ANALYSIS OF PUBLICLY AVAILABLE DATA ON ACCIDENTS

Transcription

1 ANALYSIS OF PUBLICLY AVAILABLE DATA ON ACCIDENTS INVOLVING HEAVY VEHICLES by Dnald F. Masn Transprtatin Grup University f New Brunswick Frank R. Wilsn Transprtatin Grup University f New Brunswick Albert M. Stevens Transprtatin Grup University f New Brunswick

2

3 ABSTRACT The paper reprts the results f a study n data extracted frm: Level accident reprts; files f a plice agency in the Prvince f New Brunswick fr the years 984 t 986; and, the Transprt Canada Passenger Car Study (PCS). The data selected relate t incidents invlving heavy tractrs, semi-trailer units and ther freight vehicles perating n prvincial highways. Frm the New Brunswick data, 30 cases invlving fatalities were chsen and a search was made fr additinal infrmatin frm plice investigatin files, mechanical inspectin reprts, bld alchl analyses, witness statements, accident recnstructin reprts, scene and vehicle phtgraphs and the PCS data file. The findings clearly indicate that the basic plice reprts cntained misleading infrmatin. Variables relating t truck and trailer identificatin, accident cnfiguratin, lcatin f heavy truck damage and assignment f fault were fund t cntain numerus cding errrs. The Level plice accident reprts will cntinue t be the primary surce f frequency data fr all types f vehicle accidents. The data they cntain can be made mre useful fr mni tring and cuntermeasure develpment by minr additins t the variables reprted and the establishment f imprved quality cntrl. The cding discrepancies suggests sme f the infrmatin quality limitatins and hw they might be included in the applicatin f these data. The results suggests that a methd fr quality cntrl f rutine reprts culd be develped by fllw-up studies using plice investigatin files and/r mdificatin f the PCS reprts. It appears that Level 2 and/r Level 3 accident studies are required and can prduce gd levels f detail f heavy vehicle safety factrs. The analysis cnfirms many previus reprts f such factrs as the ver invlvement f heavy vehicles in fatal accidents, under invlvement in ttal accidents, the aggressiveness f heavy vehicles and the majr rle f the driver.

4

5 .0 INTRODUCTION S called accidents can be cnsidered as ne f the failure mdes f a transprt system. Relating accidents t cmpnent design, quality, and peratinal prcedures is ne f the fundamental methds used t mnitr and imprve system quality. The ntin that large/heavy vehicles used fr highway transprt shuld include sme meaningful accident investigatin attentin was the mtivatin fr this wrk. The fcus has been n examining sme f the infrmatin which is in the public dmain fr the prvince f New Brunswick in eastern Canada. An nging prblem fr mnitring and imprving the perfrmance f accident data reprting is the scarcity f available reliable infrmatin n which t take decisins. Other prblems such as data interpretatin and the frmulatin f acceptable actins, etc. are accepted, but these are nt the fcus f this study. Detailed data are especially sparse fr invlvement rate by heavy trucks in accidents because these incidents are relatively rare ccurrences and the ppulatin f bth vehicles and their usage are highly variable. The ccurrences als tend t be lcalized. The nature f the prblem was effectively described by Sparks et al in their paper ("The Safety Experience f Large Trucks in Saskatchewan", 987) given at the 988 CTRF annual meeting. Assuming that available data n heavy truck accidents were reliable then analyses f these data culd reveal areas requiring crrective actins, research requirements, and the effectiveness f cuntermeasures and standards, etc. Mtr vehicle safety levels can be quantified in terms f such factrs as the numbers f fatalities, number and severity f injuries, amunt f prperty damage and ther claims relating t cllisins. Whatever safety dimensins are chsen they are nly useful if accurate and timely data f ccurrence are available. The primary surce f safety data fr Canada are the investigatin reprts filed by the plice. This infrmatin is available in machine readable frm fr all prvinces. The reprt frmat and cntent are nt fully standardized, but there are marked similarities. These data are cmpiled int a natinal data base by Transprt Canada which is referred t as the TRAID file (TRaffic Accident Data). These are the 'Level l' data and are the majr surce fr natinal as well as lcal strata. As 'Level l' suggests, these data prvide elemental infrmatin n all reprted accidents. Expansin f any accident sample t lcal r natinal significance will usually rely n these data. Of the many methds used t cllect data fr accidents three levels are usually identified. Level is as nted abve earlier. Level 2 is a mre detailed pst accident investigatin and Level 3 is a very detailed investigatin which includes visiting the scene as quickly as pssible subsequent t the event.

6 Because the primary purpse f 'Level l' data is in supprt f enfrcement, these data are knwn t be weak in terms f detailed cverage and accuracy. These data shuld be cst effective, and are essential because they include cmplete, and cntinuus cverage f the accident ppulatin. They are als maintained at public expense in machine readable frmat and are generally accessible because f public infrmatin legislatin. The Level data culd be made mre effective by imprving their reliability by better quality cntrl/assessment and by extending the range f detail cvered. The New Brunswick Level data were examined t determine reliability and detail n large/heavy truck accidents. Accidents invlving fatalities are rutinely investigated mre intensely. The additinal evidence which culd be matched with a sample f large truck accidents were cmpared with the Level data n these ccurrences. Level 2 data is develped by the Transprt Canada 'Passenger Car Study' (PCS). These data fr the years 984, 85 and 86 were als examined fr infrmatin n heavy truck accidents. The PCS data is fcussed n passenger car damage, especially relating t vehicle ccupant prtectin. The sample is small, abut 00 cases per year in each f ten lcatins acrss the natin and includes a cnsiderable effrt n quality cntrl. The PCS develped similarly and perated in parallel with the 'Natinal Accident Sampling System' in the United States. The PCS case data examined lacked detailed weighting. Only fatal and persnal injury accidents which invlved ne r mre passenger cars are included. The examinatin therefre reveals mstly the ptential f these data and suggestins fr hw the ppulatin cvered culd be expanded t large/heavy trucks, etc. The data bank includes vehicle and scene phtgraphs which were nt examined but are knwn t be a surce f additinal detailed infrmatin. T assist in matching the heavy vehicle cmpnents and systems which affect safety, the nature and frequency f related incidents shuld be cntinuusly mnitred. This pses prblems because f the data difficul ties. It appears that nly superficial frequency infrmatin will be available under any reasnable rutine data cllectin scheme. Level data shuld be supplemented by mre detailed infrmatin frm Level 2 and/r Level 3 studies. The data will hwever remain sparse and will have t be interpreted n an anecdtal basis. It appears especially imprtant t be able t link the available data n a case by case basis t mnitr and quantify reliability. Suggestins fr imprving the data situatin are included in sectin 5.

7 2. 0 BACKGROUND Cmmercial vehicles are ften cnsidered t be ver represented in accidents and are typically thught t be at fault. The drivers and these vehicles ften travel in excess f 60,000 km and/r 2000 hrs per year. These travel levels increase their expsure t accident situatins. The vehicle peratrs ften drive as their main ccupatin and tend t be mre experienced and qualified than passenger car drivers. The vehicles ften travel at night n high class highways when traffic is relatively light. Shuld cmmercial vehicles, especially large/heavy nes be invlved prprtinally in fewer accidents than passenger cars which ften travel during peak hurs in cngested urban areas? Previus studies have determined cmmercial vehicle accident rates t range frm 0.45 t. per millin km travelled, cmpared with 0.8 t.8 fr passenger cars (l:ix), (2:2) and (3:2). Hwever these relative relatinships d nt hld true fr fatal accidents. Cmmercial vehicles appear three times mre likely than passenger vehicles t be invlved in a fatal accident (8.2 vs 2.8) (4.). T better understand invlvement, these factrs must sme hw be apprpriately quantified and be adjusted fr expsure. Anther aspect t cnsider in cases where the heavy truck is determined t be at fault is whether it is the truck peratr's fault, r simply the characteristics f the vehicle being driven. Factrs such as relative vehicle length, width, elastic and plastic stiffness, mass, etc. cntribute t high fatal accident invlvement. Fr example the typical bumpers n trucks are relatively much strnger and stiffer than fr passenger cars and in mst instances are munted at different heights. Cnsequently impact with an ncming heavy truck is similar t striking an appraching nn-yielding wall at best r a mving narrw nearly rigid bject at wrst. The grss vehicle mass is ften ten r mre times that f a passenger car. Finally, the verall length can be ver five times the length f a passenger vehicle, implying that the prbability these vehicles will be struck n the side by ther vehicles will be relatively higher. Because the greatest part f the length f a large vehicle is at a different height than the stiff cmpnents f a passenger car, rather catastrphic results ccur frm side impact cllisins. The abve bvius situatins are well knwn but vehicle imprvements t help remedy the situatins are slw t develp. The difficulties appear t be: a) the cst effective slutin is nt bvius, b) there are questins abut the seriusness f the prblem. It is the latter questin that can be addressed by quality infrmatin. Even the nature f the slutin may be mre bvius if the nature f the prblem is better understd. The Level data base fr a jurisdictin can be enrmus cntaining many variables relating t all accidents fr a time perid. The level f detail cannt be expanded indefini tely

8 because the investigating fficers have ther respnsibilities and the main cncern is with supprting infrmatin fr plice wrk. The reliability f the variables recrded under these circumstances is f cncern. A Level study has the ptential t prduce statistically significant results fr certain variables if the segmentatin is nt excessive. Research by Sparks (:7) has shwn the enrmus sample sizes required t prduce statistically significant results fr certain vehicle ~lasses f heavy truck accidents. This is due t the many and ever changing variables assciated with these vehicles. Presently Transprt Canada maintains a Canada wide Level 2 accident study sample f serius passenger car accidents t supprt a data base fr mnitring and develping the Canadian mtr vehicle safety standards. The study was designed as a statistically significant accident sampling plan which wuld prvide data n all invlved vehicles, damage, scene, cnditins, driver and ccupants, injuries and certain special 'cmpnents f interest. All f these data are btained after the accident. This study has definite limitatins because it is nt a s called 'General Level 2' but it des include vehicles ther than passenger cars if they have been invlved. Large/heavy trucks are therefre included t a limited extent. It is the nature f large/heavy trucks t be relatively resistant t structural damage. Trucks are ften driven frm the accident scene while passenger v~hicles, if invlved, may be ttally destryed and remved t a salvage yard. Als, since time is mre valuable fr trucks and their drivers, they are back n the rad as sn as pssible. This adds t the difficulties f lcating and inspecting the vehicles and interviewing the peratrs wh generally receive less severe injuries and are therefre mre mbile. A Level 3 study as described by Wlkwicz (5:7-86), is an indepth n-the-scene investigatin in which investigatrs are n call 24 hurs per day. Often this type f study invlves a sample biased twards accidents invlving extensive vehicle damage as des the Level 2 study mentined abve. There are marked differences in cst between these three Levels f investigatin. Each data cllectin methd has its specific and apprpriate applicatin. Ideally, analysis f the Level data shuld prvide guidance and cntrl fr safety actin and research in the areas, classes f vehicles. etc. which appear t prvide pprtunities fr imprvements. It appears bvius that quality Level data are essential and that careful and systematic analysis f these data shuld be part f all vehicle perfrmance imprvement effrts. As ften happens the bvius is verlked in favr f the esteric.

9 3.0 STUDY ~~THODOLOGY T evaluate the Level data all variables were examined and a limited set was selected and cmpared. Thse selected required that additinal infrmatin be available fr fllw-up investigatin. These variables were als cnsidered t be f majr imprtance as input t safety imprvement prgrams and research. Originally vehicle identificatin variables were t be used nly fr case selectin. Hwever, early analysis shwed that they were nt necessarily cded accrding t cding specificatins. This created case selectin difficulties frm the Level data base and therefre the tw vehicle identificatin variables were added t the list. (Table ). TABLE. Heavy Truck Level and Fllw-up Investigatin Cmparisn variables. 2. Cause and Cntributing Factrs Accident Lcatin 3. Intersectin Related 4. Accident Impact Cnfiguratin 5. Number f Axles- Heavy Truck 6. Pre-Cllisin Vehicle Actin 7. Heavy Truck Identificatin 8. Lcatin f Damage -Heavy Truck The Level data, i.e. plice reprt data, were acquired frm the New Brunswick Department f Transprtatin and transferred t a mainframe cmputer. The accidents cvered 984, 985 and a prtin f 986. Using SPSSX all highway accidents invlving a truck tractr and/r a semi-trailer were selected fr analysis. By specifying "r", many accidents invlving the vehicle identificatin miscdings, as mentined previusly were selected. A ttal f 323 accidents (45 fatal), were then examined using SPSSPC n a micrcmputer. The primary resurce fr the fllw-up r simulated Level 3 investigatin was the plice accident case files. The prcedure fllwed by the plice frce which made its files available, was indepth and fllw-up investigatins f all accidents which invlved fatalities. Fifteen f the abve 45 fatal accident files were nt lcated and cnsequently drpped frm the analyses. The 30 cases, all invlving tractr semi-trailers (TST), were n average, fund t cntain the infrmatin t cnfidently determine fault and in many cases, the factrs cntributing t the accident. Infrmatin used frm the plice files included:. Witness statements 2. Driver statements

10 3. Accident scene diagrams 4. Mechanical vehicle inspectins 5. Autpsy results 6. Tachgraphs (tw ccasins) 7. Investigating fficers summaries 8. Vehicle phtgraphs 9. Scene phtgraphs 4.0 VARIABLE IDENTIFICATION, COMPARISON AND DISCUSSION 4. Y~~!~~!~~~-_~~~~~~_~~~_~~l~~_~~~!~!~~!!~~_~~~!~~~_J~~~~~~ Fur MCF' s are allcated fr each vehicle invlved in an accident. The accident reprt cding manual fr New Brunswick (6:Sect. t ) specifies the first MCF t represent the cause and the remaining three t represent the cntributing factrs. The variable is specific and disaggregated t:. human cnditins; 3. vehicular; 2. human actin; 4. envirnmental. Table 2 shws the causes and cntributing factrs fr the 30 fatal accidents as cmpiled frm the plice reprt and fllw-up data. TABLE 2. Results: Heavy Truck MCF's fr 20 TST Accidents cause/cntributing factr Level frequency cause c.f. Fllw-up frequency cause c.f. driver inattentin extreme fatigue driving t fast fr rad cnditins passing/lane usage imprper pedestrian/errr cnfusin fllwing t clsely mechanical failure (trailer tandem) jackknife lad shift surface slippery view bstructed hydrplaning uninvlved vehicle nt applicable unknwn

11 There were eight accidents in which the vehicle appraching the TST simply crssed the center line fr n apparent reasn and struck the TST. In many cases the rad surfaces were dry, autpsies revealed n alchl r drugs and subsequent mechanical inspectins cleared the vehicle f any mechanical defect. Causes appear t include driver inattentin, distractin and fatigue. In these cases, althugh fault was easily determined, the actual cause was nt and therefre required the authrs assessment based upn persnal experience and available data. On n ccasin was alchl r drug use fund t be a factr fr the truck peratr. Table 3 cntains a summary f the frequency f faul t t an invlved vehicle peratr by Level as cmpared t data extracted frm fllw-up investigatins f the 30 TST accidents. Table 3. Cmparisn: Fault Assignment 30 Fatal TST Accidents Level Level frequency TST ther unkn frequency TST ther unkn All Accidents (30) Single Vehicle Accidents (5) Multi-Vehicle Accidents (25) The Level data reprted "unknwn" as the cause fr the TST in fur cases. Further investigatin revealed that the tractr driver was nt at fault in any f these accidents and that the crrect cding wuld be "nt applicable". 4.2 Variable: Accident lcatin Since heavy truck expsure is greatest n highways it was reasnable t assume this is where mst f the accidents wuld ccur. Level results revealed that 23 f the 30 fatal accidents ccurred n rural prvincial highways and six n urban prvincial highways. The remaining accident tk place n rural private prperty. The fllw-up study revealed that the majrity f accidents (26) ccurred n rural prvincial highways. Three tk place n urban prvincial highways while ne ccurred n rural private prperty. Three cding errrs were fund when cmparisns were made fr this variable. The errrs were made cnsistently and invlved a highway sectin being cded as "highway urban" when cding shuld have been "highway rural".

12 4.3 Variable: Intersectin Related The cding manual defines an intersectin related accident as ne which ccurs at an intersectin and where at least ne vehicle was entering r exiting the intersecting radway. There are very few miles f divided, lw access highways in the study area. In fact there are ver 2000 entrances/exits nt the Trans Canada Highway in the study area, a highway that handles a significant amunt f heavy truck traffic. Results frm the accident reprt data revealed nly seven TST accidents t be intersectin related. The fllw-up investigatin identified nine accidents t be intersectin related. Six f these accidents ccurred in areas where the highway intersected with a rural rad. Three f these invlved vehicles entering the highway frm the secndary radway with the ther three invlving vehicles slwing and in the prcess f making a left turn. The ther accidents ccurred at a railway crssing, private driveway and cmmercial driveway entrance. Seven accidents intersectin related investigatin. frm Level cmpared analyses with nine were fund t fr the Leve be 3 The accident cnfiguratin variable describes the nature f the first impact in an accident. Level analyses f this variable indicated that head-n cllisins (7 ccurrences) and angle cllisins (5 ccurrences) ccurred mst ften. Sideswipe-type impacts ccurred in three cases. Three cllisins ccurred where the investigating fficer was unable t define the cnfiguratin (i.e. the cnfiguratin culd nt be represented by ne f the available cdes). The fllw-up investigatin revealed that head-n cllisins were the mst cmmn type f impact ccurring in cases. The right-angle cllisin type ccurred seven times, and the "ther" cnfiguratin cding was used fr five accidents. The remaining seven accidents invlved five difference impact cnfiguratins. The five impact cnfiguratins cded as "ther" invlved tw types f cllisins. The first ne was a jackknife type f impact. On bth ccasins the semi-trailer swung arund int the ther lane and struck an ncming vehicle. The ther three invlved TST lad shift cases. Lads spilled were 50 cm diameter steel pipe, bundled lumber. and gyprc sheets. The cding manual des nt clearly and specifically address the prper cnfiguratin cding fr these five accident types and therefre requires the use f the "ther" categry fr the accident cnfiguratin cde. The use f the "ther" categry des nt prvide sufficient data t track such imprtant factrs as lad shift. Cmparisns f these results indicated nine cding errrs. These errrs were fund t exist in nly the multivehicle accidents. Tw f these nine miscded accidents shuld have been

13 cded as head-n cllisins but were miscded as sideswipe cllisins. Tw thers cded as turning cnflict accidents invlved semi-trailer jackknife and subsequent impact with the ncming vehicle. Cding f this variable fr these fur accidents appears t be based upn the cnfiguratin schematic placed n the plice reprt instead f the cding manual definitin. 4.5 Variable: Number f Axles The plice reprt cding manual states the number f axles recrded fr a heavy truck shuld be the ttal number f axles n the vehicle including the twed semi-trailer, if ne is present. Typically a TST will have five r six axles, three fr the tractr unit and tw r three n the semi-trailer. Analyses revealed that nly 20 accidents were recrded with the TST having five r six axles n the TST. Of the remaining 0 accidents, there was ne listing tw axles, eight reprting three axles and ne listed as fur axles. The authrs investigatin, based n actual phtgraphs f the vehicles invlved revealed that all TST had five r six axles. This certainly was expected as accident selectin was made knwing a TST was invlved. Cmparisns in the cding f this variable revealed errrs in 0 cases. Eight were cded as having nly three axles. These errrs shuld nt be present in the data since selectin criteria was based upn a TST and there are very few three axle units in this regin. 4.6 Variable: Pre-cllisin Vehicle Actin Pre-cllisin vehicle actin is intended t indicate the vehicle mtin prir t an impact, an evasive maneuver r lss f cntrl f the vehicle. Level results revealed the TST t be mving straight ahead in 25 f the 30 accidents. The remainder f the cases invlved the fllwing vehicle actins; vertaking (), changing lanes (), turning (2) and slwing r stpping (). Accident files indicated that n 27 ccasins the mtin f the TST prir t impact, was "ging straight ahead". On tw ccasins the TST was passing anther vehicle and was subsequently invlved in an accident. On ne ccasin the TST was slwing in preparatin t make a left turn nt a secndary rad when it was struck frm behind by a car. Fur cding errrs were fund t exist when these variables were cmpared. Tw f the fur miscded accidents shuld have been cded as "ging straight ahead" and "slwing r stpping". The latter f the tw required familiarity with the cding manual. The tw remaining accidents invlved pre-cllisin actins that were cded with the crrect intent. They, hwever, did nt depict the precise mtin that the researcher was able t determine frm

14 the accident files. The final accident invlved a TST vertakina a mtrized tw wheel vehicle. Plice cding indicated the TST t6 be "changing Lanes" while the fllw-up investigatin revealed that the lane change had been cmpleted. 4.7 Variable: Vehicle Identificatin Fr heavy trucks this variable requires tw cdes, ne t identify the drive vehicle and ne fr the twed vehicle. Reference was made earlier in this paper t the prblems encuntered when this variable was examined. Table 4 lists the varius cnfiguratins as indicated frm the accident reprt data. TABLE 4. Level Results - Tractr Semi-Trailer Identificatin Cdes Cde Name f Vehicle Identificatin Cde Frequency 4-7" 5-8" 5-7" 5-9" 5-0" 5-97" 5-98" truck> 5000kg w/semi-trailer (flat bed) tractr semi-trailer (van) tractr semi-trailer (flat bed) tractr w/duble trailer tractr semi-trailer (tanker) tractr w/nt applicable fr twed unit tractr w/unknwn t twed unit The fllw-up investigatin revealed that 20 invlved TST (van type semi-trailer), nine invlved beds) and ne invlved a TST (tanker). accidents TST (flat I f safety researchers are t address a specific rad user grup they must be able t cnfidently identify and perhaps islate the grup fr analysis. The analyst wuld, in all prbability assume that case selectin, based upn the tw vehicle identificatin variables, selected all vehicles f cncern. The cases nt selected due t miscding, shuld nt be significant enugh t affect results. This hwever was nt fund t be the case. A significant number f rare truck cnfiguratins and bbtail units were fund t be present in the Level data. Fr example, 228 "trucks ver 5000 kg were fund t be twing semi-trailers and 50 bbtail units were identified. These researchers are cnfident based n ther traffic studies that these numbers are nt representative f the true heavy truck cnfiguratins invlved in accidents. Unfrtunately, n ther data exists t determine the actual cnfiguratins f units invlved in accidents. Using the data in the existing accident reprt files wuld result in errrs in significant magnitude.

15 4.8 Y~E!~~~~ ~~~~!!~~_~!_~~~~~~ The reprt cding manual allws recrding up t three damaged areas fr each vehicle invlved in an accident. The first area recrded is intended t represent the first area cntacted during the cllisin. The tw remaining variables lcate ther damaged areas, if applicable. T address the prper cding fr a vehicle twing anther (e. g. TST) the cding manual s ta te s that" 99 " shuld be used t indicate that a twed vehicle was struck. Analysis f the frequencies f this variable representing initial cntact revealed the frnt end f the truck tractr was struck n 6 ccasins. The remaining cases were apprximately evenly distributed amng the ther available cdes. On ne ccasin nly, was the "99" used indicating the semi-trailer t be the initial cntact bject. Examinatin f accident file data revealed the frnt end f the TST t be the primary impact lcatin n 6 ccasins. Ten accidents invlved primary impact n the TST and the remaining cases were evenly distributed amng ther available cdes representing damage t the tractr unit. Cmparisns revealed 0 cding errrs fr the "primary lcatin f damage" variables. Results are detailed in Table 5. In eight f the 0 miscded accidents primary impact frm the ther vehicle was made with the semi-trailer. The abve prcedure f cding "99" t indicate cntact with a twed vehicle is TABLE 5. Results: Lcatin f Damage t TST Damage Lcatin frequency Level frequency fllw-up * right side (truck tractr) rear end (truck tractr) left side (truck tractr) frnt end (truck tractr) semi-trailer undercarriage (truck tractr) rf (truck tractr) mre than 3 areas * -right side represents the passengers side essential t prevent damage lcatins frm the twed vehicle being misrepresented as damage t the vehicle ding the twing. These fur accidents were cded as representing primary damage t the

16 tractr. Frm inspectin, it became apparent fficers were cding withut the knwledge f the "99" cding prcedure. On tw ther ccasins the number "2" was cded as primary impact indicating mre than three areas were damaged. The manual states that this cding can nly be used fr the third f the three available cdes when mre than three areas sustain damage. 5.0 EXAMINATION OF THE PCS DATA FOR INFORMATION ON LARGE TRUCKS The ther data examined were frm the Passenger Car Study (PCS). These can be summarized as fllws: Number f accident cases by year are: The ttal number f cases in the file at time f analysis was The target set was 000 cases per year. The extent f the data base is evident by the fact that each case can ptentially cntain 24 data frms. These are listed belw t indicate the breadth f data cllected.. Case 3. Seat-back 2. Scene 4. Entrapment 3. Vehicle 5. Defect 4. Damage 6. Rear-seat 5. Nn-ccupant 7. Instability 6. Carg 8. Steering 7. Occupant 9. Fire 8. Injury 20. Tire 9. Trailed vehicle 2. Braking 0. Child restraint 22. Rllaway. Driver 23. Ejectin 2. Active restraint 24. Case management The number f cases referenced t each f the variable values in the case files als gives an indicatin f this data resurce. Table 6 presents these data.

17 TABLE 6. Number f Cases Cntaining Data n Variable Variable N.f Cases variable N.f Cases n answer ppulatin less than 2S00 ppulatin 2S00-S000 ppulatin SOOO-lOOOO ppulatin S000 ppulatin 2S000-S0000 ppulatin SOOOO-lOOOOO ppulatin S0000 ppulatin greater than 2S0000 fatal accident injury accident pedestrian/cyclist mtrcycle passenger car truck less than 5 tns truck 5-25 tns truck greater than 25 tns bus train fixed bject mveable bject ther bject unknwn bject ne vehicle tw vehicle three vehicle fur vehicle five vehicle six vehicle seven vehicle eight vehicle unknwn vehicle rllver submersin hit & run nn-cllisin fire ther nn-cllisin rear end head n rear rear angle deg. t-intersectin side swipe and side swipe directin ppsite directin ther cllisin unknwn cllisin n radway n shulder in median n radside ut f right f way parking lt bus lane ther lcatin unknwn lcatin n descriptin descriptin These data are sampled fr a number f jurisdictins by ten teams lcated in eight f the ten prvinces t attempt natinal cverage. The cncept is t expand the data using lcal sampling infrmatin and the prvincial Level data. It becmes clear that the level f precisin pssible will be 9W and the quality f the Leve data make s the accuracy fr large trucks suspect. Hwever, estimates prduced by this prcess are the best pssible

18 are the best estimates available at this time. Fr example, there are 56 cases which include data n heavy trucks. The Teighted number fr this variable is Similarly the cases that include fatalities are 933 with the weighted number being The nn fatals number was 507 with the weighted number This indicates that the sampling rate is abut 38 t. Prperty damage nly cases are nt included. The rati f fatalities by invlvement with varius vehicles was calculated frm the sample weighted pes data. The ratis are presented in Table 7. These ratis cnfirm a much higher rate f fatalities when a large/heavy truck is invlved. TABLE 7. Rati f Fatalities by Vehicle Invlvement Descriptive Invlvement Rati Pedestrian, cyclist and fatal by pedestrian cycle Mtrcycle and fatal by mtrcycle Passenger car and fatal by passenger car Truck less than 5 tns and fatal by truck less than 5 tns Truck 5-25 tns and fatal by truck 5-25 tns Truck greater than 25 tns and fatal by truck greater than 25 tns Bus and fatal by bus Train and fatal by train Fixed bject and fatal by fixed bject Mveable bject and fatal by mveable bject Other bject and fatal by ther bject Unknwn bject and fatal by unknwn bject FINDINGS 6. New Brunswick Level Data Analvsis ~--- The research n the data cllected in the study area using the Level, accident reprt data revealed numerus errrs in reprting f accidents invlving heavy trucks. The cnclusin was reached that the present quali ty cntrl fr Level accident reprt data is inadequate and the use f these data fr research and analyses f truck accidents culd yield misleading results and subsequent incrrect cnclusins. The analysis revealed 45 cding errrs n eight variables examined in 30 TST accidents invlving heavy trucks. Many f these errrs appeared t arise frm the fficer nt being familiar with the cding manual fr the accident reprt frm.

19 Five f the eight variables examined were fund t be cded inadequately and if data n these variables were used in safety research they culd nt have cntributed in a psitive manner t any study. The five variables, which wuld likely be requested fr mst research prjects, were: - accident fault assignment - heavy truck identificatin - heavy truck number f axles - accident cnfiguratin - heavy truck lcatin f damage Cuntermeasures based upn analysis f data n these variables wuld be extremely misleading and thus have ptential t direct safety research in the wrng directin. five the The peratr f the heavy truck was fund t be "at fault" in ut f 30 fatal accidents examined. The cause and cntributing factrs were determined t be human related in eght f these ccurrences. The remaining three were lad shift cases with ne being related t mechanical prblems. These cnclusins relate t the specific study area. Hwever, a review f perating prcedure in ther jurisdictins indicate that the prbability is high that the quality cntrl in these areas is basically similar t that fund in this study area. The PCS data cnfirmed the high invlvement rates in fatali ties relative t heavy vehicle/truck accidents. The data base cntains mre detail related t heavy vehicles than the study name implies. Analysis f these data cnfirms that heavy vehicles are a significant factr in verall highway safety by the frequency and nature f their invlvement in cllisins with passenger cars. The PCS technique culd be mdified t better reflect safety aspects f heavy vehicle use and an expanded analysis f the pes data is required t extract the available infrmatin (especially frm the case phtgraphic recrds). The extent f the 'unknwns', and 'thers' related t heavy vehicle data indicate the prblems with btaining infrmatin n this class f vehicle by the technique used. While the data d nt prvide reasns fr these difficulties, it is knwn that cmmercial vehicles in general tend t be returned t service very quickly s that the investigatin respnse time must be much quicker than fr passenger vehicles.

20 7.0 RECOMMENDATIONS The research reprted in this paper has generated several recmmendatins cnsidered t have the ptential t increase the quality and therefre, use f accident reprt data.. ~~!~~~~_~~E~E!_~~E~! S!~E!E!~~~!~~L~~E~~~!~~ - The need exists t clarify certain variable definitins in the accident reprt cding manual t reduce interpretatin errrs (e.g. accident cnfiguratin, pre-cllisin vehicle mtin, majr cntributing factrs, number f axles, lcatin f damage). The data cllected shuld be expanded n the accident reprt frm t enable additinal, reliable and useful data assciated with heavy truck accidents t be cllected n a large scale basis. Variables added shuld be relatively simple, factual and within the capability f traffic fficers. Pssible additins include: verall accident fault. if a vehicle crssed the centerline lad (if applicable) type and mass grss vehicle weight driving experience with this particular vehicle cnfiguratin hurs driving (e.g. less than 4, 4 t 8, mre than 8 hurs) number f lanes, divided r undivided, high/lw access driver familiarity with this rute driver relatinship t truck tractr, e.g. wner peratr, cmpany driver heavy truck distance travelled per year Natinal Safety Cde reference number - Varius jurisdictins shuld be encuraged t prduce and adpt a frm which is unifrm and mre cmplete. The cperatin shuld exist within Canada and between cuntries allwing these data t be cmbined and/r interchanged. Accident reprt data reliability shuld be cnsiderably and mnitred clsely t ensure analyses data yield wrthy results. imprved f these Plice fficers shuld receive cding manual refresher curses as ften as required t maintain a cmprehensive understanding f accident reprt variable definitins. These persns shuld be infrmed n a regular basis f cmmn cding errrs being identified by thers.

21 - An extensive aut-check system shuld be put in place t prvide quality cntrl n accident data as they are entered int prvincial cmputer files. Use shuld be made 0 f the Transprt Canada spnsred "Passenger Car Study" data t mnitr the accuracy f accident reprt cding fr a selected class f accidents. - Heavy trucks are very unfrgiving. Given, they are t imprtant t Prvincial ecnmies t reduce their mass r length, their frnt end culd be made mre friendly. Side and rear guards culd be placed n the trailer t prevent underride situatins. - Frnt brakes shuld be installed n all tractr units. - Prmtin fr the truck units shuld be braking characteristics needed, especially fr special hazard. use f anti-lck braking systems n heavy undertaken. A general imprvement in fr bth laded and unladed trucks is unladed tractrs which appear t be a 4. ~~~~_!~~_~~~~!~~~~~~_~~!~~~~~~~~~_~~~_ ~~~~X The data cntinue t shw that driver actin can be cnsidered as the majr cntributing factr t safe peratin f highway transprtatin. Sme level f driver errr has t be acceptable but situatins with serius ptential cnsequence shuld be minimized. The data als shw that high levels f detail are required t mnitr the effects f nrmal peratins, cuntermeasures, design and peratinal changes, etc. Reliable lcal and natinal Canadian data with sufficient cverage f details t imprve the levels f transprt efficiency and safety requires mre effrt and better crdinatin than nw prevails. As stated previusly, the Level I data system shuld be supprted with mre rigrus quality cntrl. It shuld be supprted by a general Level 2 system and there shuld be sme Level 3 activity t prvide special details as required. These data shuld be linked t ther infrmatin surces including medical, curt, insurance, and ther recrds s that a mre cst effective data system can be develped. In particular, the results f the recen t changes in heavy vehicle regulatins shuld be mnitred by prviding a Level 2 r 3 study f this class f vehicles. This culd be btained by mdificatins t the current passenger car Level 2 study, that is by making it a general Level 2 study with special sampling rates fr accidents invlving heavy vehicles. The ecnmies f scale fr this mve wuld be significant.

22 REFERENCES. The Safety Experience f Large Trucks in Saskatchewan, Grdn Sparks, University f Saskatchewan, prepared fr Saskatchewan Highways and Transprtatin and Transprt Canada, Truck Impact n Radway Safety, Pus, Abishai and Mahalel, David, Transprtatin Research Bard Safety Implicatins f Truck Cnfiguratins, 0iver Carsten, Transprtatin Research Bard 987, Transprtatin Research Bard. 4. Big Trucks and Highway Safety, Insurance Institute fr Highway Safety, On-the-Scene Study f Cmmercial Vehicle Accidents, Prceedings fr the Internatinal Sympsium n Heavy Vehicle Weights and Dimensins, June 8-3, 986, RTAC. 6. Mtr Vehicle Accident Infrmatin System, Reprting Manual fr New Brunswick, Secnd Editin Nvember 987.