PROPOSED REDUCTION OF CAR CRASH INJURIES THROUGH IMPROVED SMART RESTRAINT DEVELOPMENT TECHNOLOGIES Accident Analysis Methodology and Development of Injury Scenarios Report Reference: R & R5 Author(s): Richard Frampton (VSRC) Richard Morris (MIRA) Gabrielle Cross (MIRA) Marianne Page (VSRC) Date: January 006 Number of pages: 74 Number of appendices: 7 Number of figures: 4 Number of tables: 4 December 005 PRISM
December 005 PRISM
Table of Contents Report Overview / Summary...5 Introduction...7. Overview of Available Crash Injury Databases... 7. Levels of Detail Available in Crash Databases... 7. Database Limitations... 9.4 Data Sample Sizes... 0 Selection of Databases for Development of Injury Scenarios...4. Choice of Analysis Technique... 7. Real World Sample Criteria for Scenario Development... 7. Sample Relationship to the Population of Crashes... 8 4 Accident Data Characteristics...9 4. Drivers... 9 4. Front Seat Passengers... 0 5 Injury Impact Scenarios... 5. Injury Scenario Methodology... 5. Injury Scenario Results... 5. Injury Scenario Weighting... 5 6 Linking Injury Outcome Between Crash Scenarios and Simulation...7 6. Use of the Injury Severity Score... 7 6. Chest Injury Risk Predictor... 0 7 Conclusions... 7. Crash injury databases... 7. CCIS data characteristics... 7. Injury impact scenarios... 7.4 Linking simulation injury assessment to real world injuries... 8 References...5 9 Acknowledgements...7 0 Appendices...8 0. Appendix A: General Accident Information... 8 0. Appendix B: Vehicle, Occupant and Injury Specific Information... 4 0. Appendix C: Analysis of DGT Spanish Road Accidents Database (Report Reference R).... 46 0.4 Appendix D: Background to German In-depth Crash Injury Study (GIDAS)... 6 0.5 Appendix E: Case Details for Drivers with AIS + Head Injury... 69 0.6 Appendix F: Variables Related to Driver Femur Fractures... 70 0.7 Appendix F: Variables Related to AIS + Front Seat Passenger Chest Injury 7 December 005 PRISM
December 005 4 PRISM
Report Overview / Summary The main objectives of the PRISM project are to a) assess the injury reduction potential of SMART restraint technologies and b) provide guidelines on how to assess this technology with real or virtual testing protocol. By definition, SMART restraints are aimed at addressing the widest possible variety of situations, which arise in real crashes. The diversity of impact speeds, front-end overlaps, occupant sizes, ages and seating postures need to be defined in order that the restraint system knows how it needs to perform. Real-world accident data provides the necessary baseline for development of SMART systems and the means to assess their benefits. The contribution of accident data concerned; a) provision of information on crash conditions, injury outcome and injury mechanisms in production vehicles with current safety technology and b) quantification of potential injury reduction with the application of new technologies, using the current accident situation as a baseline. To provide a European perspective, it was necessary to draw on accident studies representing several European countries. Overall, accident data was to provide in-depth information on the crash conditions and vehicle/occupant variables related to crashes with serious injury outcome, that is, categorised as occupants receiving a Maximum Abbreviated Injury Score (MAIS) of at least (AAAM, 990). All this information was then to be categorised into a limited number of scenarios which could be replicated with computer simulation. SMART technologies would then be introduced into the simulations to gain knowledge of their injury reduction potential. Finally, these results would be fed back into the accident data in order to assess the potential benefits of fitting such SMART technologies. Several European databases were considered as donors of relevant crash injury data. On closer examination of those databases, the UK data was chosen because it fulfilled all the criteria needed for the PRISM analysis. It contained the level of detailed information required, it was considered representative of the general population of serious injury crashes and it had good availability of information. A sample of 75 drivers and 44 belted front passengers; all with serious/fatal injury was selected as the PRISM sample. All were in frontal crashes in cars equipped with airbags and pre-tensioned seatbelts. Of this sample, cases were selected which could be simulated using MADYMO and Human Body models. The final sample to undergo simulation consisted of 4 drivers and 5 front seat passengers. Due to the impracticality of simulating every case, it was necessary to group the crashes into a number of Scenarios. Detailed examination of each case allowed the grouping of each casualty into 0 injury scenarios. These were based on occupant kinematics and contact and load points associated with serious injury outcome to the head, neck, chest and thigh. A sophisticated and time-intensive case examination system was successfully developed, taking into account all aspects of the vehicle crash performance, occupant stature, and age, seating posture, kinematics and injuries. The scenarios were then developed as computer simulations into which SMART restraint systems could be incorporated. Each scenario was weighted according to its likelihood of occurrence in order to achieve accurate and realistic calculation of the magnitude of benefit likely to occur with the introduction of SMART restraint systems. The read-across from simulation output to real-world injuries necessitated careful consideration. The change in injury risk determined from the simulation was converted to a probable AIS value using an established technique. Chest injury outcome was considered separately for the average person and for weaker-boned occupants, by developing a set of December 005 5 PRISM
Combined Thoracic Index (CTI) risk curves based directly on cadaver data. Overall risk of mortality was assessed using the ISS. As explained in the PRISM R8 report, the ISS values derived from the simulations were based on only three of the standard five ISS body regions (PRISM ISS) since the injuries to the other two could not be assessed using the tools available. In order to validate the use of PRISM ISS as the metric for mortality assessment, overall ISS mortality risk curves were developed from,88 car occupants involved in crashes between 995 and 004 in the UK crash injury database. A comparison of standard ISS values versus the PRISM ISS for these showed no statistical difference in ISS scores. Overall it is concluded that the process of setting up crash injury data for simulation is viable, providing certain precautionary steps are taken. The necessity of having available detailed, easily available real world data was highlighted. It should be noted however that the complexities of real accidents still pose a considerable challenge for simulation. Some of the perceived requirements for similar studies in the future are highlighted in other PRISM reports (R9 and R0). December 005 6 PRISM
Introduction The overall objectives of the PRISM project are to a) assess the injury reduction potential of SMART restraint technologies and b) provide guidelines on how to assess this technology with real or virtual testing protocol. By definition, SMART restraints are aimed at addressing the widest possible variety of situations which arise in real crashes. The diversity of impact speeds, front end overlaps, occupant sizes, ages and seating postures need to be defined in order that the restraint system knows how it needs to perform. Real-world accident data provides the necessary baseline for development of SMART systems and the means to assess their benefits. The contribution of accident data concerned; a) provision of information on crash conditions, injury outcome and injury mechanisms in production vehicles with current safety technology and b) quantification of potential injury reduction with the application of new technologies, using the current accident situation as a baseline. To provide a European perspective, it was necessary to draw on accident studies representing several European countries. Overall, accident data was to provide in-depth information on the crash conditions and vehicle/occupant variables related to crashes with serious injury outcome that is, categorised as occupants receiving a Maximum Abbreviated Injury Score (MAIS) of at least (AAAM, 990). All this information was then to be categorised into a limited number of scenarios which could be replicated with computer simulation. SMART technologies would then be introduced into the simulations to gain knowledge of their injury reduction potential. Finally, these results would be fed back into the accident data in order to assess the potential benefits of fitting such SMART technologies. This report covers PRISM deliverables R and R5, Accident Analysis Methodology and Development of Accident Scenarios. The first part of the report deals with data preparation while the second describes the development of accident/injury scenarios for computer simulation. Reference material is presented in the form of appendices at the end of this report.. Overview of Available Crash Injury Databases The databases to be examined were defined in the PRISM project proposal. These were; the Fatals IDB (UK), the CCIS (UK), the DSD court database (Austria), the DGT accident database (Spain) and the GIDAS (Germany).. Levels of Detail Available in Crash Databases The initial step in data preparation was to identify what information was contained in each database and the level at which it represented the accident situation in each particular country. In this way, each database could be used according to its strengths. A template was drawn up to elicit the information. Tables and show the initial information requested for each database. Table describes the questions concerning general accident information. Table describes questions concerning data specific to the vehicles, occupants and injuries. It was initially agreed that information available from 997 onwards would be examined because of the need to commence analysis with relatively modern vehicles. The information contained in each of the databases is shown in appendices A and B. December 005 7 PRISM
DATABASE Name of database. Country. TYPE OF DATABASE National, in-depth or court cases? Who collects the data? Number of accidents (997 onward). FUNDING Who provides funding for the data collection? Table : General Accident Information DATA COLLECTION PROTOCOL Region in which data is collected How are cases chosen for investigation? What is the sampling procedure? What is the relationship of cases examined to the national population of accidents? What years of data are available? ACCIDENT INFORMATION Date and time Day Road type Road speed limit Weather conditions Road surface conditions Cause of accident Number of vehicles involved Types of vehicles (car, motorcycle etc) Bodystyle (eg: hatchback, estate) Impact type for each vehicle Object struck for each vehicle Occupant injury severity Occupant seating position Airbag fitted/deployed? December 005 8 PRISM
Table : Vehicle, Occupant and Injury Specific Information VEHICLE Make/model Age (registration year or year of manufacture) Crash Severity (Delta-V, EES ) Damage classification (collision deformation classification or other method) Number of impacts Severity of impacts (if more than ). If >, how is severity ranked according to occupant injury outcome or vehicle damage? Overlap calculated for frontal impacts? How is rollover classified? Intrusion how and where is this measured for frontal, side, rear and rollover crashes? Is the steering wheel movement measured and if so, in what directions? Object struck how are these classified? OCCUPANTS Seating position Seat track position Belt use Pretensioner (deployed/not deployed) Load limiter (deployed/not deployed) Airbag (deployed/not deployed) Type of airbag (steering wheel/facia/side head/side thorax) Age, Height, Weight, Gender Occupant loading from rear or side by other occupant/luggage INJURIES What is the scaling system used (AIS?) and if so, what version of the scale is used? How are the body regions defined? Can we get maximum injury severity for the cranium, face, neck, shoulder, arm, wrist/hand, chest, abdomen, pelvis, thigh, knee, leg, ankle, foot. In the case of limbs, can we get information for the left and right limbs separately? Are there descriptors and severity for each injury sustained by the occupant. That is, a detailed code identifying each individual injury? Is the cause of injury recorded and how many times is this known?. Database Limitations Each of the databases selected for study showed different characteristics. The DGT database provided a national picture of accidents in Spain but lacked detailed information on impact configuration, restraint fitment and performance and individual injuries (Appendix C). The IDB fatals database contained no cause of injury, no occupant height or weight and very few vehicle details. Additionally, most (90%) crashes in this database occurred between 990 and 995. Since the PRISM brief was to examine SMART restraint effectiveness from the starting point of current vehicles, due to the lack of occupant and vehicle information, it was decided to reject the IDB for further analysis. The DSD Austrian court data contained detailed accident reconstruction, occupant and vehicle information but no measure of how representative it was of the accident population in Austria. It was decided to refer to this database, only if necessary, to evaluate occupant kinematics and injury outcome at the Scenario generation phase of data analysis. CCIS and GIDAS databases were shown to contain in-depth vehicle, occupant and injury information together with methods to relate the crash samples to the population of accidents in the UK and Germany. As such, these databases held the most promise for developing the PRISM project. December 005 9 PRISM
.4 Data Sample Sizes As with any analysis of accident data, it is necessary to know how representative the data is of the population from which it is drawn. It is also important to examine datasets large enough to stand up to robust statistical analysis. The PRISM working group on accident data devised a way to broadly categorise crash types into those that would need different types of SMART solution. Tables to 5 show vehicle crash types against seatbelt use by occupants for GIDAS (German), CCIS (UK) and DGT (Spanish) data. The DGT was only able to classify 5 of the0 impact types due to limitations of the database (see Appendix C). Tables to 5 also show occupant injury severity level, based on police classifications of injury. Not injured is self-explanatory. Slight injuries involves cuts and bruises whereas serious injuries usually involve bone fracture and / or organ injury. Fatality is classed as death occurring within 0 days of the crash. December 005 0 PRISM
Table : GIDAS Data for Vehicle Model Years 997-00. Data Collection 999-00 Belted Driver OCCUPANT STATUS Unbelted Belted Unbelted Driver FSP FSP Belted RSP Unbelted RSP Occupants < yrs old IMPACT n n n n n n n Single Front head-on NI SER 59 68 84 46 8 8 7 65 5 5 8 0 5 9 7 5 6 9 Single Front oblique NI SER 67 4 4 8 7 8 7 7 4 4 0 9 9 8 9 4 4 9 4 Single Struck side perpendicular NI SER 6 6 6 8 5 6 7 Single Struck side oblique NI SER 7 70 0 6 9 5 0 40 5 7 5 6 Single Non struck side perpendicular NI SER 8 4 0 4 6 0 Single Non struck side oblique NI SER 8 75 8 4 9 4 4 0 7 4 Single Rear Roll only Roll + impact > impacts (no roll) NI SER NI SER NI SER NI SER 8 70 0 5 7 6 7 89 95 98 48 9 6 7 6 6 4 8 5 0 5 4 8 6 5 5 5 9 6 0 6 7 7 7 7 9 7 9 0 6 5 0 6 5 0 4 5 5 9 December 005 PRISM
Table 4: CCIS Data for Vehicle Model Years 997-00. Data Collection 997-00 Belted Driver OCCUPANT STATUS Belted FSP Unbelted Driver Unbelted FSP Belted RSP Unbelted RSP Occupants < yrs old IMPACT n n n n n n n Single Front head-on NI SER 8 750 4 5 60 44 5 47 0 77 6 6 9 4 0 7 6 5 5 8 Single Front oblique NI SER 76 60 8 5 8 9 0 9 75 8 6 7 6 7 0 8 Single Struck side perpendicular NI SER 8 5 6 6 7 5 4 4 Single Struck side oblique NI SER 6 45 6 9 4 9 0 4 0 Single Non struck side perpendicular NI SER 4 4 7 0 Single Non struck side oblique NI SER 40 09 7 7 6 4 4 4 8 4 Single Rear Roll only Roll + impact > impacts (no roll) NI SER NI SER NI SER NI SER 7 6 5 6 7 48 84 8 97 440 56 9 0 8 8 8 5 9 4 8 9 6 44 6 8 4 7 6 5 5 5 5 5 9 9 0 5 8 7 4 5 46 6 December 005 PRISM
Table 5: DGT National Data for Crashes 997-00 Belted Driver OCCUPANT STATUS Belted FSP Unbelted Driver Unbelted FSP Belted RSP Unbelted RSP Occupants < yrs old IMPACT n n n n n n n Single Front head-on NI SER 496 4866 45 6 64 58 8 65 84 0 9 8 4 89 9 5 4 409 6 7 5 58 4 67 8 4 44 4 Single Struck side perpendicular NI SER 94 695 460 497 09 48 8 40 66 5675 4 56 46 694 55 50 0 07 84 8 65 400 04 9 99 8 Single Struck side oblique NI SER 7977 47 57 6 65 6 6 0 0 490 45 8 4 0 9 49 76 5 4 88 87 5 5 40 5 Single Rear Roll Only NI SER NI SER 678 908 850 07 74 684 7 8 905 66 6 9 77 4 45 0 67 447 75 48 0 00 07 0 4 64 5 4 6 6 96 9 60 865 7 6 48 877 4 4 65 90 8 58 7 6 78 70 594 44 49 57 6 7 Upon initial examination, both CCIS and GIDAS appear to contain adequate numbers of data for analysis. The DGT data set is very large, as expected from a national data set. One of the benefits of national data is that it represents the whole population of crashes reported to the police. The DGT data was not however, able to classify all impact types. December 005 PRISM
Selection of Databases for Development of Injury Scenarios The development of SMART restraints from accident data requires detailed knowledge of the crash types, crash injuries and safety systems already fitted to each vehicle. Additionally, the baseline from which to develop SMART systems should consist of vehicles with current safety features added. CCIS and GIDAS data were examined for numbers of data concerning vehicles fitted with current safety features. At this point, DGT data was excluded because it could not provide detailed information on the vehicles safety features and crash injuries. Tables 6 and 7 show the numbers of cases available for analysis where full information on vehicle safety features was available in CCIS and GIDAS data. For a comprehensive description of The CCIS study methodology, the reader is referred to Mackay et al, 985. The methodology behind the GIDAS study is described in Appendix D. December 005 4 PRISM
Table 6: CCIS Data for Vehicle Model Years 997-00. Occupants with Airbag GROUP Drivers in frontal impacts with airbag + pretensioner MAIS Belted driver Unbelted driver Occupant Type Driver belt use NK Belted FSP Unbelted FSP FSP belt use NK Total MAIS 97 6 4 MAIS + alive 6 7 54 Fatal 7 9 6 Total 50 94 MAIS 6 7 44 FSP in frontal MAIS impacts with 6 4 + alive pretensioner only Fatal 4 5 Total 46 60 MAIS 0 8 8 Drivers in front MAIS + alive 8 4 oblique impacts with airbag + pretensioner FSP in front oblique impacts with pretensioner only Struck-side occupants in perpendicular impacts with side airbag Fatal 4 Total 4 54 MAIS 7 8 MAIS + alive 4 4 Fatal Total MAIS 4 MAIS + alive Fatal 4 Total 6 0 MAIS Struck-side MAIS occupants in 4 + alive oblique impacts with side airbag Fatal Total 6 0 Non-struck side MAIS 5 occupants in MAIS perpendicular + alive 6 0 impacts without Fatal 6 8 side airbag Total 4 4 MAIS Non-struck side 4 4 6 occupants in MAIS oblique side + alive 4 8 impacts without airbag Fatal 6 4 0 Total 4 8 7 44 December 005 5 PRISM
Table 7: GIDAS Data for Vehicle Model Years 997-00. Occupants with Airbag Belted Driver Unbelted Driver Occupant Type Belted FSP Unbelted FSP Belted RSP Unbelted RSP Occupants < yrs old Group n n n n n n n Single Front head-on 4 NI SER 8 4 5 7 5 Single Front oblique NI SER 99 6 8 5 8 7 5 Single Struck side perpendicular NI SER Single Struck side oblique NI SER 44 8 Single Non struck side perpendicular NI SER 0 6 Single Non struck side oblique NI SER 46 5 0 7 5 Single Rear Roll only Roll + impact > impacts (no roll) NI SER NI SER NI SER NI SER 59 6 8 0 59 68 9 4 4 0 4 8 4 0 6 7 0 5 9 4 6 4 4 6 December 005 6 PRISM
GIDAS data show very small numbers available once airbag and pretensioner fitment criteria were considered. For example, the number of seriously injured, belted drivers in single frontal crashes (head-on and oblique) fell from 55 to 0. In CCIS, only 44 equivalent cases were available for analysis. In effect, there were insufficient cases available in either GIDAS or CCIS to enable development of crash scenarios using traditional grouping methods based on vehicle crash descriptors. Additionally, numbers of side impacts with vehicles containing the latest structures and safety features were low in both CCIS and GIDAS. The PRISM sub-group on accident data therefore revised its methodology. Since there were insufficient side crashes it was decided to base the PRISM work on frontal crashes only. In order to increase the numbers of frontal crashes without including too many older vehicle designs it was agreed to extend the accident data analysis back to vehicles registered from 995 onwards. It was also decided to include the vehicles classed as multiple impact given that the majority of these will only have sustained one major impact (Lenard et al, 000) and to include them if that major impact was classed as frontal. Frontal crashes were to be chosen where belted drivers or front seat passengers had sustained any serious injury (MAIS +) since these are the injuries, which need to be addressed as the first order of priority.. Choice of Analysis Technique Since it was impractical to simulate every individual frontal crash resulting in serious injury outcome it was necessary to put crashes into groups or scenarios with a common theme related to injury outcome. A new methodology was devised for developing the crash injury data to be simulated. This was based on the premise that data numbers were too low to carry out a traditional analysis, grouping cases together by crash severity variables. Instead an injury based analysis was devised which included the benefit of classifying injuries by their causation. This method involved working out injury mechanisms by body regions and looking for causation trends. It was necessary to interrogate each case separately and the method was extremely time intensive necessitating examination of all the case details including photographs. In that respect, because CCIS was more accessible than GIDAS, it was decided to further develop the injury scenarios based on the UK data. Preparation of a new UK database to add data from 00 004 was undertaken. This, together with the inclusion of vehicle registration years 995-97 resulted in an almost doubling of case numbers over the original 997 00 sample. For example, the number of drivers with AIS + head injury rose from to 45 and those with AIS + chest injury rose from 5 to 86.. Real World Sample Criteria for Scenario Development CCIS in-depth crash injury data Passenger cars registered 995 onwards Crashes which occurred between 995 and 004 Most severe impact to vehicle front Equipped with airbag and pre-tensioner in the appropriate seat Belted driver or front passenger with AIS + injury. These cases, plus additional new variables and case photographs were put together in preparation for the generation of simulation scenarios. The variables presented were: case number, vehicle number, equivalent test speed, delta-v, airbag deployment, object hit, impact angle, percentage overlap, facia intrusion, steering wheel intrusion, make/model, registration year, occupant age, height and gender, body region injury, injury severity and December 005 7 PRISM
initial stated injury cause. Vehicles used were those manufactured from 995 onwards. An example of these starting point variables is shown in Appendix E for drivers with AIS + head injury.. Sample Relationship to the Population of Crashes In-depth crash injury data from the UK Co-operative Crash Injury Study (CCIS) was used to develop the injury scenarios and to provide information for crash simulations. The CCIS study selects cases for investigation using a stratified random sampling procedure based on injury severity (Mackay et al, 985). CCIS accident sampling gives a bias toward serious injury crashes. It examines about 80% of serious injury crashes and all fatalities, which occur in towed-away cars, less than 7 years old, in accidents occurring in the CCIS sample regions. The sample regions contain a mixture of urban and rural roads. One of the assumptions of the PRISM analysis is therefore that the crash analyses represent the majority of situations where belted front seat occupants currently receive serious injury in frontal crashes. December 005 8 PRISM
4 Accident Data Characteristics This section describes the CCIS crash injury data used to develop the injury scenarios. The study assesses injury outcome using the Abbreviated Injury Scale (AAAM, 990). The Equivalent Test Speed (ETS) was used as a measure of crash severity. ETS is the vehicle delta-v, calculated on the assumption that deformation was caused by impact with a rigid barrier. The calculation assumes the force was directed through the centre of the crush area. It does not assume the vehicle was brought to rest. ETS is different to the Equivalent Energy Speed (EES) used in other in-depth studies because the EES calculation assumes the force to be through the vehicle centre of mass and that the vehicle was brought to rest. ETS is therefore always less than or equal to EES. There are a number of factors which affect the accuracy of ETS so it is best used to place crashes into groups of similar severity rather than to compare individual crashes. Passenger compartment intrusion refers to the residual or static deformation. Dynamic deformation during the crash is usually higher than the measured static value available. 4. Drivers There were 77 drivers who had sustained MAIS + injury and/or died. All these drivers were in cars equipped with a driver s airbag and seat belt pre-tensioners. The breakdown of MAIS by survival status is shown in Table 8. Table 8: MAIS by Survival Status (drivers) MAIS Fatals (N) Survivors (N) Total 4 5 6 6 5 99 0 00 6 6 Total 57 0 77 The majority of fatally injured drivers sustained MAIS 4 or 5 (4/57). The majority of survivors sustained MAIS (99/0). One driver died with MAIS and one with MAIS. The first sustained only minor abrasions and lacerations and had drowned. The second sustained a small liver laceration, petechial haemorrhaging in both kidneys, fractures of the radius and ulna to both left and right limbs and a de-gloving of the right thigh. The cause of death was Adult Respiratory Distress Syndrome. Excluding the two fatalities with MAIS < left 75 drivers with MAIS + injury. The breakdown of body regions injured to AIS + for those drivers is shown in Table 9. AIS + head, chest, thigh injury combinations are shown for drivers with AIS + injuries to those body regions. Body regions injured to AIS + are shown for drivers without AIS + head, chest or thigh injuries. December 005 9 PRISM
Table 9: Body Regions Injured to AIS + (drivers) Drivers with AIS + to head All Drivers (n) or chest or thigh Head + Chest + Thigh 0 Head + Chest 0 Head + Thigh 8 Chest + Thigh 0 Head only 7 Chest only 46 Thigh only 4 AIS + injuries to Drivers without AIS + to head or chest or thigh Leg only Arms/Hand only Knee only Abdomen only Neck only Arms/Hand + Pelvis Pelvis only Ankle/Foot only Fatal Drivers (n) 9 9 5 5 5 5 5 Total 75 55 Of the 75 drivers with MAIS + those with AIS + head, chest or thigh injuries accounted for the majority, 4/75 (8%). This group also included all the fatalities. The remaining drivers sustained MAIS only and those injuries were mainly to the extremities only, 4/. The priority group in this sample contained the 4 drivers with AIS + injuries to the head, chest or thigh due to the risk of fatality. In this group, AIS + injuries to only the chest accounted for 46/4 drivers (0%). AIS + injuries to only the thigh accounted for 4/4 drivers (9%). AIS + injuries to only the head accounted for just 7/4 drivers (5%). AIS + head injuries were implicated in 45/4 (%) of the 4 drivers whereas AIS + chest injuries were sustained by 86/4 (6%) of those drivers. AIS + thigh injuries occurred in 69/4 (49%) of the drivers in the priority group. Of the 55 fatally injured drivers, 6% sustained serious head injury, 87% sustained serious chest injury. 4. Front Seat Passengers There were 44 front passengers who had sustained MAIS + injury and/or died. All of these passengers were in cars equipped with seat belt pre-tensioners but only 7/44 (6%) were equipped with passenger airbags. The breakdown of MAIS by survival status is shown in Table 0. Table 0: MAIS by Survival Status (front passengers) MAIS Fatals (N) Survivors (N) Total 4 5 6 5 5 5 6 0 6 Total 44 December 005 0 PRISM
The majority of fatally injured front passengers sustained MAIS 4 or 5 (8/). The majority of survivors sustained MAIS (6/44). No front passengers died with MAIS <. Of the 44 front passengers with MAIS + injury, the breakdown of body regions injured to AIS + is shown in Table. AIS + head, chest, thigh injury combinations are shown for passengers with AIS + injuries to those body regions. Body regions injured to AIS + are shown for passengers without AIS + head, chest or thigh injuries. Table : Body Regions Injured to AIS + Passengers with AIS + to All Passengers Fatal head or chest or thigh (n) Passengers (n) Head + Chest + Thigh Head + Chest Head + Thigh Chest + Thigh Head only Chest only Thigh only 6 0 6 AIS + injuries to Passengers without AIS + to head or chest or thigh Neck only Arms/Hand only Arms/Hand + Leg Leg only Abdomen only 6 Total 44 Of the 44 front passengers with MAIS + those with AIS + head, chest or thigh injuries accounted for the majority, 4/44 (77%). This group also included all the fatalities. The remaining passengers sustained MAIS only and those injuries were mainly to the extremities only, 8/0. The priority group in this sample contained the 4 front passengers with AIS + injuries to the head, chest or thigh due to the risk of fatality; a situation very similar to that with drivers. For the front passengers, AIS + injuries to only the chest accounted for 0/4 (59%). AIS + injuries to only the thigh accounted for /4 (9%). AIS + injuries to only the head accounted for 6/4 passengers (8%). AIS + head injuries were implicated in 0/4 (9%) of the 4 passengers whereas AIS + chest injuries were sustained by 5/4 (74%) of those passengers. AIS + thigh injuries occurred in 5/4 (5%) of the passengers in the priority group. Of the 0 fatally injured passengers, 40% sustained serious head injury, 00% sustained serious chest injury. December 005 PRISM
5 Injury Impact Scenarios The process for generating injury impact scenarios is described below. 5. Injury Scenario Methodology The set of CCIS crashes was analysed to determine the likely occupant kinematics, contact and load points associated with serious injury outcome to the head, chest and thigh (femur). It was necessary to sort by injury mechanism trends and consider injury severities, occupant populations, and clusters of crash types that occurred within these trends. Each kinematic trend leading to an injury trend (with other key aspects identified) was called an injury impact scenario A team of 4 complimentary experts reviewed each case together, using increasing levels of detailed information. Hypotheses were put forward about likely kinematics and injury causes which were then tested by reviewing more detailed data and evidence. The hypotheses were refined and retested, usually with several iterations before a conclusion was drawn and that conclusion clustered with similar cases. As a result of the process development the final sequence for each case was: ) Read out vehicle type, collision partner, estimated impact speed, overlap, driver age, height, gender and weight. ) Each person made a hypothesis about likely kinematics and injuries of the driver (and front seat passenger if applicable). ) The primary AIS for head and chest were revealed and the hypotheses modified or not. 4) Then the detailed injury list was read out and the hypotheses were further refined or modified to take into account known anatomical contact points including grazes, bruising etc., as well as major injuries. 5) Using the hypotheses, suspected contact sites on the interior of the vehicle were then assessed for evidence of loading and deformation. Other evidence, such as abrasion marks on the seat belt webbing, was also assessed with respect to location, to determine likely occupant kinematics. 6) If agreement was reached on the kinematics and biomechanical loading hypothesis, this was accepted and noted in summary. 7) Any other interesting or unusual aspects were also noted in case these were seen again in other cases. 8) Cases that were unclear were classed as possible in the several respective injury impact scenarios. A certain degree of re-assessment was required when particular kinematics were observed that had not been previously considered. Similar past cases were reassessed to determine if the kinematics hypothesis was still valid or if the new one was more suitable. Some practical measures were developed to assist the various participants. It was necessary to consider a wide range of data on each case, simultaneously. This was achieved using multiple PCs. The best arrangement used PCs and projectors projecting onto walls at the same time (general vehicle, crash and occupant data on one, detailed injury descriptions on the second and vehicle photos on the third.) Additional supporting material was also useful a height chart (with comparison to 5thF, 50thM and 95thM dummies), a model skeleton to identify anatomical features and positions December 005 PRISM
when seated in a car seat. A vehicle seat simulator with a length of seat belt webbing (a short and a long end) was used to allow approximate determination of occupant position when the abrasion marks were made on the belt. Also various reference books, especially for medical terms, were found useful. Of front seat occupants with MAIS + injury, 69 drivers and 5 front seat passengers sustained femur fracture. Because of the large number of drivers with femur fracture an overview analysis was carried out to establish the reasons for occurrence (Appendix F). The results show few cases with crash severity exceeding 56 km/h, no major facia intrusion and no pattern of driver age or proximity issues. The major reasons for femur fracture therefore remained unclear and further work was recommended to establish factors related to injury causation. Of front seat passengers with MAIS + injury, 5 sustained AIS + chest injury. An overview analysis was carried out to try and establish the reasons for occurrence (Appendix G). The results showed that most of the casualties were women approaching, during, or after the menopause. Crash severities were very low with static facia intrusion below 5cm in most cases. The overall conclusion to this was that loading of the seat belt webbing on the chest generated most of the AIS + injuries. 5. Injury Scenario Results The derived injury scenarios are shown in diagrammatic form in Figure. Scenarios to 9 refer to the driver, whilst scenario 0 refers to the front seat passenger. Small Driver Small drivers who naturally sit close to the steering wheel are at risk of serious chest head and neck injuries from a range of sources airbag cover contact, airbag punch-out, underchin loading and lack of distance for ride-down before steering wheel contact. Large Driver Large drivers often adopt a very reclined position to prevent roof contact. This leads to poor positioning of the diagonal belt, allowing extensive forward motion and severe submarining under the lap belt. December 005 PRISM
Very Late Deployment Similar in some respects to the small driver case, but this is a dynamic version that affects all statures. Poor crash pulse discrimination in soft impacts (pole, angled offset, shallow overlap and under-ride) allows excessive driver forward motion before airbag deployment, leading to a similar injury set to that suffered by the small driver. 4 Driver Misses Airbag Angled offset and shallow-overlap crashes cause vehicle rotation and displacement of the dashboard / steering wheel inboard (by up to 500mm). The driver then misses the airbag and has heavy head contact with either the lower A-pillar, the outboard dashboard, the driver face vent or the top of the door casing. 5 Airbag Bottoms Out Driver strikes the steering wheel indirectly through the airbag. These are of types: ) Head through the top edge of the airbag,) Chest simply overloads the airbag and penetrates through, deforming the rim and loading the hub. 6 Steering Wheel Edge Strike Driver impacts the steering wheel directly with minimal protection from the airbag. These are of types: ) Head over the top of the airbag, ) Steering wheel upward rotation followed by chest impact. Radial loading with the wheel at the on-spoke position is particularly aggressive. December 005 4 PRISM
7 Header Rail Strike Some instances of moderate under-runs with some header rail intrusion, but not a total loss of head space. Severe head injuries against deformed header rail or truck rear through the windscreen. 8 Chest Injury General Large numbers of chest injuries in crashes for no readily apparent reasons. Few cases with crash severity exceeding 56 km/h, no major steering wheel intrusion, no pattern of driver age or proximity issues. Needs more investigation to understand real loading levels, sequence, timing and effects of rib, sternum and clavicle fractures. 9 Femur Fracture General Large numbers of femur fractures in crashes for no readily apparent reasons. Few cases with crash severity exceeding 56 km/h, no major facia intrusion, no pattern of driver age or proximity issues. Off axis loading? Needs more investigation to understand direction and magnitude of real loading. 0 High Chest Injury Risk to Front Seat Passengers Lower quantities of passenger data made classification of different categories more difficult. It was clear however that most serious injuries were to the chest. Mainly sustained by older women in low severity crashes. 5. Injury Scenario Weighting The incidence of scenarios in the case population is not easy to quantify exactly as all cases are different. There are some cases that are clearly and definitely representative of a scenario but there are other cases in which the injury mechanism described by the scenario may have had some influence on the case outcome, to a greater or lesser extent. In order to assess the incidence of scenarios to allow potential benefits of smart restraints to be determined, a twin approach was adopted: The definite cases and the possible cases were identified by the methods described in section 6.. December 005 5 PRISM
Each of the possible cases was then subjectively assessed against the scenario definition and was given a rating between 0 and that attempted to quantify the degree of relevance of the scenario to that case. Since the assessment was very subjective, the possible increments were kept relatively coarse: 0.0 Not at all relevant 0. Slightly or possibly relevant, but marginal 0.4 Likely to have been relevant to a lesser degree 0.6 May have been relevant to a significant degree, or a likely low severity incidence of the scenario 0.8 Was clearly relevant, but not certain to have been the only mechanism of injury in the appropriate body region, or was not of sufficient severity to justify a score of.0 Definite case The subjective nature of these assessments implies that closer investigation of any one case may suggest some revision, but overall, it is assumed that the general levels are likely to be a reasonable indication of the incidence of the scenarios in the real world accident population. Each of the scores was then used as a weighting factor for the relevance of its aspects to the scenario and to produce a weighted population for each injury mechanism scenario. The femur and chest general scenarios are not included since it has already been established that the injury mechanisms themselves are unclear but there are some trends worthy of further investigation. The results of the weighting procedure are shown in table below. Table Incidence of Injury Scenarios Scenario Number Description Definite Cases Partial (weighted) cases Total Weighted Total Small Driver 6 8 6.6 Very large driver 4 5.4 Very late deployment 6 0 6 5.6 4 Driver misses airbag 4. 5 Bag bottoming out 4 7 9 6a Steering wheel edge strike (chest) 6 7 6b Steering wheel edge strike (head).6 7 Header rail strike 4 6 5.6 8 Chest - General 76 76 9 Femur Fractures 69 69 0 Passenger (chest) 5??? Examining the frequency of scenarios shows that the small driver, late deployment, bag bottoming out, steering wheel edge strike, chest general, femur fractures and passenger chest injuries are the most frequent situations. December 005 6 PRISM
6 Linking Injury Outcome Between Crash Scenarios and Simulation It was necessary to create a baseline simulation representing each crash scenario into which SMART restraints could be introduced and assessed. The methodology for simulation development is described in detail in report R6/R7. To be representative of real crashes, it was necessary to achieve a similar injury outcome between the baseline simulations and the real crashes. Real world injury outcome was assessed by the Abbreviated Injury Scale (AAAM, 990) whereas simulations produce measured physical parameters on the dummy. In order to read across between dummy output and real injury, it was necessary to use a common measuring system. To this end, PRISM employed a method of converting dummy outputs into a probable AIS value using injury risk curves (Kuchar et al, 00). The whole process is reported in detail in section 7. of the R6 & R7 report. This method could then be used to detect changes in injury severity after the addition of SMART restraint systems into the scenario simulations. New injury outputs could then be compared to those from current vehicles to gain insights into potential benefits. The process of evaluating benefits is reported in PRISM report R8. 6. Use of the Injury Severity Score Since changing the response of a restraint system in one area (e.g. chest deflection) is likely to have an effect on another (e.g. head loading), the ISS or Injury Severity Score (Baker et al, 974) was chosen as the general measure of system performance because it accounts for changes in injury outcome to several body regions together. Use of the ISS had the additional value of allowing a benefit assessment based on risk of fatality. The ISS is a good measure of fatality risk because it considers injuries to multiple body regions by summing the squared maximum AIS to the three most severely injured body regions out of five (Fig ). Figure ISS Body Regions The calculation of ISS from the scenarios was limited to using injury at the head, neck, chest and thigh because risk curves were only available for those body regions in the simulations. Injury reference values for the ISS body regions of face, abdomen or pelvic contents, extremities (other than thigh) and external were not available. It was hypothesised that this PRISM ISS (ISS P ) would nevertheless still give a sufficiently close value to the real ISS value whose calculation involves additional body regions. December 005 7 PRISM
The ISS P for the injury impact scenarios was calculated as follows: ISS P =max(ais HEAD, AIS NECK ) + max(ais CHEST ) +max(ais THIGH ) The simulated ISS (ISS M ) was calculated by relating the dummy output to probable AIS values from injury risk curves (Section 7. of the R6 & R7 report) for head, neck, chest and thigh injury, then using the following formula. ISS M =max(ais HIC, AIS NIJ ) + max(ais CTI ) +max(ais FFC ) Where HIC=head injury criteria, NIJ=neck injury criteria, CTI=combined thoracic index and FFC=femur force criteria. In order to ascertain how well the ISS P represented the real ISS, comparisons were made of the two, as seen in table below, for the 4 drivers in the scenarios where AIS + head, chest or thigh injuries occurred. Table Comparison of Real ISS to PRISM ISS ISS Real PRISM % % 7% 8% 8% 7% 5% 9% 6% 4% % 0% 0-0 - 0-0 - 40 4-50 5+ Total (N) 4 (00%) 4 (00%) Chi-square = 5.085, df=5, p=0.406 The ISS P calculation put more drivers into the low ISS (0-0) band and less into the -40 band. In all other bands the distribution of ISS was similar between the two calculation methods. A chi-squared test showed no statistical difference in ISS distributions at the p<0.05 level using the two methods. Therefore the ISS P body regions gave a reasonable assessment of real injury severity for these cases. In order to relate ISS values to a risk of mortality in the benefits analysis, it was necessary to build predictive models of survival rate by ISS. One factor to consider was the question of increasing fragility of occupants with age (Morris et al, 00), especially in women during or after the menopause. Additionally, Baker et al (974) showed that, for the same ISS value, the mortality risk for older casualties was higher than that for younger people. The literature describing at what age tolerance to blunt impact starts to deteriorate does not give an exact cut-off point but relates degradation of bone strength to injury tolerance levels. The research summarised in Figure, shows how skeletal mass deteriorates for males and females. Based on this, it was decided to build two predictive ISS mortality models: one to represent the population of casualties with normal bone strength and one for those where bone strength was likely to be compromised. December 005 8 PRISM
Figure Skeletal Mass Versus Age for Males and Females Using,88 occupants in the 995-004 CCIS data set, linear regression was used to build predictive models of survival rate by ISS for firstly all of the males, plus females under 55 (normal bone strength, N=,4) and secondly for all females 55 and older (compromised bone strength, N=990). Of course, these are coarse groupings but do allow at least a consideration of the difference in injury tolerance in the casualty population. Linear regression is a statistical technique which gives the predicted response of a dependent variable based on the responses to an independent (predictor) variable. In this case the dependent variable was the survival rate and the predictor variable was the real ISS. The underlying assumptions associated with the use of linear regression were validated for the data used. It should be noted that the predicted rate is only valid for the range of the predictor variable. For the model, R is a measure of the correlation between the observed value and the predicted value of the dependent variable. R is the square of the R value and indicates the proportion of the variance in the dependent variable which is accounted for by the model. Essentially, R is a measure of how good a prediction of the dependent variable can be made by knowing the value of the independent variable. If for example the regression produces an R of 0.75 then the model has accounted for 75% of the variance in the dependent variable. Ideally an R of is sought, indicating an exact prediction. The p value provides a test of whether there is a significant relationship between the independent and dependent variables. Regression is however most generally assessed on the basis of the R square value. December 005 9 PRISM
Predicted Survival Rate 00 90 80 70 60 50 40 0 0 0 0 all males/females <55 females >=55 6 6 6 6 4 46 ISS Figure Predicted Survival Rate for ) all Males, plus Females <55 and ) all Females >=55 Figure shows that generally, survival rate declines as ISS rises and that for the same ISS values, the survival rate is much lower for women over 54 years of age. The R value for all males plus females <55 is 0.899 and for females >=55 it is 0.89. For both groups the p value is <0.00. Therefore, the validity of the predictive models is high and highly significant. 6. Chest Injury Risk Predictor Of all the human body regions, the chest is the one most sensitive to occupant bone condition. Numerous studies have shown that the chest tolerance to blunt impact is highly dependent on overall bone strength, which is why older females are particularly at risk from serious chest injury (Morris et al, 00, Frampton and Mackay, 99). For the PRISM simulations, it was necessary to choose a chest injury risk predictor which accurately measured real world chest injury risk and could also predict that risk for high and low bone strength conditions. Particularly because, in the scenarios many of the serious chest injuries were sustained by older women. The Combined Thoracic Index (CTI) reported by (Kleinberger et al, 998) was chosen as the measure of chest injury severity because it was shown to be a better predictor of real-world chest injuries than chest deflection or acceleration alone (Hardy et al, 005). Two versions of the CTI were used for the PRISM simulations dependent on whether we wanted to predict chest injury for normal bone strength or for the older, more at-risk casualties in the accident population. For the normal condition the CTI risk curves developed for the 50 th percentile HYBRID III male dummy were used. For the older, more vulnerable casualties, the CTI risk curves representing the original NHTSA cadaveric test data were used (Eppinger et al, 999). Those tests used cadavers of mean age 60. There was nearly a 0 year difference in average age between the cadavers and the U.S. driver as represented by Hybrid III. The MADYMO and Human Body Models used for the PRISM simulations already contained the required CTI curves for the Hybrid III but it was necessary to derive the CTI cadaveric curve equations using information from NHTSA about how the injury risk for cadavers related to the risk predicted by the Hybrid III. In the development of CTI, the 50% probability of AIS + injury for the cadavers was adjusted to represent a 5% probability of injury for the driving December 005 0 PRISM
population represented by the Hybrid III. NHTSA suggested that the AIS +, 4+ and 5+ probabilities be adjusted by the same amount. The equations for injury risk to the Hybrid III are shown below together with those which represent the human surrogate data (Error! Reference source not found.) based on the adjustment values suggested by Eppinger et al 999 and reported in the NHTSA document on development of CTI. http://www.nhtsa.dot.gov/cars/rules/rulings/aairbagsnprm/pea/pea-iii.n.html Table 4 CTI Risk Curve Equations for Human Surrogates and HYBRID III P (AIS +) P (AIS +) P (AIS 4+) P (AIS 5+) HYBRID III 00*(/(+EXP(4.847-6.06*(CTI)))) 00*(/(+EXP(8.4-7.5*( CTI)))) 00*(/(+EXP(9.87-7.5*( CTI)))) 00*(/(+EXP(4.4-6.589*( CTI)))) Human Surrogate 00*(/(+EXP(4.847-6.06*(0.54+ CTI)))) 00*(/(+EXP(8.4-7.5*(0.54+ CTI)))) 00*(/(+EXP(9.87-7.5*(0.54+ CTI)))) 00*(/(+EXP(4.4-6.589*(0.54+ CTI)))) The resulting injury risk curves are shown in Figure 4. The dotted lines represent the chest injury risk curves for the human surrogates whereas the solid lines represent the curves for the Hybrid III. The AIS Injury severity is shown between the surrogate and dummy curves representing that level of risk. 00 90 80 % Risk of Chest Injury 70 60 50 40 0 0 0 0 + + 4+ 5+ 0 0.5.5.5. 5 CTI Figure 4 CTI Probability of Chest Injury for Human Surrogates and HYBRID III December 005 PRISM
7 Conclusions Under the European 5 th Framework project PRISM, the potential benefit of SMART restraint systems was to be assessed. By definition, a SMART system is one that can adapt to the wide range of conditions found in real crashes that are not represented in regulation or consumer crash testing. European accident data was therefore interrogated to provide indepth information on crash conditions/casualty variables related to serious injury outcome in modern cars. These cases were simulated by vehicle and occupant computer models that evaluated systems into which SMART restraints were added. The change in injury outcome measured by each model was then fed back into the accident data to assess potential overall effects of SMART systems on real crash injury outcome. 7. Crash injury databases Ideally, to fulfill the PRISM brief in-depth crash injury data would have been available for all crashes in each of the EU member states. Five European databases were examined for their level of detail (in-depth data) recorded on vehicles, occupants and injuries and for how that data represented the national population of crashes in each particular country. The DGT database provided a national picture of accidents in Spain but detailed information on impact configuration, restraint fitment and performance and individual injuries was limited. The IDB Fatals database contained no cause of injury, no occupant height or weight and very few vehicle details. Additionally, most of the vehicles were not modern. The DSD Austrian court data contained detailed accident reconstruction, occupant and vehicle information but no measure of how representative it was of the accident population in Austria. CCIS and GIDAS databases were shown to contain in-depth vehicle, occupant and injury information together with methods to relate the crash samples to the population of accidents in the UK and Germany. These held the most promise for input into PRISM. The development of injury scenarios for simulation necessitated continual access to all the crash injury data (including photographs). As the scenario development progressed, continual analysis and examination of data was necessary. This, together with greater numbers of cases in the CCIS prompted the decision to proceed using CCIS only. Neither GIDAS nor CCIS contained large enough samples of side impacts with the most current restraint technology. It was therefore decided to examine frontal crashes only, and to defer side impacts to a future project once enough data becomes available. 7. CCIS data characteristics CCIS samples approximately 80% of fatal/serious crashes in defined geographical region in Great Britain. It was therefore considered that a set of injury scenarios could be developed from CCIS which represented the situation where occupants were seriously injured in frontal crashes, in modern cars. The PRISM sample from CCIS contained belted front seat occupants, in airbag equipped cars, all with serious or fatal injuries. Because simulations could only deal with injuries to the head/neck, chest and thigh (femur) the injury scenarios needed to be based on occupants with serious injury to those body regions. The majority (8%) of the sample had sustained injuries to those body regions. The remaining 9% sustained only AIS injuries, mainly to the extremities with no fatalities. Assessing SMART restraints for occupants with December 005 PRISM
head/neck/chest/thigh injury was therefore considered an appropriate representation of the seriously injured population. 7. Injury impact scenarios Detailed examination of each case allowed the grouping of each casualty according to 0 injury scenarios based on occupant kinematics and contact and load points associated with serious injury outcome to the head, neck, chest and thigh. A sophisticated (and time-intensive) case examination system (detailed in this report) was developed which took into account all aspects of the vehicle crash performance, occupant stature, age, seating attitude, kinematics and injuries before allocating injury scenarios. Examining the frequency of scenarios show that the small driver, late deployment, bag bottoming out, steering wheel edge strike and passenger chest injuries were the most frequent cases where enough information was available to develop simulations. The procedure for developing injury scenarios was generally considered to be successful. There were some situations however where difficulties were encountered. A large number of driver chest injuries occurred where causation was unclear. This was also true of a majority of cases with femur fracture. It was concluded that insufficient information was available for the assessment of injury causation, and the further work needed to remedy this is suggested in more detail in PRISM report R9 & R0. It was important to know the incidence of scenarios in the case population in order to provide a basis for the study of potential benefits. This was not easy to quantify exactly as all cases were different. There were some cases clearly representative of a scenario but there were others where the injury mechanism described by the scenario may have had some influence on the case outcome, to a greater or lesser extent. A method of weighting was devised to quantify the degree of relevance of a scenario to a case. This was considered to address the degree of uncertainty present in scenario allocation for all situations apart from chest injury general and femur fracture. 7.4 Linking simulation injury assessment to real world injuries A conversion technique was used to convert dummy/human model output into the most probable AIS level injury to a real human. This technique had been used with success in other European projects and was considered viable. Since SMART restraints can take account of individuals varying tolerance to blunt loads it was necessary to consider varying injury tolerance for the part of the body most affected by changes in bone strength with age and gender the chest. Chest injury risk was assessed using the Combined Thoracic Index (CTI) used successfully to assess risk in previous work. In order to assess chest injury risk for both the average person and for those with weaker chests, the chest risk curves for the Hybrid III were modified to reflect a lower tolerance to chest injury in some individuals. Calculating extra risk curves based on cadaver data did this. The simulation results show substantial differences in chest injury outcome using the two sets of curves, leading to the conclusion that the decision to consider chest injury tolerance was correct. One of the ways to consider SMART restraint benefits is to examine changes in mortality risk with the introduction of new restraints. The Injury Severity Score (ISS) was used because it considers injuries to several body regions together and is a better indicator of mortality than using the Maximum Abbreviated Injury Score (MAIS) on its own. Reducing injury severity in one body region might increase the severity in another. Using the ISS had the additional benefit of monitoring that possibility. In order to relate an ISS score to risk of death, ISS mortality curves were drawn up based on,88 occupants in the 995-004 CCIS data set. December 005 PRISM
One limitation of simulations concerned available injury risk predictors. These were available for the head, neck, chest and thigh. ISS is based on additional body regions and there was concern that, without these, the simulated ISS would be inaccurate. A check was done on this situation using real-world injuries in the PRISM sample. ISS was calculated first in the conventional manner and then using injury outcome only from the head, neck, chest and thigh. There was no statistical difference between the ISS distributions using the two methods of calculation and we concluded that the simulated ISS values were accurate enough to be used in the benefits study. December 005 4 PRISM
8 References Association for the Advancement of Automotive Medicine. The Abbreviated Injury Scale 990 Revision, 990. Baker. S. P., O Neill, B., Hadden. W., Long. W. B. The Injury Severity Score: A Method for Describing Patients with Multiple Injuries and Evaluating Emergency Care", J Trauma 4:87-96; 974. Eppinger, R., Sun, E., Bandak, F., et al, Development of Improved Injury Criteria for the Assessment of Advanced Automotive Restraint Systems II, National Highway Traffic Safety Administration, November, 999. Frampton. R. J., Mackay. G. M. The Characteristics of Fatal Collisions for Belted Occupants. SAE China, Procs FISITA Congress, Beijing, China, 994. Paper 94567. Hardy, R et al. BOSCOS - Developments and Benefits of a Bone Scanning System. Procs International IRCOBI Conference on The Biomechanics of Impact, Prague, Czech Republic, 005. Kleinberger. M et al. Development of Improved Injury Criteria for the Assessment of Advanced Automotive Restraint Systems. National Highway Traffic Safety Administration report. U.S. government, September, 998. Kuchar AC, Greif R and Neat G W. A Systems Methodology for Evaluation of Vehicle Aggressivity in the Automotive Accident Environment. SAE paper 00-0-7, 00. Lenard, J. and Frampton, R.J., Multiple Two-Impact Crashes - Implications for Occupant Protection Technologies. Proceedings of the 8th International Technical Conference on the Enhanced Safety of Vehicles, US Department of Transportation, National Highway Traffic Safety Administration, ESV 00, Nagoya, Japan, 9- May, 00, pp, [CD-ROM]. Mackay, G. M., Galer. M. D., Ashton, S. J., et al. The Methodology of In-depth Studies of Car Crashes in Britain. SAE Technical Paper Number 850556, Society of Automotive Engineers, Warrendale, PA, 985. Morris, A. P., Frampton, R. J., Charlton, J. and Fildes, B. An Overview of Requirements for the Crash Protection of Older Drivers. Proceedings 46 th AAAM Conf, Tempe, Arizona, U.S.A, 00. PRISM report R6 & R7. Parametric Investigations of Impact Scenarios. PRISM report R8. Benefits of SMART Restraint Systems. PRISM report R9 & R0. Conclusions and Recommendations. http://www.nhtsa.dot.gov/cars/rules/rulings/aairbagsnprm/pea/pea-iii.n.html December 005 5 PRISM
December 005 6 PRISM
9 Acknowledgements PRISM is a DG Research project under the 5 th Framework, GRD-CT-00-00848. The authors are grateful to the European Commission for funding this work. The authors are also grateful to their fellow consortium partners of the PRISM project who have provided valuable contributions to the work. The complete list of PRISM project partners are: CIDAUT (Spain), DaimlerChrysler (Germany), DalphiMetal (Spain), MIRA (UK), TNO, (Netherlands), TRL Limited (UK), TRW (UK), Technical University of Graz (Austria) and the Vehicle Safety Research Centre (UK). December 005 7 PRISM
0 Appendices 0. Appendix A: General Accident Information NAME OF DATABASE What is the name of the database? DGT Accidents Database TYPE OF DATABASE National, in-depth or court cases? National Who collects the data? Police FUNDING Who provides funding for the data collection? Spanish government DATA COLLECTION PROTOCOL Which country? Spain Region in which data is collected? All around Spain How are cases chosen for investigation? What is the sampling procedure? All the Spanish accidents reported by the police are included in the database. What is the relationship of cases examined to the national population of accidents? All the national population of accidents is being taken account. What years of data are available? 99-00 ACCIDENT INFORMATION Date and time. Day. Road type. Road speed limit. Weather conditions. Road surface conditions. Cause of accident. (when known) Number of vehicles involved. Types of vehicles (car, motorcycle etc). Bodystyle (eg: hatchback, estate). No Impact type for each vehicle. but only for front-front impacts + roll only Object struck for each vehicle. Occupant injury severity. Occupant seating position. Airbag fitted/deployed?. No December 005 8 PRISM
NAME OF DATABASE What is the name of the database? Fatals IDB TYPE OF DATABASE National, in-depth or court cases? National Who collects the data? Police FUNDING Who provides funding for the data collection? DfT DATA COLLECTION PROTOCOL Which country? England & Wales Region in which data is collected? Most regions. A few police forces have dropped out because of data protection concerns. How are cases chosen for investigation? What is the sampling procedure? Fatal accidents only. What is the relationship of cases examined to the national population of accidents? A few geographical gaps (see above). Otherwise representative. Data is quite old. What years of data are available? (986-998. 90% of cases are between 990-995). ACCIDENT INFORMATION Date and time. Day. Road type. Road speed limit. Weather conditions. Road surface conditions. Cause of accident., using TRL's causation system Number of vehicles involved. Types of vehicles (car, motorcycle etc). Bodystyle (eg: hatchback, estate). Impact type for each vehicle. Object struck for each vehicle. Occupant injury severity. Occupant seating position. Airbag fitted/deployed?. Deployed - Fitted (but not deployed) - not reliable December 005 9 PRISM
NAME OF DATABASE What is the name of the database? DSD Austrian court database. TYPE OF DATABASE National, in-depth or court cases? Court cases. Who collects the data? DSD FUNDING Who provides funding for the data collection? DSD DATA COLLECTION PROTOCOL Which country? Austria Region in which data is collected? Mainly Upper Austria and Styria How are cases chosen for investigation? What is the sampling procedure? Cases are chosen by judge of the court, who asks independent accident reconstruction expert for accident analyses and causation. What is the relationship of cases examined to the national population of accidents? Approx. 5-0%, but has to be further investigated. What years of data are available? 995-00 ACCIDENT INFORMATION Date and time. Day. Road type. Road speed limit. Weather conditions. Road surface conditions. Cause of accident. but not coded electronically Number of vehicles involved. Types of vehicles (car, motorcycle etc). Bodystyle (eg: hatchback, estate). Impact type for each vehicle. Object struck for each vehicle. Occupant injury severity. Occupant seating position. Airbag fitted/deployed?. December 005 40 PRISM
NAME OF DATABASE What is the name of the database? Co-operative Crash Injury Study (CCIS) TYPE OF DATABASE National, in-depth or court cases? In-depth Who collects the data? universities and U.K vehicle inspectorate team (VOSA) FUNDING Who provides funding for the data collection? Majority from U.K Govt remainder from industry DATA COLLECTION PROTOCOL Which country? England Region in which data is collected? East and West Midlands, Thames Valley and a few cases from VOSA regions How are cases chosen for investigation? What is the sampling procedure? Stratified random sampling What is the relationship of cases examined to the national population of accidents? Weighting factors available for East and West Midlands for tow away cars, less than 7 years old in which there was an injury. What years of data are available? 99-00 ACCIDENT INFORMATION Date and time. Day. Road type. Road speed limit. Weather conditions. Not usually Road surface conditions. Not usually Cause of accident. Sometimes but not coded electronically Number of vehicles involved. Types of vehicles (car, motorcycle etc). Bodystyle (eg: hatchback, estate). Impact type for each vehicle. Object struck for each vehicle. Occupant injury severity. Occupant seating position. Airbag fitted/deployed?. December 005 4 PRISM
NAME OF DATABASE What is the name of the database? GIDAS TYPE OF DATABASE National, in-depth or court cases? In-depth Who collects the data? Medical university of Hannover and Technical university of Dresden FUNDING Who provides funding for the data collection? Forschungsvereinigung Automobiltechnik (Automotive Industry Research Association) and BASt (Bundesanstalt für Straßenwesen or the Federal Road Research Institute DATA COLLECTION PROTOCOL Which country? Germany Region in which data is collected? Hannover and surroundings, Dresden and surroundings How are cases chosen for investigation? What is the sampling procedure? Stratified sampling What is the relationship of cases examined to the national population of accidents? Weighting factors available What years of data are available? 999-00 ACCIDENT INFORMATION Date and time. Day. Road type. Road speed limit. Weather conditions. Road surface conditions. Cause of accident., using police report system Number of vehicles involved. Types of vehicles (car, motorcycle etc). Bodystyle (eg: hatchback, estate). Impact type for each vehicle. Object struck for each vehicle. Occupant injury severity. Occupant seating position. Airbag fitted/deployed?. December 005 4 PRISM
0. Appendix B: Vehicle, Occupant and Injury Specific Information Detailed Information - Vehicle CCIS Fatals IDB Make and model Age of vehicle Crash severity Delta-v and ETS Estimate of pre-crash speed Damage classification SAE CDC No Number of impacts up to Severity of impacts (if>, how ranked?) Accoring to injury severity According to injury severity Overlap calculated, only when damage extends from a corner No Rollover classified Flagged if rotation>90 0 Flagged if occurred Intrusion - where/how measured Static bodyshell intrusion measured Steering movement measured. No Object struck Detailed Information - Occupants CCIS Fatals IDB Seating position Seat track position No No Belt use - 75% Pretensioner (deployed/not deployed) No Load limiter (deployed/not deployed) No Airbag fitted/deployed? Deployed - Not deployed - not reliable Age Height - 60% No Weight - 60% No Gender Rear/side loading by occ or luggage Presence of lugg/occs noted, but not loading Detailed information - Injuries CCIS Fatals IDB Scaling system AIS 90 AIS90 where PM If no PM, just body region/cause of death noted For non-fatalities region only recorded Definition of body regions All anatomical regions Head, neck, chest, abdomen, defined arms, legs Individual injuries recorded Cause of injury, known for around 60% of injuries Only for non-minor injuries noted in PM Not often December 005 4 PRISM
Detailed Information - Vehicle DSD Austrian court database DGT Accidents Database Make and model No Age of vehicle Crash severity Delta-v and ETS No Damage classification Not coded, but could be investigated from photo records No Number of impacts No Severity of impacts (if>, how According to EES and delta-v No ranked?) Overlap calculated No Rollover classified Flagged if rotation>90 0 No Intrusion - where/how measured Static intrusion measured only in severe cases, but maybe could be estimated from photos No Object struck Detailed Information - Occupants DSD court database DGT Accidents Database Seating position Seat track position No No Belt use - 85% Pretensioner (deployed/not deployed) No Load limiter (deployed/not deployed) No Airbag fitted/deployed? No Age Height - 50% No Weight - 50% No Gender Rear/side loading by occ or luggage Presence of occs recorded but not loading Detailed information - Injuries DSD court database DGT Accidents Database Scaling system Medical reports Fatal, serious, slight, unhurt Definition of body regions Medical reports Head, face, neck, chest, back, abdomen, arms, legs and all body Individual injuries recorded Medical reports No Cause of injury, known for around 80% of injuries No December 005 44 PRISM
Detailed Information - Vehicle GIDAS Make and model Age of vehicle Crash severity Delta-v and EES Damage classification SAE CDC Number of impacts up to 99 Severity of impacts (if>, how ranked?) Overlap calculated According to injury severity and/or Delta-v, only when damage extends from a corner Rollover classified Flagged if rotation >90 0 Intrusion - where/how measured Object struck Detailed Information - Occupants Seating position Seat track position Belt use Pretensioner (deployed/not deployed) Load limiter (deployed/not deployed) Static intrusion measured according a defined template GIDAS No fitted fitted Airbag fitted/deployed? / Age Height - 80% Weight - 80% Gender Rear/side loading by occ or luggage Detailed information - Injuries GIDAS Scaling system AIS 90 - Update 98 Definition of body regions cranium, face, neck, shoulder, arm, wrist/hand, chest, abdomen, pelvis, thigh, knee, leg, ankle, foot - left/right Individual injuries recorded Cause of injury, known for around 85% of injuries December 005 45 PRISM
0. Appendix C: Analysis of DGT Spanish Road Accidents Database (Report Reference R). Report Overview / Summary This report has been developed within the framework of the European PRISM (Proposed Reduction of car crash Injuries through improved SMart restrain development technologies) project and co-ordinated by VSRC as Task.: Analysis of existing accident data. Cidaut and Dalphimetal have worked together analysing DGT Spanish Road Accidents database. Every car accident between 997 and 00 has been taking into account. In this study, some problems have been found as the information requested for the PRISM project differs a lot from the information collected in the Spanish Road Accident database. Thus, only a baseline picture of overall injury to all occupants in all crash types has been carried out in this report and even these data are different from those desirable for the PRISM project. Finally, some conclusions about trends and specifics instances of accident scenarios are obtained. This will come in useful to develop smart restraint systems that could have some effect in the injury level of the casualties. December 005 46 PRISM
Introduction DGT: Directorate General of Transport Spanish Road Accidents database is carried out by a public organisation called DGT, dependent of the Ministry of the Interior. DGT competencies related to traffic management in fields like: interurban traffic control and management, establishing directresses for agent education and action, direction and co-ordination in order to improve road safety, information and road assistance to drivers DGT Spanish Road Accidents Database contains the whole population of accidents with casualties in Spain. Limitations of DGT Spanish Road Accidents Database Information contained in DGT Spanish Road Accidents Database is collected by police forces, when an accident occurs. No medical or engineering team takes part in this process. The information is general and these data can not be considered as an in-depth database. So, specific information as MAIS injury classification or presence or absence of some restraint systems as airbags, seat belt pretensioners or AWD s is not presented in this database. The goal of this task is to identify the impact scenarios depending on the injury level of the casualties. So it is necessary to assess the impact supported by each of the occupants of the vehicle. Another problem arises from this, as DGT Spanish Database does not categorise impacts according to the impact types proposed for the PRISM project. Although some impact configurations are taking into account in DGT database, it is impossible to distinguish which kind of impact (front, lateral, rear ) is supported by each vehicle involved in the accident, or in the case of an impact against an obstacle, to know impact angle for example. Some restrictions have been used to fill the tables (they will be explained in point. Used methodology ). Analysis of data The analysis of the DGT Spanish Road Accident Database has been made taking into account all car accidents (vans are not included) between 997 and 00 for vehicles whose registration date is 997 onwards. As the DGT database does not offer information about the date of car manufacture, taking into account registration date is a way to solve this problem. Data contained in the database gives the number of dead people at the moment of the crash (killed within 4 hours of the accident). In this report, these results are presented, although, in agreement with the standard definition given by UN/ECE (Geneva, 995), fatalities within 0 days have been considered. They have been calculated using weighting factors officially used by DGT, which have been obtained by the Ministry of the Interior after statistic studies. These factors are periodically updated. December 005 47 PRISM
Methodology Scientific and technical objectives The primary scientific and technical objective considered in Task., to which this report refers, is: - To identify trends and specific instances of accident scenarios from various data sets. Used methodology In a first step, all Spanish accident records from 997 to 00 were obtained from the DGT Database. As it was explained before, there were some difficulties to assign each impact configuration in the Spanish National data to the impact types proposed under the PRISM project. Data had to be filtered to obtain only car accidents and to remove cases with missing data. Lastly, calculations to know fatalities within 0 days of the accident were performed. CLUSTERING DATA: Restrictions about accident type must be done. In Spanish accident database, when an accident is classified as Head-on impact it means an impact in which each one of the two cars involved in the crash, hits the other in its front. So really, this implies that we have two Single Front Head-on collisions for the PRISM project, but only one Head-on Impact for the DGT. Front-side Impact means that the front side of one of the cars hits the lateral side of the other, without specifying any angle collision. So, for the PRISM project, we should count one Single Front Head-on and one Side Struck perpendicular or one Side Struck oblique, but in the Spanish accident database it is codified as only one Front-side Impact. DGT considers an accident as Side Impact when it only implies the lateral side of both vehicles, independently of the angle between the major component force and the lateral axis of the car. Rear Impact means that the front of one vehicle impacts against the rear part of the other car. So we have one Single Front Head-on and one Single Rear for the PRISM classification, but we have only one Rear Impact in the DGT database. Moreover, since in the database is not possible to distinguish occupants that were travelling in the struck car and those who were in the other car, we can not separate correctly the impact type as it is required for the PRISM project. Thus, we have done the next equivalence: December 005 48 PRISM
Spanish national data Categories Head-on impact Front-side impact. Side impact. Rear impact. Rollover on the road. Run off road by the left side and rollover. Run off road by the right side and rollover. PRISM - Impact types Single Front Head-on Single Struck Side perpendicular Single Struck Side oblique Single Rear Roll only Table. Equivalence between crash types. It is important to keep in mind that this classification is not quite proper for the PRISM project purposes. Only the cases Head-on Impact and Rollover on the road / Run off road by the left side and rollover / Run off road by the right side and rollover behave exactly as it is required for the PRISM project ( Single Front Head-on and Roll only, respectively). Besides, it must be considered that inputs in the table are not referred to the whole population of accidents. This is due to the fact that there are other impact types within Spanish database that contains impacts suitable for the PRISM project (i.e. collisions against fixed objects, multiple collisions between vehicles ) but it is impossible to classify them according to the project criteria. FILTERING DATA All car accidents (vans are not included) with casualties from 997 to 00 in Spain have been selected. Only vehicles whose registration date is 997 onwards have been considered. Furthermore, we have removed from the study cases in which information concerning variables as Age, Injuries or Use of belt was unknown. ALITIES WITHIN 4 HOURS AND WITHIN 0 DAYS In agreement with the standard definition given by UN/ECE (Geneva, 995), Spain adopted in 99 the definition of person killed in a road accident as a fatality that happened at the moment of the accident or within 0 days of the accident. The best method to determine the number of people died in car accidents would be to establish a plan to follow the evolution of every injured person involved in a car accident within 0 days of the accident. In practise, this is impossible for the whole population of casualties. The way to calculate fatalities within 0 days of the accident is using correction factors that are applied on the collected data. Values of these correction factors are presented in table : December 005 49 PRISM
HIGHWAY AREAS CATEGORIES Correction Factors: 997-000 Driver.44.4 Passenger.7.4 Correction Factors: 00-00 URBAN AREAS CATEGORIES Correction Factors: 997-000 Driver.9.7 Passenger.80.5 Correction Factors: 00-00 Table. Correction factors. Calculus is carried out using the following expression: K ( 0 days) = K(4hours) + SI CF where: K( 0 days) : Dead people within 0 days of the accident. K( 4 hours) : Dead people within the following 4 hours after the accident. SI : Severely Injured people at the moment of the accident. CF : Correction factor. Once again, these weighting factors are provided by the DGT after exhaustive studies. They are updated periodically. December 005 50 PRISM
Results / Findings Accident Data As it was said in point, only Single Front Head-on and Roll only cases are fully coincident in the DGT definition and in the PRISM classification. These are the unique fully valid values for the purpose of the project. The results obtained are presented in tables A and B. Letter A means that the results are referred to killed within 0 days of an accident and B means that the results concern to killed within 4 hours of an accident. However, if equivalence assumed in table is taken into account, it is possible to include more data in the study, which are just an approximation to the reality. These results are presented in tables 4A and 4B (letters A and B have the same meaning they had before) Abbreviations list: - NI: Not injured. - : Slightly Injured. - SER: Seriously Injured. - : Fatal. - FSP: Front Seat Passenger. - RSP: Rear Seat Passenger. December 005 5 PRISM
OCCUPANT STATUS 997-00 IMPACT INJURY(0 days) Belted Unbelted Belted FSP Unbelted Belted RSP Unbelted Occupant<= driver Driver FSP RSP yrs old Single Front NI 496 64 84 4 4 5 8 head-on 4866 58 0 89 409 58 4 SER 45 8 9 9 6 4 44 6 65 8 5 7 67 4 Roll only NI 74 77 0 6 48 58 49 684 4 00 6 877 7 57 SER 7 45 07 96 4 6 6 8 0 0 9 4 78 7 Table A. Accident data (within 0 days). December 005 5 PRISM
OCCUPANT STATUS 997-00 IMPACT INJURY(4 hs) Belted Unbelted Belted FSP Unbelted Belted RSP Unbelted Occupant<= driver Driver FSP RSP yrs old Single Front NI 496 64 84 4 4 5 8 head-on 4866 58 0 89 409 58 4 SER 485 6 944 64 49 46 57 57 6 49 4 60 Roll only NI 74 77 0 6 48 58 49 684 4 00 6 877 7 57 SER 7 464 097 0 48 67 8 5 98 06 87 8 64 5 Table B. Accident data (within 4 hours). December 005 5 PRISM
OCCUPANT STATUS 997-00 (total) IMPACT INJURY(0 days) Belted Unbelted Belted FSP Unbelted Unbelted Occupant<= Belted RSP driver Driver FSP RSP yrs old Single Front NI 496 64 84 4 4 5 8 head-on 4866 58 0 89 409 58 4 SER 45 8 9 9 6 4 44 6 65 8 5 7 67 4 Single Struck side perpendicular Single Struck side oblique Single Rear Roll only NI 94 09 66 46 0 9 695 48 5675 694 07 65 99 SER 460 8 4 55 84 400 497 40 56 50 8 04 8 NI 7977 65 0 4 49 4 5 47 6 490 0 76 88 5 SER 57 6 45 5 87 40 6 0 8 9 5 NI 678 905 67 4 60 65 70 908 66 447 64 865 90 594 SER 850 6 75 5 7 44 07 9 48 4 6 8 NI 74 77 0 6 48 58 49 684 4 00 6 877 7 57 SER 7 45 07 96 4 6 6 8 0 0 9 4 78 7 Table 4A. Accident data (within 0 days). December 005 54 PRISM
OCCUPANT STATUS 997-00 (total; 4 hours) IMPACT INJURY(0 days) Belted Unbelted Belted FSP Unbelted Unbelted Occupant<= Belted RSP driver Driver FSP RSP yrs old Single Front NI 496 64 84 4 4 5 8 head-on 4866 58 0 89 409 58 4 SER 485 6 944 64 49 46 57 57 6 49 4 60 Single Struck side perpendicular Single Struck side oblique Single Rear Roll only NI 94 09 66 46 0 9 695 48 5675 694 07 65 99 SER 59 90 6 58 87 409 48 9 47 5 95 6 NI 7977 65 0 4 49 4 5 47 6 490 0 76 88 5 SER 586 64 50 5 88 40 49 9 9 5 NI 678 905 67 4 60 65 70 908 66 447 64 865 90 594 SER 87 9 8 5 7 45 86 6 40 5 6 NI 74 77 0 6 48 58 49 684 4 00 6 877 7 57 SER 7 464 097 0 48 67 8 5 98 06 87 8 64 5 Table 4B. Accident data (within 4 hours). December 005 55 PRISM
PRISM Accident Analysis Although it is impossible to identify trends and specific cases of accidents where the application of smart restraint systems could have had some effect, since such information is not present in the DGT Spanish database, a brief analysis of the accident data obtained has been made. Fatalities within 0 days of the accident have been used for this study.. TOTAL CASUALTIES DEPENDING ON OCCUPANT POSITION: In figure, number of casualties (serious injured and fatal) depending on occupant position in the car can be seen. All accident types have been considered in this figure. Casualties vs. Occupant Position Number 000 0000 8000 6000 4000 000 0 999 4 64 95 86 476 Driver FSP RSP Occupant Position Fatal Serious injured Figure. Casualties depending on occupant position. The figure shows that both driver fatalities and serious casualties cope with the majority rate (60%) of the total fatalities and serious casualties respectively, followed by front seat passengers (6%). December 005 56
. INJURY LEVEL DEPENDING ON USE OF SEAT BELT: In this item, belt influence is studied in relation with the injury level of casualties. Again, all impact configurations are taken into account and each occupant position in the vehicle is analysed separately. It has been done a comparison between occupant using seat belt or not. Occupant position: driver 0000 8000 8579 6000 4000 000 580 684 60 0 Belted driver Unbelted driver Fatal Serious Injured Figure. Injury level depending on occupant position: driver. When driver does not use seat belt, the rate of fatalities grows near twice the rate for drivers that use this restraint system (from 5% to %). In figure, it also can be seen that the proportion of drivers that do not use seat belt is very low. Occupant position: FSP 4000 750 000 000 000 0 645 Belted FSP 6 56 Unbelted FSP Fatal Serious Injured Figure. Injury level depending on occupant position: FSP. A similar situation can be found in figure for the front seat passengers. The rate of fatalities decreases from 8% to 5% when the occupant uses the seat belt. Again the seat belt use rate is very high for these occupants. December 005 57
Occupant position: RSP 000 600 58 00 800 400 0 87 7 Belted driver 89 Unbelted driver Fatal Serious Injured Figure 4. Injury level depending on occupant position: RSP. Once more, the effectiveness of using seat belt can be probed in figure 4 for rear seat passengers. The rate of fatalities decreases from 0% without using seat belt to % with seat belt. But in this occasion, the seat belt use rate is quite much lower, as it reflects the real situation for Spain. Children 600 495 400 00 97 0 Children Fatal Serious Injured Figure 5. Injury level for children Lastly, figure 5 shows the casualties (fatal and serious injured) in case of children younger than years. As conclusion, it can be said that it is obvious that, when the occupant does not use the seat belt, the rate of fatalities grows near twice the rate for occupants that use it. This growth is totally independent of the occupant position in the vehicle. December 005 58
. CASUALTIES DEPENDING ON IMPACT TYPE FOR EACH OCCUPANT POSITION Dependence between number of casualties (serious injury and fatal) and impact configuration is analysed. This study has been done depending on occupant position and distinguishing whether the occupant was using the seat belt or not. A) Driver position Belted Driver fatalities Unbelted Driver fatalities 8% 7% 4% % 40% Single Front head-on Single Struck side perpendicular Single Struck side oblique Single Rear Roll only 46% 6% 4% 4% 0% Single Front head-on Single Struck side perpendicular Single Struck side oblique Single Rear Roll only Figures 6 Driver fatalities. In the case of drivers, and talking about fatalities, the most dangerous impact type is the Single Front Head-on impact when the driver is using the belt, but if the driver is not using it, the worst impact is Roll only impact. Belted Driver serious injured Unbelted Driver serious injured 6% 0% 8% Single Front head-on Single Struck side perpendicular Single Struck side oblique Single Rear % 4% Single Front head-on Single Struck side perpendicular Single Struck side oblique Single Rear 7% 9% Roll only 0% 5% 8% Roll only Figures 7 Driver serious injured. If the study is focused on serious injured casualties, the most frequent accidents are Single Front Head-on, Single Struck side perpendicular and Roll only with rates quite similar between them, for belted occupant. For unbelted drivers, Roll only rate is slightly superior. B) Front Seat Passenger position Belted FSP fatalities Unbelted FSP fatalities 7% 4% 0% 4% 8% Single Front head-on Single Struck side perpendicular Single Struck side oblique Single Rear Roll only 4% 6% 4% 4% % Single Front head-on Single Struck side perpendicular Single Struck side oblique Single Rear Roll only Figures 8 Front Seat Passenger fatalities. December 005 59
For the belted front seat passengers, the worst impact type is Single Struck Side perpendicular, but in case of unbelted front seat passenger, the most dangerous is again- Roll only. Belted FSP serious injured Unbelted FSP serious injured 9% 0% 7% 5% 9% Single Front head-on Single Struck side perpendicular Single Struck side oblique Single Rear Roll only 4% 9% 6% % 8% Single Front head-on Single Struck side perpendicular Single Struck side oblique Single Rear Roll only Figures 9 Front Seat Passenger serious injured. In the case of serious injured, rates are similar in both cases (belted and unbelted). Rates for Single Front Head-on, Single Struck Side perpendicular and Roll only are the highests and are also similar between them. C) Rear Seat Passenger position Belted RSP fatalities Unbelted RSP fatalities 7% 8% % % % Single Front head-on Single Struck side perpendicular Single Struck side oblique Single Rear Roll only 45% 5% 6% 7% 7% Single Front head-on Single Struck side perpendicular Single Struck side oblique Single Rear Roll only Figures 0 Rear Seat Passenger fatalities. The impacts, which produce the main quantity of fatalities in case of belted occupants, are Single Front Head-on, Single Struck Side perpendicular and Roll only with a similar rate. For unbelted passengers Roll only is the worst configuration, as usual. Belted RSP serious injured Unbelted RSP serious injured 4% 0% 7% % 6% Single Front head-on Single Struck side perpendicular Single Struck side oblique Single Rear Roll only 40% 8% 5% % 5% Single Front head-on Single Struck side perpendicular Single Struck side oblique Single Rear Roll only Figures Rear Seat Passenger serious injured. In figure, it can be seen that Roll only is the main cause of serious injured casualties for rear seat passengers when they are using seat belt. Roll only once again, for unbelted occupants. December 005 60
D) Children Children fatalities Children serious injured 8% % 5% 5% 9% Single Front head-on Single Struck side perpendicular Single Struck side oblique Single Rear Roll only 7% 9% 8% 6% 0% Single Front head-on Single Struck side perpendicular Single Struck side oblique Single Rear Roll only Figures Children casualties. Roll only is the first cause of fatalities for children, but the worst impact type in case of serious injured is Single Front Head-on impact. December 005 6
Conclusions Since DGT Spanish Road Accident Database is not an in depth accident database, the results of the analysis are rather limited for the purposes of the PRISM project. Some approximations have been made. Only one of the tables requested in Task. has been filled and there are not data about MAIS injury level or restraint systems in detail. In spite of these difficulties, some studies have been done with the data contained within DGT database, which allows us to approach a little the Spanish accidental situation in relation to the purposes of the PRISM project. When the occupant does not use the seat belt, the rate of fatalities grows near twice the rate for occupants that use it. This growth is totally independent of the occupant position in the vehicle. For all unbelted occupants the worst impact type is Roll only, and finally, for belted drivers the worst one is the Single Front Head-on impact, for front seat passengers is the Single Struck Side perpendicular and for the other occupant positions (including children) the worst one is the Roll only impact. December 005 6
0.4 Appendix D: Background to German In-depth Crash Injury Study (GIDAS) GIDAS German In-Depth Accident Study Cooperative project between Bundesanstalt für Straßenwesen (Federal Road Research Institute) and the Forschungsvereinigung Automobiltechnik e. V. (Automotive Industry Research Association) Introduction In Germany, accident trends are presented annually based on the official accident statistics of the Federal Institute of Statistics, Wiesbaden. These accident statistics use the data from police accident reports. Although these statistics are useful, a limitation is that very little information about how accidents occur, the cause of the accident and the injury mechanisms is available. This limitation can be overcome by carrying out specialist in-depth accident investigations, collecting more detailed information than available in the police records. Such investigations begin immediately after the accident occurs. Specialist teams go directly to the scene of the accident to collect the necessary information to complete detailed accident reconstructions as well as the medical data about how the involved people were injured and treated. In this way, extensive information about a wide range of fields of research such as vehicle design for passive and active safety, biomechanics, "driver behaviour", "trauma medicine", rescue services, road design and road conditions can be collected. Development of accident research studies on site The first so-called In-Depth Investigation Teams were initiated in the 970s by German automakers. In 97, the Federal Road Research Institute established an independent team at the Medical University of Hanover (in cooperation with the Technical University of Berlin). By 984, this developed into a long term on-scene accident research study, based in a defined geographical area surrounding and including Hanover, which collected representative results. As of 985, a target of 000 accidents per year was set to form the basis for future evaluations. A statistical sample plan was used for selecting accidents for investigation and extensive information about the various aspects of the pre-accident, collision, and postaccident phases was collected and compiled into a database. The value of in-depth accident research studies has been recognized internationally and many other countries also have such teams. Since such detailed information is essential for improving the safety of cars, a strong collaboration with automakers developed. This resulted in a joint project between (Forschungsvereinigung Automobiltechnik or Automotive Industry Research Association) and BASt (Bundesanstalt für Straßenwesen or the Federal Road Research Institute) in 999. In this project, the geographical area was extended and a second team was set up in the Dresden area providing additional cases from a different part of Germany. Both December 005 6
teams function in the same manner using the same systems, procedures and databases. The joint project started on July, 999. Design of the research studies Geographical area of the research studies: Hanover The geographical area covers both the municipality of Hanover and the surrounding rural areas. There are. million residents in this area and the surface area is approximately,89 km². 0% is designated as urban. Dresden The area includes the city of Dresden as well as parts of the counties of Meißen, Riesa-Großenhain, Weißeritzkreis, Sächsische Schweiz, Bautzen and Kamenz. There are approximately 95,000 residents in the area and the surface area is approximately,575 km². Sample plan: Accidents involving personal injury are investigated according to a statistical sampling process. In both areas, the respective police, rescue services, and fire department headquarters report all accidents continuously to the research team. The December 005 64
team then selects accidents according to a strict selection process and investigates these cases following detailed procedures contained in a handbook and coding manual. In order to avoid any bias in the database, the data collected in the study is compared to the official accident statistics for the respective areas and weighting factors are calculated annually. This process explains why the data captured by the research teams can be seen as representative for their areas. Statements about the national situation are only possible for those accident features that are relatively independent of regional influences. This is true for the variables which have an effect on the injuries sustained in crashes and therefore the findings from the study can be considered as representative for most aspects of passive safety. Research studies timing: Accident investigation takes place daily during two six-hour shifts following a -week cycle as follows: First week: from 00:00 to 06:00 and from :00 to 8:00 Second week: from 06:00 to :00 and from 8:00 to 4:00. This makes it possible to cover all periods of the day throughout the whole year. Accident team and special vehicles: During each shift, a team consisting of two technicians, a doctor, and a coordinator is on duty. Each centre has two specially equipped vehicles available. These are equipped with flashing blue lights, sirens, special signals and emergency radio equipment. Various cameras and instruments are available for measuring and recording purposes. Accurate scale sketches of the scene of the accident are created using a technique known as "photogrammetry". December 005 65
Scope of data: Between the two centres, about 000 accidents are investigated annually. The studies include such information as: Environmental conditions Road design Traffic control Accident details and cause of the accident Crash information e.g. driving and collision speed, Delta-v and EES, degree of deformation - Vehicle deformation - Impact contact points for passengers or pedestrians - Technical vehicle data - Information relating to the people involved, such as weight, height etc. The information collected "on the scene" is complemented by more detailed measurement of the vehicles (usually on the following day), further medical information about injuries and treatment and an extensive accident reconstruction generated from evidence collected at the accident scene. By applying established physical principles, the impact events are reconstructed (e.g. collision speed) using proven software such as PC-Crash. The output can be graphically displayed to allow a full understanding of the crash events. Approximately 500 to,000 pieces of information per accident are obtained in total. Any personal data included is processed according to data protection regulations. Medical confidentiality and the rights of the individuals are guaranteed. All information is stored anonymously in a data base produced using SIR (Scientific Information Retrieval) software and is available for evaluation. Use of statistical data The data collected is used in various ways; Legislators can study the accident cases in detail to identify and quantify future areas for legislation and to recognize negative developments in advance. The detailed December 005 66
documentation available (eg detailed information about vehicle deformation, causes of injury to passengers and other road users such as pedestrians and cyclists) can form the basis for future legislation. This type of data has served as the basis for developing suitable test methods for type approval (e.g. EU Directives). In this way, research studies conducted by traffic accident researchers has already had an effect on legislation and can be seen as having a positive effect on the accident and casualty situation. Special notice should be given to studies on the effectiveness of safety belts, on the necessity of safety helmet for cyclists, and the protection offered to motorcyclists wearing crash helmets. Both the automotive industry and the BASt can compare real accident situations to crash tests. Structures causing injuries can be recognized at an early stage. The statistical data is also used for developing crash test programs, for supporting and validating computer simulations, recognizing and assessing potential areas of future safety developments and evaluating vehicle safety performance in real world accident situations. Feedback regarding road traffic engineering (such as assessing the severity of collisions between vehicles and objects at the side of the road) can also be obtained. In terms of impact severity and injury risk, trees are ranked as very high, safety barriers as medium and free space at the side of the road as very low. Measures can be identified from the data which lead to improved design of roadside objects (e.g. poles, posts, and stakes) such as erecting protective barriers around trees. In addition, statistical data on essential factors (such as speed, mass and angle) serves as the basis for defining standards for impact tests (e.g. DIN EN 7). Integration into international research Cooperation with research projects in other countries enables the statistical data collected and analyzed to be used for international comparisons and research activities. One example is a European project called "STAIRS" (Standardisation of Accident In-Depth Research Studies) which was completed in 998. In STAIRS, a standardized method for collecting crash injury data was developed. The recommendations from STAIRS have been implemented into the data collection procedures used by both centres. Within the framework of future European research activities, a feasibility study on the subject Accident Research Studies on Site at a European Level and the possibility of a pan-european accident data base are both being discussed. Reporting MHH and TUD produce monthly and yearly reports about the data collected. These reports give an overview of the cases investigated and allow comparisons to be made between the centres regarding types of vehicle and road users, the severity of the accident, etc. In addition, special reports are produced on topical safety issues based on requests containing recommendations for improving safety. December 005 67
Coordination A steering committee, consisting of representatives from, BASt, MHH and TUD, coordinate and manage the wide range of activities involved in the project such as: content issues, data compatibility and quality, joint publications. December 005 68
0.5 Appendix E: Case Details for Drivers with AIS + Head Injury ETS delta v obj hit DOF overlap facia intrusion s/wheel intrusion model driver age height gender 0 999 78 0 0 MICRA 6 9.99 59 999 89 0 BRAVO 4 9.99 4 5 0 5 VECTRA 6.77 4 4 5 7 9 45 405 59 9.99 4 50 4 00 7 ACCORD 66 9.99 5 999 999 4 00 999 0 NEXIA 64.6 6 999 999 00 7 40 00.7 5 999 999 4 00 0 0 NEXIA 46 9.99 4 999 4 56 0.77 4 9 999 00 0 0 8 75 9.99 4 6 0 6 00 0 ESCORT 67.7 4 999 999 65 0 4 KA 6.8 84 78 76 0 0 44 57.78 4 54 54 48 45 VECTRA 8.7 4 5 58 74 0 0 MICRA 65.6 6 6 88 0 6 STARLET 8 9.99 4 5 58 8 999 5 KA 4 9.99 4 59 89 00 4 0 FOCUS 5.6 4 999 999 4 00 7 6 ESCORT 7 9.99 4 999 999 4 8 06 45 9.99 4 4 5 00 0 MONDEO 5.8 5 56 00 0 0 MATIZ 6 9.99 4 999 999 4 00 4 6 69 9.99 6 999 999 4 COROLLA 68.8 4 999 999 00 60 999 PUNTO 9.99 4 6 66 80 8 999 ESCORT 7.75 8 999 58 0 0 KA 5.6 999 999 4 00 6 5 S40 54.7 66 44 8 4 8 600 5 9.99 5 999 999 4 00 8 0 PUNTO 7 9.99 6 6 999 5 0 ASTRA 48 9.99 6 44 999 5 0 46 70.5 5 5 69 00 0 0 ASTRA 45 9.99 48 50 00 6 0 5 4.6 4 57 5 999 999 Carisma 6.65 4 76 65 5 5 4 Focus 4.8 5 999 999 4 00 999 999 Nubira 9.99 5 7 999 7 44 79 9.99 5 9 8 49 6 06 6 9.99 5 Chest AIS airbag December 005 69
0.6 Appendix F: Variables Related to Driver Femur Fractures Gender of Drivers with Femur Fracture 80 70 60 percent 50 40 0 0 0 0 male female gender 7% were male and 7% were female. Age of Drivers with Femur Fracture cumulative % 00 90 80 70 60 50 40 0 0 0 0 0 0 0 0 40 50 60 70 80 age (years) The inter quartile range of ages was 7 49 years with a median age of 6. All Ages were known. December 005 70
Driver Right/Left limbs with Femur Fracture 80 percent 70 60 50 40 0 0 0 0 both left right limb 70% had sustained only right femur fracture, % sustained left femur fracture and 9% had fractured both femurs. Mortality of Drivers with Femur Fracture 80 percent 70 60 50 40 0 0 0 0 fatal mortality survivor 7% with femur fracture survived whilst 7% died. December 005 7
Equivalent Test Speed for Drivers with Femur Fracture cumulative % 00 90 80 70 60 50 40 0 0 0 0 0 0 0 0 40 50 60 70 80 90 00 ets (km/h) The Equivalent Test speed was known for 55/69 drivers with femur fracture (80%). The inter quartile range of ETS was 4 59 km/h with a median ETS of 44 km/h. 8% of cases occurred with speeds over 56km/h and 0% occurred over 64 km/h. Static Facia Intrusion for Drivers with Femur Fracture cumulative % 00 90 80 70 60 50 40 0 0 0 0 0 0 0 0 40 50 60 70 80 facia intrusion (cm) The median facia intrusion was 5 cm. 6% of cases occurred with less than 5cm of intrusion. December 005 7
0.7 Appendix F: Variables Related to AIS + Front Seat Passenger Chest Injury Gender of Front Passengers with AIS + Chest Injury 80 70 60 percent 50 40 0 0 0 0 male female gender % were male and 88% were female. Ages of Front Passengers with AIS + Chest Injury cumulative % 00 90 80 70 60 50 40 0 0 0 0 0 0 0 0 40 50 60 70 80 90 00 age (years) The inter quartile range of ages was 5 74 years with a median age of 5. /5 front passenger ages were unknown. December 005 7
Equivalent Test Speed for Front Passengers with AIS + Chest Injury cumulative % 00 90 80 70 60 50 40 0 0 0 0 0 0 0 0 40 50 60 70 80 90 00 ets (km/h) The Equivalent Test speed was known for 9/5 front passengers with AIS + chest injury. The inter quartile range of ETS was 9 5 km/h with a median ETS of 6 km/h. % of cases occurred with speeds over 56km/h and 0% occurred over 64 km/h. Static Facia Intrusion for Front Passengers with AIS + Chest Injury cumulative % 00 90 80 70 60 50 40 0 0 0 0 0 5 0 5 0 5 0 5 facia intrusion (cm) 65% of cases occurred with no intrusion. 80% of cases occurred with less than 6cm of intrusion. December 005 74