Kellerman pyramid paper

INJURY PREVENTION/ORIGINAL RESEARCH
The Pyramid of Injury: Using Ecodes to
Accurately Describe the Burden of Injury
Michael C. Wadman, MD
Robert L. Muelleman, MD
J. Arturo Coto, MPH
Arthur L. Kellermann, MD, MPH
From the University of Nebraska
Medical Center, Department of
Surgery, Section of Emergency
Medicine, Omaha, NE
(Wadman, Muelleman); the
Office of Epidemiology,
Nebraska Health and Human
Services System, Lincoln, NE
(Coto); and the Department of
Emergency Medicine, Emory
University, Atlanta, GA
(Kellermann).
Study objective: Although much is known about injury-related deaths from the use
of external cause of injury codes (ecodes) on death certificates, the contribution of
nonfatal injury is unknown, with most information based on estimates from national
surveys. Some states mandate ecoding of charts for hospitalized patients, but few
require ecode assignment for emergency department (ED) records. Missouri and
Nebraska mandated ecoding of ED records in 1993 and 1994, respectively, allowing
for a more complete description of injuries in those states. We describe fatal and
nonfatal injury frequencies in Missouri and Nebraska by using ecodes, with graphic
representation in the form of injury pyramids.
Methods: Ecode frequencies for 1996 to 1998 for all injury causes in Missouri and
Nebraska were reported directly from their respective health departments. The
ecode frequencies were grouped according to the Centers for Disease Control and
Prevention’s ecode matrix for presenting injury and mortality data.
Results: During the study period, 13,052 deaths, 131,210 hospitalizations, and 1,914,140
ED visits occurred as the result of injury. The most frequent lethal injuries were unintentional motor vehicle crashes (32.3% of total deaths), self-inflicted gunshot wound
(13.2%), unintentional falls (11.3%), gunshot wound from an assault (7.7%), and unintentional poisoning (4.3%). The leading causes of injury-related hospitalization were
unintentional falls (47.8% of total hospitalizations), unintentional motor vehicle crashes
(15.5%), self-inflicted poisoning (6.5%), and overexertion or strenuous movements
(2.4%). Of 1.9 million ED injury visits, unintentional falls accounted for 24.3%, unintentionally being struck by an object or person for 14.6%, unintentional motor vehicle
crashes for 11.4%, unintentionally being cut or pierced for 10.7%, and overexertion or
strenuous movements for 8.5%.
Conclusion: Ecoding in Missouri and Nebraska provides a comprehensive data
retrieval system that allows for a graphic depiction of the burden of injury derived
from real patient encounters within specific geographic regions.
[Ann Emerg Med. 2003;42:468-478.]
Copyright © 2003 by the American
College of Emergency Physicians.
0196-0644/2003/$30.00 + 0
doi:10.1067/mem.2003.306
4 6 8
ANNALS OF EMERGENCY MEDICINE
42:4
OCTOBER 2003
THE PYRAMID OF INJURY: USING ECODES
Wadman et al
INTRODUCTION
Injury threatens the well-being of all Americans and
weighs heavily on the US health care system. Injury
results in approximately 147,000 deaths in the United
States annually and an estimated 37.6 million emergency department (ED) encounters, with 2.6 million
patients requiring hospitalization.1-3 Although heart
disease and cancer cause more deaths annually, injuries
account for 3.5 million years’ potential life lost before
65 years of age. Cancer is a distant second, with 2 million years’ potential life lost.4
Injury estimates derived from current data sources
contain potential limitations, making accurate description of the injury epidemic difficult and possibly misrepresenting the total effect of injury on our society.
Although death certificate data provide an accurate
description of the various causes of injury fatalities,
national surveys consisting of self-reported patient
information and trauma registries that reflect only local
injury events currently serve as the major sources of
nonfatal injury data. National surveys are limited in
terms of the data they collect, and they may not accurately reflect local injury concerns.5 Trauma registries
are usually limited to specific urban locales and may not
provide a representative sample of injuries in the population. Furthermore, trauma registries fail to provide
information for patients not requiring hospital admission.6 Although some individuals question the relevance of these data because of the relatively small cost
of ED care per case, others provide objective evidence
that outpatient encounters account for approximately
one third of health care expenditures for injury.7 Also,
ED visits for injuries to the hand, back, and brain may
result in significant disability not accounted for in acute
care expenditures alone. A full understanding of the
spectrum of injury requires a consistent data source
describing not only fatal injuries but also injuries requiring hospitalization or ED care.
The International Classification of Diseases, Ninth
Revision, Clinical Modification (ICD-9-CM) is currently
the accepted standard for defining and classifying the
nature of various health conditions. It requires an exter-
OCTOBER 2003
42:4
ANNALS OF EMERGENCY MEDICINE
nal cause of injury code, or ecode, for all ICD-9-CM
diagnoses from 800 to 999 (also known as N-codes),
describing the nature of injuries, poisonings, and adverse effects.8 All states ecode death certificates, and 23
states require ecoding of hospital discharges, but only
11 states mandate ecoding of ED records.9 Missouri and
Nebraska, among the first states to routinely assign
ecodes to ED records (since 1993 and 1994, respectively) provide an opportunity to obtain nonfatal injury
data.
The relative effect of fatal and nonfatal injuries is
often graphically depicted in the form of an “injury
pyramid,” an adaptation from Heinrich’s occupational
injury prevention research conducted in the 1930s.10
The pyramid represents, from top to bottom, the numbers of fatalities, hospital admissions, and ED visits
resulting from injuries (Figure 1). The size and shape of
the pyramid reflect the magnitude and nature of the
injury cause, allowing injury specialists to develop and
evaluate injury control programs according to the effect
of each injury mechanism on society. Ecode data ob-
Figure 1.
Injury pyramid.
4 6 9
THE PYRAMID OF INJURY: USING ECODES
Wadman et al
tained from death certificates compose the top of the
injury pyramid, but the lack of information about hospitalizations and ED visits and hospitalizations may
lead to erroneous estimates of the total burden of injury
and suboptimal prioritization of injury-prevention
activities.11
The purpose of this study is to describe a comprehensive data retrieval system for fatal and nonfatal injuries
by using ecoding of death certificates, hospital discharge records, and ED charts in Missouri and Nebraska,
providing data for a more complete and targeted determination of injury frequencies for each mechanism and
intent within a well-defined geographic region. The
depiction of the obtained data in the form of injury
pyramids may facilitate analysis of the relative effect of
each mechanism and intent on the health care systems
of the states included in the study.
METHODS
For this study, an injury record consisted of hospital
discharge record or ED record from all acute care civilian hospitals in Missouri and Nebraska (n=233) with an
ICD-9-CM diagnosis from 800 to 999. The Missouri
Center for Health Statistics–Missouri Department of
Health and the Nebraska Health and Human Services
System provided the requested ecode frequencies for
1996 to 1998.
Missouri and Nebraska require ecode reporting for
all inpatient hospital admissions and outpatient ED visits as mandated by state statutes passed in 1993 and
1994, respectively. Missouri requires hospitals to provide date of birth, sex, race, zip code, county of residence, admission date, procedures, total billed charges,
and expected source of payment for all hospital admissions and ED encounters to the Missouri Center for
Health Statistics–Missouri Department of Health. To
facilitate the implementation of ecode reporting in
Nebraska, the Nebraska Health and Human Services
System uses the Nebraska Hospital Information System
of the Nebraska Hospital Association by using DataTrac
software (SKC, Eighty Four, PA), rather than requiring
hospitals to report directly to the Nebraska Health and
4 7 0
Human Services System. The Nebraska Hospital
Information System uses the electronic version of the
UB-92 (HCFA-1450), which includes the following
data: procedures, admission date, discharge date, disposition, date of birth, sex, county of residence, and zip
code. Ecoded death certificates from both states provided cause of injury fatality data. The grouping of
ecode data for injury-related deaths, hospitalizations,
and ED visits followed the Centers for Disease Control
and Prevention’s ecode matrix for injury cause or mechanism and injury intent.12
US Bureau of the Census population data for 1997
allowed the calculation of annual rates of deaths, hospitalizations, and ED visits per 100,000 individuals.
Further demographic data, including racial make-up,
age distribution, and population density for Missouri,
Nebraska, and the United States, were also obtained
from the US Bureau of the Census.13
Injury mechanism and intent category rates per
100,000 population determined from Missouri and
Nebraska ecode data were compared with National
Electronic Injury Surveillance System–All Injury Program estimated rates by using χ2 analysis.14
The study received exempt status from the University of Nebraska Medical Center institutional review
board before initiation and received assistance from
personnel at the Missouri Department of Health and the
Nebraska Health and Human Services System.
R E S U LT S
The population demographic characteristics for the
states included in this study and for the United States as
a whole are presented in Table 1. A comparison of the
population demographics in the United States with the
combined population for Missouri and Nebraska
revealed differences that were statistically significant
by χ2 analysis (P<.001).
During the 3-year period from 1996 to 1998, injuries
accounted for 13,052 deaths, 131,210 hospitalizations,
and 1,914,140 ED visits in Missouri and Nebraska. The
combined annual rates for the 2 states under consideration were 62/100,000 injury deaths, 620/100,000 injury-
ANNALS OF EMERGENCY MEDICINE
42:4
OCTOBER 2003
THE PYRAMID OF INJURY: USING ECODES
Wadman et al
related hospitalizations, and 9,039/100,000 injured
patients treated and released from an ED.
Overall, the injury mechanisms most commonly
fatal in this study were motor vehicle crashes, firearms,
and falls. Among the nonfatal injuries, falls, motor
vehicle crashes, and poisonings resulted in the most
hospitalizations. The leading causes of injury-related
visits to the ED were falls, motor vehicle crashes, and
struck by or against an object or person (Figure 2; Table
2).
Specifying intent, as well as mechanism, the most
frequent causes of injury-related death were unintentional motor vehicle crashes (4,211; 32.3% of total
injury deaths), self-inflicted gunshot wound (1,721;
13.2%), unintentional falls (1,471; 11.3%), assaultive
gunshot wound (1,002; 7.7%), and unintentional poisoning (562; 4.3%). The leading causes of injuryrelated hospital admissions were unintentional falls
(62,717; 47.8% of total injury-related hospitalizations), unintentional motor vehicle crashes (20,332;
15.5%), self-inflicted poisoning (8,540; 6.5%), and
overexertion or strenuous movements (3,223; 2.4%).
Of the more than 1.9 million ED injury visits, unintentional falls accounted for 24.3% (465,800), unintentionally being struck by an object or person accounted
for 14.6% (279,501), unintentional motor vehicle
crashes accounted for 11.4% (217,273), unintentionally being cut or pierced accounted for 10.7% (204,601),
and overexertion or strenuous movements accounted
for 8.5 % (163,464; Figure 3; Table 3).
Violent injury events, combining assault-related and
self-inflicted injuries, accounted for 4,218 deaths (32.3%
of total injury-related deaths), 14,720 hospitalizations
(11.2%), and 94,999 ED encounters (5.0%) from 1996
to 1998. Self-inflicted gunshot wounds, assault-related
gunshot wounds, and intentional poisonings were the
most common causes of violent-injury deaths in the
Table 1.
Table 2.
Study population demographic characteristics.
Demographic
Characteristics
Nebraska
Combined United States
Population
5,402,000
Urban (MSA), %
68.0
Black, %
11.2
Native American, %
0.4
Asian, %
1.1
White/Non-Hispanic, %
87.3
Hispanic/Latino
1.5
(any race), %
1,657,000
48.7
4.0
0.9
1.3
93.8
4.1
7,059,000
63.5
9.5
0.5
1.1
88.8
2.1
OCTOBER 2003
42:4
ANNALS OF EMERGENCY MEDICINE
Injury totals: All intents.
Injury mechanisms: ED visits, hospitalizations, and deaths.
Missouri
MSA, Metropolitan statistical area.
Figure 2.
268,930,000
79.8
12.7
0.9
3.7
82.7
11.0
Injury
Mechanism
ED Visits
(n=1,914,140)
Hospitalizations
(n=131,210)
Deaths
(n=13,052)
Falls
MVCs
Firearms
Struck
Poisonings
Cut/pierced
Other
Unspecified
466,174
219,730
4,801
333,282
12,449
212,356
159,556
123,883
62,839
21,050
2,271
5,436
12,994
3,144
4,460
4,938
1,504
4,385
2,847
106
1,184
229
264
334
MVC, Motor vehicle crash.
4 7 1
THE PYRAMID OF INJURY: USING ECODES
Wadman et al
study. Intentional poisonings, being struck by or against
an object or person, and being cut or pierced represented the leading causes of violence-related hospitalization and ED use (Table 4).
Injury pyramids constructed from the data obtained
for each injury cause revealed different sizes and shapes,
representing the varying magnitude and lethality of
each mechanism of injury and intent. Analyzing the
size and shape of each injury pyramid facilitates an
assessment of the relative frequency and lethality of the
injury mechanism and intent under consideration,
leading to widely varying rates of death, hospitalization,
and ED use (the x axis, or width, scales vary with each
injury mechanism).
The pyramid for unintentional motor vehicle crashes,
the leading overall cause of injury death, reveals not
only a large number of deaths (nearly one third of total
deaths) but also huge numbers of patients with nonfatal
injuries requiring either hospitalization or ED treatment. The size and shape of the pyramid confirms that
unintentional motor vehicle crashes are not only a leading mechanism of injury with lethal potential but also a
major cause of nonfatal injuries requiring hospitaliza-
tion or ED visits. Therefore, they affect use of the health
care system to a greater extent than might be appreciated by analysis of mortality data in isolation (Figure 4).
The firearm injury pyramid shapes vary markedly
because of the lethality of each intent category. The
intentional self-inflicted firearm injury pyramid seems
inverted because of the high case-fatality ratio inherent
in this injury cause. Suicide attempts with a firearm
account for a large percentage of total deaths, but they
represent a far smaller percentage of hospitalizations
and ED visits because there are few survivors. The
assault-related gunshot wound pyramid has a broad
top, again reflecting the fact that this cause of injury is
highly lethal. Assault-related gunshot wounds produce
similar rates of death, hospitalization, and ED visits.
The shape of the unintentional gunshot wound pyramid seems similar to other unintentional injury causes
with a smaller percentage of patients requiring hospitalization (Figure 5).
Injuries from falls were almost entirely caused by
unintentional events; only 0.04% of total falls were
coded as resulting from violence. Unintentional falls,
the third most common cause of injury deaths in this
Figure 3.
All injury mechanisms and intents.
4 7 2
ANNALS OF EMERGENCY MEDICINE
42:4
OCTOBER 2003
THE PYRAMID OF INJURY: USING ECODES
Wadman et al
study, also represented the leading cause of injuryrelated hospitalizations and ED visits. The injury pyramid for unintentional falls graphically demonstrates
the large number of hospital admissions and ED visits
for each fatality (Figure 6).
Patients struck by an object, either unintentionally
or by assault, produced a proportionately large number
of ED encounters relative to injury-related deaths.
Unintentionally being struck by an object accounted
for 14.6% of all injury-related ED visits in our study but
only 0.6% of total injury deaths. The low case-fatality
ratio for this cause of injury results in an ED impact
more than 3,000-fold higher than that of fatal cases and
80-fold higher than that of hospitalized patients. Likewise, being struck by an object or person during an
assault accounted for 54.9% of violence-related ED visits but only 0.6% of violent injury fatalities (Figure 7).
Unintentional poisonings led to a large number of
ED visits, hospitalizations, and deaths. Intentional
ingestions of poison affected the ED less but produced
relatively more hospitalizations. Rates of death from
poisoning were similar, regardless of intent (Figure 8).
Injuries resulting from cutting or piercing led to a
large number of ED evaluations, unintentional injuries
accounting for the fourth leading cause of ED visits in
this study. The pyramids generated by this injury mechanism reflect the relatively low case-fatality ratio of this
mechanism, especially when these injuries are unintentional (Figure 9).
Comparing the results from our study with the
National Electronic Injury Surveillance System–All
Injury Program data reveals statistically significant differences (P<.001) between estimated injury mechanism
and intent rates derived from a national cross-sectional
survey and data obtained from actual ED patient numbers from the defined region under study14 (Table 5).
DISCUSSION
The injury pyramids constructed in this study validate
the notion that different causes of injury produce
markedly different rates of death, hospitalization, and
ED visits, which explains why the pyramids we constructed differed widely in size and shape. In general,
injury data result in 3 distinct pyramid types according
to the lethality of the cause of injury. Injury mechanisms with low case-fatality ratios generate the classically shaped pyramid (type A), whereas causes with an
intermediate case-fatality ratio produce a rectangular
pyramid (type B). A third pyramid type, an inverted
pyramid, results from mechanisms with high casefatality ratios (type C; Figure 10). Viewing injury from
the limited perspective of fatal outcomes may lead to
disproportionate attention to the highly lethal injury
mechanisms that produce type C pyramids, such as
firearm suicides and homicides, whereas injury causes
with relatively lower case-fatality ratios (type A and B
Table 3.
Injury mechanisms and intents: ED visits, hospitalizations,
and deaths.
Injury Intent/Mechanism
Unintentional MVC
Self-inflicted gunshot wound
Unintentional fall
Unintentional poisoning
Self-inflicted poisoning
Overexertion/ strenuous
movement
Unintentionally struck by object
Unintentionally cut/pierced
OCTOBER 2003
42:4
ED Visits Hospitalizations Deaths
(n=1,914,140) (n=131,210)
(n=13,052)
217,273
198
465,800
22,738
9,847
163,464
20,332
191
62,717
3,837
8,540
3,223
4,211
1,721
1,471
562
508
1
279,501
204,601
3,330
1,708
81
6
ANNALS OF EMERGENCY MEDICINE
Table 4.
Violent injury mechanisms and intents: ED visits, hospitalizations, and deaths.
Injury Intent/Mechanism
Self-inflicted gunshot wound
Assault-related gunshot wound
Self-inflicted poisoning
Assault-related struck by/against
Assault-related cut/pierced
Self-inflicted cut/pierced
ED Visits Hospitalizations Deaths
(n=94,999)
(n=14,720)
(n=4,218)
198
1,609
9,847
52,111
5,292
2,203
191
1,090
8,540
2,076
977
427
1,721
1,002
508
24
193
29
4 7 3
THE PYRAMID OF INJURY: USING ECODES
Wadman et al
pyramids) could be underemphasized. In fact, injuries
that produce type A pyramids, such as motor vehicle
crashes and falls, have a far greater cumulative effect on
Figure 4.
Motor vehicle crashes.
Figure 5.
Firearms.
4 7 4
health care system use and costs than high-lethality
injuries such as gunshot wounds. Graphic depiction of
the burden of injury in the form of pyramids allows
direct comparison between different injury mechanisms and intents, allowing program planners to consider the full spectrum of fatal and nonfatal injuries
before making decisions that prioritize the prevention
of one type over another.
This study demonstrates the feasibility and potential
value of ecode reporting on a statewide level of all injured patients requiring hospital admission or ED evaluation. States such as Missouri and Nebraska mandate
reporting of ecode data for all patients hospitalized or
receiving care in the ED, which allows for a full understanding of injury from real patient numbers within
these specific geographic regions and facilitates tracking of injury causes over time. In the absence of ED and
hospital discharge ecoding, trauma center registries can
provide estimates of the numbers of patients hospitalized for injury, with some important limitations. The
resources required to abstract data from medical records
limit this type of data collection to large metropolitan
Figure 6.
Falls.
ANNALS OF EMERGENCY MEDICINE
42:4
OCTOBER 2003
THE PYRAMID OF INJURY: USING ECODES
Wadman et al
trauma centers, leading to an overrepresentation of
injuries common to the urban setting, such as firearm
injuries, and possibly neglecting suburban and rural
injury concerns.6 Injured patients hospitalized at
non–trauma center community hospitals are missed by
trauma registries, as are patients who do not require
hospital admission. These groups represent a vast number of health care encounters, as demonstrated in this
study.
Nationally, several cross-sectional surveys provide
estimates of injury frequency but may not reflect local
injury patterns. The variability of injury patterns between different geographic areas was demonstrated in a
previous study reporting ecode data from Missouri,
which revealed important differences in leading injury
mechanisms between 2 locales within the same state.7
Injury-prevention programs must consider local injury
problems to maximize the effect of planned intervention activities, and reliance on national survey data may
lead to inefficient targeting of increasingly scarce resources.
National surveys addressing patients treated and
released from the ED include the National Hospital
Ambulatory Medical Care Survey, the National Health
Interview Survey, and the National Electronic Injury
Surveillance System, but only the National Electronic
Figure 8.
Poisonings.
Figure 9.
Cut or pierced.
Figure 7.
Struck.
OCTOBER 2003
42:4
ANNALS OF EMERGENCY MEDICINE
4 7 5
THE PYRAMID OF INJURY: USING ECODES
Wadman et al
Injury Surveillance System follows a format similar to the
ICD-9-CM ecode matrix.14-16 The National Electronic
Injury Surveillance System, a hospital data collection
system previously focusing on consumer product–
related injuries, now includes all types of external
injuries and poisonings. The National Electronic Injury
Surveillance System–All Injury Program collects data
from 66 of the 100 National Electronic Injury Surveillance System sample hospitals for approximately
600,000 injury and consumer product–related ED visits
annually. Trained coders classify each injury event into
intent and mechanism of injury groups, consistent with
ICD-9-CM guidelines, but fail to assign actual ecodes
because of limited resources.14 Comparison of findings
from our study with National Electronic Injury Surveillance System–All Injury Program data demonstrates
significant differences between injury rates for certain
mechanism and intents. Violence-related injuries occur
at lower rates in the study states than on the national
level, whereas unintentional injuries approximate the
national trends. This variance from the national injury
rates may lead program planners in this specific region
to allocate resources differently than would be the case
in the absence of ecode data.
Considering patients admitted to the hospital, the
National Hospital Discharge Survey collects data from a
national probability sample of approximately 270,000
inpatient records representing about 500 nonfederal
“short-stay” (average length of stay <30 days) hospitals.
The 1997 National Hospital Discharge Survey reported
injury and poisoning as the first listed diagnosis for
8.2% of all patients discharged from survey hospitals,
which represents 2.5 million patients hospitalized for
injury in 1997, an admission rate of 932/100,000 population.17 We found an injury admission rate of 620/
100,000, somewhat less than reported in this survey.
Ecoding of injury records provides a valuable means
of obtaining data and allows a characterization of the
burden of injury. This study, however, is subject to important limitations. Miscoding is a problem inherent to
any system that seeks to assign categories to the seemingly infinite number of mechanisms by which people
injure themselves, and underreporting may occur, espe-
Figure 10.
Pyramid types.
Table 5.
Annual ED visits: combined ecode data versus National
Electronic Injury Surveillance System–All Injury Program
estimates.
Combined Ecode
Injury Intent/Mechanism
Unintentional
Fall
Struck by/against
MVC
Cut/pierced
Overexertion
Violence-related
Assault-related struck
by/against
Self-inflicted poisoning
Assault-related cut/pierced
Self-inflicted cut/pierced
Assault-related gunshot wound
*
NEISS-AIP
No.
Rate
No.
Rate*
155,267
93,167
72,424
68,200
54,488
2,200
1,320
1,026
966
772
7,021,456
5,565,289
3,298,869
2,428,021
3,243,702
2,555
2,025
1,200
884
1,180
17,370
246
1,307,649
476
3,282
1,764
734
536
46
25
10
8
149,898
120,388
54,231
45,632
55
44
20
17
NEISS-AIP, National Electronic Injury Surveillance System–All Injury Program.
*
Per 100,000 population.
4 7 6
ANNALS OF EMERGENCY MEDICINE
42:4
OCTOBER 2003
THE PYRAMID OF INJURY: USING ECODES
Wadman et al
cially with initial ecoding efforts. Each state recognizes
the potential for these problems and uses a system of
quality checks to ensure accurate and complete reporting. Missouri verifies ecoding of medical records by
ensuring documentation of a valid ecode on at least 99%
of a hospital’s injury records. Missouri data from a representative 1997 sample reviewed by expert coders
indicated that for injury records, 88% of inpatient and
98% of outpatient records included an accurate ecode.
For Nebraska, 90% of records with an injury code, Ncode, as the first listed diagnosis also displayed a valid
ecode in 1996, and 86% of such records in 1997 and
1998 contained a valid ecode. Nebraska estimates, with
the Nebraska Health and Human Services System 1998
annual survey as the criterion standard, determined
that 91% of inpatient and 65% of ED records had an
accurate complete ecode. The lower percentage of agreement between ecode determined by expert reviewer and
ecode contained in the hospital discharge record represents the complete ecode, and any discussion of what
constitutes an “accurate” ecode must include what level
of coding is under consideration. One study comparing
ecodes assigned to computerized hospital discharge
records with those assigned by an expert reviewer found
agreement for the complete ecode ranging from 57% for
firearm injuries to 72% for poisonings. The same study
found agreement of 87% for mechanism of injury and
95% for intent.18
Although the ICD-9-CM ecode matrix provides a format to describe the injury patterns of a region, it is not
yet the ideal system because of certain inherent deficiencies. The coding rule requiring the determination
of a single cause of injury is problematic. When the initial event leading to the injury takes precedence over
any subsequent events, the true cause of injury is often
obscured. An example described in a recent study found
a 17.7% increase in drowning deaths determined by
records review over deaths reported by submersion
ecodes alone. One description of missed drowning
deaths in the report involved motor vehicles that left
the highway and entered water and received motor
vehicle crash rather than submersion ecodes.19 The
implementation of International Classification of
OCTOBER 2003
42:4
ANNALS OF EMERGENCY MEDICINE
Diseases, 10th Revision may improve the relevancy of
coding in this example.
The potential for incomplete data, a third limitation
to the use of ecodes for injury data reporting, arises
from the method of data collection. The use of the UB92 billing form for ecode reporting in Nebraska leads to
the omission of some data elements that may serve an
important role in defining the nature and impact of
injury, such as third-party source of payment and patient
race and ethnicity. In addition, Missouri and Nebraska
do not obtain information about the victim’s level of
education, income, alcohol or substance use, first
admission or subsequent admissions because of the
same primary injury, source of admission (ED versus
clinic), and long-term disability. Addressing long-term
disability and subsequent health care encounters is of
particular importance because some authors assert that
nonfatal injuries producing an ED encounter but not
requiring hospital admission represent a relatively
minor component of health care expenditures because
of the small cost for each case of ED care.20 Many ED
visits, however, involve mechanisms (such as falls and
being struck by or against an object or person) that lead
to injuries of the hand, lower extremity, brain, and back,
all associated with significant long-term impairment
and associated loss of productivity.
The results obtained from this study underscore the
massive effect of injury on the overall health status of the
residents of Missouri and Nebraska. Despite the enormousness of this toll, injury prevention and trauma care
receive little attention relative to other health issues. The
lack of quality data may suggest one possible explanation
for the mismatch between health impact and health care
dollars spent. Although ecoding of hospital records for
inpatients takes place in about half of all states, only 11
states routinely assign ecodes to ED visits. Without accurate data, decisionmakers in health care, business, insurance, and managed care cannot measure the financial
impact of injuries caused by falls, motor vehicle crashes,
and other potentially preventable accidents.
Improving data capture and providing payers with
information addressing the contribution of injuries on
premature death, hospitalization costs, and ideally, dis-
4 7 7
THE PYRAMID OF INJURY: USING ECODES
Wadman et al
ability rates may serve to spotlight the public health
impact of various causes of injury. This information
may lead citizens and policymakers to more fully
understand the potential benefits of directing a more
proportionate share of health care resources to injury
prevention and trauma care.
In conclusion, ecoding of records for injured patients
hospitalized or treated and released from the ED, as well
as injury deaths, provides local data from real patient
encounter numbers to characterize a state or community’s burden of injury. Analysis of nonfatal injuries, as
well as fatalities, allows for a more informed prioritization of injury control efforts and may lead to more targeted approaches to injury prevention.
We thank Mark Van Tuinen of the Missouri Center for Health
Statistics–Missouri Department of Health for his assistance in the
acquisition of data for the State of Missouri.
Author contributions: MCW, RLM, and ALK conceived the study.
MCW and JAC supervised data collection. MCW, RLM, JAC, and
ALK analyzed the data. MCW drafted the manuscript; RLM, JAC, and
ALK contributed substantially to its revision. MCW takes responsibility for the paper as a whole.
6. Waller JA, Skelly JM, Davis JH. Trauma center-related biases in injury research.
J Trauma. 1995;38:325-329.
7. Muelleman RL, Watson WA, Land GL, et al. Missouri’s emergency department ecode data reporting: a new level of data resource for injury prevention and control.
J Public Health Manage Pract. 1997;3:8-16.
8. US Department Of Health and Human Services. International Classification of
Diseases, 9th Revision, Clinical Modifications, 6th ed. Washington, DC: US Department
Of Health and Human Services; 1997. Publication (PHS) 96-1260.
9. Annest JL, Fingerhut LA, Conn JM, et al. How states are collecting and using
cause of injury data. Washington, DC: Injury Control and Emergency Health Services
Unit, American Public Health Association, 1998:1-14.
10. Heinrich HW. Industrial Accident Prevention. New York, NY: McGraw-Hill; 1936.
11. Rivara FP, Grossman DC. Injury Prevention. N Engl J Med. 1997;308:88-91.
12. McCloughlin E, Annest JL, Fingerhut LA, et al. Recommended framework for presenting injury mortality data. MMWR Recomm Rep. 1997;46(RR-14):1-30.
13. US Census Bureau. Profile of General Demographic Characteristics: 1997.
Washington, DC: US Census Bureau; 1998.
14. National Electronic Injury Surveillance System All Injury Program. National estimates of non-fatal injuries treated in hospital emergency departments. MMWR Morb
Mortal Wkly Rep. 2001;50:340-346.
15. Nourjah P. National Hospital Ambulatory Medical Care Survey: 1997 emergency
department summary. Advance Data from Vital and Health Statistics; No. 304. Hyattsville, MD: National Center for Health Statistics; 1999:1-24.
16. Warner M, Barnes PM, Fingerhut LA. Injury and poisoning episodes and conditions:
National Health Interview Survey, 1997. Vital Health Stat Series No. 10(202);2000:1-25.
17. Kozak LJ, Lawrence L. National Hospital Discharge Survey: annual summary, 1997.
Vital Health Stat Series No. 13(144);1999:1-41.
18. Lemier M, Cummins P, West TA. Accuracy of external cause of injury codes in
Washington States hospital discharge records. Inj Prev. 2001;7:334-338.
19. Smith GS, Langely JD. Drowning surveillance: how well do ecodes identify submersion fatalities? Inj Prev. 1998;4:135-139.
20. Robertson LS. Child injury control: surveillance and research questions. Am J Med
Sci. 1994;308:88-91.
Received for publication September 27, 2002. Revision received
February 19, 2003. Accepted for publication April 30, 2003.
Presented in poster format at the 6th World Conference on Injury
Prevention and Control, Montreal, Quebec, Canada, May 2002.
The authors report this study did not receive any outside funding or
support.
Reprints not available from the authors.
Address for correspondence: Michael C. Wadman, MD, Section of
Emergency Medicine, 981150 Nebraska Medical Center, Omaha, NE
681198-1150; 402-559-6948, fax 402-559-9659; E-mail mwadman@
unmc.edu.
REFERENCES
1. Murphy SL. Deaths: final data for 1998. Hyattsville, MD: National Center for Health
Statistics; 2000. Vital Health Stat Series No. 48 (11).
2. Popovic JR, Hall MJ. 1999 National Hospital Discharge Survey. Advance Data From
Vital and Health Statistics; No. 319. Hyattsville, MD: National Center for Health
Statistics; 2001.
3. McCraig LF, Burt CW. National Hospital Ambulatory Medical Care Survey: 1999
Emergency Department Summary. Advance Data From Vital and Health Statistics; No.
320. Hyattsville, MD: National Center for Health Statistics; 2001.
4. National Center for Health Statistics. Years of potential life lost (YPLL) before age
65. Hyattsville, MD: Vital Statistics System; 1999.
5. Annest JL, Conn JM, James SP. Inventory of Federal Data Systems in the United
States for Injury Surveillance, Research, and Prevention Activities. Atlanta, GA: National
Center for Injury Prevention and Control, Centers for Disease Control and Prevention;
1996:1-76.
4 7 8
ANNALS OF EMERGENCY MEDICINE
42:4
OCTOBER 2003