Injury Prevention 1996; 2: 55-60

INJURY CLASSIC

Theory and methods of epidemiologic study ofhome accidents

Reuel A Stallones

MethodsProblems in the design of epidemiologicstudies may be categorized conveniently intothose related to ascertainment and those con-

cerned with sampling. Sampling problemsconfronted in home accident studies are notdifferent, in any essential way, from thoseencountered in any other epidemiologicstudies. With reference to sampling, theepidemiologic critics are far in advance ofepidemiologic practitioners, so that the latterhave yet to propose a study which the formercannot demolish by attacking the samplingprocedures. The critics apparently havealready discovered a whole host ofbiases whichwill only be demonstrated by studies yet to bedevised. For these reasons, sampling problemswill be treated only briefly in this report, wherethey are inextricably involved with the businessof ascertainment.

Problems of ascertainment in home accidentstudies do present some special, though per-

haps not unique, features. Ascertainment is theprocess of acquiring desired information. Theterm carries the connotation that this is an

active quest, involving a search for the data.Each item of information sought must bespecified, defined, the need for it justified, andthe method by which it is to be collectedestablished. Generally, more attention is paidto the disease diagnosis than to other charac-teristics of the host. On logical grounds, one

may hold that a correlation of the occurrence ofdisease with some other factor, such as age, may

be rendered invalid as certainly by errors inrecording age as by errors in recording thedisease, but this has little emotional appeal andin practice we continue to expend the mostconcern on those things that interest us most.Therefore, this paper will be devoted largely toproblems of ascertainment of cases. Many ofthe ideas referable to this characteristic of thehost can be generalized to the ascertainment ofother host and environmental characteristics.

To study home accidents requires thathome accidents be defined and immediatelyhome accident studies are in trouble. To my

knowledge there is no satisfactory definition ofan accident, and no satisfactory means ofseparating completely home accidents fromother forms of accidents.While number of recent publications on

home accidents have failed to present a

definition, Suchman and Scherzer,' in an excel-lent review, discussed a variety of definitionswhich have received some acceptance. Most of

them incorporate a statement of recognizableinjury or loss. Some refer to 'chance' events,which connotes that accidents are distributedrandomly, and this is clearly not true. TheUnited States Public Health Service publica-tion Uniform Definitions of Home Accidents?refers to an accident as a sequence, chain ofevents, or series of interactions between theperson and environment or agent, including themeasurable or recognizable consequences.This definition is perfectly satisfactory if onealready knows what an accident is; if not, itcould just as well define typhoid fever or cancerof the lung. Most definitions share this lack ofselectivity. Part of the difficulty comes from thegreat number of diverse events included asaccidents and this is compounded by theinability to speak very confidently aboutcauses, for specificity of definition often isattained by reference to cause.

Epidemiologic studies may reasonably haveas an aim more precise definition of diseasesyndromes. Data from studies already done areadequate to propose that much of the neededclarification can be realized by characterizingeach event independently along four axes, andsubjecting each subset obtained in that way tomore detailed analysis. Although this approachdoes not permit generalization to the wholeclass of events commonly called accidents, Ideem this an advantage, since these generaliza-tions lump such diverse events that more is lostthan is gained. The four axes are:A. The nature of the event.B. The nature and severity of the injuryC. The responsibility for the event.D. The place of occurrence.

A. THE NATURE OF THE EVENTThis classification is the most important step insorting out from the general melange groups ofaccidents which can be handled, and thisclassification permits satisfactorily specificdefinition of the events to be studied. Thesubcategories used most often employ suchterms as falls, struck by or against an object,bums, poisonings, etc. A more generalclassification is as follows:

1. Injury resulting from the absorption ofkinetic energy. It may be important tospecify whether the object or the injuredperson or both were in motion at the time ofimpact, but I know of no instance in whichthis has been tried.

This is the sixth paper ina series of Injury Classics.Our goal is to reprint onesuch paper in each issue toinitiate newcomers to thefield of these old, oftenquoted, and importantcontributions. As manyare difficult to find, itshould help all of us tohave a copy at hand.Your suggestions aboutfuture articles arewelcome. Write to theeditor with details ofyourfavourite, most quotedpaper.

Professor of PublicHealth, University ofCalifornia, Berkeley,California

This paper first appeared inthe Annals of the New YorkAcademy of Sciences (1963;107: 647-58) and isreproduced by kindpermission of the New YorkAcademy of Sciences.

55

on January 4, 2020 by guest. Protected by copyright.

http://injuryprevention.bmj.com

/Inj P

rev: first published as 10.1136/ip.2.1.55 on 1 March 1996. D

ownloaded from

Stallones

2. Injury resulting from the absorption ofradiant energy.3. Injury resulting from chemical toxins.

(a) Corrosive.(b) Metabolic.

4. Other: suffocation, drowning, electro-cution, animals and insect bites.

The heterogeneity ofhome accidents may beexemplified by some data from a study ofchildhood accidents in Alameda and ContraCosta Counties, California,3 although theclassification scheme presented above was usedonly in part. In this instance, the cases werecollected by a reporting system from theemergency services of all of the non-federalhospitals in the two counties, and the popula-tion base was estimated by the California StateDepartment of Finance. In figure 1 the agedistribution of all reported accidents is showntogether with the age distribution of twocategories of accidents classed by the nature ofthe accident. Classification by characteristicsother than age shows similar variation indifferent categories of accidents and illustratesthe necessity of sorting home accidents intomore homogeneous subgroups.

B. THE NATUREAND SEVERITY OF THE INJURYA number of studies have used a scheme ofclassification based on the nature of the injury,using categories such as contusions, abrasions,lacerations, sprains, strains, fractures, headinjuries, burns, ingestions, etc. This is justifiedon epidemiologic grounds by consideration of afurther analysis of data from the Alameda-Contra Costa Counties study. In figure 2, theage distributions are given for three classes ofaccidents: falls, automobile accidents, andinjuries resulting from being hit by or against

200-0

100.0

a0

*. 10-0

10-0

._1.C.

C.

~0

o 0-1

:3 .

F All accidents

dV

0-o_ --\Y Sports and games

0- accidents

9/ 0--

,9'~~Poisonings

IIIII

0-01 I±0 1 2 3 4 5 10 15

Age in years

Figure I Age distributions of all reported accidentalinjuries and of poisonings and injuries incurred in sportsandgames; Alameda and Contra Costa Counties, Califor-nia, 1957.

an object. In each of these kinds of events, anumber of different kinds of injuries mightresult. Here, the head injuries have beenselected and, despite the variation in the agedistributions of the accident groups, show aremarkably similar age selection. That is, thereappears to be an age selection factor affectingthe occurrence of head injury which is partlyindependent of the age selection factors whichaffect the occurrence of the different kinds ofaccidents in which heads were injured. Thusthe nature of the injury is a second characteris-tic which must be accounted for beforehomogeneous groups of cases can be obtained.The severity of the injury, as well as the

nature of the injury, also surely must have anindependent effect, but even more importantlyis intimately associated with the method ofascertainment of cases. Epidemiologists favorviewing problems in an holistic way, studyingthe whole spectrum of a phenomenon, and thishas been a rewarding approach in a largenumber of diseases where the picture becameclear only after the whole gamut of the diseaseprocess came under scrutiny. In accidentstudies this laudable aim leads only to chaos.The spectrum is there, right enough, and canbe treated theoretically, but practically a largesegment of the spectrum cannot be reduced toidentifiable, measurable phenomena.The spectrum starts at the far right (figure 3)

with death resulting from an untoward, unex-pected, non-infectious, traumatic event. Mov-ing to the left on the spectrum, a series ofpointscorrelated with severity of injury can beselected, marking identifiable events which canbe ascertained by one means or another. Thisholds true down to the far left of the spectrumbut there things fall apart. Previous problemswith definitions pale beside the difficulty ofdefining an accident which did not happen.The requirement for an epidemiologist is to

establish some criteria which will define a pointalong the spectrum and apply these criteria in aconsistent and comprehensive fashion. Thecutting point which is selected will restrict theselection of methods of ascertainment of cases,for some are obviously inappropriate. Some

0 1 2 3 4 5 10 15Age in years

Figure 2 Age distributions of three classes of accidentsand ofthe head injuries associated with them; Alameda andContra Costa Counties, California, 1957.

56

on January 4, 2020 by guest. Protected by copyright.

http://injuryprevention.bmj.com

/Inj P

rev: first published as 10.1136/ip.2.1.55 on 1 March 1996. D

ownloaded from

Theory and methods of epidemiologic study of home accidents

generalizations may be made concerning themore common methods of ascertainment.

1. Mortality analysesAnalyses of mortality data not only arerelatively inexpensive and relatively easilyaccomplished, but also usually permit relatingthe cases to a definable population without toogreat difficulty. However, the information onthe base population (usually from the nationaldecennial census) is collected intermittentlyand in a different fashion. While the recordingof deaths may be reasonably complete, thedeaths do not provide a satisfactorily completerecord of accidents. In many accidentcategories the proportion of events terminatingin death is exceedingly small, and even whenthis is not true, the deaths cannot reasonably betreated as a representative sample of theaccidents. The consequences of a fall down aflight of stairs vary greatly by age, for example.Where conditions occur in low fequency orwhere death is not often associated with acondition under study, the collection of casesmay have to extend over a long time or cover alarge population. The data available directlyfrom death certificates are limited and augmen-tation of them may be difficult. Finally, theinvestigator has little or no control over thequality and variability of the information.

If mortality analyses are based on othersources of data, such as hospital records, theinformation may be more complete in someparticulars, but the population of hospitaldeaths is very difficult to relate to a definablepopulation base and comprises an indefinitesample of all accident deaths.

2. Morbidity analyses(a) Hospital admissions - Admission to hos-pital constitutes an easily defined event whichcan serve to identify a study group. As with

Figure 3 Relation of severity of injury incurred in home accidents to the portion ofcases ascertained by different methods.

deaths, the hospital admissions are a verybiased sample of all accidents and the inferencedrawn from such studies must be interpreted inthis light. Also, since the data were recorded forpurposes other than epidemiologic study andusually by a number of different people, thereare serious deficiencies of completeness andconsistency. If a study is designed to collectcases over a period oftime in the future, some ofthese difficulties may be reduced.

Comparability between hospitals is difficultto secure, since the selection biases operatingon hospital populations are expected to differfrom one hospital to another.

(b) Medical attendance- The collection of allaccidents coming to medical attention is feasi-ble under some circumstances. Ordinarily, thisinvolves review of the records of practitioners,which is even less satisfactory than review ofhospital records, or a survey of currentexperience of the practitioners. In the lattercase, the problem of non-response becomessignificant.

(c) Disability - Accidents causing disabilitymay be determined by population surveys orthrough such means as school or workabsenteeism records. Population surveys willbe considered in the next section. School andwork absence records serve as a valuableresource, but rarely are they useful other thanas a starting point, for the information concern-ing the event to be studied is likely to be verysketchy. This may be supplemented by inter-view, however, and then this group has themajor advantage of being related immediatelyto a base population, in which many individualcharacteristics are already known.

(d) Population surveys - The most compre-hensive way in which to collect cases is bypopulation survey. This may be done once,repeated at intervals, or, by a diary system,maintained over a period of time. Howeverdone, the survey is at the mercy ofrespondents,who, through lack of interest, memory lapse, orinborn contrariness, make life precarious forinvestigators. Population surveys are generallyexpensive, for sampling procedures and inter-viewing cause costs to mount rapidly. In thisform of study, adherence to a predetermineddefinition of accidents is very difficult, for eachevent sought must be evaluated by each respon-dent, communicated to the interviewer, andinterpreted by him. Persons knowledgeable insurvey methods have provided us with meansfor reducing and handling interviewer varia-tion with some success, but there is no way toassure that the respondents will adopt thatcommonality of viewpoint which we desire.We may conclude that not one of the usual

methods of collecting a series of cases is entirelysatisfactory. The investigator must weigh therelative values of the methods available to him,and having made his choice, proceed with thestudy with full awareness of its shortcomings.

... ...................

57

on January 4, 2020 by guest. Protected by copyright.

http://injuryprevention.bmj.com

/Inj P

rev: first published as 10.1136/ip.2.1.55 on 1 March 1996. D

ownloaded from

Stallones

C. THE RESPONSIBILITY FOR THE EVENTTwo kinds of assessment of responsibility maybe considered. One calls for a differentiationbetween acts of God and acts of man, and theother attempts to separate injuries incurred as aresult ofthe action ofanother person or personsfrom those resulting directly from the activitiesof the injured person. In either case, thedifferentiation is often difficult and sometimesimpossible. A golfer struck by lightning duringa thunder shower may be thought of as a victimof celestial caprice, but he may have consideredthe risk and accepted it in the hope that anunfavorable weather report would clear thefairways of many of his less dedicated com-petitors. There must be some accidents that aretruly unaffected by human design, but theutility of characterizing these seems to me notto have been demonstrated. However, there isutility in determining whether the injuredperson played an active or a passive role in theaccident leading to his injury. If a mother dropsher baby and the baby is injured, the accidentmay be considered to have occurred to themother, although the injured infant is likely tobe the index person in a study. This distinctionis hopelessly blurred in many instances, butsome attempt to account for responsibility isneeded since certain control procedures mustbe directed toward the responsible persons.

D. THE PLACE OF OCCURRENCEThe classification of accidents by place ofoccurrence poses many problems for the inves-tigator. The common categories ofhome, farm,occupational, transportation, and recreationalaccidents are not mutually exclusive. Ourgolfing injury may be classed as a recreationalaccident but if the golfer were an insurancesalesman he might consider it as an occupa-tional hazard. If a pedestrian is struck down bya motorcar while crossing a public thorough-fare, this is clearly a motor vehicle accident;however, if the car invades a front yard andassaults a pedestrian peacefully watering hislawn, it may be classed as a home accident.Perhaps in the great bulk of events, theseaberrant occurrences are of a frequency lowenough that they may be ignored, or at leasthandled by a series of arbitrary decisions. Thisdoes offer an easy way out, and considering thevariety ofproblems confronted in these studies,

the easy solution to one like this may well be thebest solution.Thus, a multiple classification by type of

event, by type of injury, by person responsible,and by place of occurrence appears to be a

necessary step in the analysis of accidents. Theclassification option which is adopted will serveto define the events to be studied in a mannermore precise than has been forthcoming fromany general definition of accidents. Thespecification of events to be studied also maydictate, or at least restrict, the method ofascertainment of cases and thus profoundlyaffect the study design. The difficulty whichthis poses for epidemiologic study is for-midable, for the frequency of occurrence ofmany accidents classified in this way is so lowthat very large populations are required togenerate sufficient cases to permit meaningfulinterpretations. This is illustrated in the table.In this study, previously described, althoughthe number of events recorded was large, as theclassification became more complex thenumbers in the cells were too small to permitdetailed analysis, especially for some of themore interesting categories.

TheorySome general consideration of the theory ofaccidental injuries and accident preventionmay help to guide this discussion.There are two absolute requirements for an

accidental injury to occur. (1) There must bepresent in the environment a situation whichthreatens injury to a person exposed to it in aspecified way. (2) There must be a personsusceptible to the injury and exposed to thehazard. Accident prevention may be viewed asa task for environmental engineers, and this is auseful approach. Many of the rough edges havebeen knocked off of our environment withundoubted salutary results. However, themyriad aspects of the environment which arepotentially harmful overwhelm the student ofthese problems and it is neither feasible nordesirable to send a human being through life ina sterile, stimulus-free, shock-resistantpackage. An existence free of danger and with-out stress is devoid of interest or purpose.Thus, the control of accidental injury in thehome by environmental manipulation islimited to attacking a small fraction of the

All childhood accidents reported by hospitals, by type of accident, and by type of injury; Alameda and Contra Counties, California, 19573'Type of injuryContusion or Sprain or Head Other and Per cent

Type of accident abrasion Laceration strain Fracture injury Burn Poisoning not stated Total of total

Automobile accident 657 222 35 122 114 0 0 390 1540 5 6Fall 1291 4096 832 1385 529 50 0 939 9122 33 0Caught in device or object 736 466 26 83 1 3 0 145 1460 5 3Hit by object 753 2818 181 216 181 404 4 582 5139 18 6Stepped on object 69 425 4 0 0 30 0 912 1440 5 2Fire or explosion 2 4 1 1 1 56 0 7 72 0-3Firearms 2 1 0 1 1 3 0 20 28 0.1Ingestion 0 8 0 0 0 0 2109 509 2626 9-5Strangling or suffocation 0 1 0 1 0 0 1 18 21 0.1Other and not stated 792 2031 272 241 53 235 22 2529 6175 22 4

Total 4302 10 072 1351 2050 880 781 2136 6051 27 623 100 0Per cent of total 15 6 36 5 4 9 7 4 3 2 2 8 7-7 21 9 1000 -

58

on January 4, 2020 by guest. Protected by copyright.

http://injuryprevention.bmj.com

/Inj P

rev: first published as 10.1136/ip.2.1.55 on 1 March 1996. D

ownloaded from

Theory and methods of epidemiologic study of home accidents

outstanding hazards, redesigning appliancesand machinery, and altering the structure andconfiguration ofobjects which seem most likelyto cause injury. Epidemiologic studies have a

definite, but limited, place in this area ofendeavour, in the recognition and designationof the nature of the environmental hazards.Much information ofthis kind has been derivedfrom contemplation ofcase series only casualperusal of mortality data or the records ofpoison control centers is required to show thatsomething needs to be done to limit the acces-

sibility ofkerosene or aspirin to small children.The second popular approach to accident

prevention is that of education. This ispredicated on the notion that a large proportionof accidents are the result of failure of theaccidentee to appreciate the nature of a hazar-dous material or dangerous situation. One ofthe more serious problems here is to distin-guish between conscious awareness andintuitive perception of danger. Many of our

avoidance reactions operate below the level ofconscious thought and are presumablyaccumulated through experiencing repeatedpainful stimuli. Whether instruction at a cons-

cious level can ever satisfactorily substitute forthe conditioning process remains to be demon-strated, but seems to me to be an uncertainpresumption. Much more work is needed toassist us to understand the mechanismswhereby we develop the capacity to recognize

and avoid danger and the relation of these to

conscious thought and overt behavior. Onemay conjecture whether the conditioning pro-

cess must result from the uncertain aggregationof natural events, occasionally marked by a

tragic result, or whether some artificial proce-

dure could be developed to produce the same

result more predictably and more safely.Epidemiologic study has a considerable con-

tribution to make in the field of education. Itpermits the identification of specific events inthe accident complex which require attention.It identifies the subgroups of a population inwhich the specific event is of greatest conse-

quence and toward which the prevention prog-

ram should be directed. It provides a means forassessing the effectiveness of the preventionprogram. The design of studies for this pastpurpose is perhaps the most seriouslyunderdeveloped aspect of home accidentepidemiology. I think it has never been done at

all satisfactorily, although it is evident that noreasonable statement of the value of any

preventive activities can be made until theprograms can be assessed with some rigor.The application of epidemiologic methods to

home accidents presents the promise not onlyof furthering preventive activities of the kindjust discussed, but also of development ofmeans of prevention not presently apparent to

us.

Epidemiologic studies of home accidentswould be furthered by the development ofsome general theoretic concept ofthe ocurrenceof accidental injury in human populations. I donot believe that any satisfactory formulation ofthis kind has been presented. The simplehost: agent: environment model developed

from epidemiologic studies of infectiousdisease has some value, but calls for a series ofarbitrary distinctions between host, agent, andenvironment which have little basis in realityand little utility as abstractions. Perhaps thechief use of this model is contained in theimplication that an holistic approach is desired,but this view is hardly unique to this concep-tualization.The 'web of causation' concept propounded

by MacMahon and coauthors4 also seems to meto be of little assistance in developing arationale for studying accidents. With littleeffort we can weave a very tangled web. Theidea of a 'web' of causation, while implying thatthe interaction of a large number of factors maybe involved in the causation of an accident,leaves their identification and the assessment oftheir interactions to be discovered empirically,merely providing a scheme into which they maybe fitted after the fact.Suchman and Scherzer' have proposed that

accidents be studied as phenomena reflecting abalance between subjective assessment of theprobability of injury and risk-taking behavior.These factors are to be considered against anenvironmental background which containssituations representing varying degrees of riskof accident. In these terms, prevention mightbe achieved by reducing the accident potentialof the environment, by increasing the accuracyofjudgment ofthe probability ofan accident, orby decreasing the willingness to take chances.This concept certainly has the advantage ofindicating a number of specific approaches toaccident research, using the methods ofbehavioral science. It also emphasizes thatsome accidental injuries are properly studiednot by treating the person injured as the studysubject, but rather the person who was capableof exercising some control over the situation,the person whose judgment was involved.

Epidemiologic study is based upon one cent-ral axiom, that disease is not randomly dist-ributed in human populations. If it were, then agroup of persons with the disease would cons-titute a random sample of the general popula-tion from which they came and study of theircharacteristics would be unrewarding. Insteadof this, disease shows aggregation in time, inspace, and according to measurable humantraits. It is this aggregation observed as varia-tions in the risk of occurrence of disease, whichis properly the substance of epidemiologicstudy. Variations in the risk of disease aremanifestations of variations in individualexposure or in individual susceptibility to thecausative factors.Thus, once a properly defined condition has

been selected, an epidemiologic study proceedsaccording to the following steps.

1. Measurement of variations in the risk of adisease in human populations. These varia-tions may be temporal, spatial, or accordingto any measurable attributes of the studysubjects; usually a combination of all ofthese.2. Development of inferences relating to thesignificance of these observations in terms ofexposure and susceptibility.

59

on January 4, 2020 by guest. Protected by copyright.

http://injuryprevention.bmj.com

/Inj P

rev: first published as 10.1136/ip.2.1.55 on 1 March 1996. D

ownloaded from

Stallones

3. Development of inferences explaining thevariations in exposure and susceptibility interms of other environmental or host factors,which may then be considered to play a partin the causation of the disease.At any step, the substitution of direct obser-

vation for inference or the confirmation ofinference by direct measurement greatlyenhances the strength of the chain. It was thisprocess which Frost described so well in theintroduction to the Delta Omega publication ofJohn Snow's work on cholera:5

Epidemiology at any given time issomething more than the total of itsestablished facts. It includes theirorderly arrangements into chains ofinference which extend more or lessbeyond the bounds of direct observa-tion. Such of these chains as are welland truly laid guide investigation tothe facts of the future; those that are illmade fetter progress.

This concept, that disease is not randomlydistributed, that the aggregations of disease areattributable to variations in exposure andsusceptibility, and these, in turn, are related tomeasurable differences in the characteristics ofpeople and in the human environment, pro-vides a theoretic framework of quite generalapplicability to epidemiology. It serves as a

guide to the design and conduct of

epidemiologic investigations, stimulates thepresentation of epidemiologic findings inprecise, quantitative terms, and encouragesstudy of the relation between findings and theirarrangement in ordered arrays.From such general theoretic considerations

should come imaginatively designed andcarefully conducted studies, and from these wemay expect useful specific formulations of thecauses of home accidents to emerge, togetherwith new insights into the nature and interrela-tions of those personal and environmentalfactors which modify an individual's risk ofexposure to those causes and his degree ofsuceptibiliy to injury following exposure. This,we may hope, will permit the development ofprograms for prevention that are scientificallysound, administratively feasible, and capable ofevaluation.

1 Suchman EA, Scherzer AL. Current research in childhoodaccidents. New York: Association for the Aid of CrippledChildren, 1960.

2 Suchman EA, Scherzer AL. Uniform definitions of homeaccidents. Washington, DC: US Department of Health,Education, and Welfare, Public Health Service, 1958.

3 Stallones RA, Corsa L Jr. Epidemiology of childhoodaccidents in two California counties. Public Health Rep1961; 76: 25-36.

4 MacMahon B, Pugh TF, Ipsen J. Epidemiologic methods.Boston: Little, Brown and Co, 1960.

5 Snow J. Snow on cholera. New York: The CommonwealthFund, 1936: ix.

You be the judge - preventable or 'freak accidents'?

'Freak accidents' can happen at a time or place when you least expect them. Here are someexamples to educate parents of hidden dangers.* A 2 5 year old boy was drinking from a straw when he tripped over a cat. The straw sliced

open the back of his throat, requiring surgery.* A 1 year old girl sitting in a high chair with the seatbelt secured slid down in the chair,

strangling her to death on the seatbelt. The chair was not equipped with a T-strap (thestrap which goes between the child's legs and secures to the seatbelt).

* A 2 year old girl was trying to step over a velvet rope divider at a bank. She stepped on itinstead, causing the metal supporting pole to tip over and strike the child on the forehead.The child required several stitches.

* A teenager was pushing shut a door by the glass. When the door shut, her hand wentthrough the glass. The glass just missed slitting her artery, and required 30 stitches.

* A 3 year old boy was shutting a corral gate and caught his thumb in it. The thumb wassevered; however, it was able to be surgically reattached.

* A teenager was sitting on her bike, waiting to cross the street. A tractor trailer attempted anarrow right hand turn, causing the trailer to jump the curb. The trailer pinned the girlagainst a telephone pole, killing her instantly.

* A toddler was brushing his teeth, and somehow managed to swallow his toothbrush. Along procedure was required to extract the toothbrush.

* A 6 year old boy playing on the beach was digging a pit in the sand. When the pit reachedapproximately 5 feet deep, he jumped in it. The pit collapsed, suffocating the child beforerescuers could reach him (from CCSN BBS).

(Editor's note: it seems to me all of these are preventable in one way or another. But othersmay disagree. If you do, do write and say why.)

60

on January 4, 2020 by guest. Protected by copyright.

http://injuryprevention.bmj.com

/Inj P

rev: first published as 10.1136/ip.2.1.55 on 1 March 1996. D

ownloaded from