© International Epidemiological Association 1999
International Journal of Epidemiology 1999;28:1124–1129
Printed in Great Britain
Assessing disaster-attributed mortality:
development and application of a definition
and classification matrix
Debra L Combs,a Lynn E Quenemoen,b R Gibson Parrishc and Joseph H Davisd
Background A useful step in developing and implementing sound policies to prevent disasterattributed mortality is to classify the relationship between disasters and mortality.
While there are classification methods for specific health outcomes, there is no
standard method that includes all potential outcomes from exposure to a natural
disaster. Without standards, our ability to assess health effects from disasters and
implement prevention programmes is limited.
Methods
We present a method for ascertaining and classifying disaster-attributed mortality
which includes a case definition, flow chart, and matrix. The matrix is used for
coding, reporting, and evaluating information about manner, cause, and circumstance of disaster-attributed deaths and geographical location and time of the
disaster. To illustrate its use, two readers determine and classify deaths attributed
to Hurricane Andrew (1992, USA).
Results
Of 322 deaths investigated by the Dade County Medical Examiner’s Office, our
readers showed 97% (313/322) agreement on case status and 83% (35/42) agreement on case classification.
Conclusions Our definition allows for a liberal interpretation of what constitutes disasterrelated circumstances and the conditions or diseases that might arise from these
circumstances. The inclusion of the flow chart and matrix provides a framework
for consistent case classification and reporting. It also provides information
about relationships between exposures and health effects, thereby identifying
prevention policy needs.
Keywords
Case ascertainment, classification system, disaster epidemiology, hurricane, natural
disaster, disaster prevention
Accepted
22 June 1999
A useful step in developing and implementing sound policies
to prevent disaster-attributed mortality is to classify the
relationship between disasters and mortality. While there are
classification methods for specific types of health outcomes,
such as injuries, there is no standard classification method that
includes all of the potential outcomes of exposure to a natural
a Surveillance and Programs Branch, Division of Enviromental Hazards and
Health Effects, National Center for Environmental Health, Centers for
Disease Control and Prevention, Atlanta, GA, USA.
b Work Right, Duluth, MN, USA.
c Surveillance and Programs Branch, Division of Enviromental Hazards and
Health Effects, National Center for Environmental Health, Centers for Disease
Control and Prevention, Atlanta, GA, USA. Current address: Epidemiology
Program Office, Centers for Disease Control and Prevention, Atlanta, GA, USA.
d Office of the District 11, (Dade County) Medical Examiner, Miami, FL, USA
(now retired).
Reprint requests to: Debra L Combs, (F-47), Centers for Disease Control and
Prevention, 4770 Buford Highway NE, Atlanta, GA 30341-3724, USA.
disaster. This may be due, in part, to the lack of a uniformly
accepted definition for the exposure itself.
There are many definitions for disasters. The term ‘natural
disaster’ is often used to describe natural forces such as
hurricanes and earthquakes. Others—recognising that natural
phenomena are not, by themselves, disasters—define a disaster
in terms of the adverse effects resulting from the exposure and
the efforts required to correct these effects.1–5
Differences in the definition of a disaster lead to differences in
the definition of disaster-attributed health effects. While most
agree that direct exposure to environmental forces contributes
to morbidity and mortality, many disagree about the contribution
of indirect exposures to adverse health outcomes.6–11 For
example, in 1994, at the annual meeting of the National Association of Medical Examiners (NAME), we asked 63 attendees
to determine the relative contribution of a disaster to several
deaths that, although unbeknown to these attendees, had been
attributed to disasters by other medical examiners. These
1124
DISASTER-ATTRIBUTED MORTALITY
attendees had a wide range of opinion, with agreement ranging
from 27% to 100%, depending on the circumstances, causes,
and manner of death. They were less likely to agree if death was
from natural causes or if death resulted from indirect exposures
associated with the disaster.
Without standards, inconsistent data collection and reporting
limit our ability to assess health effects for a given disaster as
well as compare the effects associated with the same type of
phenomenon over time.8,10,12–17 This, in turn, limits our ability
to evaluate the effectiveness of prevention policies. Ultimately,
the lack of a standard definition and classification system may
lead to the implementation of costly, unnecessary programmes
as well as limit our ability to implement programmes aimed at
reducing mortality from future disasters.18
Since 1989, we (DLC, RGP) have been conducting mortality
surveillance in the US through the Medical Examiner and
Coroner Information Sharing Program (MECISP), which relies
on the opinion of the medical examiner or coroner (ME/C)
in the affected jurisdiction. To facilitate consistent classification
and coding of mortality attributable to disasters, we developed
the case definition and classification method described in this
paper. Although we illustrate the use of this method by assessing
mortality attributed to Hurricane Andrew in Dade County,
Florida (USA, 1992), the method could be used to assess
hurricane-attributed morbidity or health effects from other
types of disasters.
Method for Determining
Disaster-Attributed Mortality
For purposes of this paper, we defined a natural disaster as a
time- and place-specific event that originates in the natural
environment and the resulting disruption of the usual functions
and behaviours of the exposed human population.
Case Definition (Figure 1, Appendix)
We defined disaster-attributed deaths as those caused by either
the direct or indirect exposure to the disaster. Directly related
deaths are those caused by the physical forces of the disaster.
Indirectly related deaths are those caused by unsafe or unhealthy
conditions that occur because of the anticipation, or actual
occurrence, of the disaster. These conditions include the loss or
disruption of usual services, personal loss, and disruption of an
individual’s lifestyle.
Flow chart (Figure 1)
To reduce subjectivity and enhance consistent case selection, we
developed a flow chart that includes the general categories outlined in the definition. Using these categories, the flow chart
directs the user to sections of a classification matrix that provides
additional details about causes and circumstances of death.
Classification matrix (Figure 2, Appendix)
The classification matrix is used to identify the disaster and the
affected geographical location, circumstances leading to death,
cause and manner of death, and prevalence and case ascertainment periods.
Disaster identification and geographical location
Located in the upper left corner of the matrix, the name of the
disaster, if applicable, and the affected geographical area need to
1125
be clearly defined by those collecting data as well as by those
who report the results.
Circumstances of death
The circumstance (rows) is the initial activity or event leading
to death. This matrix contains three general types of circumstances, identical to those outlined in the definition and flow
chart. Circumstances that are directly related to the disaster
appear in Part I, and those that are indirectly related appear in
Parts II and III. We separated the circumstances related to
indirect effects because they have different policy implications.
Circumstances in Part II reflect disruption of organized community services, whereas circumstances in Part III reflect
disruption of a more personal nature.
Cause and manner of death a
Cause of death categories (columns) are grouped by manner of
death. Non-natural causes and manner are on the left side of
the matrix and natural causes and manner are on the right.
Placed in between these two general groups is an ‘Other’
column, providing a place for deaths from causes that cannot be
grouped elsewhere. Because one of the purposes of the matrix
is to provide information at a glance, deaths from non-natural
causes are placed toward the left side of the cells in this column
and natural deaths are placed toward the right side.
Time periods
Our case definition does not include recommended time
periods. Both the prevalence period during which deaths or
incidents leading to death occurred and the case ascertainment
period during which data were collected should be determined
by the investigator. These time periods should also be included
in any reports of final results.
Case summary codes
A disaster-attributed death is summarized by use of the manner
code followed by the individual case number. This number may
be the one assigned by the ME/C or it may be arbitrarily assigned
by the individual reviewing the cases. The summary code is
entered in the cell of the classification matrix that corresponds
to the circumstance and cause of death for that particular case.
For an occupationally related death, the code is underlined. If
death occurred in a volunteer providing relief efforts, parentheses are placed around the summary code.
Summarizing mortality information
In addition to a final summary, it may be useful to produce reports
specific to selected time periods or geographical locations. Regardless of who conducts surveillance, mortality numbers should never
be reported without a case definition, geographical reference, the
time periods during which the death and data collection occurred,
and the agency(s) responsible for conducting the surveillance.
Coding guide (Appendix)
To further improve coding consistency, we developed a guide
that provides more detailed information than is possible to
include in the matrix. Whereas the matrix contains a brief
description of circumstances and causes of death, the coding
guide offers additional detail for each circumstance. For
additional detail and clarification of cause of death categories,
coders should reference international classification standards.19
a Manner of death is determined by the ME/C. Some assign manner without
making judgements about intent, but most consider intent before making a
final decision.
1126
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
Figure 1 Mortality-related to natural disasters
Figure 2 Mortality attributed to natural disasters: Classification and coding matrix
DISASTER-ATTRIBUTED MORTALITY
1127
1128
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
Results of Application of Method
Acknowledgements
We (LEQ) reviewed investigation, autopsy, and toxicology
reports for the 322 decedents whose deaths were investigated
by the Dade County Medical Examiner (ME) Office from
12:01 a.m. on 22 August through midnight on 21 September
1992.a Using the definition and flow chart, we found 46 (14%)
deaths that were hurricane-related. We then classified these
deaths using the coding matrix (Figure 2).
Although the purpose of this paper is merely to introduce and
demonstrate the use of our method, we also conducted a preliminary assessment of the reliability of this method by comparing the results obtained from a second reader who used the
same method as the first. Our two readers showed 97% (313/
322) agreement on case status. Of the 42 disaster-attributed
deaths on which they agreed, the readers showed 83% (35/42)
agreement on case classification. For the seven cases for which
classification did not agree, disagreement was over cause of
death for three cases and circumstances for four.
For their review and comments, we thank the following CDC/
ATSDR staff: Brent Burkholder, MD, Scott Lillibridge, MD, Mike
McGeehin, PhD, Eric Noji, MD, Lee Sanderson, PhD,; and NOAA
staff Rainer Dombrowski, PhD, National Weather Service; and
Harold Brooks, PhD, Severe Storms Laboratory. For use of the
matrix to evaluate death records, we thank Deborah Hiatt, MPH.
References
1 de Boer J, Brismar B, Eldar R, Rutherford WH. The medical severity
index of disasters. J Emerg Med 1989;7:269–73.
2 de Ville de Goyet C, Lechat MF. Health aspects in natural disasters.
Trop Doctor 1976;6:152–57.
3 Lechat MF. Accident and disaster epidemiology. Public Health Rev
1993;21:243–53.
4 Noji EK. The nature of disaster: general characteristics and public
health effects. In: Noji EK (ed.). The Public Health Consequences of
Disasters. New York: Oxford University Press, 1997, pp.3–20.
5 Rutherford WH. The definition and classification of disasters. Injury
Discussion
Because disasters are complex events, a standard case definition
and classification system for disaster-attributed mortality must
be flexible enough to include all potentially preventable deaths
and detailed enough to exclude deaths that are not related to
the disaster. Our definition is similar to others in that it allows
for a liberal interpretation of what constitutes disaster-related
circumstances and the conditions or diseases that might arise
from these circumstances.20,21 The disadvantage of this flexibility is the wide range of interpretation. The inclusion of the
flow chart and classification matrix provides a framework for
identification, classification, and reporting of disaster-attributed
deaths, thus minimizing the problems associated with the flexibility of the definition itself.
Preliminary results suggest that our method enhances consistent case selection and classification, especially for assessing
indirectly attributed deaths. Four of the deaths presented to the
1994 NAME attendees were also evaluated by our two readers.
For these four deaths, our readers showed 100% agreement on
both case status and general classification. Without using a
classification method, the NAME attendees showed 52–100%
agreement on case status and 48–100% agreement on general
classification, with 100% agreement being only for directly
attributed deaths.
In addition to enhancing consistent case selection and classification, our matrix also displays relationships between specific
circumstances and health outcomes. This graphic depiction helps
identify not only clustering of events but sentinel events as well.
Because this information can be acquired at a glance, without
extensive analysis, responsible parties can develop or modify
prevention policies and public education to direct efforts toward
emerging problems or to prioritize policy implementation.
Disaster prevention policies are dependent upon our understanding of the relationship between humans and their
environment. We hope that our proposed method will provide
a framework for standardized case selection and classification
that will help us better understand this relationship.
a Hurricane Andrew made landfall in Dade County on the morning of
24 August 1992.
1983;15:10–12.
6 Centers for Disease Control and Prevention. Heat-related deaths—
United States, 1993. MMWR 1993;42:558–60.
7 Centers for Disease Control and Prevention. Surveillance of deaths
attributed to a nor’easter—December 1992. MMWR 1993;42:4–5.
8 Combs DL, Parish RG, McNabb SJN, Davis JH. Deaths related to
Hurricane Andrew in Florida and Louisiana, 1992. Int J Epidemiol
1996;25:537–44.
9 McNabb SJ, Kelso KY, Wilson SA, McFarland L, Farley TA. Hurricane
Andrew-related injuries and illnesses, Louisiana, 1992. Southern Med J
1995;88:615–18.
10 Philen RM, Combs DL, Miller L, Sanderson LM, Parrish RG, Ing R.
Hurricane Hugo-related deaths: South Carolina and Puerto Rico,
1989. Disasters 1992;16:53–59.
11 Smith SK, McCarty C. Demographic effects of natural disasters: a case
study of Hurricane Andrew. Demography 1996;33:265–75.
12 Kilbourne EM. Heat waves and hot environments. In: Noji EK (ed.).
The Public Health Consequences of Disasters. New York: Oxford University
Press, 1997, pp.262–63.
13 Lillibridge SR. Tornadoes. In: Noji EK (ed.). The Public Health Con-
sequences of Disasters. New York: Oxford University Press, 1997, p.242.
14 Malilay J. Floods. In: Noji EK (ed.). The Public Health Consequences of
Disasters. New York: Oxford University Press, 1997, pp.297–98.
15 Malilay J. Tropical Cyclones. In: Noji EK (ed.). The Public Health Conse-
quences of Disasters. New York: Oxford University Press, 1997, p.223.
16 Noji EK. The use of epidemiologic methods in disasters. In: Noji EK
(ed.). The Public Health Consequences of Disasters. New York: Oxford
University Press, 1997, pp.21–36.
17 Wetterhall SF, Noji EK.Surveillance and epidemiology. In: Noji EK
(ed.). The Public Health Consequences of Disasters. New York: Oxford
University Press, 1997, pp.58–59.
18 Lechat MF. Disaster epidemiology. Ann Soc Belge Med Trop 1976;56:
193–97.
19 World Health Organization. International Statistical Classification of
Diseases and Related Health Problems, 10th Revision. Geneva, Switzerland:
World Health Organization, 1992.
20 National Weather Service. Storm data and related reports. In: National
Weather Service Operations Manual. Silver Spring, MD, 1994;F42:1–12.
21 McMichael AJ, Haines A, Sloof R, Kovats S. Introduction. In: McMichael
AJ, Haines A, Sloof R, Kovats S (eds). Climate Change and Human Health.
Geneva, Switzerland: World Health Organization, 1996, pp.1–15.
DISASTER-ATTRIBUTED MORTALITY
Appendix: Coding Guide
I. Direct effects of disaster (forces originating
from the natural environment)a
A. Environmental forces originating from atmospheric
disturbance
1. Tropical cyclones (i.e. hurricane, typhoon) [X37]
2. Storm surge, tidal wave caused by tropical cyclone [X37]
3. Rapidly moving surface water (e.g. torrential rainfall [X38],
flash flood [X37] (excludes floods caused by collapse of dam or
man-made structure)
4. Floods from slowly rising waters [X38]
5. Blizzard (snowstorm), nor’easter, ice storm [X37]
6. Avalanche from snow storm [X39]
7. Fog [X39]
8. Precipitation, not otherwise specified [X39]
9. Tornados, twisters [X37]
10. Dust storm, sand storm [X39]
11. Wind, not otherwise specified [X39]
12. Extreme temperature (heat [X30], cold [X31])
13. Lightning [X33] and related fires [X00-X09]
B. Environmental forces originating from geophysical events
(e.g. ground movement and eruptions)
1. Earthquake [X34]
2. Volcanic eruption [X35]
3. Avalanche, mudslide, landslide [X36] (excludes avalanche
from snow storm)
4. Tsunami, tidal wave caused by earthquake [X39] (excludes
tidal wave caused by tropical cyclone)
5. Sinkhole [X39]
C. Environmental forces originating from other natural
occurrences
1. Drought [X39]
2. Fire, not otherwise specified (excludes fires caused by lightning) [X39]
3. Sunlight (excludes other forms of natural radiation exposure)
[X32]
4. Radiation exposure (e.g. radon, ultraviolet light) [X39]
5. Other, not otherwise specified [X39]
II. Indirect effects: loss or disruption of services
A. Loss or disruption of transportation-related services
(excludes evacuation)
1. Traffic signals and guards
2. Public transportation
3. Roadway access
4. Bridge, tunnel or other transportation-related structural
collapse
B. Loss or disruption of public utilities
1. Electricity, natural gas
2. Water (e.g. inaccessible water supply, contaminated water
supply)
a Adapted from ‘Exposure to forces of nature’ codes (in brackets) of the
World Health Organization’s International Classification of Diseases,
10th Revision.19 Because of the complexity of ICD coding schemes, we
have not provided ICD codes for indirect effects of a natural disaster.
1129
3. Waste disposal (e.g. exposure to sewage, exposure to toxic or
hazardous waste)
C. Exposure to industrial toxics, hazards
1. Untreated industrial effluent
2. Chemical, biological, and radioactive substances used in
industrial processes
D. Loss or disruption of usual access to medical or mental
health care
1. Treatment for pre-existing physical or emotional conditions
2. Emergency services
3. Prescription drugs
4. Social services (e.g. well-baby care, nutritional supplement
services)
E. Social disruption, anarchy
1. Crimes associated with loss of usual police protection
2. Crimes committed by opportunistic criminal
3. Crimes committed by prisoners providing clean-up in the
community
4. Crimes committed by formerly incarcerated prisoners freed
because of damaged structures
III. Indirect effects: personal loss, lifestyle disruption
A. Evacuation (excludes transportation-related services in
Part II)
1. Motor vehicle collisions while evacuating from either mandatory or non-mandatory evacuation areas)
2. Stress, anxiety during evacuation
B. Use of temporary sheltering, provisions
1. Physical aspects of shelter
2. Provisions such as food, water, and other essential supplies
3. Stress, anxiety associated with use of temporary shelter,
supplies
C. Preparation for disaster, clean-up following disaster
1. Disaster-proofing homes
2. Clearing debris
3. Rebuilding structures
4. Stress, anxiety associated with preparation or clean-up
D. Return to unsafe or unhealthy structure/environment
1. Unstable structures
2. Structures without window coverings, guard rails on porches
and stairs, roofs
3. Structures that enable criminal access
4. Structures, environments that enhance exposure to extreme
heat, cold, moisture/humidity, and other unhealthy environmental conditions.
E. Psycho-social stress or anxiety, including pre-existing stress
or anxiety (if there is evidence that the effects of the disaster
resulted in circumstances leading to acute stress or anxiety or
exacerbated chronic stress or anxiety, which ultimately led to
either a natural or violent death)
1. Associated with personal loss, hopelessness
2. Associated with disaster-related fears
3. Delayed stress, post-traumatic stress disorder
4. Generalized, non-specific disaster-related stress