more deeply into the events leading up to an incident and the reasons than do the crash databases. (TIFA is unique in that researchers contact persons involved for extra detail including, for example, the results of drug tests and not simply that a drug test was performed.) Internal car- rier information is sensitive, however, and arrangements to access this data source could not be made. 2.1.1 TIFA—Trucks Involved in Fatal Accidents TIFA is maintained by UMTRI for FMCSA. It is a census file on the fatal accident experience of medium and heavy trucks nationwide and is essential to any evaluation of truck safety issues. The database compilation begins from the files in the Fatality Analysis Reporting System (FARS) database. Information on any fatal truck accident in FARS is extracted and then enhanced by UMTRI by calling the carriers, medical institutions, and law enforcement organizations to confirm information reported in FARS and in PARs generated as a result of the accident. UMTRI extensively supplements the information obtained from FARS using additional fields in the TIFA database. TIFA is the only accident database admin- istered by the government that follows up on the drug and alcohol tests administered and records the results of these tests. While the TIFA identifies only fatal crashes, it is a good source for information to identify potential causes. TIFA provided some indications of driver-related factors that con- tributed to the fatal crash. The research team conducted a thorough analysis of crash records in TIFA from 2006 to 2008. While the focus of this C h a p t e r 2 Root Causes of Cargo Tank Truck Rollovers

8was developed to represent operating characteristics (see Table 1). The PARs were requested from eight states (see Table 2) based upon the database showing sufficient numbers of inci- dents representing the characteristics in Table 1 and known quality of reporting. All states responded with nearly 100% of the requested records. Some states employed unique code- books for interpretation of the reports. 2.2 Analysis Framework An important prerequisite for conducting this type of root cause analysis is to develop a consistent framework for iden- tifying and classifying relevant crash factors. Comparing the available information with an existing framework of crash factors can facilitate identifying the role that each factor may have played in the crash and other contributing factors that logically would have been present, yet may not have been included in the report. The two primary sources of informa- tion are the PARs and TIFA. Similar analysis frameworks were developed for each, but are not identical. This has been done intentionally to mine as much information as possible from the separate analyses. Developing a framework that fit both sources would have minimized the result. The amount of detail provided from each of these sources does not allow for identification of corporate and organizational factors. assessment is cargo tank truck crashes, which represent a sub- set of the crashes reported in TIFA, it is useful to analyze a broader set of reports for comparison. Parameters that might contain driver factors were identified and the database was then queried to identify the driver factors at three levels: 1. All fatal truck crashes for four vehicle configurations— trucks with three or more axles, trucks with a trailer, trac- tors and semitrailers, and doubles; 2. A subset of fatal truck crashes involving cargo tank trucks; and 3. A subset of fatal truck crashes involving cargo tank trucks where a rollover occurred as part of the crash sequence. The analyses captured 6,570 records overall, with 599 records in the second level and 163 records in the third. The first case was analyzed to provide a set of data that might identify dif- ferences between fatal truck crashes and fatal cargo tank truck crashes. Rather than use the entire TIFA dataset for Case 1, by using the subset of vehicle configurations that contain cargo tank trucks, accurate differences might be identified. 2.1.2 HMIRS—Hazardous Material Information Resource System HMIRS is maintained by PHMSA and covers all reportable hazardous material incidents in the United States as desig- nated in Section 171.16, 49 CFR. Changes to the structure of the HMIRS database in 2005 have made it more difficult to identify rollovers. The database was analyzed for indicators of driver-related contributing factors, but was found unsuit- able for that purpose. HMIRS was used to identify hazardous material incidents when combined with the MCMIS database for the purposes of selecting a sample set of PARs. 2.1.3 MCMIS—Motor Carrier Management Information System MCMIS is maintained by FMCSA and contains informa- tion on the safety fitness of commercial motor carriers (truck and bus) and hazardous material shippers subject to Federal Motor Carrier Safety Regulations (FMCSR) and the Hazard- ous Materials Regulations. MCMIS was felt to be particularly useful to identify rollover events. It contained sufficient fields from which to gather a representative sample of more than 400 incidents involving potential driver-related factors. 2.1.4 PARs—Police Accident Reports Police reports on over 400 individual rollover crashes were requested from selected state reporting agencies. In order to obtain a thorough sample set of crash incident data, a list Table 1. Characteristics used in PAR selection. Hazardous materials Region & state Vehicle configuration No. vehicles involved Weather Trafficway Accessway Table 2. PAR requests from states. State Number Requested Colorado 30 Florida 15 Louisiana 77 New York 53 Oklahoma 5 0 Pennsylvania 50 Texas 80 Virginia 57 TOTAL 412 Recei ve d 407

9 Researchers have analyzed 407 PARs against the frame- work, reviewing both data and narrative fields contained in the reports. The assignment of reports has been overlapped so that the results of each researcher can be compared and reviewed with senior project researchers to ensure consis- tency of analysis. Likely contributing factors are correlated to the unsafe driver acts based upon the report of the investi- gating officer. The researchers have sought not to instill their own opinions or guesses on what the contributing factors may have been; rather, they have interpreted and recorded what the officer documented in the report. The PAR analy- sis focuses on driver behaviors; therefore, for the records in which contributing factors related solely to the vehicle or to the environment (and not to the driver) or in which no clear driver-related contributing factors were identified, the con- tributing factors are listed as “none specified” in the summary table. If a driver-related factor accompanied a vehicle-related or environment-related factor, then the record is included in the summary table and classified under the appropriate driver-related contributing factor. In all, 26% of the records analyzed have been classified. The remainder could not be classified as the reports did not contain sufficient informa- tion to clearly determine factors. The TIFA analysis evaluates several driver-related factors: age, speeding, hours driven prior to accident, overall health, alcohol involvement, drug involvement, avoidance maneu- ver, violations, and other driver-related factors. Each of these factors will be discussed in the following section. The conceptual framework for PAR analysis has been developed as a matrix that matches identified unsafe driver acts to contributing factors. Unsafe driver acts include • Driving too fast for conditions, • Following too closely, • Illegal maneuvering or improper turning, • Failing to signal, • Inadequate evasive action, • Panic or freezing, • Overcompensating, • Poor directional control, • Failing to heed, and • Unknown reason. Contributing factors and their defining characteristics are shown in Table 3. Table 3. Definition of contributing factors in PAR reviews. Contributing Factors Defining Characteristics Personal Factors Training Years of experience Driver age Ph ys iological Factor s Driver health (heart attack or other physical impairment) Visual capabilities [visual acuity, useful field of view (UFOV)] Cognitive abilities (decisionmaking, information processing) Strength Fitness to drive Attitudinal Factors Attitudes toward safety Moderate or severe crash history Driving habits Information Gathering Distraction (internal or external) Poor situation awareness Failure to recognize hazard Inadequate visual surveillance Driv er State Impaired (alcohol or medications) Aggressive Drowsy Asleep Capacity limited Organizational Factors Stop work (vehicle condition) Always swerve to avoid collision Get the work out Productivity incentives Onboard computer (OBC) monitoring Coaching and positive reinforcement Family education to support the driver getting proper rest and nourishment Pre-shift screening of “fitness for duty” Hours of Service regulations Second jobs not always monitored Multiple paper logs

10 2.3 Summary of Findings from the Root Cause Analysis 2.3.1 PAR Findings Table 4 shows the results from the PAR reviews and indi- cates which contributing factors are associated with the spe- cific unsafe driver acts across all crash reports. The unsafe acts that are most frequently identified are driving too fast for conditions, illegal maneuvering or improper turning, inad- equate evasive action, and poor directional control. In each of these unsafe acts, information gathering is identified as the chief contributing factor (7 of 11 for unsafe acts); 17 of 19 for illegal maneuvering or improper turning; 8 of 12 for inad- equate evasive action; and 36 out of 55 for poor directional control. In all, information gathering accounts for 72% of identified contributing factors, followed by driver state, which accounts for 19% of identified contributing factors. Information gathering includes such characteristics as distrac- tion, poor situational awareness, failure to recognize a hazard, and inadequate visual surveillance—in short, instances of not paying attention. Driver state includes such characteristics as impairment (e.g., alcohol, drugs, or medications), aggressive behavior, drowsiness, being asleep, or having limited capacity— in short, not being fit for duty or in the proper condition or state of mind at the time of the crash. Of course, there are numer- ous contributing factors, and the accident reports do not pro- vide any further details to uncover further root factors such as training, fitness for duty, effectiveness of training, fatigue, and so forth. The research will show that motor carriers and others need to successfully employ a range of good practices to reduce the likelihood of their drivers finding themselves in a harmful or fatal situation as a result of not paying attention. In fact, even the best drivers will attest to finding themselves in such situ- ations, but were simply lucky enough that they did not have harmful or fatal results. In addition to the summary Table 4, a report annotation table that holds key descriptive information from the narrative data field is provided online in Appendix A. 2.3.2 TIFA Findings TIFA again shows that driver-related factors are significant contributors to fatal cargo tank truck crashes. The 3 years of Unsafe Driver Acts Contributing Factors Personal Physio-logical Attitudinal Driver State Organiza- tional Info Gathering Subtotal Specified None Specified Row Total Too fast for conditions – unsafe speed – uncontrolled speed – turning too fast 4 7 11 112 123 Too slow for traffic stream 0 Following too closely – sudden slow or stop 5 5 False assumption of other road user's actions 2 2 Illegal maneuver or improper turning – other improper driving action – turned when unsafe – wrong side – fail to yield 1 1 17 19 37 56 Failure to turn on head lamps – turning signal 2 2 Inadequate evasive action 2 2 8 12 17 29 Panic or freezing Overcompensation 1 6 7 22 29 Poor directional control (careless driving) – drifting – passing – veering – parking 1 5 13 36 55 93 148 Failed to take heed to signage – road signs – yield signs – traffic lights 3 3 7 10 Unknown 3 3 Total for each contributing factor 2 8 0 20 0 77 107 300 407 Table 4. Contributing factors associated with the unsafe driver acts identified across crash reports.

11 TIFA data made it possible to analyze multiple- and single- vehicle fatal truck crashes separately and obtain some signifi- cant findings. Driver-related factors are much more likely to be associated with single-vehicle fatal truck crashes than they are in multiple-vehicle crashes involving trucks. The analyses clearly showed that driver factors such as driving too fast or failure to control the rig (e.g., over correcting) were impor- tant contributors. Physical or mental condition of the driver was also found to be important. The analyses clearly showed that use of alcohol and the use of drugs (whether taken legally or illegally) are associated with single-vehicle cargo tank truck rollover fatal crashes more often than in multiple- vehicle accidents involving trucks. Similarly, extremely obese drivers—perhaps an indication of being more prone to sleep apnea and, therefore, sleep deprivation—were also more fre- quently associated with these single-vehicle cargo tank truck rollover fatal crashes. Clearly there are preventable driver fac- tors that are significant contributors such as driving too fast for conditions. The analyses of the TIFA data show that there is room for improved performance through effective driver training and safety programs. Many more details came from the TIFA analysis than are presented here (see the tables and discussion in Appendix B online). The PAR and TIFA analyses used different contributing factor definitions, due in part to the different information present. The PAR analysis contributing factors were shown in Table 3. The relationships between the two are shown below (see Table 5). 2.3.3 Summary of Key Findings The separate analyses of TIFA and PARs did yield correla- tions in potential driver-related root factors. The sources of information do not yield enough to identify absolutely and conclusively the root factors. This would require the type of detailed analysis performed by insurance companies and carriers following major crashes, or the effort that was con- ducted for the Large Truck Crash Causation Study (LTCCS) (FMCSA, 2006). Data available from MCMIS, TIFA, HMIRS, and PARs are not sufficiently detailed to conclusively determine driver- related influences. Crash data and accident reports focus more on what happened at the time of the crash and the immediate factors. Combining likely contributing factors, the expertise of the research team, and lessons learned from the interviews, the team has constructed a table of possible influences to the key critical factors. Using the analysis framework, significant areas of potential driver-related contributing factors include the following: • Driver state, • Physiological condition, • Information gathering, • Obesity and health, • Alcohol or drug involvement, and • Vehicle control. A number of these areas relate to, or contribute to, the others. Certainly any of the first five areas can result in poor vehicle control, as well as alcohol or drug involvement being considered a characteristic of driver state. Driver state, in turn, can be a factor in—but not the sole causal factor of— information gathering. Complete information for a thorough root cause analysis is best obtained by thorough investigation. It remains cost- prohibitive to conduct such analysis under the public sector purview for each rollover. Carriers and insurance companies hold the most complete set of information for this analysis, but business reasons prohibit their information being released into the public domain. A process that would allow for root PAR Analy sis Contributing Factors TIFA Driv er Factors Personal Age, race not listed as a driver factor, compiled elsewhere Physiological Physical or mental condition Attitudinal Operating the vehicle in careless or inattentive manner Aggressive driving or road rage Driver State Physical or mental condition Organizational Not considered Info gathering Possible distractions within vehicle Vehicle Blown tire listed under skidding and sliding, brake failure not tied to any driver factor Environment Skidding, swerving, and sliding, and also visual obstructions Table 5. Comparison of PAR and TIFA contributing factors.

12 causes at an aggregate level to be obtained, that would allow for valuable lessons to be shared to improve safety across the industry, and that would provide legal protection and ensure confidentiality to those providing the data is likely the most effective solution to root cause identification of driver-related factors in cargo tank truck rollovers. The analysis did show that it might be worthwhile to study one subset of cargo tank truck rollover crashes: those involving single vehicles. Several of the driver factors associated with these crashes are more prevalent compared with multiple-vehicle crashes. Investigat- ing single-vehicle cargo tank truck rollover crashes would also have the additional advantage that the root and contributing causes of these crashes are more likely to be associated with the driver of the truck and not the driver of another vehicle shar- ing the roadway. Identifying ways to lessen the role of driver factors in single-vehicle cargo tank truck rollover crashes might translate well to improving highway safety and reduc- ing the incidence of truck crashes involving cargo tank trucks.