The Scope of Injuries as Public·Health and Research Problems

The narrow epidemiological view of a public-health problem such as in-
juries focuses on incidence (how many injuries occur), their severity (death
or length and type of disability) and factors that place one at greater or lesser
risk of these outcomes. In this book that narrow view is confined mainly to
the first three chapters. The basic theme of the book is that, technically, we
know enough about injuries to prevent many and reduce the severity of the
vast majority of them. There remains some basic epidemiological and other
research work to be done and the problems associated with it will be ad-
dressed. But the major focus is on the conceptual, behavioral, social, eco-
nomic, and legal barriers to injury control.

The term injury is used interchangeably with trauma throughout to
refer to damage to the body caused by exchanges with environmental energy
that are beyond the body's resilience.(1) Mechanical-energy exchanges in
motor vehicle crashes, shootings, and falls are the most common causes of
severe and fatal trauma. The result of acute exposure to large concentra-
tions of energy is usually called injury, while the result of long-term, less
concentrated exposures, such as to low-level ionizing radiation, is usually
classified as disease.

The definition of injury in terms of its necessary and specific cause,
energy exchange, avoids the issue of fault that has so pervaded scientific in-
vestigation as well as injury-control efforts. The attribution of fault or
human error is of prime concern to many persons in cases where the injury
was unintended, while the intent of the injured is usually the major focus of
attention in the case of suspected suicide or that of the one or more assail-
ants in the case of homicide. This obsession with blame is at least partly the
result of a legal system that focuses on allocation of compensation and pun-
ishment according to the intent or fault of the persons immediately involved.
Prior intent in the attribution of fault is not measured directly nor often
scientifically inferred; it is concluded from statements of the persons involved
and any physical evidence of the injury having been planned. While the
assessment of intent may be relevant to compensation of the injured party or
punishment of the party blamed for the injury, there is substantial doubt that
this process has much effect on the prevention of injuries. In contrast, the ap-
plication of public-health principles has shown great promise in the limited
instances in which they have been used. This book discusses those principles
and their applicability to specific energy sources and the injuries they cause.

It is interesting that notions of fault and negligence of individuals im-
mediately involved in damaging transfers of mechanical, thermal, chemical,
and radiation energy have seldom been applied to interpersonal transfers of
harmful biologic organisms. In medieval times, persons thought to be car-
riers of the plague, but who actually were not, were persecuted and in some
instances murdered.(2) But in modern times people seldom if ever think of su-
ing someone who conveys bacteria or viruses that result in disease. Surely
the person who knowingly has a disease that is transmitted by sneezing in
crowds, kissing, sexual intercourse, or whatever and who then infects others
by engaging in those activities is no less negligent than the alcoholic who
drives while intoxicated and injures someone. Why do we believe that the
latter is somehow more subject to control by legalistic fault finding and
punishment than the former? Infectious-disease epidemiologists seldom if
ever concern themselves with blame assignment, although carriers of the
more serious diseases may be pursued by public-health physicians for the
purpose of treating the disease and stopping the chain of transmission. Yet
the primary purpose of police and often of expert investigation of car
crashes is to assign fault in reports or to testify in lawsuits for damages.

Traditionally, and in much current popular parlance, people refer to
"accidents" when they mean unintentional injuries. The emphasis on intent
reinforces the blaming approach, and the broader rubric, accident, includes
a large set of phenomena in which no damage occurs. Most people daily ex-
perience unintended events that are called accidents (a tie in one's soup,
mistakes in typing, locking keys in one's car) but that are not physically in-
jurious. It is only those events that result in human damage that concern us
here. If for no other reason than the magnitude of the problem, it makes
more sense to think about injury control than accident prevention.

The significance in terms of public health of an injury is not whether it
was intended but the extent and ultimate outcome of the damage. Concepts
of blame, including blame attributed to chance, acts of God, and the like in
the unintentional case, are barriers to injury control. The fact that injuries
are not often considered as a public-health problem is one of the reasons
that they are a public-health problem.

Incidence and Severity

Because many injuries are minor cuts and bruises that are not medically
treated and are soon forgotten, estimates of the total incidence of injuries
are questionable. The National Health Survey reports that in 1979 about 74
million injuries occurred to 69 million persons--more than a third of the U.S.
population. This estimate is based on recall in interviews of a random sam-
ple of the population.(3) Obviously, only a percentage of the minor cuts and
bruises were probably included.

Injuries in the survey report are classified by where they happened, age
and sex of the injured, and numbers of days in bed as well as days of re-
stricted activities associated with the injuries. Per population in each group,
injuries were reported more frequently among males (38.6 percent) than
females (25.9 percent) and more often among children and adults up to age
forty-four (35 to 39 percent) than among adults aged forty-five to sixty-four
(23.3 percent) and older (18.2 percent). The responses indicate that 12.5 per-
cent of the population was injured in the home in 1979, compared to 8.9
percent at work and 2.7 percent by motor vehicles. An additional 14 percent
were in the "other" category.

The cliche that "accidents occur most often in the home" may be true,
but such statements are misleading with respect to severity. If the reported
bed disability days are divided by the numbers of persons injured, a quite
different perspective emerges. The bed disabilty days per person injured is
highest among those injured in or by motor vehicles, more than twice that in
either the work place or home (Table 1-1). This statement is qualified
somewhat by the fact that for some people the motor vehicle is the work
place. Also, injuries are much more likely to result in bed disability the
older the person injured, particularly so among the elderly.

At least three factors could account for these patterns. First, persons be-
ing interviewed in their homes may selectively recall injuries that happened
there more often and/or selectively remember those that happened to the
children or other members in the household. Second, the level of violence in
motor vehicle crashes is more often severe compared to that in home and
work incidents. And, third, resilience declines with age.

Assessment of injury severity based on clinical evidence is usually more
accurate than recall of incidents such as bed disability. Substantial progress
has been made toward objectively classifying injury severity and relating it
to predictability of outcome. For research purposes, though seldom used
clinically, the Abbreviated Injury Scale (AIS) is used to rank injuries in five
categories, exclusive of death: (1) minor (for example, ache or stiffness);
(2) moderate (such as simple rib fracture); (3) severe, not life-threatening
(such as multiple rib fracture affecting respiration); (4) severe, life-threat-
ening, survival probable (for example, flail chest); and (5) critical, survival
uncertain (such as aortic laceration).(4) In multiple injury cases, each injury is
scored but the most severe (MAIS) is often the only one reported.

Follow-up study of 2,128 persons who suffered motor vehicle injuries
and who were hospitalized or who died during a two-year period in Balti-
more led to an important refinement of the AIS. The researchers noted that
survival declined exponentially as the MAIS increased. Also, people who
had two injuries scored as 3 and 4 respectively had about the same survival
rate as those with one injury scored (5). Further work with the data led to the
Injury Severity Score (ISS), "the sum of the squares of the highest AIS grade
in each of the three most severely injured areas."' The proportion of per-
sons who died was a linear function of injury severity scores above 25. Also,
an elderly person with the same ISS as a younger person was more likely to
die.

The injury-severity-scoring procedure has been simplified into a stan-
dardized instrument that can be used in case abstraction from medical
records and subsequently translated into AIS, MAIS, or ISS from a com-
puterized dictionary.(6) The researchers who developed this procedure are ap-
plying it to all injuries seen in acute care facilities in five northeast Ohio
counties--up to now a unique opportunity to compare injuries from a variety
of sources on the same severity scales.

Most studies of injury incidence and severity are limited to a specific
source of injury, site of injury, or disability outcome. The AIS was originally
developed to score motor-vehicle injuries and it has been most widely used
in classifying such injuries. Although a nationally representative sample of
all injuries of any kind has never been scored by the AIS or other clinical
criteria, national estimates of motor vehicle injury distributions have been
derived from samples of crashes in which a vehicle was towed away in a
wide variety of areas around the country. The total incidence is estimated to
be about 2 percent of the population per year. This means that about 1 in 50
Americans (some 4.2 million) is injured annually in motor vehicles; and
more injuries of a similar or greater severity may occur to pedestrians in
situations where the vehicle is not sufficiently damaged to be towed.

Since the case findings for these estimates are from official police
reports, the actual incidence is probably higher. The Northeastern Ohio
Trauma Study of emergency-room trauma cases found almost 46,000 cases
related to motor vehicles compared to 32,000 in police records for 1977. The
discrepancy for aggravated assault and rape was even larger, the rate in
emergency-room cases being four times that in official police statistics.(7)
Higher injury rates based on hospital data compared to official police
records also have been reported in Britains and California.(8)

The distribution of the nonfatal injuries in the vehicle towaway study by
MAIS, age, sex, and age-sex specific population rates is presented in table
1-2. Although most of the injuries are in the less severe categories, more
than 1 per 500 population is categorized as severe (MAIS 3) or worse and
more than 1 per 2,000 population is classified as life-treatening (MAIS 4
and 5). The rates are much higher in the teenage and young-adult groups,
particularly among males. About 1 of every 21 males aged 15 to 24 in the
population is injured annually in or by motor vehicles in a towaway crash
and 1 of every 565 males of that age sustains a life-threatening motor vehicle
injury annually, excluding those who die.

Studies of potentially life-threatening and permanently disabling injuries
such as spinal cord and head trauma find similar distributions. Almost 56
percent of traumatic spinal cord injuries and consequent deaths or para-
plegia and quadraplegia occur in or by motor vehicles. These, along with
those associated with firearms (12 percent), diving (5 percent), and other
recreation peak in teenaged and young adult males.(10) The exception is spinal
trauma associated with falls (19 percent), which is higher in the elderly
population. Approximately 2,500 to 3,000 new cases of permanent dis-
ability from spinal cord injury are added annually.(11)

In addition to paralysis, trauma to the head can result in seizures,
amnesia, personality changes, psychiatric disorders,(12) and disfigurement(13) if
the person survives. Using minimum criteria of loss of consciousness, post-
traumatic amnesia, or skull fracture for case finding, one study found an
annual incidence of 270 per 100,000 population in males and 116 per 100,000
population in females.(14) Forty-six percent of these injuries were associated
with motor vehicles or bicycles, the latter often to people struck by motor
vehicles. Falls accounted for 29 percent, and recreational incidents (led by
falls from horses and football injuries) contributed to 9 percent, of the head
trauma. As in the case of spinal-cord injury, head-injury rates peaked
among males in their mid- to late teens and early twenties in the motor-
vehicle and recreational cases. Falling from horses was the only case of
higher rates for women among the various activities of the young. Head in-
juries from nonrecreational falls were highest among the elderly. These data
come from a northern state where the severity of the winters reduces the ex-
tent of activities that contribute to head injuries, such as driving, motor-
cycling, diving, and horseback riding. In areas where these activities are
more frequent, the incidence of injuries is probably worse.

Persons with severe head trauma (brain contusion, intracerebral or intra-
cranial hematoma, or twenty-four hours of unconsciousness or amnesia)
suffered subsequent epilepsy many times more frequently than the general
population. Epilepsy occurs in less than 0.1 percent of the population; but,
excluding those with pretrauma seizures, 7.1 percent of persons who survived
severe head trauma had seizures within one year and 11.5 percent had them
within five years.(15)

Persons with less severe head trauma nevertheless have substantial
problems, including headache and memory problems three months after
the injury in the vast majority of moderate injury cases.(16) Among those clas-
sified as suffering minor head injury (loss of consciousness for twenty min-
utes or less), 34 percent who had been employed before the injury had not
returned to work, 79 percent had frequent headaches, and 59 percent were
experiencing memory losses at a three-month post-injury examination.(17)

The author is unaware of research that systematically separates out by
cause the loss of sight, hearing, and various lengths of limbs and other ap-
pendages from injuries. Surveys of disability suggest substantial increases in
five years of age; specific causes, however, are not adequately specified.(18) A
full accounting of the consequences of trauma would have to include these
as well as internal organs,(19) the removal or alteration of which often dis-
ables survivors to some degree. Documentation of the extent of mental retar-
dation,(20) pain, suffering, phobias, and impaired human relations involving
the injured, disabled, and those around them--including the survivors of
those who die prematurely--is less possible in any quantitative sense.

Occasionally persons bereaved by the loss of a parent, spouse, or child
are motivated to constructive action. One man whose son was murdered
with a handgun in San Francisco quit his job with a major corporation and
founded Handgun Control, Inc., an organization that is growing in effec-
tiveness as a lobby for gun-control legislation. But for every one such per-
son, there are tens of thousands annually whose grief, guilt, fear, revenge,
or other emotion may lead only to long-term psychological and social prob-
lems.

Little is known about the types and extent of such sequelae of trauma.
There is evidence that surviving children are treated differently (to their
detriment) when a child in the family dies. Children who lose one or more
parents are more often divorced and have criminal records when they grow
up. Inference of causation is difficult because of the delay in time between
the loss and the subsequent behavior, as well as the possibility of commonly
shared inheritance and environments that might contribute both to the
death and the child's subsequent problems.(21) Nevertheless, the anecdotal ac-
counts of permanently altered emotions and behavior(22) suggest that scien-
tists who develop ways of separating out nontraumatic contributing factors
will find that trauma is an important component in the etiology of psyche-
logical and social problems.

These data characterize periods during which the nation is not at war.
The nature of battle injuries changes from one war to the next, as weapons
are invented and others are abandoned or banned from use by international
agreements. Because of such changes as well as differences in classification
systems, comparisons of injuries among wars and among services in a par-
ticular war are difficult.(23) Such comparisons will not be attempted here. The
major focus is on severe and fatal injury in peace time.

Fatal Injury

We have greater knowledge about the circumstances of fatal injury in cases
where the death occurs soon after the injury than about the causes of non-
fatal injuries. More than 152,000 people in the United States died from in-
juries each year in recent years, about 1 death for every 2,000 people in the
population per year. In numbers, injuries are the third leading cause of
death after cardiovascular diseases (960,000) and malignant neoplasms
(397,000). Numbers of deaths, however, are not the complete story.

Table 1-3 presents, in addition to total numbers of deaths in the six ma-
jor categories, the median age at death and the approximate number of
years lost prior to the age sixty-five, the most frequent age of retirement in
1978. Deaths from cardiovascular diseases and malignancies are concen-
trated in the elderly: half of the deaths from the former occur to people
older than seventy-six years of age and half of the fatal malignancies involve
people older than sixty-eight. In contrast, injury is the predominant cause
of death for the young. Half of motor vehicle-related deaths occur among
persons twenty-seven years old or younger. Because of deaths from falls
among the elderly, the median age at death from other unintentional in-
juries is higher (fifty) but, nevertheless, substantially lower than the median
age of the leading causes of death in numbers alone. The median age at
death of persons who die from trauma associated with homicide or suicide
is thirty-one and forty-two, respectively. The conclusion to be drawn from
these figures is that the leading cause of loss of preretirement years of life is
injury. More than 2.8 million preretirement years are lost annually from un-
intentional injury and an additional 1.3 million years are lost from inten-
tional injuries. In total, these injury deaths take more preretirement years
than cardiovascular diseases and malignant neoplasms combined.

Such a conclusion, it should be pointed out, does not imply that the
retirement years of life are any less desirable or valuable than the preretire-
ment years. Nor does it imply that reductions in cardiovascular diseases and
malignancies are less desirable than injury control. It does emphasize the
enormity of the cost in economically productive years of life lost--indeed a
loss that potentially reduces the quality of life for the elderly because of
reduced contributions to Social Security and other programs by the income-
earning younger population. In individual cases among the elderly, the loss
of support by a child or the added responsibility of a disabled one would
multiply these burdens.

Without corrective action, the situation is likely to worsen in the 1980s.
Current trends in correlates of motor-vehicle fatalities led analysts at the
National Highway Traffic Safety Administration to project 70,000 motor
vehicle-related deaths per year by 1990, mainly because of reductions in
vehicle size, the effects of which more than offset the decline in the youthful
population.(24) A projected increase in the elderly population is likely to
result in more fatal and disabling falls.

As might be expected from the distributions of injury severity, fatalities
are concentrated in particular age and sex groupings. In recent years, about
37 percent of motor-vehicle fatalities occur among fifteen to twenty-four
year olds and more than three-fourths of these are males.(25) About 1 in every
120 adolescent males who reaches his fifteenth birthday will die of motor-
vehicle injuries before his twenty-fifth birthday. About 27 percent of homi-
cides occur in that age group; 78 percent of those are males.(26) One in every
410 males who reach age 15 dies by homicide, mainly gunshot wounds,
before age twenty-five. In summary, injuries from motor vehicles, other
unintentional injuries, homicide, and suicide, cause approximately 1 of
every 60 males to die of injuries in the decade after his fifteenth birthday.

The only decade of life in which the percent of women who die from in-
jury is comparable to that of males is the seventy-five to eight-four-year-old
age group. About 13 percent of the deaths from unintentional injuries un-
related to motor vehicles occur in that age group; and women are involved
in a larger proportion than men--52 percent--but their greater numbers of
the population at that age keeps the rate per population lower than that for
men. Most of these deaths occur from injuries in falls or medical complica-
tions, such as pneumonia, that sometimes accompany post-fall immobility.

Economic Costs

In the absence of knowledge of the incidence and severity of trauma of all
types, it is not possible to estimate with any exactness the economic losses in
even relatively easily calculable direct costs such as medical care, court
costs, and funerals. Recent comparison of direct and indirect costs (un-
realized wages and so forth) due to coronary heart disease, cancers, and
motor vehicle injuries suggests that, if all trauma had been included, in-
juries would have been found to be the most costly health problem in the
United States. These cost estimates do not include amounts for pain and
suffering or secondary consequences of trauma or the diseases in damage to
psyches and interpersonal relationships.

The costs in medical care are greater for cancers and about the same for
coronary heart diseases as for motor vehicle injuries; but because of the
large differences in age distributions of those involved, the foregone pro-
ductivity of those disabled or killed in or by motor vehicles results in motor
vehicle injuries being second only to cancers in calculable costs. In 1975
dollars, the economic cost of cancers was $23.1 billion. Using the same ac-
counting methods, motor-vehicle injuries cost $14.4 billion and coronary
heart disease cost $13.7 billion.(27) Because motor vehicles are involved in only
half of the deaths from unintentional injury, and only half of severe dis-
abilities from spinal cord injury, it is not stretching the matter to attribute
the highest total societal costs to injuries. If homicide and suicide were in-
cluded, the total cost would be even more staggering. Those who would
argue that apples and oranges are being compared should remind them-
selves that cancer is not a homogeneous category: there are several hundred
known carcinogens.

Researching Injuries

At numerous points in the following chapters, it is necessary to qualify con-
clusions because of issues in research methodology. Too much of what
passes for research on injuries and injury control is of questionable value.
This is partly due to the fact that this field of investigation has only emerged
in recent years as a discipline perceived by competent researchers as worthy
of full-time endeavors. Almost twenty years ago, Haddon et al. warned
researchers that use of concepts and methods from any single one of the
traditional academic disciplines often resulted in too narrow an application
to this emerging field.(28) Nevertheless, some researchers based in traditional
academic disciplines venture occasionally into the field because they believe
that some theory or method in their disciplines can be applied easily to an
injury problem. In some cases the belief is correct but, like all research, the
field is mined with boobytraps for the unwitting who are unwilling to spend
the time and energy necessary to find them.

Based on more than a decade of reviewing research proposals and
research reports in the injury field, the author finds that most such research
endeavors can be classified in one of three groupings:
1. the researcher's main interest is focused on a phenomenon that may
have an effect on injuries;
2. the researcher has a research method or ameliorative approach that
she/he thinks applies to almost any issue, including injuries; or
3. the researcher is interested in a particular set of injuries and is trying to
explain their incidence and severity, and/or how they can be reduced.
Researchers with the first or second of these orientations are the more likely
to get into trouble.

Emphasis on a favorite hypothesized cause often leads to neglect of
alternative explanations and naivete regarding the appropriate categoriza-
tions as well as reliability and validity of data on injuries. For example, the
researcher primarily interested in alcohol as a cause of various problems
may develop an interest in injuries because of alcohol's potential effect on
behavior that increases the probability of injury. Repeatedly one sees
studies by such researchers that use police reports of alcohol use in "acci-
dents." Yet, police reports are notoriously unreliable indicators of alcohol
use."(29) Furthermore, when care is taken to measure alcohol in breath, blood,
or other body fluids, the extent of its presence is highly related to the degree
of violence that produced the injury. Illegal blood alcohol concentrations
are found in about 15 percent of drivers in all motor vehicle crashes
reported to police,(30) in about one-half of fatally injured drivers, and in two-
thirds of fatally injured drivers in single-vehicle crashes.(31) Also, the drink-
ing individual whose behavior contributed to the injury, but who is not in-
jured, such as a driver who strikes a pedestrian, often is not tested for blood
alcohol concentration. Thus, the extent of estimated "causation" is depen-
dent on the injured population investigated.

Reliance on a favored method often results in neglect of methods that
would produce more reliable and definitive results. Recently, the analysis of
trends in fatality rates from illness and injuries during long periods of years
in relation to other changes in society during that time, has been favored by
some epidemiologists and economists. Such analyses are often oriented to
debunking the efficacy of deliberate intervention to reduce harm. One epi-
demiologist has claimed that because many death rates attributable to cer-
tain infectious diseases were in substantial downtrends before the use of
antibiotics, their importance is insignificant.(32) Aside from the questionable
reliability of diagnosis in earlier years and the experimental evidence that
antibiotics kill several important microorganisms, the analysis ignores some
obvious points. The factors that affect a trend during one period are not
necessarily those that affect it during a subsequent period. Improvements in
sanitation or housing can reduce exposure to an organism during a period to
the point where there is no longer exposure from that source; however,
other exposures can continue to occur or increase that, when penicillin or
other therapy is used, reduces severity as the therapy is more widely applied.
It is even more illogical to infer that, had the therapy been available years
earlier, it would have had no effect on the trend because it was already slop-
ing downward. Historical data on relevant variables in the population is
usually inadequate to make a strong inference reearding the relative effects

A few economists and others have attempted to infer a lack of effect of
motor vehicle safety standards, introduced in the 1960s, in a similar
fashion.(33) But trends in motor vehicle fatalities and their correlates did not
behave in the same way in the 1940s and 1950s as they did in the 1960s.(34)
The urbanization and suburbanization of large segments of the population
had important effects on the vehicles used, the kinds of uses of vehicles, and
the age and sex groups using them that are not reflected in correlations of
gross death rates to aggregated statistics of miles driven on freeways, per-
cent of population aged fifteen to twenty-four, economic trends, and the
like. Much more refined data are necessary to make reasonable causal in-
ferences regarding ameliorative actions (see chapter 7).

A research method that has been too frequently used by psychologists,
sociologists, and epidemiologists is the questionnaire or interview. Despite
the well-established fact that people do not always do what they claim to
do--such as use seat belts--much less think what they claim to think, the
use of questionnaires and interviews is a common source of data supposedly
indicative of behavior and thought. This problem is illustrated by data
showing that 23 percent of the drivers who claimed to always use belts on
local trips were not using them when observed near their homes and that 54
percent who claimed to always use them on long trips were not doing so
when observed at a long distance from home.(35)

The estimates of the effectiveness of seat belts in reduction of injury
severity in crashes has ranged from 7 percent to 85 percent because of the
biasing effect of claimed use and the extent of severity of the injuries con-
sidered. The hypothetical data in Table 1-4 · illustrate how these factors
could affect estimates of effectiveness. In more severe crashes, use of belts
could be verified by belt marks on the persons involved or possibly by
studies of damage to belt fibers; but in the less severe cases, the only indica-
tion of belt use would be in response to interview. Notice in the table what
happens to the belt effectiveness estimate if 500 (5 percent) of those inter-
viewed falsely claimed to use belts. The belt-effectiveness estimate increases
from 40 to 53 percent. If one were to try to correct that error by using only
injury cases, however, the bias is in the opposite direction--belt effec-
tiveness would be grossly underestimated, 22 to 23 percent rather than 40
percent. The latter bias has long been known(36) and both have recently been
illustrated in studies of belt effectiveness.(37)

Researchers who are not pushing a single cause or a single methodolog-
ical or ameliorative approach also may overlook issues of reliability and
validity of data and methods. Focus on the end results in human damage,
however, seems to lead to a more systematic review of the range of possible
contributing factors as well as of ameliorative strategies and better, if not

Data Sources

When considering the collection of data, use of existent data, or the logic of
a research report, one must keep in mind the factors that contribute to
potentially invalid or biased results. One way to view the reporting of injuries
is to think of the filters or contingencies that increase or decrease the
probability that that type of injury will be reported and where.

Responses to questionnaires and interviews that solicit the recall of in-
juries are filtered by the memory and attitudes of the interviewees. An in-
jury that is forgotten or thought too trivial to mention by one respondent
may be considered important by another. The extent of bleeding, pain, time
lost from work or school, and the like may contribute to the degree to which
an injury is remembered or thought important enough to mention in
response to a question. To the extent that propensity to report particular
types of injuries is correlated to other personal characteristics, a false in-
ference of a causal role for such personal attributes may be drawn.

Injuries that are embarrassing or suggestive of culpability may not be
reported or, if severe enough to require treatment, may not be presented as
they happened. This selectivity is especially true of spouse and child abuse,
assaults, and altercations. Some people will go to an emergency room or
physician and/or call the police for injuries that are relatively trivial com-
pared to those that others treat themselves. Reports to police may be predi-
cated on relationship to others involved, social characteristics and de-
meanor of the injured and other involved persons, and responsiveness of
the police. Concern for economic consequences, such as liability suits or in-
surance surcharges, can influence how and to whom an injury is reported.

Another limitation of data sources is the filtering of information on in-
juries by the interviewers, police, physicians, and insurance adjustors who
gather it. Preconceptions regarding the physical and social characteristics
and the behavior of the injured as well as interest or disinterest in particular
injuries and/or their circumstances may result in unreliable or invalid re-
cording of information. Standards for reporting, legal requirements, and
reporting forms that neglect relevant variables are important factors in the
usefulness of official records. Until recently, for example, police report
forms used the nonsensical category "noncollision" for motor vehicle
crashes in which the vehicle left the road and hit a fixed object.

Police reports that concern factors on which they are not expert or that
must depend on the memory and veracity of the injured or witnesses have
often been found unreliable. For example, few police are medically quali-
fied to judge the severity of injuries even though they are required to do so.
Occasionally a person who is reported by police to have a trivial injury sub-
sequently dies. Even some physicians inexperienced in treating injuries may
misdiagnose severe internal injuries.(38) Likewise police reports of factors
such as precrash speeds of vehicles and, alcohol, drug, and seat belt use are
subject to serious biases.

Records of emergency-room visits and hospitalizations may contain
reliable data on clinical aspects or nonfatal injuries, but little or no good
data on the agents of circumstances of injuries. Some fatal cases do not
is admitted to the hospital or treated in the emergency room depends on
variable hospital and physician criteria.

Furthermore, most violent deaths are investigated by medical examiners
or coroners, but the quality of the data varies among jurisdictions. In areas
where well-trained forensic pathologists conduct the investigations, the
available data are often extensive and reliable. Where there is lack of exper-
tise or selectivity in choice of cases for autopsy and alcohol and drug
testing, the potential for unreliable or biased data exists. In some cases,
cause of death (such as suicide) may be altered or left off death certificates
to avoid embarrassment to families.

The extent of standardized, systematic reporting systems has been ex-
panded as federal and state agencies have been given the authority to inves-
tigate injuries and develop countermeasures. Inclusion of types and severity,
as well as relevant variables on sources of injury and persons involved,
varies according to the mission of the agencies, the feasibility of obtaining
the data, and the orientations of the persons who developed the systems.
Files of injury data related to motor vehicles,(39) consumer products (ex-
cluding important ones, such as motor vehicles and guns),(40) occupations,(41)
and boats,(42) among others, are maintained by federal agencies. The agencies
involved also conduct or support special investigations that may contain
more information than in their surveillance-type files. The data are public
information and are available to researchers, often at a fee for computer
processing of tables or cases that are not ordinarily published.

In some agencies the data are extensive in terms of both persons and
products involved. Although one must be cautious about the biases in police
reporting of alcohol use, belt use, and the like, the data on fatal motor vehi-
cle crashes are relatively rich in information. In contrast, the Consumer
Product Safety Commission reports nothing about the people involved in
the injuries in its summary data on products and the Bureau of Labor
Statistics does not collect data on the machinery or vehicles involved in
worker injuries in its general data-gathering effort. Some good research has
been done with these and other extant files; but there is often no substitute
for designing research to fit the questions to be answered and gathering
fresh data for that specific purpose.

Research Strategies and Designs

Much of the injury-control research that is done in private and government
laboratories by physicists, chemists, and engineers to establish human
tolerances and to test materials and products is not discussed in technical
materials and designs that result in improvements in areas such as energy ab-
sorption and reduced flammability have been and can be adopted for use
without field testing after they have undergone extensive laboratory tests. It
would have been foolish, for example, to fail to use energy-absorbing steering
assemblies in motor vehicles until field studies of their effects are completed.
Only a nihilist could argue that an energy-absorbing steering assembly of
whatever design is less preferable than a steel shaft that spears drivers in the
chests in frontal crashes. Field studies can be implemented after a product is
generally introduced in order to refine it for maximum benefit.

Most of the physical science necessary for application to injury control
(briefly outlined in Chapter 2) is well understood by those trained in dis-
ciplines such as physics, chemistry, biomechanics, and engineering. How-
ever, that does not mean that those so trained always pay appropriate atten-
tion to the injurious consequences of the materials and products that they
research or design for use in the human environment. One of the reasons
that injuries remain such a large public-health problem is that those who are
in a position to understand the injury problems their work produces usually
are not alerted to such considerations in any systematic way during their
training. Moreover, they often work in organizations where management is
either unfamiliar or unconcerned with such matters.

The behavioral sciences are relatively underdeveloped and their useful-
ness in injury control is therefore limited. Some of what is known about
behavior that contributes to injury is discussed in chapter 3. Most quantita-
tively trained behavioral scientists do have the skills to analyze experimental
and nonexperimental data when they understand enough of the substance
of a problem to ask the right questions and obtain the appropriate data. The
evaluation of attempted behavior changes to reduce injuries or the possibility
of inadvertent responses to changes in materials or products make the in-
volvement of behavioral scientists or epidemiologists cognizant of behav-
ioral issues in field studies useful.

It is important to recognize that field studies do not preclude the use of
tightly controlled experimental designs. For the nonscientist reader, the
controlled experiment, in which a change is introduced in one group but not
in an otherwise comparable control group, is the least questionable method
for inferring a possible cause or consequences of the change. If possible, the
experiment's subjects should be randomly assigned in order to decrease the
chances that some unrecognized difference between them will make the con-
clusion about the introduced change invalid.(43) Several controlled ex-
periments in driver education, television advertising, and use of incentive
systems to change behavior related to injuries are described in chapter 5.

When the scientist cannot control introduced changes or the groups into
which they are introduced (such as the introduction discussed),
that can simultaneously affect injury incidence or severity is necessary in
order to infer the relative effects of various factors.(44)Where the conse-
quences of a law should be abrupt and major in scope, the before-after
comparison of trends using appropriate statistical analysis may be adequate
to observe and infer the presence or absence of a major effect.(45) The further
examination of reasonably comparable groups of jurisdictions in which
changes in law or other introduced factors did not occur increases con-
fidence in these conclusions.(46) This process may seem relatively simple but
repeated examples are noted in chapters 5 through 7 in which public offi-
cials, other interested parties, and more than a few scientists inferred that
some program or law had an effect that closer examination indicated could
be reasonably attributed to expected fluctuation in other trends or factors.

It should be evident from the discussion of age and sex differences in in-
juries per population that comparison of groups that differ on these and
other factors must include a statistical correction for the differences or com-
parison of specific age-sex or other comparable subgroups. Also the choice of
denominator when calculating rates or the consideration of alternatives to
calculation of rates can make a remarkable difference in the inferences that
are made.

In illustration, the author and a colleague examined differences in in-
juries among workers in metal-working plants and found that the rate per
person-years worked did not peak in the youngest workers as might be ex-
pected (a person-year is one person working a year, two working six months
each, and so on). Instead, as shown in the top section of table 1-5, average
injuries per person-year were highest among workers in their thirties and
forties.

Suspicious because of the incongruity with other injury studies and
aware that hazards varied greatly among the departments in the plants, we
used the average injuries per person-year in each department to calculate an
expected number of injuries for each individual based on the number of
years or fraction of years he or she worked in a given department. When
this expected number of injuries was subtracted from the actual number of
injuries, the average difference was highest among teenagers, declined with
age through the fifties, but increased somewhat among the oldest workers--
a pattern similar to that for other types of injuries (see the bottom half of
table 1-5). The variations around the averages, indicated by the standard
deviations, were much more uniform than in the case of rates per person-
year. We concluded that management perceived differences among hazards
and placed younger, less experienced workers in departments where hazards
were fewer. When that differential exposure was accounted for, however,
the effects of inexperience and other correlates of age, such as impulsive-
ness, appeared.(47)

The researcher or the consumer of research should by now be aware
that the choice of a denominator for an injury rate, such as miles driven or
flown or years worked, involves assumptions about the denominator's lack
of correlation to an unplanned or programmatic change that is thought to
affect only the injuries in question. If the factor also has an effect on the
miles driven or flown, years worked, and/or when and where that activity
occurs--in other words, the quantity and quality of exposure to the hazards
involved--an appropriate means of indicating that fact must be found for
the research to accurately reveal the causal process.

References

1. W. Haddon, Jr., "Advances in the Epidemiology of Injuries as a
Basis for Public Policy," Public Health Reports 95(1980):411-21.
2. L.F. Hirst, The Conquest of Plague: A Study of the Evolution of
Epidemiology (Oxford: Clarendon Press, 1953).
3. National Center for Health Statistics, Current Estimates from the
National Health Interview Survey: United States, 1979 (Hyattsville, MD:
U.S. Department of Health and Human Services, 1981).
4. Committee on Medical Aspects of Automotive Safety, "Rating the
Severity of Tissue Damage I. The Abbreviated Scale," Journal of the
American Medical Association 215(1971):277. Committee on Medical As-
pects of Automotive Safety, "Rating the Severity of Tissue Damage II. The
Comprehensive Scale," Journal of the American Medical Association
220(1972):717. E. Petrucelli, J.D.States, and L.N. Hanes. "The Abbreviated
Injury Scale: Evolution, Usage and Further Adaptability," Accident Analysis
and Prevention 13(1981):29.
5. S.P. Baker, et al., "The Injury Severity Score: A Method for
Describing Patients With Multiple Injuries and Evaluating Emergency
Care," The Journal of Trauma 14(1974):187.
6. J.I. Barancik and B.F. Chatterjee, "Methodological Considerations
in the Use of the Abbreviated Injury Scale in Trauma Epidemiology," Jour-
nal of Trauma 21(1981):627.
7. J.I. Barancik, et al., "Northeastern Ohio Trauma Study I. The
Magnitude of the Problem," American Journal of Public Health, in press.
8. J.P. Bull and B.J. Roberts, "Road Accident Statistics: A Com-
parison of Police and Hospital Information," Accident Analysis and
Prevention 5(1973):45-53.
9. J.F. Kraus, et al., "Some Epidemiologic Features of Motorcycle
Collision Injuries: I. Introduction, Methods and Factors Associated With
Incidence," American Journal ofEpidemiology 102(1975):74-98.
10. J.F. Kraus, et al., "Incidence of Traumatic Spinal Cord Lesions,"
Journal of Chronic Diseases 28(1975):471.
Il. C.N. Smart and C.R. Sanders, The Costs of Motor Vehicle Related
Spinal Cord Injuries (Washington, DC: Insurance Institute for Highway
Safety, 1976).
12. H.S. Levin and R.G. Grossman, "Behavioral Sequelae of Closed
Head Injury. A Quantitative Study," Archives ofhreurology 35(1978):720.
13. T. Karlson, "The Incidence of Hospital-Treated Facial Injuries
From Vehicles," The Journal of Trauma 22(1982):303.
14. J.F. Annegers, et al., "The Incidence, Causes and Secular Trends
of Head Trauma in Olmstead County, Minnesota, 1935-1974," Neurology
30(1980):912.
15. J.F. Annegers, et al., "Seizures After Head Trauma: A Population
Study," Neurology 30(1980):683.
16.R.W. Rimel, et al., "Moderate Head Injury: Completing the
Clinical Spectrum of Brain Trauma," Neurosurgery 11(1982):344.
17. R.W. Rimel, et al., "Disability Caused by Minor Head Injury,"
Neurosurgery 9(1981):221.
18. A. Colvez and M. Blanchet, "Disability Trends in the United States
Population 1966-76: Analysis of Causes," American Journal of Public
Health 71(1981):464.
19.W.F. Northrup and R.I. Simmons, "Pancreatic Trauma: A
Review," Surgery 71(1972):27.
20. J. Caffey, "The Whiplash Shaken Infant Syndrome: Manual Shak-
ing By the Extremities With Whiplash-Induced Intracranial and Intraocular
Bleedings, Linked With Residual Permanent Brain Damage and Mental
Retardation," Pediatrics 54(1974):396-403.
22. R.J. Lifton, The Broken Connection: On Death and the Continuity
of life (New York: Simon and Schuster, 1979).
23. J.V. Henderson, "The Importance of Operational Definitions in
Design of a Combat Casualty Information System," Journal of Medical
Systems, in press.
24. National Highway Traffic Safety Administration, Traffic Safety
Trends and Forecasts (Washington, DC: U.S. Department of Transporta-
tion, 1981).
25. National Highway Traffic Safety Administration, Fatal Accident
Reporting System, 1979 (Washington, DC: U.S. Department of Transpor-
tation, 1980).
26. National Center for Health Statistics, Vital Statistics of the United
States--1978, Volume II, Mortality (Hyattsville, MD: U.S. Department of
Health and Human Services, 1981).
27. N.S. Hartunian, C.N. Smart, and M.S. Thompson, The Incidence
and Economic Costs of Major Health Impairments (Lexington, MA: Lex-
ington Books, 1981), p. 370.
28. W. Haddon, Jr., E.A. Suchman, and D. Klein, (eds.), Accident Re-
search: Methods and Approaches (New York: Harper and Row, 1%4), pp.
14-15.
29. Haddon, et al., (eds.), Accident Research, p. 208.
30. R.F. Borkenstein, et al., The Role of the Drinking Driver in Traffic
Accidents (Bloomington, IN: Indiana University Department of Police Ad-
ministration, 1964).
31. W. Haddon, Jr., et al. (anonymously), 1968 Alcohol and Highway
Safety Report (Washington, DC: Committee Print, Committee on Public
Works, U.S. House of Representatives, 1968).
32. T. McKeown, "Determinants of Health," Human Nature 1(1978):60.
33. S. Peltzman, "The Effects of Automobile Safety Regulation,"
Journal of Political Economy 83(1975):677.
34. L.S. Robertson, "A Critical Analysis of Peltzman's 'The Effects of
Automobile Safety Regulation,' " Journal of Economic Issues 11(19'1'1):
587.
35. P.F. Waller and P.Z. Barry, Seat Belts: A Comparison of Observed
and Reported Use (Chapel Hill: University of North Carolina Highway
Safety Research Center, 1969).
36. Haddon, et al., (eds.), Accident Research, pp. 705-6.
37. L.S. Robertson, "Estimates of Motor Vehicle Seat Belt Effec-
tiveness and Use: Implications for Occupant Crash Protection," American
Journal ofPublic Health 66(1976):859.
38. H.R. Gertner, et al., "Evaluation of Management of Vehicular
39. National Highway Traffic Safety Administration, Fatal Accident
Reporting System, 1979.
40. 1979 Annual Report (Washington, DC: U.S. Consumer Product
Safety Commission, 1979).
41. Bureau of Labor Statistics, Occupational Injuries and Illnesses in
the United States by Industry, 1978 (Washington, DC: U.S. Department of
Labor, 1980).
42. U.S. Coast Guard, Boating Statistics 1979 (Washington, DC: U.S.
Department of Transportation, 1980).
43.D.T. Campbell and J.C. Stanley, Experimental and euasi-
experimental Designsfor Research (Chicago: Rand-McNally, 1963).
44. D.G. Kleinbaum, L.L. Kupper, and H. Morgenstern, Epidemio-
logic Research: Principles and Quantitative Methods (Belmont, CA: Life-
time Learning Publications, 1982).
45. H.L. Ross, "Law, Science and Accidents: The British Road Safety
Act of 1967," The Journal of Legal Studies 2(1973):1.
46.L.S. Robertson, "An Instance of Effective Legal Regulation:
Motorcyclist Helmet and Daytime Headlamp Laws," Law and Society
Review 10(1976):467.
47. L.S. Robertson and J.P. Keeve, "Worker Injuries: The Effects of
Worker's Compensation and OSHA Citations,'' Journal of Health Politics,
Policy and Law, in press.