ACCIDENTS STUDIES AND COLLISION CHARACTERISTICS By i r . H .G . Paar and i r . L . T . B . van Kampen INSTITUTE FOR ROAD SAFETY RESEARCH SWOV, THE NETHERLANDS

1 . ACCIDENT STUDIES 1 . 1 . Def in i t i on o f acci dent study The inve s t i gation o f accidents i s the gathering, storing, adapting and analysing o f a c c ident data wi th the purpose to improve road safety. Accidents can bappen to every kind of road-user but in thi s paper mainly car acci dent studi e s are menti oned, s inc e car acc ident s , i n most c ountri e s , are re spons ib l e f or the greater part o f the traff i c casual t i e s . The human , vehi c l e and envi ronmental e l ements o f acci dents are covering the complete area of all factors concerned in acc i dent s . Result s o f a c c ident s tud i e s are directly dependant o n whi ch el ement or combinati on o f e l ements i s s tudied . A far more important and interest ing di st inct i on can be made i n the sequence of facts during actual acc idents ( c o l l i s i ons ) and therefore in the approacb of acci dent s tudi e s , as menti oned in the next paragraph .

1 . 2 . Pre crash versus crash s tudi es

The sequence o f what i s happening i n acci dent s , can be divided i n t o three main phase s : pre crash , crash and po st crash phase . In actual c o l l i s i ons these pbase s can be defined very accurately , for whi ch we have to thank the former USA Nati onal Hi ghway Safety Bureau, who i nvented the so cal l ed Program Matrix for Hi ghway Safety Research (Figure 1 ) . In thi s c e l l matrix the important areas can be easily detected. The main di st inction as far as acci dent studi e s are concerned l ays between the pre crash and the crash phase . In the pre crasb phase one i s c oncerned with accident avo i dance or rather acc i dent causat i on , since we are talking about real-acc ident studi e s . In tbe c rash phase one i s concerned with injury preventi on or rather injury causat i on . In the po st crash phase one i s concerned with injury severi ty reduct i on , or rather wi·th c i rcumstance s that have influence on injury severi ty. In thi s paragraph only the main di stincti ons between pre crash accident studi e s and crash accident stud i e s wi l l be considered.

As already stated, i n pre crash accident s tud i e s tbe i nvest igators are l o oking for acc ident causati on . Thi s i s a parti cularly tri cky f i e ld since the definition o f the a c c ident cause i s n o t a simpl e one , i f there i s any. And even then , the most important part o f accident causat i on i s contributed to the human e l ement, so far mo st studi e s agree , but thi s e l ement i s the tri cki est o f al l . Pre crash studi es

1 5 3

have to deal with the most unpredi ctabl e factor i n traf fi c : the human being and i t s behavi our . Even i f those stud i e s were concerned mainly wi th vehi cLes or envi ronmental factor s , the human influence can always spo i l everything, unl e s s robot s are used for driving. Be s i de s , another very important factor in pre crash accident studi es i s the need of accurate expo sure data. Tho se are by no means easy to gather , but are abs o lute ly nece ssari ly for comparing the acc ident popul ation wi th the non acci dent popul ati on.

Crash acci dent s tud i e s are concerned wi th injury causati on . Of course there i s a human factor again in thi s area ; in fact these stud i e s are concerned mostly wi th the human e l ement , but i t i s qui te di fferent from the pre crash human e l ement . Human behavi our during the so cal l ed second c o l l i s i on (the crash phase i n whi ch the o ceupant moves rel ative­ly to the car interior and h i t s parts of i t and becomes injured ) i s so predi ctable that already many mathemati cal ( computer ) mode l s , de-scribing behavi our , are exi stent. For the s tudy of injury causat i on , the acci dent i nve stigator needs : a description o f the acc i dent general ; a description o f the exterior and i nteri or damage to the car s ; a descripti on o f the injuries o f oc cupants or o ther road users in­volved . The se f i e lds are covered i n the paragraphs 2 . 1 t o 2 . 3 o f thi s paper .

lt might be strange for the study o f human tol erance data, f or whi ch IRCOBI stand s , c rash acci dent studi e s cannot give conclusive data, s ince the mi s s i ng factors in most crash studi e s are the exact decel er­ati ons and forces that actual ly caused the injur i e s or fatal i t i e s . In paragraph 3 "Final remarks and recommendati on s " s ome poss ible s o lut i ons to thi s probl em are menti oned.

However , crash acci dent s tud i e s do give answers to many important questi ons : 1 . They give the i njury pattern for the studi ed accident sample and

thi s points straight to the areas o f the human body that have to be protected. They a l s o give the parts of the car ( interior ) that are respons ib l e for the injuries and the se parts al s o wi l l form a pattern with main c ontributors ( l ike steering co lumn and windsh i e ld ) , that shoul d be improved.

2. Crash ac c ident s tud i e s wi l l give effects o f certain new features and improvements l ike energy absorbing steering co lumn s , padding material and the use of seat be lt s .

Mo st o f the stud i e s menti oned in the f o l l owing paragraphs wi l l be crash acc i dent s tudi e s . The authors acknowl edge however , that future pre crash acci dent studi e s , i f properly organised , might be more effect ive than crash stud i e s , s i nce the resul ts are bound to take away the cause s o f acc ident s . But for the near future , crash acci dent s tud i e s are more effective , s inc e they do n o t have to deal with . the sti l l un­predi ctable human behavi our . And even i f human behavi our during the second col l i s i o n i s not qui te uni form due to di fferences in age , sex , 1 ength, seat p o s i ti on etc . , the se data can be recorded eas i ly and the i nfluence can be analysed.

lt i s also true that, whatever ( legal ) measures wi l l be taken i n the near future , acci dents wi l l go on happening and therefore studi e s into injury reduction are neces sari ly.

1 . 3 . Class i f i cation o f acci dent stud i e s by gual i ty/guantity rat i o

The perfect acci dent study covers a l l acc idents o f a certain country in a certain t ime and gives complete indepth informati on on all o f these acci dent s . Both quantity ( the amount o f di fferent cases or accident s ) and qual i ty (the amount o f d i fferent data for each case and the ir degree o f rel i ab i l i ty ) have t o b e 1 00% for the perfect acci dent study. Of course thi s i s a non-exi stent s i tuati on, even in the smal l e st countr i e s . Generally spoken qual ity and quantity are inversely proporti onal to each other . A study with the h i ghest quanti ty (a l l acci dents that happened during a c ertain time ) wi l l have the l owe st qual ity . The studying o f a nat i ons acci dent stati sti c s might be cal l ed such a high quanti ty/l ow qua l i ty study. In such a i nve stigati on the. gathering o f the data i s n ot part o f the acci dent study, s i nce i n most countri e s tho s e data are gathered primarely for general purpose s . I n most real­acci dent stud i e s the se general acci dent stat i st i c s wi l l form the reference material to whi ch the study sample can be compared and adjusted. An example of the other extreme i s the indepth mult i-d i s c i pl i nairy study o f a smal l number o f spe c i f i c acc i dents wi th teams go ing to the scene , us ing all techni cal means to gather all ava i lable data ( l ike tape-recorders for interviewing and reporting, and camera ' s for the recording o f car damage etc . ) , thus resulting i n a l ow quanti ty/high qual i ty study. Even the best teams , or better just the best teams , cannot gather much more than about 50 accident cases a year , as pract i ce o f many acci dent teams wi l l have learned .

Now here we have the main prob�em related to h igh qual i ty ( i ndepth ) studi e s : the sample i s too small for any kind o f stat i st i cal analysi s . And s ince the main purpose o f acc ident studi e s i s to improve traff i c safety by it s re sults , i t i s imperative that the se resul ts are based on a stati sti cally rel i able analys i s . Even though the data themselves mi ght be very rel i abl e , the sample wi l l be nearly always too small for overall rel i abil ity and tho s e who are responsible for l egal or other traff i c safety measures need r e l iable informati on . Of c ourse thi s i s rec ogni sed by mo st acci dent i nvesti gators and some­time s increas ing the sample by studyi ng l e s s i ndepth or be i ncreas ing the number o f teams , whi ch i s a very expens ive method, i s cons i dered. For stati sti cal analysi s a minimum sample of 1 000 to 1 0 . 000 cases have to be i nve st igated, dependant on how many data relat i ons the analyser wi l l consider s imultaneously, and what l evel of r e l iabi l i ty wi l l be accepted for s i gn i f i cant di fferences . Therefore the mo st valuable acci dent stud i e s of the world have sample s of over 1 0 . 000 cas e s .

However , the i ndepth study o f acci dents mi ght b e give answers to spe c i f i c quest i ons , resul ting from l arger acci dent studi e s . Espe c ially when tho s e studies are concerned wi th one make and model

1 5 5

car , sometime s even with one product i on year and with spe c i f i c defined c o l l i s i on types ( frontal only, or rol lovers only ) they might be give the exact answers that the i nvest igators ( o ften manufacturers ) are l o oking for , for instance an answer to the quest i on wether their steering column i s working properly in actual co l l i s i ons or not .

The qua l i ty o f big sample acc ident studi es ( 10 . 000 cases or more ) can j . vary, s ince i t i s dependent on the way the acci dent data are gathered and by whom. Data can be gathered by pol i c e , by other road patrol s , by spe c i al teams even, or they can be supp l i ed by drivers themselves as we wi l l s e e i n the examp l e s . The more di stance there i s between analyser and acci dent in terms of different reporters , the l e s s rel i able the data mi ght become . Qual i ty can s t i l l be fairly h igh as compared to i ndepth i nf ormati on , especially i f the organi sati on has been careful ly p lanned and not more i s asked for than can be gathered rel iably.

1 . 4 . Clas s i fi cation on acci dent i nvestigators

The i nves t i gati on and data reporting on one band and the data a�alysing on the o ther can be done by pebple o f d i fferent d i s c ipl ine s , as we wi l l s e e i n the examples (paragraph 1 . 5 . ) . But o f course tho se who f inally analyse the data and draw the ir conclusi ons are the ones responsibl e for the s tudy as a who l e . Obvi ous categor i e s for acci dent i nve stigating inst itute s are : 1 . Scient i f i c insti tute s , inc luding hospi tals and univers i t i e s ; 2 . Manufacturers (mostly car manufacturers ) ; 3 . Pol i c e ; 4 . Governmental resort s , excluding pol i ce and univers i t i e s ; 5 . Other inst itut e s . Another c lass i f ication i s p o s s ible a s far a s the d i s c ipl ines o f the inve st igators are concerned . Mo st current i nvest igators can be devided into three main di s c ipl ine s : 1 . Techni cal (veh i c l e and traff i c engineering ) ; 2 . Medi cal ( surgery ) ; 3 . Psycho l ogi cal (also s o c i o l ogi cal ) . In mult i di scipl inairy teams one wi l l f ind at l east i nve st igators of d i s c ipl ine 1 . and 2 . , no matter whi ch category o f inst itutes i s c oncerned . Psycho l og i st s are needed for pre crash s tudies only.

1 . 5 . Exampl e s o f important acc ident studies

The example s menti oned i n thi s paragraph are chosen for d i fferent reasons . In the f irst place all of them were completed or nearly completed and reported. In the second place they all show some interes t i ng character i s t i c s both in their organi sation and in the ir resul t s . They have been chosen out o f di fferent countri e s ; s t i l l the cho i c e i s l imited, due to l imited space for thi s paper. Six d ifferent stud i e s are described shortly i n the next paragraphs . In paragraph 2 . 4 . s ome o f the i r co l l i s i on characteri sti c s are compared.

Thi s we l l known study (ref . 1 ) i R not a current s tudy anymore , although the

results can be s t i l l very useful . Data were gathered in Sweden during 1965 and 1966 . Only Vo lvo car acci dents were cons i dered and even more than 28 . 000 cases were examined. Thi s Volvo study i s an example of a crash s tudy using compl eted que sti onaires o f drivers as bas i c material and al s o using p o l i c e and ho sp i tal reports for addit i onal information. The sample was qui te good for stat i st i cal analysi s .

The German Insurance Group , called HUK Verband, has studi ed accidents that o c curred in 1 969 and were reported to the German car insurance compani es (ref . 2 ) . A f i rst sample o f 1 0 . 271 accidents wi th o ccupant injur i e s were considered. Data were analysed by experts on car damage and injur i e s . Apart f rom data o f insurance report s , data from pol i c e and ho spital s were used. The reports yet publ i shed ( I and I I ) can be cons i dered as examples of a typi cal crash study. The materi al i ts e l f i s al s o usable for pre crash analysing. No data about the amount of work needed for thi s s tudy and the re l iabi l i ty have been publ i shed yet .

Thi s on the spot survey of acci dents i s the only typi cal pre crash acc ident invest igat i on menti oned wi thin thi s set of example s (ref . 3 ) . l t shows a l l the characteri sti c s o f such stud i e s . The sample (gathered over a seven month;; peri od ) i s too smal l f or stati sti cal purposes but qui te b ig for on the spot work wi th a team. Thi s i s the due to the fact that more than one team was ava i lable and that the working hours were 2 4 hours a day , and that not a l l a c c idents were studi ed compl etely indepth. The area was a mainly rural area around the laboratory.

Thi s acci dent s tudy o f injury and noninjury crashe s in rural Utah (ref . 4 ) i s i ntere sting becaus e the effect o f the use o f seat belts coul d be analysed in re lati on to al l crashes . The ac c idents happened from September 1966 through August 1 967 and were reported by po l i ce on spec ial forms to Cornel l ; photographs o f vehi c l e s and medi cal reports from medi cal and hospital authori t i e s were included. ACIR devel opped a system of coding the se data ; the pro c e s s ing f o l l owed the general patterns already devel opped. The samp l e contains 7 . 125 car drivers .

The re sults of thi s stati sti cal re search on the effects o f safety belts and on injury causati on have not been publ ished yet . A f inal report i s expected at the end o f 1973 . During the end of 1968 and the years 1 969 and 1 970 accidents involving passenger cars were reported t o SWOV by Dutch pol i ce and hi ghway patrol o f f i cers . The area was sel ected to represent most parts of the country. Rural and urban acci dents were reported. General acci dent data and general o ccupant data were gathered by means o f forms sent to the drivers who completed and returned the s e . Overal l response was more

1 5 7

than 50% and the working sample consi sts o f 22 . 082 dr1 vers . Medi cal data were c o l l ected after the accident by a co-operating medical insti tute . Car damage data were reported by special teams who vi s ited the cars after the acc ident whenever thi s car could be found , whi ch happened in 35% o f all cas e s . In all 2 2 . 082 cases a damage report was made by the spe cial teams from the general driver data, for whi ch purpose the f orm was also devel oped. An analys i s us ing multivariate analys i s te chni c s i s carried out.

The internat i onal study whi ch i s part of the big Road Safety Programme o f the Committee on the Chal l enges o f Modern Soc i e ty o f NATO i s a p i l ot study, whi ch means that the results are not directed towards improve­ment of road safety primarely, but towards improvements of acci dent inve stigations method s . The purpose in the study i s the gathering o f acc i dent data (mainly crash data, but pre crash data can be examined voluntary ) by means o f on the spot examination with multid i s c ipl inairy teams . About 9 countri e s are partic ipating. Data are reported in a special c o l l i s i on forn and the study i s already on the point that f inal dec i s i ons about how to carry on in a f o l l ow up study can be made . The mo st important i tem o f thi s study, apart from the internati onal character i s the co l l i s i on analys i s report f orm, which i s s imi lar f or all parti cipating team s .

Worth menti oning are several current studie s o f the world . 1 . There is in the USA a nati on wi de tri-l evel study, in whi ch three

type s , each wi th a d i fferent qual i ty/quantity relation, are combined al l the advantage s of each seperate study. Since in 1966 the Hi ghway Safety Act and the Traff i c and Motor Vehi c l e Safety Act became operati onal , many valuabl e acci dent investigation programme s have been started i n the USA, and thi s tri-l evel programme i s one o f them . Reports on i t s indepth mul tidi sc ipl inairy team studi e s are currently avai lab l e all over the world ; mo st of the reports are deal ing with pre crash , crash and post crash factors s imultaneously. 2 . Another interesting "mul tidi sciplinairy" study i s formed by the

c ombined e fforts of Wayne State University , USA ( Pro f . Patri ck ) and Volvo Sweden. From thi s investigati on i s not much informati on avai labl e . Purpo s e o f thi s study i s the determinati on o f human tol erance data by the combined methods o f actual accident inve sti­gati on and tests in the laboratory. 3. NATO - CCMS, already menti oned, i s carrying more pro j ects in whi ch

many countri e s o f the wor ld ( including Japan and Sweden, being no NATO countri e s ) are partic ipating . Worth menti oning i s the Experimental Safety Vehi c l e ( ESV) programme, in whi ch many acci dent studi e s were carried out by manufacturers or others . 4 . Many other institutes , all over the world , are working on the i r own

acci dent inve stigati on programme and i t i s one o f the aims o f thi s paper that more co-operation wi l l be reached � since i t i s proved that i t can be done . ( See the NATO - CCMS example ) .

2 . COLLISION CHARACTERI STICS In the first three paragraphs of thi s chapter , data needed for accident investigation are discussed. In paragraph 2 . 4 s ome of the actual c o l l i­s ion characteristic s o f the 6 main examples ,ment ioned in 1 . 5 , wil l be compared, or at l east put next to each other .

2 . 1 . Data c la s s i f i cation

Data needed for accident investigat i on can easily be' div±ded into three main categories : 1 . General accident and occupant data; 2 . Car ( damage ) data ; 3 . Injury data. For crash ac c ident studies a l l three categorie s of data are needed ; for pre crash s tudies detailed data are needed o f category 1 , including a des cription of the maintenance condition of the car before the accident . In the next two paragraphs , data needed for crash accident s tudies are dealt with .

2 . 2 . Review of important data and the ways of obtaining them

General accident data date and l ocation of acc ident overal l c o l l i s ion type ( frontal , s i de , etc . ) c o l l i s ion s equence description ( c onfiguration ) col l i s ion speed and/or driving speed before the accident .

General occupant data ( for each occupant ) seat position age sex description of availab l e restraint system use o f restraint system.

Car (damqge) data general car characteri sti c s {make , mode l , ei c . ) car damage � every part o f the car s eparately, interior and exterior ) damage c ausation.

Injury data type and s everity of every s ingl e injury to each occupant .

The ways to obtain a l l thes e data are various and dependent on how the study i s organised . Most general accident and occupant data are easy to record. Police re­ports can be used and also driver reports ( such as insurance reports or special enquiry forms ) .

1 5 9

The greatest prob l em i s the speed determinati on o f whi ch the c o l l i s ion speed is the most important . Onc e this factor is known , there is a reli­abl e parameter for the acc i dent s everity . Many methods of rec onstructing c o l l i s i on speed have been created , us ing skid marks and other physi cal eviden c e . Dr iving s�eed can b e determinated more easy , ( for instance by drive r , by wi tnes ses ) but gives l e s s information about the resulting col l is i on se­ver ity and the information might be bi ased�

Obtaining damage data o f the case car(s) i s mostly work for special i s t s , though there are some indirect ways that can b e qu ite useful . For in­stanc e , s ince in many ac tual accident cases the spe c ia l i sts are not ava i l ab l e , or wil l c ome too late on the s c ene , it can b e neces sarily that the damage data are recorded by others ( for ins tance po l ic e o f f i c ers , towing s ervic e people ) by means of photographing . The responsible acci­dent inveatigators and analysers may use the se photographs or slides to determinate the damage o f the car . Even i f spe c ia l i sts d i d see the car ( s ) , in many cases sl ides and photographs proved to be very useful . In Figure 3 a c ompar i son i s made for the mentioned study examples as far as the obtaining and description of the important a c c i dent data are c oncerned ( see also paragraph 2 . 3 ) . In s ome studies damage data were pro­vided c ompl etely or partly by drivers ( or in case of fatal i t i e s ) by po­l ice-officers . It also might be useful , i f driver enquiries are being used , to ask questions about the parts in the interior that were hit and relate the answers to the same data from reports of damage special i sts .

Accurate injury data can eas ily be obtained, assuming that a phys i c ian or surgeon made a report on the injuries . Co-operation of medi cal authorities proved to be rather wel l in most studies yet reported. Since in most countries various legal problems arise i f someone i s kil led in an acc ident , the autopsy reports ( if available ) and injury de ·-scriptions might be very difficult to obtain , thus caus ing a big lack of information for those cas e s . Hospital teams , organis ing ac c i dent studies , have an advantage there , s ince they wi l l start their cases on detai led injury descriptions even of fatal injured occupants . The only source for injury description of minor injured occupant s , are the occupants themselve s . Thi s information and also the knowledge that certain occupants were not injured at all , are important data in every crash acc i dent study , because thi s information might help to answer the questions about tbreshold values for injury causation, c o n c e rn i n g interior structures and human tol erance data .

2 . 3 . Data description in accident forms

2 . 3 . 1 . General

In every ac c i dent study some kind of accident form is used for the uni­form rec ording and storage of acc ident data . Since for the studies with bigger samples almost always ( digital ) computer analys i s or at least computer counting is appl ied , the forms wil l show already s igns for that purpos e . Thi s means that numerical codes have to be used and code l i sts have to be availab l e , e ither on the form itsel f , or on s eparate l i sts , for every separate variable .

Simple data l ike sex (male , f emale , unknown ) are easy to describe numeri­cal , for they need only one c omputer card column. In cases with more than 9 or 10 pos sib i l ities within one variable it i s possible to use two or more columns ( digits ) , as with "speed" or "age" or "year" .

In most studies codes have been created for severity ratings both o f the car damage and the injuries to each occupant. These severity ratings can be very important for s impl i fying the analysi s and for comparing acc i ­dents . A wel l known car damage severity index is the Vehi�le Deformati on Index (VDI ) , created in the USA . Thi s index i s among others described in the SAE draft Recommendati on Practice , nr J 224a , 19 7 1 . Apart from a classi­f i cation for the damage severity thi s index gives a description of the main c o l l i s ion events and the damaged parts of the car (us ing a total of 7 digits ) . Thi s index has been used in the NATO-CCMS study but unti l now, on its interpretati on no f inal agreement has been reached.

Simple , but very useful severity ratings for car damage c lassify the damage into 5 or 6 categorie s , ranging from "0" = no damage to " 5 " or 1 16 11 = complete destruction o f car compartment . To avo id mis interpretation it i s important that a severity scale i s s impl e enough . But at the same time , to allow the analyser to determi­nate the important di fferences between categories , it has to be detailed enough .

Another important severity scale i s the injury severity scale , of which many different kinds are in existance today . A well known American one i s the Abbreviated Injury Scale (AIS ) , also been used in the NATO-CCMS study. All these scales give in more ar less categories the overall sever-

ity of the injuries to an occupant , ranging from not injured to fatal injured. The s impl est way is a three category scal e : not injured, injured , dead . A more sophisticated one i s : not injured , s l ightly injured , severe injured , fatal injured . Of course all depends on the proper description o f the d ifferent catego­ries , but practice proved that less difference b etween analysers existed in interpretation of injury severity than in interpretation of car dam­age severi ty . Both the damage severi ty and the injury severi ty need sup­port of detailed damage and injury descriptions for detai l ed analysi s .

2 . 4 . Examples o f character istics o f accident studies

In this paragraph some of the most interesting characteristics of the before mentioned studies wi ll b e compared , as far as comparison is possi­ble and allowed . Important characteristics o f ac c i dent studies are : di stributions o f c o l l i �i on types ; di stributions o f c o l l i s i on speeds ; di stributions of injuries to body regions ; distr ibuti ons o f injury severity ; effects o f use o f s eat be lts . In Figure 4 these characteristics are given.

1 6 1

In some studies no c o l l i s i on speeds had been determinated but driving speeds by means of driver report s . The effect of the use o f seat belts can be described i n terms of the number ( or percentage ) of people that would have been saved from fatali­ties or injuries with regard to the number of peopl e actual ly kill ed or injured , if they had been us ing their be lts .

3 . FINAL REMARKS AND R�COMMENDATIONS

Crash acc ident studies have proved to be adequate means of improving traffic safety by applying the ir results in terms of improving exteri or and interior parts of cars ( espec ial ly by applying restraint systems ) and environment . Pre crash studies have to deal with the almost inpredictab l e human be­haviour factors and their results might therefore be difficult to apply ; although recommendations for improvement of cars and roads , as far as acc ident prevention i s concerned , seems to be necessary.

Human tolerance data , the determation of which is part of the IRCOBI programme , cannot be gathered only from crash accident studies . Suppl e­mentary information i s needed on the forces and decelerations that caused actual injuries or fatal itie s . Two important ways exi sting today are : 1 . The use o f l aboratory ( s led ) test in which dummies or human cadavers are b eing used and in which the c o l l i s ion conditions can b e accurately measured . By comb ining these results with those of actual acc i dent in­formation on injuries , it i s possible to determinate threshold values for injury causation. Many different pro j ects in this area are carried out all over the worl d at thi s moment ( one being mentioned in paragraph 1 . 5 . 7 . , item 2 ) . 2 . By providing a big sample o f cars ( preferably of one owner ) with so called crash recorders , in whi ch the decelerations of the ca� during the accident are recorded , the acc ident inve stigator has got much actual information about the accident severity ; but even then he wi l l miss the forces and dece lerations to the occupants themselve s . Again i n the USA a promis ing project had b een started, involving cars of f l eet owners . But this project seems to be abandonned due to lack of money .

lt has been stated that there i s need for big sample acc ident studie s . Since gathering indepth information ( o ften on scene studie s ) means sma l l samples , there are two ways l e ft : 1 . Organis ing l arge statistical accident studies a s given in some o f the examples , in whi ch qual itatively fair information was gathered without vi s it ing the s c ene of the accident . A prel iminary study of the differen­ces between on-s cene and off- scene information as far as quality c oncerned ( c arried out by SWOV) did not show that off scene information was neces­sarily worse than on scene information. 2 . The combining of indepth studies . If organisers of indepth crash studies use the same accident report forms , and if the ways of gathering the data are the same , it i s possi.b l e to compare and/or to c ombine re­sults , provided that differences in spec if i c local fac tors can be e l imi­nated .Up t i l l now it proved to be almost impossible to compare results of different acc ident studie s . The examples of Figure 4 will prove this .

The NATO-CCMS Acc ident Investigation Program seems to be one of the first in whi ch these facts are acknowledged, and we hope that all parties concerned wi l l agree upon ways to go on with this kind of comb ined investigation.

No crash acc ident studies have yet been mentioned concerning improvement of the environment ( the third factor in the acc ident area of Figure 1 ) . However , many studies are exi sting and it has been proved that studies on roads ide obstacles and guard rai l s are important contributors to road safety .

REFERENCES

1 . N . I . Bohlin . A stati stical analysis of 28000 acc ident cases with emphas is on occupant restraint value . AB VOLVO, Gothenburg , Sweden

2 . Innere Sicherheit im Auto . Eine Untersuchung des HUK-Verbandes über PKW-Unfälle mit Insas senverl etzung . Bericht über 10 . 27 1 Fäll e , nr I und II HUK-Verband , Hamburg , Germany , 1972 .

3 . R .N . Kemp . I . D . Neilson, G . C . Staughton and H .A . Wilkin s . A prel imi­nary report on a on the spot survey of acc ident s . TRRL report LR 434. TRRL, Crowthorne , Berkshire , U .K . , 1972 .

4 . Jaako K . Kihlberg . Efficacy of seat belts in injury and noninjury crashes , in rural Utah . ACIR-CAL report VJ-2721-R3 . C .A . L . , Buffalo N .Y . , USA , 1969 .

1 6 3

pre crash crash post crash

l:::l 1 2 3 "' � �

� � 4 5 6 t) 'M �

� I>

;.J l:::l � 7 8 9 � 0 �

'M I> l:::l �

Figure 1 . Programme Matrix for Highway Safety Research

�

O"')

U'1

samp

le y

ear

num.be

r of

ca

ses

VOLV

O HUK

TR

RL

ACI

R

swov

NA

TO-C

CMS

1 2

3 4

5 6

1965/6

6 --

----

1969

-

-19

68/69

1966/6

7 19

69/

70

1970

/71

2878

0 dr

iver

s 10

271

driv

ers

247

driv

ers

7125

dr

iver

s 22

082

driv

ers

500

driv

ers

quan

tity

/ h

igh/

hig

h/ lo

w/ fa

ir/

hig

h/ ��

low/

qua

lity

fa

ir

fair

h

igh

fair

fa

ir

hig

h

data

ana

lyse

d by

inst

itu

te c

ode

(see

1.4

.)

area

typ

es o

f ca

rs

typ

es o

f ac

cide

nt

VOLV

O,

HUK,

Swed

en

Germ

any

2 5

rura

l m

ostl

y

unkn

own

Vol

vo

's o

nly

a

ll

all

on

ly d

riv

ers

who

wer

e "q

uil

ty"

only

inj

ury

pr

odu

cing

a

ccid

ent

s

TRRL

, U

.K.

1 rura

l m

ostl

y

all

all

CORN

ELL,

U

SA.

1

rura

l

swov

, N

ethe

rlan

ds

1 rura

l an

d u

rban

all

a

ll

inju

ry

pro

-al

l du

cing

and

no

n in

jury

pr

odu

cing

a

ccid

ents

Figu

re 2

. Ch

arac

teri

stic

s of

6 d

iffe

rent

acc

iden

t st

udie

s.

SOUT

H W

ESTE

RN

RESE

ARCH

IN

ST.

USA

.

1

diff

eren

t fo

r d

iffe

rent

te

ams

dif

fere

nt f

or

dif

fere

nt

team

s

dif

fere

nt f

or

diff

eren

t te

ams

Rep

orti

ng

of

gen

eral

dat

a by

car

damag

e da

ta b

y in

jury

dat

a by

damag

e se

ve:r

­i t

y s

cale

use

d

damag

e r

epor

­ti

ng

inju

ry s

ever

­i t

y s

cale

use

d

inju

ry r

epor

­ti

ng

amou

nt

of

data

or

com

pute

r ca

rds

for

one

VOLV

O

driv

ers

driv

ers

driv

ers

own

syst

em

repa

ir c

osts

an

d es

tim

a­te

d co

lli­

sion

spe

ed

not

de

tail

ed

HUK

TRRL

driv

ers

team

dam

age

grou

p te

am

doct

ors

and

hos

pit

als

own

syst

em

5 ca

tego

ries

no

n li

nea

ir

deta

iled

team

+

hosp

ital

s

own

syst

em

5 po

int

sca

le

+ li

nea

ir

not

de

tail

ed

own

syst

em

USA

syst

em

STA

TS s

yst

em

4 ca

tego

ries

3

and

4 ca

te-

7 ca

tego

ries

go

ries

grou

ps

o:f

inju

ries

co

ded

at l

east

80

dif

fere

nt

data

deta

iled

cod

e no

ne

syst

em f

or

sev

erit

y,

type

an

d b

ody

regi

on

abou

t 20

0 di

ffer

ent

data

unkn

own

ACI

R

pol

ice

pol

ice

pol

ice

+ h

osp

ital

s

own

syst

em

esti

mat

ed

impa

ct s

peed

in

to 4

cat

e­go

ries

not

deta

iled

swov

driv

ers

+ p

olic

e

NATO

-CCM

S

dif

fere

nt f

or

dif

fere

nt

team

s

spec

ial

team

s se

e ab

ove

spec

ial

phy-

see

abov

e si

cian

+

driv

-er

inf

orm

atio

n

own

syst

em

VDI

5

cate

gori

es

non

lin

eair

deta

iled

v

ery

deta

iled

pol

ice

syst

em o

wn

syst

em

AIS

5

cate

gori

es

7 ca

tego

ries

none

abou

t 80

di

ffer

ent

data

deta

iled

cod

e de

tail

ed c

ode

syst

em

("N

code

") fo

r se

ver

ity

, ty

pe

and

body

re

gion

abou

t 28

0 di

ffer

ent

data

(a

ver

age

of

4 co

mpu

ter

card

s)

syst

em f

or

sev

erit

y,

type

an

d bo

dy

regi

on

abou

t 30

00

diff

eren

t da

ta

(av

erag

e of

50

co

mpu

ter

card

�

Figu

re 3

. Co

mpar

ison

of

the

ways

da

ta w

ere

repo

rted

and

des

crib

ed i

n 6

acc

iden

t st

udie

s

VOL

VO

num

ber

of

2878

0 dr

iver

s dr

iver

s 87

31 fr

ont

seat

an

d pa

s-53

02 r

ear

seat

se

nger

s 42

813

occu

pant

s di

stri

bu-

.fro

ntal

35

,7

tion

of

si

de

33,5

co

llis

ion

off

road

17

,2

type

s·

roll

ov

er

4 ,9

rear

and

ot

her

8 7

1 otr

2878

0 ca

ses

dist

rib

u-n

ot a

vail

able

t

ion

of

spee

ds

dis

trib

u-no

t av

ail

able

ti

on o

f da

mage

se-

ver

ity

_.,.

Fig

ure

4.

Part

1

cn

......

HUK

TRR

L A

CIR

1027

1 dr

iver

s n

ot r

e-71

25 d

riv

ers

5175

fron

t se

at p

orte

d 41

63 f

ront

sea

t 19

09 r

ear

seat

25

64 r

ear

seat

1735

5 oc

cupa

nts

409

unkn

own

1426

1 oc

cupa

nts

fron

tal

30,2

no

t re

-fr

onta

l 40

,6

side

41

,3

port

ed

side

11

,3

rear

28 %5

rear

15

,3

1 ro

ll o

ver

8,

6 10

271

case

s un

know

n 24 %2

1 14

261

case

s

only

ava

ilab

le

not

re-

hig

hes

t im

pact

fo

r se

pera

te

port

ed

spee

d (m

ph)

coll

isio

n ty

pes

<

20

23,7

20

-39

35,2

40

-59

22,6

)

60

7,8

unkn

own

10 %7 10

0 14

261

case

s

only

ca

t to

car

0-

8,9

not

avai

lab

le

coll

isio

n 1

. 1-

17,0

(82

26 c

ase

s)

1-38

,5

ligh

t 5,

9 4-

22,3

m

ediu

m

31,3

1-

8,9

hea

vy

46,6

1t-

2�8

ver

y h

eavy

13,

0 ov

er

. 1

6 to

tal

des-

14-

1,

tru

ctio

n

10�%2

swov

N

ATO

-CCM

S

2208

2 dr

iver

s n

ot y

et

9088

fro

nt s

eat

repo

rted

53

47 r

ear

seat

14

90 o

ther

+ u

nkn

3820

7 oc

cupa

nts

fron

tal

41j 5

see

si

de

25,5

abo

ve

rear

13

,6

rol

l ov

er

4,0

unde

rrid

e 0,

2 re

ar+f

ront

7,

3 ot

her

4,0

unkn

own

10&%9

2208

2 ca

ses

driv

ing

spe�

d se

e ( km

/h)

abov

e 0

1 3, 5

1-

25

19,2

29

-50

34,4

51

-75

1 4, 7

76

-100

12

,2

101-

125

2,2

> 12

5 0,

8 un

know

n 10&%0

m

inor

63

,3 se

e m

oder

ate

14,0

abo

ve

sev

ere

11, 3

con

side

rab

le

com

part

men

t da

mag

e 3 ,

3 co

mple

te

com

part

men

t 0,

9 de

stru

ctio

n

unkn

own

7 2

100%

VOLV

O HUK

TR

RL

ACI

R sw

ov

NATO

-CCM

S

dist

ribu

-no

t av

ail

able

no

in

juri

es 2

2,8

not

re-

not

avai

lab

le

no

inju

ries

63

,7 n

ot y

et

tion

of

fo

r al

l 28

780

ligh

t 60

,2 p

orte

d li

ght

20,4

rep

orte

d in

jury

se-

case

s m

oder

ate

12,3

m

oder

ate

8,7

ve

riti

es

sev

ere

3,0

se

ver

e 3

,6

(dr

iver

s da

nger

ous/

da

nger

ous/

on

ly)

1 if

e th

reat

-li

fe t

hrea

t-en

ing

o,6

en

ing

o,6

fa

tal

0,9

fa

tal

1,4

un

know

n 0

2 10

0% un

know

n 0

2 10

0% di

stri

bu-

not

ava

ilab

le

hea

d 39

,6 n

ot r

e-no

t av

ail

able

br

ain

14,1

se

e ti

on o

f th

orax

17

,2 p

orte

d h

ead

14,5

abo

ve

inJu

ries

ab

dom

en

5,6

n

eck

6

,8

to b

ody

nec

k

19,5

th

orax

7

, 7

reg

ion

s sp

ine

abdo

men

1,

2

(dr

iver

s (l

ower

) 3

,0

back

2

,8

only

) ar

ms

32,4

p

elv

is

0,3

le

gs

1&�%3

arm

s 2

1,3

10

271

=

legs

..

..lL. 2 (n

umbe

r of

16

046

=

100%

dr

iver

s)

(num

ber

of

inju

rie

s)

effe

ctiv

e -3-

poin

t be

lts

: co

nsi

dera

ble

n

ot r

e-la

p be

lts

, 50

-80%

red

uct

ion

see

abov

e n

ess

of

40-9

0%

effe

ctiv

enes

s p

orte

d al

l oc

cupa

nts

: of

fat

ali

ties

se

atbe

lts

inju

ry

50%

redu

ctio

n;

for

all

seat

bel

t re

duct

ion

any

inju

ry:

type

s to

geth

er

depe

nden

t on

40

% m

ore

risk

sp

eed

if

seat

bel

t no

fat

ali

ties

no

t u

sed

bel

ow 6

0 mp

h a

ccid

ent

spee

d

Fig

ure

4.

Comp

aris

on o

f so

me

char

acte

rist

ics

of 6

a

ccid

ent

stud

ies.

Par

t 2

Summary

Thi s paper give s a review o f real-ac c i dent stud i e s in general , the requi red efforts , the problems and the bene f i t s . Example s o f spe c i f i c stud i e s from different countri es o f the world are menti oned and some o f the i r characteri sti c s are compared, both in terms o f organi sation and of result s . Though much o f the data used in thi s paper are based on vari ous l i terature , on�y in spe c i f i c cases references are given. Emp'has i s wi l l be la id on so called crash a c c ident s tudie s , of whi ch a definition wi l l be given.

1 6 9