EPIDEMIOLOGIC REVIEWS Vol.

10, 1988
Copyright © 1988 by The Johns Hopkins University School of Hygiene and Public Health Printed in U.S.A.
All rights reserved

PUBLIC HEALTH SURVEILLANCE IN THE UNITED STATES

STEPHEN B. THACKER1 AND RUTH L. BERKELMAN2

In 1963, Alexander D. Langmuir defined Epidemiologic surveillance is the ongoing sys-

tematic collection, analysis, and interpretation
disease surveillance as "the continued of health data essential to the planning, imple-
watchfulness over the distribution and mentation, and evaluation of public health prac-
trends of incidence through the systematic tice, closely integrated with the timely dissemi-

Downloaded from http://epirev.oxfordjournals.org/ at Pennsylvania State University on March 5, 2016

nation of these data to those who need to know.
collection, consolidation and evaluation of The final link in the surveillance chain is the
morbidity and mortality reports and other application of these data to prevention and con-
relevant data" and the regular dissemina- trol. A surveillance system includes a functional
capacity for data collection, analysis, and dis-
tion of data to "all who need to know" (1, semination linked to public health programs (3,
pp. 182-183). Langmuir was careful to dis- p. ii).
tinguish surveillance both from direct re-
sponsibility for control activities and from A critical word in this definition is "on-
epidemiologic research, although he recog- going"; one-time surveys or sporadic stud-
nized the important interplay among epi- ies do not constitute surveillance. An on-
demiologic studies, surveillance, and con- going system of data collection and colla-
trol activities. In 1968, the 21st World tion is also not sufficient to constitute
Health Assembly held technical discussions public health surveillance, because to be
on the National and Global Surveillance of useful the data must be integrated into the
Communicable Disease and identified these conduct and evaluation of specific public
main features of surveillance: 1) the sys- health programs, which may include epi-
tematic collection of pertinent data; 2) the demiologic research leading to prevention.
orderly consolidation and evaluation of The purpose of this review is to describe
these data; and 3) the prompt dissemina- the historical and current practice of public
tion of the results to those who need to health surveillance, to discuss new direc-
know, particularly those who are in a posi- tions for surveillance both in terms of new
tion to take action (2). public health priorities and new methodo-
Subsequently, the applications of sur- logical tools, and to assess the limitations
veillance concepts have broadened to in- of surveillance.
clude a wider range of health data—risk
HISTORICAL OVERVIEW
factors, disability, and health practices—as
well as disease. This is reflected in the 1986 Current concepts of public health sur-
Centers for Disease Control (CDC) defini- veillance have evolved from public health
tion of epidemiologic surveillance: activities developed to control and prevent
disease in the community. In the late Mid-
Abbreviations: AIDS, acquired immunodeficiency dle Ages, governments in Western Europe
syndrome; CDC, Centers for Disease Control; NCHS, assumed responsibility for both health pro-
National Center for Health Statistics; NIOSH, Na-
tional Institute for Occupational Safety and Health; tection and health care of the population of
WHO, World Health Organization. their towns and cities (4). A rudimentary
1
Center for Environmental Health and Injury Con- system of monitoring illness led to regula-
trol, Centers for Disease Control, Atlanta, GA 30333.
(Reprint requests to Dr. Stephen B. Thacker.) tions against polluting streets and public
2
Epidemiology Program Office, Centers for Disease water, instructions for burial and food
Control, Atlanta, GA. handling, and the provision of some types
The authors thank Drs. Philip S. Brachman, Mi-
chael B. Gregg, Alexander D. Langmuir, and R. Gibson of care. In the 17th century, John Graunt
Parrish for their contributions to the manuscript. used the Bills of Mortality to monitor dis-
164
 US PUBLIC HEALTH SURVEILLANCE 165
ease in London (5). In 1766, Johann Peter It was not until 1925, however, following
Frank advocated a more comprehensive markedly increased reporting associated
form of public health surveillance with his with the severe poliomyelitis epidemic in
system of police medicine in Germany, 1916 and the influenza pandemic of 1918-
which covered school health, injury preven- 1919, that all states were participating in
tion, maternal and child health, and public national morbidity reporting (11). After a
water and sewage (4). In addition, the gov- 1948 PHS study led to the revision of mor-
ernmental measures to protect the public bidity reporting procedures, the National
health were delineated (4). Office of Vital Statistics assumed the re-
William Farr (1807-1883) is recognized sponsibility for morbidity reporting. In
as the founder of the modern concepts of 1949, weekly morbidity statistics that had
surveillance (6). As Superintendent of the appeared for several years in Public Health

Downloaded from http://epirev.oxfordjournals.org/ at Pennsylvania State University on March 5, 2016

Statistical Department of the Registrar Reports were published by the National
General's Office of England and Wales Office of Vital Statistics. In 1952, mortality
from 1839 to 1879, Farr concentrated his data were added to what is now known as
efforts on collecting vital statistics, on as- the Morbidity and Mortality Weekly Report.
sembling and evaluating those data, and on Since 1961, this publication has been the
reporting them to both responsible health responsibility of CDC.
authorities and to the general public. The Malaria Eradication Program was
In the United States, public health sur- undertaken by CDC and state health de-
veillance has focused primarily on infec- partments in 1946 to address endemic ma-
tious diseases. Basic elements of surveil- laria in the United States at a time when
lance were found in Rhode Island in 1741 World War II veterans were returning from
when the colony passed an act requiring Africa and from the Mediterranean and
tavern keepers to report contagious disease Pacific theaters and introducing Plasmo-
among their patrons. Two years later, the dium vivax to the population (1). Spraying
colony passed a law requiring reporting of of dichlorodiphenyltrichloroethane (DDT)
smallpox, yellow fever, and cholera (7). had begun before surveillance was initiated.
National disease monitoring activities By 1947, it was clear that earlier reports of
did not begin until 1850 when mortality morbidity and mortality had been erro-
statistics based on the decennial census of neous. Mississippi, South Carolina, and
that year were first published by the federal Texas had the highest reported incidences
government for the entire United States of malaria, but because there was no diag-
(8). In 1878, Congress authorized the fore- nostic verification, the reported occurrence
runner of the Public Health Service (PHS) was exaggerated. A change in reporting re-
to collect morbidity reports for use with quirements that included case reports with
quarantine measures against pestilential diagnostic verification was illuminating. In
diseases such as cholera, smallpox, plague, Mississippi, for example, the reported in-
and yellow fever (9). In 1893, an act pro- cidence of provisional cases dropped from
vided for the collection of information each 17,764 to 914 in the first year, only a very
week from state and municipal authorities few of which could be confirmed. Such new
throughout the United States. By 1901, all criteria revealed that malaria had disap-
state and municipal laws required notifi- peared as an endemic disease from the
cation (i.e., reporting) of selected commu- South. The malaria experience was a major
nicable disease to local authorities such as factor emphasizing the necessity of a more
smallpox, tuberculosis, and cholera (10). In current and comprehensive system of sur-
1914, PHS personnel were appointed as veillance.
collaborating epidemiologists to serve in The critical demonstration in the United
state health departments to telegraph re- States of the importance of surveillance
ports weekly to the Public Health Service. was made following the Francis Field Trial
166 THACKER AND BERKELMAN

of poliomyelitis vaccine in 1955 (12, 13). ment measures that included isolating pa-
Within two weeks of the announcement of tients at home and rapidly vaccinating per-
the results of the Field Trial and initiation sons in surrounding houses. Key population
of a nationwide vaccination program, six contacts for surveillance included not only
cases of paralytic poliomyelitis were re- government officials and religious leaders
ported through the notifiable disease re- but also school children, tea shop owners,
porting system to state and local health people in markets, nomads, and refugees.
departments; case investigations revealed The Conference (now Council) of State
that these children had received vaccine and Territorial Epidemiologists was au-
produced by a single manufacturer. The thorized in 1951 by its parent body, The
Surgeon General requested the manufac- Association of State and Territorial Health
turer to recall all outstanding lots of vac-

Downloaded from http://epirev.oxfordjournals.org/ at Pennsylvania State University on March 5, 2016

Officials, to determine what diseases should
cine and directed that a national poliomy- be reported by states to the Public Health
elitis surveillance program be established Service and to develop reporting proce-
at CDC. Intensive surveillance and appro- dures. The Council currently meets an-
priate epidemiologic investigations by fed- nually and, in collaboration with CDC, rec-
eral, state, and local health departments ommends to its constituent members ap-
found 141 vaccine-associated cases of par- propriate changes in morbidity reporting
alytic disease, 80 of which were found in and surveillance, including what diseases
family contacts. Daily surveillance reports should be reported to CDC and published
were distributed by CDC to all persons in the Morbidity and Mortality Weekly Re-
involved in these investigations. This na- port.
tional common-source epidemic was ulti- Until 1950, the term surveillance was
mately related to a particular brand of vac- restricted in public health practice to
cine that had been contaminated with live watching contacts of serious communicable
virus. Had the surveillance program not diseases, such as smallpox, to detect early
been in existence, many and perhaps all symptoms so that prompt isolation could
vaccine manufacturers would have ceased be instituted (15). Langmuir has been cred-
production. ited with broadening the application of sur-
Surveillance was critical to the contain- veillance to populations (1), and in 1968,
ment strategy for global eradication of the 21st World Health Assembly focused
smallpox, and the success of the program on national and global surveillance of com-
demonstrated to the international commu- municable diseases, applying the term to
nity the practical value of surveillance (14). diseases rather than to the monitoring of
To facilitate early outbreak detection, sur- individuals with selected communicable
veillance teams actively investigated re- diseases (2). Over the intervening years, a
ported cases, sought nearby cases, and ini- wide variety of health events, such as child-
tiated rapid containment measures. Sur- hood lead poisoning, leukemia, congenital
veillance was intensified in areas in which malformations, abortions, injuries, and be-
cases were confirmed. When outbreaks de- havioral risk factors have been brought un-
creased, these teams continued to search der surveillance. In 1976, recognition of the
high-risk areas for cases until independent breadth of surveillance activities through-
assessment confirmed that transmission out the world was made evident by a special
had been interrupted. Routine reporting of issue of the International Journal of Epi-
cases and the work of the surveillance demiology devoted to papers specially com-
teams were further supplemented in some missioned to examine health surveillance
settings by one-week, village-level, inten- (16).
sive case identification. These surveillance In 1986, CDC, in collaboration with the
activities were clearly linked to contain- Council of State and Territorial Epide-
 US PUBLIC HEALTH SURVEILLANCE 167
miologists, published its first Comprehen- broadened applications. The use of the
sive Plan for Epidemiologic Surveillance term epidemiologic, however, also engen-
(3). In this document, CDC explicitly delin- dered both confusion and controversy. In
eated its policies and goals in surveillance, 1971, Langmuir noted that some epide-
specified plans to establish and evaluate miologists tend to equate surveillance with
surveillance systems, and described rele- epidemiology in its broadest sense, includ-
vant activities in research and training. ing epidemiologic investigations and re-
Since the term surveillance was first ap- search (15, p. 12). He found this "both
plied to a disease rather than to an individ- etymologically unsound and administra-
ual in 1950, it has assumed major signifi- tively unwise," favoring a definition of sur-
cance in disease control and prevention. Its veillance as "epidemiological intelligence."

Downloaded from http://epirev.oxfordjournals.org/ at Pennsylvania State University on March 5, 2016

specific connotations, however, have not Surveillance activities, however, have
been universally understood. In 1963, frequently led to epidemiologic investiga-
Langmuir clearly limited surveillance to tions of etiology. After the initiation of the
the collection, analysis, and dissemination National Influenza Immunization Program
of data (1). The term did not encompass in October 1976, cases of Guillain-Barre
direct responsibility for control activities. syndrome were reported to CDC through a
In 1965, the Director General of the World nationwide surveillance system established
Health Organization (WHO) established to monitor illnesses occurring after influ-
an Epidemiological Surveillance Unit in enza vaccination (19). Subsequent epide-
the Division of Communicable Diseases at miologic studies demonstrated a relation of
WHO (17). The Division Director, Karel Guillain-Barre syndrome to the swine in-
Raska, defined surveillance much more fluenza vaccine that was in use, which re-
broadly than Langmuir and included in it sulted in the cessation of the vaccination
"the epidemiological study of disease as a program for the year (20). To test whether
dynamic process." In the case of malaria, the syndrome could result from use of other
he saw epidemiologic surveillance as en- influenza vaccines, a special surveillance
compassing control and prevention activi- system was established in 1978 which used
ties. Indeed, the WHO definition of malaria 1,813 neurologists as reporters (21). The
surveillance included not only case detec- data collected for that and several subse-
tion but also taking of blood films, drug quent years showed no association between
treatment, epidemiologic investigation, and influenza vaccines and Guillain-Barre syn-
follow-up (18). drome.
The 1968 World Health Assembly dis- In addition, public health surveillance
cussions reflected the broadened concept of systems are often the source of cases for
epidemiologic surveillance and addressed case-control studies. For example, in re-
the application of the concept to public sponse to concerns expressed by Vietnam
health problems other than communicable veterans about the possibility of increased
diseases (2). In addition, epidemiologic sur- risk for fathering children born with birth
veillance was said to imply "the responsi- defects, CDC conducted a case-control
bility of following up to see that effective study using as cases with serious structural
action has been taken" (2, p. 9). birth defects infants identified by the Met-
The use of epidemiologic to describe sur- ropolitan Atlanta Congenital Defects Pro-
veillance first appeared in the mid-1960s gram (22). This surveillance system at-
and was associated with the establishment tempted to ascertain all infants with de-
of the WHO unit of that name. This was fects diagnosed during the first year of life
done both to distinguish this activity from born to mothers who resided in the Atlanta
other forms of surveillance, such as for area. Cases and controls were selected from
military intelligence, and to reflect its infants born alive in the Atlanta area dur-
168 THACKER AND BERKELMAN

ing the years 1968 through 1980. This study meaning of surveillance in the public health
found that Vietnam veterans did not have setting, having led in the past to the inap-
an increased risk of fathering children with propriate incorporation of research into the
defects. Other examples of epidemiologic definition of surveillance (18). For this rea-
research facilitated by case ascertainment son, in this paper, we will not adhere to the
through surveillance include the demon- current practice of using the term epide-
stration of the association of tampon use miologic to modify surveillance. We pro-
with the development of toxic shock syn- pose that a more appropriate term is public
drome (23), the relation between salicylate health surveillance, because its use retains
use and Reye's syndrome (24), the risk of the original benefits of the term epidemio-
breast cancer associated with long-term logic cited previously and removes some of

Downloaded from http://epirev.oxfordjournals.org/ at Pennsylvania State University on March 5, 2016

oral contraceptive use (25), and quantifi- the confusion surrounding current practice.
cation of the risk of acquired immunodefi- Surveillance is more correctly an element
ciency syndrome (AIDS) from certain sex- of public health, and persons encountering
ual practices (26). the term should understand this.
We believe that there are two issues to
be addressed in this discussion. First, what SURVEILLANCE PRACTICE
are the boundaries of surveillance practice?
Second, is epidemiologic an appropriate Data collection
modifier of surveillance as it is used in Surveillance data are collected from mul-
public health practice? To address these tiple sources. Physicians, laboratories, and
questions, we must first examine the struc- other health care providers are required to
ture of public health practice. One can di- report all cases of those diseases or health
vide public health activities into surveil- conditions specified by state law to be no-
lance, epidemiologic and laboratory re- tifiable (or reportable); most of these con-
search, service (including program ditions are of infectious origin. Typically, a
evaluation), and training. Surveillance data case report form is completed for each case
should be used to identify areas needing by the health care provider or laboratory
research and service which, in turn, help to and mailed to the local or state health
define training needs. Unless data are pro- department. In some states, the authority
vided to those who set policy and imple- to change the list of notifiable diseases is
ment programs, their use is limited to ar- granted to the state health authorities; in
chives and academic pursuits and are ap- other states, each change must be newly
propriately considered to be health legislated. Penalties for failure to report a
information rather than surveillance data. notifiable condition may include suspen-
Surveillance, however, does not encompass sion of a physician's license (27), but in
research or service. These are related but practice such penalties are rarely enforced.
independent public health activities and Physician reporting is influenced by disease
may be based on surveillance. Hence, the severity, availability of public health mea-
boundary of surveillance practice is drawn sures, public concern, ease of reporting, and
before the actual conduct of research and physician appreciation of public health
the implementation of delivery programs. practice in the community.
Given this context, the use of the term A disease traditionally is notifiable only
epidemiologic to modify surveillance is mis- when there is a clear link between a case
leading. Epidemiology is a broad discipline report and a public health action. For many
that incorporates research and training of these diseases, case investigations are
that is distinct from a public health process performed by the state or local health de-
that we call surveillance (table 1). Because partment. Individual names and other per-
of the much broader content of epidemiol- sonal identifiers are often required for pur-
ogy, the use of epidemiologic confuses the poses of contact identification or treat-
 US PUBLIC HEALTH SURVEILLANCE 169
TABLE 1
Distinctions between public health surveillance and epidemiologic research
Public health surveillance Epidemiologic research
Reason for initiating data Problem detection Hypothesis testing
collection Problem description Problem description
Identify cases for epidemiologic studies
May be legally required
Monitor geographic and temporal trends
in disease occurrence

Frequency of data collec- Ongoing Usually time-limited

tion

Method of data collection Established systems or procedures Special procedures tailored to hypotheses or

Downloaded from http://epirev.oxfordjournals.org/ at Pennsylvania State University on March 5, 2016

questions of interest
Many persons involved Fewer persons involved
Traditionally depends on voluntary par- Depends on paid, supervised employees
ticipation

Amount of data collected Can be considerable and usually detailed

per case Usually minimal

Completeness of data col- Usually complete

lected Often incomplete

Analysis of data Traditionally simple Can be complex

Primarily to detect change in incidence Hypothesis testing often requires statistical
methods
Usually historical comparison groups Concurrent controls

Dissemination of data Timely Not timely

Regular Sporadic
Review in public health agency External review
Targeted to public health and clinical au- Targeted to academic as well as public health
dience and clinical audience

Use of data Identifies a problem Describes a problem in detail

Triggers intervention Provides etiologic information
Suggests hypotheses Tests hypotheses, suggests additional hy-
potheses
Commonly used to evaluate programs Less often used to evaluate programs
Estimates magnitude of a problem

ment. In addition, collecting names aids in ments to establish specialized disease re-
identifying duplicate reports. Because of porting systems. Other federal agencies are
the need to identify individuals, however, involved in the collection of surveillance
concerns about confidentiality affect noti- data; for example, the Food and Drug Ad-
fiable disease reporting, and individual ministration (FDA) conducts postmarket-
identifiers are not usually collected at the ing surveillance of adverse reactions to
national level. These concerns have been drugs (29), and the Consumer Product
heightened by the epidemic of AIDS (28). Safety Commission conducts surveillance
The Council of State and Territorial Ep- of product-related injuries (30).
idemiologists determines which notifiable For many health events, national sur-
diseases should be reported from the state veillance systems rely on data collection
health department to CDC. In addition, the efforts by the National Center for Health
quarantinable diseases—yellow fever, chol- Statistics (NCHS) of CDC, including the
era, and plague—are reportable by inter- National Health Interview Survey (31), the
national regulation. To obtain information National Hospital Discharge Survey (32),
(case reports) on specific topics such as and the National Health and Nutrition Ex-
birth defects, influenza, low birth weight, amination Survey (33) (table 2). Although
and nosocomial infections, CDC has collab- such surveys do not constitute public health
orated with state and local health depart- surveillance systems, data obtained in these
 o

Downloaded from http://epirev.oxfordjournals.org/ at Pennsylvania State University on March 5, 2016

-3
TABLE 2
Selecteiyiational data sources that support public health surueillance, United States o
X
m
Title Scope Responsible organization Sources of data Dates w
>
Ambulatory Sentinel Practice National Ambulatory Sentinel Practice Family physicians 1981-present Z
Network for North America (unrepresentative) Network D
W
M
Boating Accident Reporting National Coast Guard Boat operators 1961-present 33
System
ts
f
Fatal Accident Reporting Sys- National Department of Transportation Police records, vital records, medical exam- 1975-present
tem iners, coroners, hospital records

Hazardous Materials Informa- National Department of Transportation

tion System Highway patrol 1971-present

McAuto National McDonnell-Douglas Corpora-

(unrepresentative) tion Hospital discharge abstracts 1982-present

National Accident Sampling National

System (unrepresentative) Department of Transportation Police, hospitals

National Ambulatory Medical National

Care Survey NCHS (CDC)* Office-based, medical practices 1973-1981,1986

National Burn Registry National National Institute of Burn

(unrepresentative) Medicine Burn centers 1964-present

National Disease and Thera- National IMS, Inc.

Office-based, medical practices 1960-present
peutic Index
 US PUBLIC HEALTH SURVEILLANCE 171
surveys can be used as part of surveillance
systems that are more clearly linked to
public health practice (table 3). Similarly,
hospital abstracting services, such as the
Commission of Professional and Hospital
£ S Activities (34), provide information on
a
more than 50 per cent of all acute-care
civilian hospital discharges in the United
States. In addition, more than half of the
states have enacted legislation placing hos-
pital discharge or claims data into the pub-

Downloaded from http://epirev.oxfordjournals.org/ at Pennsylvania State University on March 5, 2016

lic domain (35). Again, such data collection
activities do not constitute surveillance sys-
tems, but may provide useful data for sur-
£
veillance.
i
.2 a)
There are relatively few national data
sets in the area of ambulatory care, and
£ £
these are used only rarely for surveillance
o —• purposes (36-38). National data on diag-
11
> K
nosis and drug therapy from office-based
practices are available from the National
Ambulatory Medical Care Survey of NCHS
(36) and the commercially available Na-
tional Disease and Therapeutic Index (37).
2 5
2-2 tn •— National influenza surveillance efforts
have been complemented by a convenience
a- <
1
5 ^~- (0 -—.
sample of family practitioners that provides
S3
§ 6 u
Q o o
Q a a £ CDC with demographic data and culture
O
a is
o o Q
sS specimens for all cases of influenza-like
tice

CO CO CO

K c c illness seen in their offices during influenza

I5 O
z
o
z
CJ
z 1°
O
season (38). Emergency room data are
found in the National Electronic Injury
Surveillance System maintained by the
Consumer Product Safety Commission (39)
and the Drug Abuse Warning Network sup-
ported by the National Institute on Drug
I Abuse (40).
Registries are also useful sources of in-
u formation (41). Unlike national surveys
conducted by NCHS, registries are de-
signed to collect information on a specific
topic and are usually limited in scope. Like
the NCHS surveys, registries are not sur-
£ E 11 g veillance systems, but data from registries
can be used for public health surveillance.

z
III
S oco
a 3

CO
if
o 5
s c
CDC's Metropolitan Atlanta Congenital
Defects Program is an example of a registry
Z a. that has been developed into a system of
public health surveillance (42). The best
 Downloaded from http://epirev.oxfordjournals.org/ at Pennsylvania State University on March 5, 2016
TABLE 3
Selected national public health surveillance systems, United States
Title Scope Responsible organization Types of data Dates
Adverse drug reaction program National Food and Drug Administration Pharmaceutical manufacturers, 1969-present
physicians

Behavioral Risk Factor Survey National CDC* via state health departments Household telephone survey 1981-present
(unrepresentative)

Birth Defects Monitoring Pro- National Commission of Professional and Hospital Hospital discharge abstracts 1970-present
gram (unrepresentative) Activities

Disease- and condition-specific National CDC via state health departments Physicians, hospitals, laboratories Various, by con-
surveillance systemst dition (since
1954)

Fatal Accident Circumstances National National Institute for Occupational Safety Medical examiners 1980-present
and Epidemiology and Health (CDC)

National Electronic Injury Sur- National Emergency rooms 1972-present

veillance System Consumer Product Safety Commission

National Notifiable Disease National Physicians hospitals, laboratories 1920-present

Surveillance System CDC via state health departments

Nutrition (pediatrics and preg- National Public clinics 1975-present

nancy) (unrepresentative) CDC

Surveillance, Epidemiology, and National Cancer registry 1972-present

End Results (unrepresentative) National Cancer Institute

• CDC, Centers for Disease Control.

t This includes approximately 80 independent surveillance systems for specific diseases or conditions, mostly infectious.
 US PUBLIC HEALTH SURVEILLANCE 173
known and most widely used registries are to those who implement or influence public
those for cancer. There are population- health practice. Public health surveillance
based cancer registries in 38 states; 11 of data can be used to inform policymakers
these are part of the National Cancer In- and the public about the nature and extent
stitute's Surveillance, Epidemiology, and of health problems and to persuade these
End Results Program (43). audiences to address particular issues. In
this way, a health agency can develop a
Data collation and analysis constituency to support public health pro-
Data on infectious diseases are collated grams and to justify the expenditure of
and analyzed in local and state health de- public funds.
partments as well as at CDC. Descriptive More than half of all state health depart-

Downloaded from http://epirev.oxfordjournals.org/ at Pennsylvania State University on March 5, 2016

statistics, including sex, age, race, and ments and 40 per cent of county health
county of occurrence, have been the most departments publish surveillance data in a
useful for analyzing infectious disease data routine bulletin or newsletter for the local
with emphasis on total number of cases for medical and public health community (48).
a defined time period (e.g., weekly for no- State-specific notifiable disease data are
tifiable diseases). Additional analyses for presented weekly in tabular format in
trends over time and summary statistics on CDC's Morbidity and Mortality Weekly Re-
demographic information on cases may be port. Infectious disease data are also pub-
performed, and rates of disease may be lished annually in the CDC's Summary of
calculated. An exception to these limited Notifiable Diseases and in similar publica-
analyses has been the application of regres- tions by state health departments. In 1982,
sion and time series analyses to mortality CDC began publishing the CDC Surveil-
data for the surveillance of influenza (44- lance Summaries, which contains surveil-
46). lance reports on specific health events for
Public health surveillance of noninfec- which CDC has program responsibilities.
tious conditions emphasizes population- Surveillance data from other agencies may
based rates of disease. Linkage of data be published in special publications (e.g.,
the FDA Drug Bulletin). State and local
sources has facilitated calculation of rates
health department reports, federal publi-
and improved reporting (e.g., birth-weight-
cations like the Morbidity and Mortality
specific death rates linked by birth and
Weekly Report, and peer-reviewed public
death certificates) (47). Typically, health
health and clinical journals, however, are
department statistical staff calculate dis- the most common forms of disseminating
ease rates by sex, race, and age. In addition, surveillance data. The "Rainbow Series" of
trends over time (by year for most chronic NCHS publications reflects data collected
conditions) are determined. In the past, and analyzed from vital statistics and na-
only national and regional data have been tional surveys (49). Although NCHS data
available for estimates of morbidity related sets are not specifically linked to public
to many noninfectious conditions, and few health programs, they are frequently used
small-area comparisons have been made. to establish policy and monitor the effect
As increasingly large morbidity data sets of national intervention programs (50).
are being used (e.g., hospital discharge ab-
stracts), the number and variety of appli-
cations are increasing (35). Application to program
The concept of public health surveillance
Dissemination of data has evolved from primarily an archival
An important purpose of data analysis function prior to 1950 to one in which there
and dissemination is to provide easily is timely analysis of the data with an ap-
understood information in tabular or propriate response. Because surveillance is
graphic formats (in contrast to raw data) part of public health practice, it should be
174 THACKER AND BERKELMAN

used to guide control and/or prevention infectious diseases (54-57), although there
measures (or relevant research). No public have been some efforts to assess the appro-
health surveillance system is complete priateness of various data sources for the
without being linked to action. The uses of surveillance of other kinds of health prob-
surveillance include detecting new health lems (58-60).
problems (e.g., antibiotic-resistant strains A surveillance system is useful if it can
of bacteria), detecting epidemics, docu- be applied to a public health program to
menting the spread of disease, providing control and prevent adverse health events
quantitative estimates of the magnitude of or to better understand the process leading
morbidity and mortality, describing the to an adverse outcome. The simplest way
clinical course of disease, identifying poten- to assess usefulness is to ask those involved

Downloaded from http://epirev.oxfordjournals.org/ at Pennsylvania State University on March 5, 2016

tial factors involved in disease occurrence, in public health practice by means of inter-
facilitating epidemiologic and laboratory views or surveys (48, 61, 62). A more rig-
research, and assessing control and preven- orous approach to defining usefulness is
tion activities (51). through the assessment of the impact of
Surveillance has been vital to developing surveillance data on policies and interven-
hypotheses and stimulating epidemiologic tions (63), but there are no published stud-
research. Historically, acute infectious dis- ies of this kind. Decisions affecting public
ease problems have almost always been de- health surveillance programs are more
fined by epidemiologists in terms of their often based on changes in more general
temporal and geographic patterns. The program directions than on detailed analy-
need to define chronic as well as acute sis of a particular system (e.g., directing
diseases in terms of temporal and geo- resources away from routine contact trac-
graphic trends is being increasingly recog- ing for gonorrhea control to programs for
nized (52). preventing AIDS).
Public health surveillance efforts have The economic analysis of surveillance
often been intensified when the means for systems has received little systematic at-
primary prevention of most or all cases is tention apart from the accounting of direct
at hand (e.g., vaccine for measles or small- costs to health agencies. In a 1983 report
pox) or when the disease is severe and from Vermont, the authors reasoned that
newly emerging, with major efforts being costs were too high to justify active, health-
made to develop control and prevention department-initiated surveillance of se-
measures (e.g., toxic shock syndrome). Ad- lected acute infectious diseases unless un-
ditionally, public health surveillance has quantified subjective benefits, such as im-
served as the means for identifying persons proved relations with practicing physicians,
with a health problem who can participate were great (56). In a 1985 report from Ken-
in epidemiologic studies for developing pre- tucky, on the other hand, the benefits as-
vention strategies (53). Even before AIDS sociated with health-department-initiated
was documented to have a viral etiology, surveillance of hepatitis A were found to
for example, measures to lower a person's outweigh the costs (64).
risk of disease were suggested by studies of CDC has proposed a systematic method
cases detected through public health sur- to evaluate surveillance systems on the ba-
veillance (26). sis of usefulness and cost as well as seven
attributes of quality: sensitivity, specificity
Evaluation of surveillance programs and predictive value positive, representa-
Established surveillance systems require tiveness, timeliness, simplicity, flexibility,
regular review and modification based on and acceptability (51, 65). These attributes
explicit criteria of usefulness, cost, and of a surveillance system are interdepen-
quality (51). Most published evaluations of dent, and the improvement of one may
surveillance systems have been limited to improve or compromise another. Increasing
 US PUBLIC HEALTH SURVEILLANCE 175
the sensitivity of a system to detect a directly linked to public health prevention
greater proportion of a given health event programs. For example, since 1945,
in a population may also improve repre- population-based community studies on
sentativeness and usefulness of the system, the natural history of stroke have been
yet may lead to greater cost, lower specific- conducted on data collected from medical
ity, and more false positive events. This records and death certificates in Rochester,
method of evaluation is currently being Minnesota (67). Similar community studies
used to assess all surveillance systems at have been conducted on cardiovascular dis-
CDC at least once every three years. Such ease (68, 69). For cancer, the most success-
an approach for evaluating surveillance ful approach to community-based surveil-
systems should enable public health prac- lance has been the use of registries (43, 70),

Downloaded from http://epirev.oxfordjournals.org/ at Pennsylvania State University on March 5, 2016

titioners to efficiently assess their surveil- an approach that has also been used for
lance practices and thus improve the deliv- stroke (71) and hypertension (72).
ery of public health services. For various chronic conditions, efforts
have been made to obtain comprehensive
NEW PUBLIC HEALTH PRIORITIES data not only from medical records and
death certificates but also from special sur-
Chronic diseases veys (73, 74). Cancer has become a notifi-
Better data are essential for progress in able disease in at least 36 states in an effort
chronic disease prevention and control, to broaden the scope of data collection for
particularly incidence data to establish that condition (T. Aldrich, Oakridge Na-
priorities and to evaluate programs (66). tional Laboratories, personal communica-
These data should describe the burden and tion, 1988).
the determinants of disease and help to At the state and national levels, large
evaluate programs. data sets are available for application to
Three aspects of chronic diseases make the surveillance of chronic diseases (table
surveillance difficult. First, for some dis- 1). For example, national stroke mortality
eases (e.g., mesothelioma following asbes- (75) and cancer deaths (76) have been mon-
tos exposure), the latency between a precip- itored using death certificate data available
itating event or exposure and the eventual from the NCHS. An alternative approach
chronic disease not only hinders linkage to the use of such national data bases is the
between exposure and outcome but also pooling of information from state and local
complicates development and evaluation of sources to monitor national trends, as has
prevention programs. Second, the multifac- been done with nutrition data. Data on
torial etiology of many chronic conditions height and weight obtained from publicly
often prevents accurate linkage between supported health and nutrition programs
exposures, risk factors, or interventions demonstrated high prevalences of growth
and outcomes. Third, because the public stunting in Native American and Hispanic
health community is often interested in children (77). This finding, together with
arresting or reversing the progression of a high weight-for-height in these popula-
chronic condition, surveillance of various tions, suggests that the diet of these chil-
stages of disease is important. dren is adequate in quantity but inadequate
There has been extensive experience in in quality of nutrient intake. If confirmed,
data collection and analysis of chronic dis- such findings call for nutritional programs
ease occurrence. Indeed, at the community focused on high quality protein and in-
level, there have been many examples of creased essential vitamins rather than sim-
monitoring of heart disease, stroke, and ply increased calories. Although pooling of
cancer, although these programs are typi- state data is less efficient than conducting
cally not ongoing, are usually limited to national samples, the close linkage at the
data collection and analysis, and are rarely state level of data collection and analysis
176 THACKER AND BERKELMAN

to program intervention is an important hensive national data base on workplace

consideration. In addition, there have been hazards existed (86).
efforts to bring together national data from Major efforts are currently under way to
various sources for chronic health prob- perform surveillance of occupational dis-
lems, such as esophageal cancer (78) and eases both at the national and state levels.
alcohol abuse and alcoholism (79). Although past efforts have focused on data
The Surveillance, Epidemiology, and gathering and analysis, current efforts are
End Results Program of the National Can- motivated by attempts to collect data in a
cer Institute is an elaborate registry of in- way that will lead directly to intervention.
cident cancers in 11 geographic areas in the NIOSH first developed a list of the "Ten
United States which provides detailed Leading Work-related Diseases and Inju-

Downloaded from http://epirev.oxfordjournals.org/ at Pennsylvania State University on March 5, 2016

population-based information on mortality ries" and is now identifying those occupa-
due to malignant neoplasms (43). It is the tions and industries at high risk for adverse
principal source of national estimates of health events (87).
site-specific cancer incidence and trends, A survey of states conducted by the Iowa
documenting increases in cancer of the lung Department of Health in 1985-1986
and bronchus and declines in the incidence showed that at least 30 (60 per cent) states
of gastric cancer (80). This program, which had either voluntary or mandatory report-
costs about $5,800,000 annually, is partic- ing programs for selected occupational
ularly useful for national trend estimates health conditions (88). The states have not
and epidemiologic research. Its uses for lo- been uniform, however, concerning the
cal prevention and control activities, how- conditions for which they require reporting,
ever, have been limited. Less elaborate although lead poisoning, silicosis, and as-
state cancer registries may be more closely bestosis are frequently included on the re-
linked to state cancer control efforts (70). portable disease list. Also, the reporting
None of these data collection activities criteria for these occupational health
constitute public health surveillance sys- events are not uniform across states (e.g.,
tems. The usefulness of existing data sets Texas requires reporting of blood lead lev-
for chronic disease surveillance has not els >40 mg/ml, whereas New York requires
been proven. Such data may, nonetheless, reporting of all blood lead levels >25 mg/
prove useful for public health and are es- ml) (P. Honchar, CDC, personal commu-
sential to assess the completeness and ac- nication, 1987).
curacy of existing chronic disease data and Although many state health departments
their appropriateness for this purpose. To have reporting laws, few have maintained
date, there has been a limited effort to a professional staff that could respond to
apply the principles of public health sur- the incoming reports. Fortunately, this gap
veillance to specific chronic conditions (81) in surveillance is being addressed. For ex-
or to assess alternative approaches to col- ample, the Texas Department of Health
lecting chronic disease data for public performs case investigations routinely in
health surveillance (82-84). response to reported cases of occupation-
ally acquired lead poisoning (P. Honchar,
Occupational safety and health CDC, personal communication, 1987). Case
In 1984, J. Donald Millar, the Director investigation includes 1) assuring proper
of the National Institute for Occupational clinical management of the affected person
Safety and Health (NIOSH), told Congress and 2) offering an evaluation of the work-
that federal surveillance of occupational site to detect factors potentially responsible
illness was "70 years behind that of com- for the case. This evaluation is accom-
municable disease surveillance" (85, p. 11). panied by recommendations for preventing
Before the enactment of the 1970 Occupa- further cases. Screening for elevated blood
tional Safety and Health Act, no compre- lead levels in coworkers may be conducted.
 US PUBLIC HEALTH SURVEILLANCE 177
As another example, NIOSH is conducting spills and had inadequate data on injury,
the Fatal Accident Circumstances and Ep- death, and cost (91). The state of California
idemiology Project, which focuses upon se- compared data from the Hazardous Mate-
lected electrical-related and confined rials Information System with similar in-
space-related fatalities (89). The purpose formation collected by the California High-
of the program is to identify factors influ- way Patrol related to hazardous material
encing the risk of fatal injuries in the work- spills to determine number and nature of
place. incidents (59). Of 941 incidents involving
In addition to data gained from case re- highway transport of hazardous materials
ports, 30 states now include occupational and related exposures and injuries, only 18
information on death certificates; only 18 were reported in both systems. Despite such

Downloaded from http://epirev.oxfordjournals.org/ at Pennsylvania State University on March 5, 2016

states collected such information in 1981 limitations, the Hazardous Materials Infor-
(90). Fourteen health departments include mation System could be integrated into a
parents' occupations on the birth certifi- system of public health surveillance be-
cate. Reporting to health departments will cause it offers useful data on place, cause,
be expanded through the Sentinel Event and mode of spill.
Notification System for Occupational The most extensive public health sur-
Risks, a NIOSH-sponsored health event veillance system developed for outcomes
reporting system based on reporting se- related to an environmental hazard evolved
lected occupational disease and injury out- from 62 childhood lead-poisoning preven-
comes amenable to control and prevention tion programs (92). Over a 10-year period
(E. Baker, CDC, personal communication, ending in 1981, 247,000 children with lead
1987). Other data used by state health de- toxicity were identified among nearly 4 mil-
partments for surveillance include hospital lion screened. The data were disseminated
discharge data, workmen's compensation at both the local and national levels and
data, and cancer registries (11 states report were applied to program planning and im-
occupational history on all cancer cases). plementation. This routine reporting sys-
At the national level, questions on health tem was complemented with data from the
outcomes and health risks of relevance to Second National Health Assessment and
occupational health have been incorporated Nutrition Examination Survey (92). These
into the National Health and Nutrition data sources documented the decrease in
Examination Survey and the National blood lead levels associated with the reduc-
Health Interview Survey.
tion of lead used in gasoline. When federal
funding was discontinued in 1981, the na-
Health effects of environmental toxic tional program stopped, and most local ac-
exposures tivities were curtailed or eliminated.
Public health surveillance in environ- More typically, public health surveillance
mental health includes both hazard (expo- of specific environmental health outcomes
sure) and health effects monitoring. An is established, often in the form of regis-
example of an ongoing national system for tries, and then attempts are made to relate
collecting data on potential exposures is the these outcomes to particular exposures or
Hazardous Materials Information System etiologies. Important examples are the sys-
of the Department of Transportation, tems established for the surveillance of con-
which was established in 1971 by a federal genital malformations. Widespread interest
law that seeks voluntary reporting of spills in birth defects followed the epidemic of
occurring during interstate commerce (91). limb reduction deformities which was as-
Comparisons of these reports with inde- sociated with women taking thalidomide
pendently collected data from the state of during early pregnancy. This event, coupled
Washington, however, indicated that the with epidemiologic patterns for several
federal system missed over 80 per cent of malformations indicative of an environ-
178 THACKER AND BERKELMAN

mental etiology, led to the establishment of Public health surveillance in a disaster

the Metropolitan Atlanta Congenital De- setting is critical to the optimal allocation
fects Program and the nationwide Birth of scarce and often poorly organized re-
Defects Monitoring Program in 1967 and sources. Surveillance systems were estab-
1974, respectively (42). These two systems lished, for example, to monitor exposure to
are used to monitor trends in specific birth radiation following the incident at the nu-
defects or combinations of defects and to clear reactor at Three Mile Island (96).
stimulate epidemiologic investigations Surveillance systems were also developed
when increases are identified. The data to monitor the health effects of the volcanic
have been used to demonstrate the lack of ash plume created by the eruption of the
teratogenicity of exposures of serious public Mount St. Helens volcano in 1980 (97) and

Downloaded from http://epirev.oxfordjournals.org/ at Pennsylvania State University on March 5, 2016

concern such as spray adhesives (93), vinyl the health effects of exposure to toxic waste
chloride (34), airport noise (94), and mili- at Love Canal (98). Similar short-term, lo-
tary service in Vietnam (22). Similarly, cal environmental monitoring systems have
cancer registries have been used to study been established in response to chemical
possible relations between specific cancers spills (99). Although these are examples of
and environmental exposures (43). ad hoc surveillance established in an acute
The first challenge in the public health situation, there are few examples of ongo-
surveillance of environmental hazards is to ing systems of public health surveillance
determine which hazards warrant ongoing linked to public health programs of control
programs of surveillance. The major con- and prevention. Occasionally, emergency
straint of the outcome approach is the lim- preparedness plans include surveillance,
ited knowledge of the health effects of spe- such as during the 1984 Olympics when the
cific toxins (i.e., natural) and toxicants (i.e., potential for terrorist activities was consid-
man-made), which inhibits our ability to ered high (100). Currently, CDC is working
detect unexpected associations between with the American Red Cross to organize
disease and exposure. Humans have re- disaster surveillance and to establish an
leased thousands of toxins and toxicants international activity in this area (P. Du-
into the environment, but both the health clos, CDC, personal communication, 1987).
impact and the exposure potential of most Finally, there have been efforts to com-
of these substances are unknown or, at best, bine environmental monitoring data with
established only in laboratory animals. The health outcome information. After the se-
Agency for Toxic Substances and Disease vere heat wave of 1980, for example, CDC,
Registry has been given the responsibility in collaboration with medical examiners,
of ranking the leading priority chemicals in state and local health departments, and the
terms of risk to human health (95). Yet, National Weather Service, developed a sys-
even when this task has been accomplished, tem of surveillance of mortality related to
policies for establishing systems of public summer heat waves (101).
health surveillance will need to be formu-
lated by local, state, and federal agencies. Injuries
Once priorities are established, data must The recognition of both intentional (e.g.,
be identified for both exposures and out- homicide) and unintentional (e.g., falls) in-
comes. Fortunately, many data sets (e.g., juries as major public health problems has
the Birth Defects Monitoring Program) al- led to the need for developing systems of
ready exist for both and need only to be public health surveillance (102-104). Be-
integrated into public health programs (42). cause of the acute nature of injury events,
It is simpler and less costly to use existing surveillance principles learned from expe-
data and data systems than to establish rience with acute infectious diseases are
new ones. Additionally, historical data en- often readily adaptable to injuries (105).
able one to analyze long-term trends. The current approach to establishing public
 US PUBLIC HEALTH SURVEILLANCE 179
health data bases for injury has been to Fatal Accident Reporting System (111) and
adapt data, such as medical examiner re- the National Accident Sampling System
ports and vital statistics, to public health (112) maintained by the National Highway
needs (103, 105). This approach has been Traffic Safety Administration, the Na-
used most widely at the state level where tional Burn Registry initiated by the Na-
vital statistics, hospital discharge data, tional Institute of Burn Medicine (113), the
emergency room data, and household sur- National Fire Incident Reporting System
veys have been used to measure the extent established by the Federal Emergency
and nature of the unintentional injury Management Agency (114), the US Coast
problem in particular populations, as well Guard investigations of boating incidents
as to assess the impact of prevention pro- (115), and the National Spinal Cord Injury

Downloaded from http://epirev.oxfordjournals.org/ at Pennsylvania State University on March 5, 2016

grams (102, 106-108). Medical examiner Network (116). As with chronic diseases,
data have also been used in the surveillance the usefulness of many of these data
of injuries and associated risk factors such sources for public health surveillance re-
as alcohol and drug use (109). mains to be assessed.
Approaches to injury surveillance vary at The challenge of surveillance of inten-
the state and local levels. During one year, tional injuries is even more complex. Data
the Statewide Childhood Injury Prevention are available from vital records and medical
Program in Massachusetts detected 5,953 examiners, but information on the circum-
fatal and nonfatal injuries in 87,022 chil- stances of homicide and suicide is often
dren and adolescents through a public absent or limited in these data sets. Public
health surveillance system based on hospi- health surveillance of intentional injuries
tal and emergency room records from 23 will require the collaboration of the public
hospitals in 14 communities (102). Using health community with a new array of ex-
these data, program personnel focused pre- perts, especially in the fields of law enforce-
vention resources on the injury problems ment and sociology (113). At the state and
of highest incidence in particular commu- local levels, data from criminal justice agen-
nities. In 1987, the Council of State and cies, medical examiners and coroners, and
Territorial Epidemiologists adopted a res- medical and social service agencies are
olution to recommend that spinal cord in- being explored for use in the surveillance
jury be made reportable in all states (168). of intentional injury (117). Illinois insti-
North Dakota has already made notifiable tuted mandatory uniform crime reporting
all injuries resulting in at least one day of in 1972; the state maintains the data on
disability (J. Pearson, North Dakota State computer and publishes a report each year
Department of Health, presented at the (118). Few data exist on morbidity related
annual Council of State and Territorial Ep- to assault or child abuse, and only rarely
idemiologists meeting, May 1987). Trauma have epidemiologic studies been conducted
registries can also be adapted for surveil- in this area (119). Efforts are under way to
lance (110). assess the feasibility of alternative ap-
Several national data sets are available proaches to the surveillance of domestic
for the surveillance of unintentional inju- violence (120).
ries (table 2). NCHS compiles and analyzes On the basis of data from the national
mortality statistics, hospital discharge mortality files of NCHS and population
data, office-based physician utilization estimates of the US Bureau of the Census,
data, and data collected in an ongoing a 40 per cent increase in youth suicide was
health interview survey of the general pop- documented in the decade ending in 1980
ulation. Other sources for national data (121). This increase was found primarily in
include the National Electronic Injury Sur- white males 15-24 years of age. These sur-
veillance System maintained by the Con- veillance data documented the dramatic
sumer Product Safety Commission (39), the change of suicide as a problem of the elderly
180 THACKER AND BERKELMAN

to a problem of the young. Current efforts (127). The sample for each survey, con-
in suicide surveillance have demonstrated ducted by telephone, is selected generally
the importance and difficulty of arriving at with a multistage cluster design based on
uniform definitions, a problem complicated the Waksberg method (128).
by the interdisciplinary nature of this en- Interview surveys can obtain personal,
deavor. Uniform definitions of child and health-related information with only minor
spouse abuse, problems for which incidence differences in the prevalence of various
data are sparse, are also needed. Yet, as health conditions when conducted by tele-
such data bases are developed, surveillance phone or in person (127). Telephone inter-
will play a crucial role in public health views have the advantages of lower cost
programs aimed at controlling and pre- (about one-third to one-half the cost of
venting these and other injuries. Other na- personal interviews) and the ease of super-

Downloaded from http://epirev.oxfordjournals.org/ at Pennsylvania State University on March 5, 2016

tional data sources, such as the Uniform vising interviewers. Although there are
Crime Reports of the Federal Bureau of problems of bias related to omitting those
Investigation and the annual National households without telephones, telephone
Crime Survey of the US Department of coverage exceeds 93 per cent in the United
Justice, have proven to be useful (121-125). States (127).
Results of the surveys are published by
Personal health practices both CDC and state health departments
At a national level, the Health Interview (129, 130). They are also distributed to the
Survey conducted by NCHS has provided press and to a variety of local and state
the most information on personal health organizations, including voluntary health
practices such as alcohol use and smoking. agencies, hospitals, health maintenance or-
The prevention supplements to the 1982 ganizations, and state legislators. In 1986,
and 1985 surveys have provided more de- the 43 states that had conducted these sur-
tailed information in this area (126). As the veys reported that these data were fre-
role of personal health behavior in the de- quently used by the health department to
velopment of chronic diseases and injuries prepare state planning documents and to
has become more fully recognized, state- establish state-level health objectives (62).
based programs to reduce the prevalences Sixty-five per cent of these states reported
of unhealthy behaviors have been estab- using the data to support legislative initia-
lished. In turn, interest in providing a sys- tives, especially seat belt and anti-smoking
tematic means of collecting population- legislation (62). Limitations, however, exist
based prevalence data on a state-specific when these data are used; many states cited
basis resulted in the initiation of the Be- a circumscribed authority to disseminate
havioral Risk Factor Surveys in 1981 (127). the findings. In addition, because the sur-
National estimates can be obtained more veys only recently have been initiated at
efficiently, but local programs benefit from the state level, not enough time has elapsed
involvement in data collection as well as to adequately analyze trends.
from the ability to adapt the collection
process to their particular needs. As of Preventive health technologies
1987, 35 state health departments are con- Health technology includes the drugs, de-
ducting ongoing surveys of behavioral risk vices, and medical and surgical procedures
factors in persons aged 18 years or older. used in health care, and the organizational
Each state uses a standardized question- and supportive systems within which such
naire to determine the prevalence of a va- care is provided (131). The implementation
riety of personal health practices including of new technologies is a prominent growth
cigarette smoking, smokeless tobacco use, industry in health. Dramatic examples,
alcohol consumption, exercise, seat belt such as carotid endarterectomy, artificial
use, dieting, and hypertension control hearts, osteoporosis screening, and AIDS
 US PUBLIC HEALTH SURVEILLANCE 181
testing, are very much in the public eye. undergoing mammography are those most
Concerns regarding premature diffusion or likely to benefit from screening.
misapplication of health technologies have The need for surveillance of technology
highlighted the need for routine surveil- use is evident, but the process of gathering
lance of the application of the technologies, the primary data is not established cur-
particularly as these new technologies are rently for most technologies other than
used in healthy or asymptomatic popula- drugs—the latter being a responsibility of
tions to prevent disease (132). Currently, the Food and Drug Administration. As il-
efforts are under way in several state health lustrated by the surveillance of tubal steri-
departments to assess the effectiveness of lization, some hospital data sets can be
both cervical cancer and breast cancer helpful in tracking inpatient procedures.

Downloaded from http://epirev.oxfordjournals.org/ at Pennsylvania State University on March 5, 2016

screening programs. Systems of public There is a lack of state and national sur-
health surveillance are an integral part of veillance information, however, to track
these assessments. diffusion of technologies in the outpatient
There are, however, few examples of sur- setting, where complex and expensive tech-
veillance of health technologies despite this nologies are being used increasingly (135).
widespread diffusion of new devices and Although surveillance is usually under-
practices. Immunization against selected taken by public health agencies at the local,
infectious diseases is probably the most state, and federal levels in collaboration
effective and well-known technology used with the medical community, efforts to es-
in public health. More recently, public tablish surveillance systems at all levels
health surveillance of selected medical have faltered in recent years. In its lead
technologies has been developed by CDC in federal role in health care technology, the
response to concerns in the public health National Center for Health Services Re-
community. For example, in response to search and Health Care Technology As-
state health officials during a perceived cri- sessment should be encouraged to develop
sis in 1982 concerning the use of insulin priority-setting criteria for bringing tech-
pumps, CDC established a short-term sur- nologies under surveillance and subse-
veillance system to determine the fre- quently for analyzing the impact of tech-
quency and severity of complications asso- nologies in terms of their effect on morbid-
ciated with these devices (133). Using phy- ity, mortality, disability, and cost.
sician reporting, the investigators iden-
tified previously unrecognized adverse NEW TOOLS FOR PUBLIC HEALTH
events associated with pump use as well as SURVEILLANCE
35 deaths among pump users. The data
were used to assist the American Diabetes Computers
Association in developing a policy state- The introduction of computer hardware
ment for clinicians that included new cri- and software has provided public health
teria for initiating pump use (134). professionals with the capability to perform
Public health surveillance of technology surveillance more efficiently on common
use provides a mechanism for monitoring conditions. Large data bases may be better
the use of a practice or device and, together managed and analyzed, and in some in-
with data on morbidity and mortality, pro- stances may be linked. In addition, the
vides an ongoing measure of its effective- microcomputer has empowered the public
ness and safety in the populations being health professional with an increased abil-
monitored. Surveillance will also indicate ity to organize, communicate, tabulate, and
whether an effective technology is being analyze data. Use of the computer has in-
applied to the population that is likely to creased the timeliness of both data collec-
benefit from such technology. It is not tion and analysis and has decreased the
known, for example, whether the women epidemiologist's reliance on programmers
182 THACKER AND BERKELMAN

and biostatisticians for data analysis and tional sample of hospitals currently aid in
interpretation. collecting information on nosocomial infec-
The Public Health Foundation initiated tions (140).
an electronic mail system in 1983. Several
federal health agencies, including CDC, and Statistical methods
44 state health departments are now on- The increased sophistication of statisti-
line. In addition, three states have enrolled cal methods, the availability of computers,
their local health departments and can and the development of statistical software
telecommunicate with them. In 1984, this for analysis have broadened the potential
network was used in six states to pilot-test of statistical analysis in day-to-day public
the transfer of notifiable infectious disease health practice and have led to the inves-

Downloaded from http://epirev.oxfordjournals.org/ at Pennsylvania State University on March 5, 2016

data weekly to CDC (136). By early 1988, tigation of new methods of data analysis.
37 reporting areas were transferring indi- The usefulness of time series analysis (45),
vidual case data on over 40 notifiable dis- of detecting clusters of adverse health
eases to CDC each week. The ability to events in time and place (141-144), and of
transfer binary file will allow the telecom- mathematical models to forecast epidemics
munication of graphics, which facilitates based on surveillance data (145) remains to
review of aggregate data. Surveillance at be fully assessed.
the state level has been hampered by a lack Although detecting temporal and spatial
of microcomputer software for managing clusters of disease has always been a goal
and analyzing large numbers of disease rec- of public health surveillance, formal statis-
ords. Currently, software developed for use tical testing for clusters has rarely been
in epidemic investigations has been applied to routinely collected surveillance
adapted for surveillance and used in 20 data. The statistical problems associated
states (A. Dean, CDC, personal communi- with determining whether an "outbreak"
cation, 1987). Use of such software in Geor- has occurred were addressed in depth in the
gia has enabled early detection of an epi- 1960s, and a variety of alternative analyses
demic of illness due to Salmonella havana, were proposed. Two commonly used meth-
facilitating efforts to identify the environ- ods for space-time clustering, proposed by
mental source of the organism (137). Knox and Lancashire (143) and by Ederer
Programs at CDC for vaccine- et al. (146), are based on the number of
preventable diseases, tuberculosis, AIDS, "close" pairs of cases and the sum, over all
and diabetes have also developed computer space divisions, of the maximum number
networks with state health departments to of cases in any time unit within a space
enhance their surveillance capabilities. In division. For example, Ederer et al. em-
addition, state health departments have in- ployed a summary statistic to detect both
itiated computer linkage with selected local clusters of leukemia over time and out-
health departments for disease reporting breaks of polio and hepatitis.
(138). In Wisconsin, for example, case data The SCAN statistic, based on such sum-
from sexually transmitted diseases clinics mary statistics as those used by Mantel
are telecommunicated to each other and to (144) and Ederer et al. (146), was proposed
local and state health departments, improv- by Naus (147) and has recently been ap-
ing the efficiency of follow-up of patients plied to a cluster of trisomies in three New
(A. Dean, CDC, personal communication, York City hospitals (148). This statistic is
1987). computed by plotting points over time, tak-
Use of microcomputers has also ex- ing a "moving window" of a fixed length of
panded surveillance activities to nontradi- time, and then finding the maximum num-
tional reporting sources. Computers in ber of observations revealed through the
medical examiners' offices will aid in injury window as it scans or slides over the entire
surveillance (139); microcomputers in a na- time period. The statistic is based on the
 US PUBLIC HEALTH SURVEILLANCE 183
assumption that the size of the population (152) has clearly demonstrated the impor-
at risk remains fairly constant and that the tant role graphs can play in visual decoding
condition shows no seasonal or cylical pat- of large quantities of data. Although Tu-
tern over the time period plotted. Other key's methods have not yet been widely
analytic methods have been suggested for applied to surveillance data, his pioneering
using environmental data to predict the work together with the introduction of
occurrence of Rocky Mountain spotted fe- computer graphics has laid the foundation
ver, but such methods have not been used for graphic analysis of surveillance data
routinely (142). (153). Microcomputer graphics, in particu-
The Chandra Sekar-Deming method de- lar, have also made the results of data
veloped by demographers has been used to analysis far more useful to private and pub-

Downloaded from http://epirev.oxfordjournals.org/ at Pennsylvania State University on March 5, 2016

estimate completeness of reporting by com- lic policymakers in their planning and man-
parison of two independent surveillance agement of health care resources (35). Al-
systems with individual identifiers so that though simple data still are incorporated
the data may be linked (149). This method best into textual material or a tabular for-
has recently been applied to AIDS data mat, a graphic display can give the reader
reported through the notifiable disease sys- an understanding of large and complex data
tem and through death certificate registra- sets that cannot be conveyed easily in other
tion (150). It has also been applied to esti- ways (154).
mate the sensitivity of two systems for de- The interest in computer mapping in
tecting vaccine-preventable diseases (151). public health is strong. A 1976 workshop
Surveillance systems are subject to both sponsored by NCHS featured several ap-
selection and information biases. Notifi- plications of automated cartography to ep-
able disease reports, for example, are likely idemiology (155). In the area of surveil-
to come from a nonrepresentative sample lance, mapping of disease rates by county,
of practicing physicians who may report sex, age, and race based on large comput-
specific diseases because of personal inter- erized data sets first proved its usefulness
est. Private practitioners, for example, may when the cancer atlases were developed by
be less likely than physicians at public the National Cancer Institute in the 1970s
health clinics to report certain conditions (76). The Environmental Protection
(e.g., sexually transmitted diseases). At the Agency has also produced maps on cancer
same time, certain kinds of data are less (156). Injury maps have been used to con-
likely to be reported than others because of vey visually the race- and sex-specific dif-
ease of ascertainment (e.g., age or sex vs. ferences in rates of various injuries (157).
pathologic diagnoses). Analytic models are Although it has been common practice to
required to measure the impact of bias on plot individual cases or rates of disease on
surveillance data. Other important re- geopolitical maps, population-based maps
search issues on the statistics of surveil- have been produced to account for popula-
lance include the development of methods tion size. More recently, exploded popula-
to handle incomplete or missing data, the tion maps have been considered for use in
use of multiple subset sampling, modeling surveillance. These maps are developed by
of timeliness, and the combination of data the isomorphic reduction of geographic en-
from independently collected data sets. tities in relation to the entity with the
greatest population density, with or with-
out an overlap of the geopolitical map (158).
Graphic methods for data analysis and Other mapping of surveillance data for pro-
display grammatic use has included the develop-
Graphics have the potential to serve as ment of probabilistic contouring, with maps
powerful tools for displaying data both for demonstrating the estimated probability of
analysis and for communication. Tukey a health event or an exposure, a technique
184 THACKER AND BERKELMAN

that has proven particularly useful in pro- voting even a chapter to the subject (161,
gram planning (159). 162). The only substantive training for sur-
veillance in the United States is as part of
LIMITATIONS IN THE PRACTICE OF the actual practice of public health. The
SURVEILLANCE public health community is only now begin-
The variety of uses of public health sur- ning to approach surveillance in a more
veillance is not widely appreciated. For scientific manner, looking beyond case
some, the concept of surveillance is limited counting and simple descriptive epidemiol-
to reporting notifiable communicable dis- ogy. Sophisticated statistical tools such as
eases to state and local health departments. time series analysis (45) and the SCAN
Others think in terms of laboratory- or statistic (148), for example, have been ap-

Downloaded from http://epirev.oxfordjournals.org/ at Pennsylvania State University on March 5, 2016

hospital-based surveillance, particularly for plied successfully to surveillance data. Ex-
nosocomial infections. Another interpreta- pansion of public health into new fields
tion is seen in recent legislation establish- such as chronic disease demands more rig-
ing the Agency for Toxic Substances and orous scientific scrutiny of surveillance
Disease Registry, which limited health sur- methods as well as different approaches to
veillance to medical screening of individ- public health epidemiology (52). To date,
uals (160). however, the communities of both health
Other perspectives limit the potential care providers and teachers of medicine,
scope of public health surveillance. The nursing, and public health remain unin-
most common is that surveillance is limited formed about needs in public health sur-
to data collection and collation. It is impor- veillance. Their involvement, in the future,
tant for a system of public health surveil- could contribute significantly to the prac-
lance to include analysis and interpretation tice and development of public health sur-
of data, as well as dissemination of those veillance.
data to the relevant persons. Finally, to be
complete, a public health surveillance sys- Data gaps
tem requires linkage to programs. When Even in communicable disease reporting,
this broad perspective is not understood, data are often incomplete, unrepresenta-
the practice of surveillance and of epide- tive, and untimely. Depending on the se-
miology in public health is constrained verity and perceived importance of a dis-
short of its potential. ease, rates of reporting notifiable diseases
have been estimated to vary from 6 per cent
Inexperience with surveillance methods to 90 per cent (54, 163-165). Both measles
Except in state and local health depart- and AIDS programs, for which many re-
ments, relatively few persons have been sources have been targeted toward surveil-
involved in a complete program of public lance efforts, attain greater than 90 per cent
health surveillance. Most persons are in- sensitivity (150, 166). In a study of Shigella
volved with only one portion of a surveil- surveillance in Washington, DC, however,
lance system (e.g., data collection) or with investigators found that persons with dis-
only a limited array of health events (e.g., ease were more likely to be reported if they
communicable diseases). The lack of famil- were treated by private physicians—a prac-
iarity with public health surveillance is tice that leads to unrepresentative surveil-
even more pronounced in medical schools lance data (55). Efforts to improve the qual-
and schools of public health, where it is ity of reporting have been shown to have
rarely discussed and is almost never the some effect—improving sensitivity at the
subject of careful analysis. Textbooks of local level as much as ninefold for selected
epidemiology and public health are simi- acute infectious diseases—but the ultimate
larly remiss in addressing the scope of pub- impact of such improvements in surveil-
lic health surveillance, with few texts de- lance remains to be assessed in terms of
 US PUBLIC HEALTH SURVEILLANCE 185
improved health and reduced cost (56, 57, CONCLUSIONS
64, 167).
Surveillance in rapidly evolving areas of Public health surveillance provides a
public health, such as injuries and chronic quantitative basis for other distinct facets
diseases, often relies on existing data sets, of public health practice, including epide-
because of the usefulness of historical data miologic research and control and preven-
and the prohibitive costs of new systems tion services. Public health surveillance in-
(167). Evaluation of the use of such data cludes not only data collection and analysis
sets for public health surveillance repre- but also the application of these data to
sents an important new challenge. control and prevention activities by dis-
seminating information to practitioners of

Downloaded from http://epirev.oxfordjournals.org/ at Pennsylvania State University on March 5, 2016

Policy public health and others who need to know.
Effective public health practice requires Although surveillance has been conducted
the following: 1) an accurate assessment of in some form for more than a century, its
the public health; 2) definition of specific uses and practices have evolved most dra-
public health priorities; 3) development and matically over the past 40 years. A signifi-
implementation of research and control cant change has been the extension of sur-
programs to improve health; and 4) an eval- veillance beyond infectious disease to in-
uation of these programs (2). Public health clude the spectrum of public health
surveillance data can provide a quantitative problems in chronic disease, occupational
basis for policy decisions and allocation of health, injury, the environment, personal
scarce resources. Furthermore, the infor- behaviors, and preventive health technolo-
mation gained from surveillance programs gies. A second significant change has been
can significantly contribute to the contin- the effort to put public health surveillance
uous redefinition of public health priorities on a more quantitative basis.
as problems are resolved and other needs Public health surveillance has been per-
emerge. In short, good surveillance data can ceived by most as an early warning system,
and should be used to guide public health a crude indication of the occurrence of un-
practice. usual disease patterns. Because of a focus
Policy should be based on accurate data. on timeliness and simplicity, there has
The quality and limitations of both sur- often been less concern for data quality. In
veillance data and their interpretation recent years, however, there has been an
must be recognized by those communicat- increased use of data obtained outside of
ing the information and by those establish- public health practice and a concomitant
ing policy. Ideally, policymakers, in re- increased concern with the quality of sur-
sponding to questions related to health pol- veillance data and methods used to collect
icy, will know to turn to the surveillance and analyze these data (51). It is appropri-
program. ate, therefore, for the epidemiologist to ex-
Similarly, public health surveillance amine this tool carefully and to ascertain
should not be seen as an end in itself, but how one can efficiently improve the collec-
rather as a tool for use in promoting health tion, analysis, and dissemination of sur-
and preventing and controlling disease and veillance data. In other words, application
disability. Surveillance data should not be of a scientific approach to this method
acquired at the cost of privacy, nor should should improve its usefulness.
the quest for precise numbers or exquisite Several current activities will have a sig-
analyses lead to costs that outweigh the nificant impact on the practice of public
benefits of such information to the public health surveillance. We need to identify
health. Again, the need for data must be data sets relevant to specific health prob-
kept in perspective in relation to their in- lems in the most rapidly evolving areas of
tended use. public health. In some cases, this will re-
186 THACKER AND BERKELMAN

quire creating new data sets such as the 7. Hinman AR. Surveillance of communicable dis-
eases. Presented at the 100th annual meeting of
Behavioral Risk Factor Survey, which is the American Public Health Association, Atlan-
jointly conducted by CDC and state health tic City, NJ, November 15, 1972.
departments (62). More often, ongoing data 8. National Office of Vital Statistics. Vital Statis-
tics of the United States, 1958. Washington, DC:
collection efforts, such as the notifiable dis- NOVS, 1959.
ease data systems maintained by state 9. Centers for Disease Control. Manual of proce-
health departments, and data surveys con- dures for national morbidity reporting and public
health surveillance activities. Atlanta, GA: CDC,
ducted by NCHS, will be adapted to sur- 1985.
veillance needs. Statistical and graphic 10. Chapin CV. State health organization. JAMA
techniques will improve utilization and un- 1916;66:699-703.
11. National Office of Vital Statistics. Reported in-
derstanding of available data. Computers cidence of selected notifiable diseases: United

Downloaded from http://epirev.oxfordjournals.org/ at Pennsylvania State University on March 5, 2016

will play an increasingly large role, not only States, each division and state, 1920-50, vital
in analysis but also in graphic display statistics special reports. National Summaries.
1953;37:1180-l.
methods and electronic data dissemination. 12. Langmuir AD, Nathanson N, Hall WJ. Surveil-
The critical challenge in public health lance of poliomyelitis in the United States in
surveillance today, however, remains the 1955. Am J Public Health 1956;46:75-88.
13. Nathanson N, Langmuir AD. The Cutter inci-
ensurance of its usefulness. For this pur- dent: poliomyelitis following formaldehyde-
pose, therefore, we need regular, rigorous inactivated poliovirus vaccination in the United
evaluation of public health surveillance sys- States during the spring of 1955. I. Background.
Am J Hyg 1963;78:16-28.
tems. Even more basic is the need to regard 14. Global Commission for the Certification of
surveillance as a scientific endeavor. To do Smallpox Eradication. The global eradication of
this properly, one must fully understand smallpox. Geneva: WHO, 1980.
15. Langmuir AD. Evolution of the concept of sur-
the principles of surveillance and its role in veillance in the United States. Proc R Soc Med
epidemiologic research and other aspects of 1971;64:681-9.
the overall mission of public health. What 16. Int J Epidemiol (entire issue). 1976;5:3-91.
17. Raska K. National and international surveil-
is necessary now is to develop the epide- lance of communicable diseases. WHO Chron
miologic methods relevant to public health 1966;20:315-21.
surveillance; to apply computer technology 18. World Health Organization. Report for drafting
committee. Terminology of malaria and of ma-
for efficient data collection, analysis, and laria eradication. Geneva: WHO, 1963.
graphic display; to apply surveillance prin- 19. Retailliau HF, Curtis AC, Storr G, et al. Illness
ciples to practice; and to routinely assess after influenza vaccination reported through a
nationwide surveillance system, 1976-1977. Am
the usefulness of surveillance systems. J Epidemiol 1980;lll:270-8.
20. Schonberger LB, Bregman DJ, Sullivan-Bolyai
REFERENCES JZ, et al. Guillain-Barre syndrome following vac-
1. Langmuir AD. The surveillance of communicable cination in the national influenza immunization
diseases of national importance. N Engl J Med program, United States, 1976-1977. Am J Epi-
1963;268:182-92. demiol 1979;110:105-23.
2. World Health Organization. Report of the tech- 21. Centers for Disease Control. Guillain-Barre Syn-
nical discussions at the twenty-first World drome Surveillance Report, January 1978-
Health Assembly on "National and Global Sur- March 1979. Atlanta, 1980.
veillance of Communicable Diseases." A21/ 22. Erickson JD, Mulinare J, McClain PW, et al.
Technical Discussions/5. Geneva: WHO, May Vietnam veterans' risks for fathering babies with
1968. birth defects. JAMA 1984;252:903-12.
3. Centers for Disease Control. Comprehensive 23. Shands KN, Schmid GP, Dan BB, et al. Toxic-
plan for epidemiologic surveillance: Centers for shock syndrome in menstruating women. N Engl
Disease Control, August 1986. Atlanta, GA: J Med 1980;303:1436-42.
CDC, 1986. 24. Waldman RJ, Hall WN, McGee H, et al. Aspirin
4. Hartgerink MJ. Health surveillance and plan- as a risk factor in Reye's syndrome. JAMA
ning for health care in the Netherlands. Int J 1982;247:3089-94.
Epidemiol 1976;5:87-91. 25. The Centers for Disease Control Cancer and
5. Anonymous (Editorial). Surveillance. Int J Epi- Steroid Hormone Study. Long-term oral contra-
demiol 1976;5:3-6. ceptive use and the risk of breast cancer. JAMA
6. Langmuir AD. William Farr: founder of modern 1983;249:1591-5.
concepts of surveillance. Int J Epidemiol 26. Jaffe HW, Choi K, Thomas PA, et al. National
1976;5:13-18. case-control study of Kaposi's sarcoma and
 US PUBLIC HEALTH SURVEILLANCE 187
Pneumocystis carinii pneumonia in homosexual genital malformations surveillance: two Ameri-
men: epidemiologic results. Ann Intern Med can systems. Int J Epidemiol 1981;10:247-52.
1983;99:293-8. 43. Horm JW, Asire AJ, Young JL Jr, et al. SEER
27. California Department of Health Services. Dis- Program: cancer incidence and mortality in the
ciplinary action by Board of Medical Quality United States, 1973-1981. Bethesda, MD: US
Assurance for failure to report a reportable in- Department of Health and Human Services,
fectious disease. California Morbidity Aug 11, 1984. (NIH publication no. 85-1837).
1978 (no. 31). 44. Serfling RE. Methods for current statistical
28. Health and Public Policy Committee, American analysis of excess pneumonia-influenza deaths.
College of Physicians, The Infectious Diseases Public Health Rep 1963;78:494-506.
Society of America. Acquired immunodeficiency 45. Choi K, Thacker SB. An evaluation of influenza
syndrome. (Position Paper). Ann Intern Med mortality surveillance, 1962-1979.1. Time series
1986;104:575-81. forecasts of expected pneumonia and influenza
29. Faich GA, Knapp D, Dreis M, et al. National deaths. Am J Epidemiol 1981;113:215-26.
adverse drug reaction surveillance: 1985. JAMA 46. Lui K-J, Kendal AP. Impact of influenza epidem-

Downloaded from http://epirev.oxfordjournals.org/ at Pennsylvania State University on March 5, 2016

1987;257:2068-70. ics on mortality in the United States from Oc-
30. Coleman PJ, Sanderson LM. Surveillance of oc- tober 1972 to May 1985. Am J Public Health
cupational injuries treated in hospital emergency 1987;77:712-16.
rooms—United States, 1982. In: CDC Surveil- 47. McCarthy BJ, Terry J, Rochat RW, et al. The
lance Summaries (MMWR) 1983;32(no. 2SS): underregistration of neonatal deaths: Georgia,
31SS-7SS. 1974-1977. Am J Public Health 1980;70:977-81.
31. National Center for Health Statistics. The Na- 48. Centers for Disease Control. Survey of measles
tional Health Interview Survey design, 1973-84, surveillance activities in state and local health
and procedures, 1975-83. Washington, DC: US departments. MMWR 1980;29:160, 165-6.
GPO, 1985. (Vital and health statistics. Series 1, 49. Jekel JF. The "Rainbow Reviews" Publications
no. 18) (DHHS publication no. (PHS)85-1320). of the National Center for Health Statistics. J
32. National Center for Health Statistics. Develop- Chronic Dis 1984;37:681-8.
ment of the design of the NCHS Hospital Dis- 50. National Center for Health Statistics. Health,
charge Survey. Washington, DC: US GPO, 1977. United States, 1986. Washington, DC: US GPO,
(Vital and health statistics. Series 2, no. 39) 1986. (DHHS publication no. (PHS)87-1232).
(DHEW publication no. (HRA)77-1199). 51. Thacker SB, Parrish RG, Trowbridge FL. A
33. National Center for Health Statistics. Plan and method to evaluate systems of epidemiologic sur-
operation of the Second National Health and veillance. World Health Stat Q 1988;41:11-18.
Nutrition Examination Survey, 1976-80. Wash- 52. Kuller LH. Relationship between acute and
ington, DC: US GPO, 1981. (Vital and health chronic disease epidemiology. Yale J Biol Med
statistics. Series 1, no. 15) (DHHS publication 1987;60:363-76.
no. (PHS)81-1317). 53. Riley LW, Remis RS, Helgerson SD, et al. Hem-
34. Edmonds LD, Anderson CD, Glynt JW, et al. orrhagic colitis associated with a rare Escherichia
Congenital central nervous system malforma- coli serotype. N Engl J Med 1983;303:681-5.
tions and vinyl chloride exposure: a community 54. Marier R. The reporting of communicable dis-
study. Teratology 1978;17:137-42. eases. Am J Epidemiol 1977; 105:587-90.
35. Caper P. The epidemiologic surveillance of med- 55. Kimball AM, Thacker SB, Levy ME. Shigella
ical care. Am J Public Health 1987;77:668-9. surveillance in a large metropolitan area: assess-
36. National Center for Health Statistics. National ment of a passive reporting system. Am J Public
Ambulatory Medical Care Survey: background Health 1980;70:164-6.
and methodology, United States. Washington, 56. Vogt RL, LaRue D, Klaucke DN, et al. Compar-
DC: US GPO, 1974. (Vital and health statistics. ison of active and passive surveillance systems
Series 2, no. 61) (DHEW publication no. of primary care providers for hepatitis, measles,
(HRA)74-1335). rubella and salmonellosis in Vermont. Am J Pub-
37. Blount JH, Reynolds GH, Rice RJ. Pelvic in- lic Health 1983;73:795-7.
flammatory disease: incidence and trends in pri- 57. Thacker SB, Redmond S, Rothenberg RB, et al.
vate practice. In: CDC Surveillance Summaries A controlled trial of disease surveillance strate-
(MMWR) 1983;32(no. 4SS):27SS-34SS. gies. Am J Prev Med 1986;2:345-50.
38. Centers for Disease Control. Influenza—United 58. Sondik EJ, Young JL, Horm JW, et al. 1985
States, 1985-1986. MMWR 1986;35:470, 475-9. Annual Cancer Statistics Review: Bethesda, MD:
39. Rivara FP, Bergman AB, Lo Gerfo JP, et al. Department of Health and Human Services,
Epidemiology of childhood injuries. Am J Dis 1986. (NIH publication no. 86-2789).
Child 1982;136:502-6. 59. Shaw GM, Windham GC, Leonard A, et al. Char-
40. National Institute on Drug Abuse. Annual Data acteristics of hazardous material spills from re-
1985. Data from the Drug Abuse Warning Net- porting systems in California. Am J Public
work. Rockville, MD: National Institute on Drug Health 1986;76:540-3.
Abuse, 1986. (DHHS publication no. (ADM)86- 60. Kircher T, Nelson J, Burdo H. The autopsy as a
1469). measure of accuracy of the death certificate. N
41. Weddell JM. Registers and registries: a review. Engl J Med 1985;313:1263-9.
Int J Epidemiol 1973;2:221-8. 61. Centers for Disease Control. Survey of viral hep-
42. Edmonds LD, Layde PM, James LM, et al. Con- atitis surveillance activites in state and local
188 THACKER AND BERKELMAN
health departments. MMWR 1981;30:164, 169- veillance Summaries, August 1986 (MMWR)
170. 1986;35(no. 2SS):1SS-6SS.
62. Remington PLS, Smith MY, Williamson DF, et 80. Bailar JC III, Smith EM. Progress against can-
al. Design, characteristics and usefulness of cer? N Engl J Med 1986;314:1226-32.
state-based behavioral risk factor surveillance: 81. Burgess HJL. Surveillance of the population at
1981-1986. Public Health Rep (in press). risk: the community. In: Beaton GH, Bargon
63. Thacker SB, Osborne EJ, Salber EJ. Health care HM, eds. Nutrition in preventive medicine. Ge-
decision making in Southern County. J Com- neva: WHO, 1976:256-67.
munity Health 1978;3:347-56. 82. Deutscher S, Robertson WBC, Smith AP. Age
64. Hinds MW, Skaggs JW, Bergeisen GH. Benefit- and sex trends in ischaemic heart disease, cere-
cost analysis of active surveillance of primary brovascular disease, hypertension, and diabetes:
care physicians for hepatitis A. Am J Public a comparison between hospital discharge and
Health 1985;75:176-7. mortality data. Br J Prev Soc Med 1971;25:84-
65. Centers for Disease Control. Guidelines for eval- 93.
uating surveillance systems. MMWR supple- 83. Habicht J-P. Some characteristics of indicators

Downloaded from http://epirev.oxfordjournals.org/ at Pennsylvania State University on March 5, 2016

ment 1988;37(no. S-5):l-20. of nutritional status for use in screening and
66. Mason JO, Koplan JP, Layde PM. The preven- surveillance. Am J Clin Nutr 1980;33:531-5.
tion and control of chronic diseases: reducing 84. Swerdlow AJ. Cancer registration in England
unnecessary deaths and disability—a conference and Wales: some aspects relevant to interpreta-
report. Public Health Rep 1987;102:17-20. tion of the data. J R Stat Soc (A) 1986;149:146-
67. Matsumoto N, Whisnant JP, Kurland LT, et al. 60.
Natural history of stroke in Rochester, Minne- 85. US House of Representatives. Occupational
sota, 1955-1969: an extension of a previous Health Hazard Surveillance: 72 years behind and
study, 1945 through 1954. Stroke 1973;4:20-9. counting. Sixty-First Report by the Committee
68. Kuller LH, Cooper M, Perper J, et al. Myocardial on Government Operations together with Addi-
infarction and sudden death in an urban com- tional and Supplemental Views, October 8,1986.
munity. Bull NY Acad Med 1973;49:532-43. Washington, DC: US GPO, 1986:11-23. (Publi-
69. Gillum RF, Feinleib M, Margolis JR, et al. Com- cation no. 63-9690).
munity surveillance for cardiovascular disease: 86. Sundin DS, Pedersen DH, Frazier TM. Occupa-
the Framingham cardiovascular disease survey. tional hazard and health surveillance. Am J Pub-
J Chronic Dis 1976;29:289-99. lic Health 1986;76:1083-4.
70. Hisserich JC, Martin SP, Henderson BE. An 87. Centers for Disease Control. Leading work-
areawide cancer reporting network. Public related diseases and injuries—United States.
Health Rep 1975;90:15-17. MMWR 1983;32(2):24-6, 32.
71. Harmsen P, Tibblin G. A stroke register in Go- 88. Muldoon JT, Wintermeyer LA, Eure JA, et al.
teborg, Sweden. Acta Med Scand 1972;191:370- Occupational disease surveillance data sources,
463. 1985. Am J Public Health 1987;77:1006-8.
72. Smith DA, Schnall PL. Improved hypertension 89. Centers for Disease Control. Worker fatalities
control using a surveillance system in a neigh- due to excavation cave-ins. MMWR 1986;35:49-
borhood health center. Med Care 1980;18:766- 50.
74. 90. Dubrow R, Sestito J, Lalich N, et al. Death
73. Parrish HM, Payne GH, Allen WC, et al. Mid- certificate-based occupational mortality in the
Missouri stroke survey: a preliminary report. United States. Am J Ind Med 1987;l:329-42.
Missouri Med 1966;63:816-21. 91. US Congress, Office of Technology Assessment.
74. Fortmann SP, Haskell WL, Williams PT, et al. Transportation of hazardous materials. Wash-
Community surveillance of cardiovascular dis- ington, DC: US GPO, July 1986. (OTA publica-
eases in the Stanford Five-City Project. Am J tion no. SET-304).
Epidemiol 1986; 123:656-69. 92. Centers for Disease Control. Annual Summary
75. Kuller LH, Bolker A, Saslaw MS, et al. Nation- 1981: reported morbidity and mortality in the
wide cerebrovascular disease mortality study. I. United States. MMWR 1982;30:112-13.
Methods and analysis of death certificates. Am 93. Hanson JW, Oakley GP Jr. Spray adhesives and
J Epidemiol 1969;90:536-44. birth defects. JAMA 1976;236:1010.
76. Hoover R, Mason TJ, McKay FW, et al. Cancer 94. Edmonds LD, Layde PM, Erickson JD. Airport
by county: new resource for etiologic clues. Sci- noise and teratogenesis: a negative study. Arch
ence 1975;189:1005-7. Enrivon Health 1979;34:243-7.
77. Trowbridge FL. Prevalance of growth stunting 95. US Department of Health and Human Services,
and obesity: pediatric nutrition surveillance sys- Environmental Protection Agency. Notice of the
tem, 1982. In: CDC Surveillance Summaries first priority list of hazardous substances that
(MMWR) 1983;32(no. 4SS):23SS-26SS. will be the subject of toxicological profiles (FRL-
78. Chilvers C, Fraser P, Beral V. Alcohol and esoph- 3174-9(a);52(74)) (OPTS-400003) Friday, April
ageal cancer: an assessment of the evidence from 17, 1987, 12866-74.
routinely collected data. J Epidemiol Community 96. Falk H, Caldwell GG, Stein GF. Presentation of
Health 1979;33:127-33. incident—Three Mile Island. In: Finberg L, ed.
79. Berkelman RL, Ralston M, Herndon J, et al. Report of the Eighty-Fourth Ross Conference on
Patterns of alcohol consumption and alcohol- Pediatric Research. Columbus, OH: Ross Labo-
related morbidity and mortality. In: CDC Sur- ratories, 1982:74-8.
 US PUBLIC HEALTH SURVEILLANCE 189
97. Baxter PJ, Ing R, Falk H, et al. Mount St Helens 1987. (FEMA publication no. 772-629/60498).
eruptions. May 18 to June 12, 1980. JAMA 115. US Coast Guard. Coding Instructions for the
1981;246:2585-9. Automated File of Commercial Vessel Casualties.
98. Janerich DT, Burnett WS, Feck G, et al. Cancer Washington, DC: US Coast Guard, 1984.
incidence in the Love Canal Area. Science 116. Goldberg M, Gelfand HM, Mullner R. An eval-
1981;212:1404-7. uation of the Illinois Trauma Registry—the com-
99. Reich MR, Spong JK. Kepone: a chemical dis- pleteness of case reporting. Med Care
aster in Hopewell, Virginia. Int J Health Serv 1980;5:520-31.
1983;13:227-46. 117. University of California at Los Angeles, Centers
100. Gunby P. A medical team goes to Olympics for Disease Control. The epidemiology of homi-
(news). JAMA 1984;252:453-4. cide in the city of Los Angeles, 1970-1979. At-
101. Von Allmen SD. Summer mortality surveillance lanta, GA: Centers for Disease Control, 1985.
from selected city and county medical examiners. 118. Illinois Criminal Justice Information Authority.
In: CDC Surveillance Summaries (MMWR). Research Bulletin—Introduction to Illinois Uni-
1983;32(no. 1SS):1SS-6SS. form Crime Reports. Chicago: Illinois Criminal

Downloaded from http://epirev.oxfordjournals.org/ at Pennsylvania State University on March 5, 2016

102. Gallagher SS, Finison K, Guyer B, et al. The Justice Information Authority, 1985.
incidence of injuries among 87,000 Massachu- 119. National Center on Child Abuse and Neglect
setts children and adolescents: results of the (NCCAN). Study findings: national study of the
1980-1981 statewide childhood injury prevention incidence of child abuse and neglect. Washing-
program surveillance system. Am J Public ton, DC: NCCAN, 1981. (DHHD publication no.
Health 1984;74:1340-7. (OHDS)81-30325).
103. Ing RT, Baker SP, Frankowski RR, et al. Injury 120. Sedlak AJ, Bowen GL, Straus MA. Domestic
surveillance systems—strengths, weaknesses, violence surveillance system feasibility study.
and issues workshop. Public Health Rep 1985; Report to the Centers for Disease Control, 1985.
100:582-6. Westat, Inc, Rockville, MD, and Family Re-
104. Rosenberg ML. Surveillance for suicide, homi- search Laboratory, Atlanta, GA, 1986.
cide, and domestic violence: strengths, weak- 121. Centers for Disease Control. Youth suicide in the
nesses, and issues. Public Health Rep 1985; United States, 1970-1980. Atlanta, GA: CDC,
100:593-5. November 1986.
105. Guyer B, Gallagher SS, Azzara CV. Injury sur- 122. Mercy JA. Homicide surveillance, 1970-1978. In:
veillance—a state perspective. Public Health CDC Surveillance Summaries (MMWR) 1983;
Rep 1985;100:588-91. 32(no. 2SS):9SS-13SS.
106. Rockett IRH. Program perspective on injury sur- 123. US Department of Justice. Crime in the United
veillance: Rhode Island's experience. Public States: Federal Bureau of Investigation—Uni-
Health Rep 1985;100:591-3. form Crime Reports of the United States, 1983.
107. O'Connor MA, Boyle WE Jr, Prum DM. An Washington, DC: US Department of Justice,
analysis of childhood injuries in New Hamp- 1984.
shire—1982-1983. Presented at the Meeting of 124. Cantor D, Cohen LE. Comparing measures of
the American Public Health Association, Las homicide trends: methodological and substantive
Vegas, NV, September 1986. differences in the Uniform Crime Report time
108. Standfast SJ, Glebatis D, Stacy A. Building a series. Soc Sci Res 1980;9:121-45.
statewide injury surveillance system: progress in 125. US Department of Justice, Bureau of Justice
New York. Presented at the Meeting of the Statistics. National crime surveys: national sam-
American Public Health Association, Las Vegas, ple, 1973-1979. Ann Arbor, MI: Inter-University
NV, September 1986. Consortium for Political and Social Research,
109. Berkelman RL, Herndon JL, Callaway JL, et al. 1981.
A surveillance system for alcohol- and drug- 126. National Center for Health Statistics. Health
related fatal injuries. Am J Prev Med 1985;1:21- promotion data for the 1990 objectives, estimates
8. from the National Health Interview Survey of
110. Graitcer PL. The development of state and local Health Promotion and Disease Prevention,
injury surveillance systems. J Safety Res 1988; United States, 1985. Hyattsville, MD: NCHS,
18:191-8. 1986. (Advance data from vital and health statis-
111. National Highway Traffic Safety Administration tics. Series no. 126) (DHHS publication no.
(NHTSA), Fatal Accident Reporting Systems (PHS)86-1250).
(FARS), user's guide. Washington, DC: NHTSA, 127. Marks JS, Hogelin GC, Gentry EM, et al. The
1981. behavioral risk factor surveys. I. State-specific
112. National Highway Traffic Safety Administra- prevalence estimates of behavioral risk factors.
tion. National Accident Sampling System Am J Prev Med 1985;l:l-8.
(NASS), analytical user's manual. Washington, 128. Waksberg JS. Methods for random digit dialing.
DC: NHTSA 1981. J Am Stat Assoc 1978;73:40-6.
113. Ing RT. Surveillance in injury prevention. Public 129. Centers for Disease Control. Smokeless tobacco
Health Rep 1985;100:586-8. use in the United States—behavioral risk factor
114. Federal Emergency Management Agency, United surveillance system, 1986. MMWR 1987;36:337-
States Fire Administration. Fire in the United 40.
States (1983). 6th ed. Washington, DC: FEMA, 130. Becker C, Eichelberger BJ, Small A. Statewide
United States Fire Administration, GPO, July health survey reveals risky behavior. Penn Med
190 THACKER AND BERKELMAN
1986;89:60-2. Review of death certificates to assess complete-
131. Office of Technology Assessment. Assessing the ness of AIDS case reporting. Public Health Rep
efficacy and safety of medical technologies. 1987;102:386-90.
Washington, DC: US GPO, 1978. (OTA publi- 151. Orenstein W, Bart SW, Bart KJ, et al. Epide-
cation no. 052-003-00593-0). miology of rubella and its complications. In:
132. Thacker SB, Berkelman RL. Surveillance of Grunberg EM, Louis C, Goldson SE, eds. Vacci-
medical technologies. J Public Health Policy nating against brain syndromes: the campaign
1986;7:363-77. against measles and rubella. New York: Oxford
133. Teutsch SM, Herman WH, Dwyer DM, et al. University Press, 1986. (Monographs in Epide-
Mortality among diabetic patients using contin- miology and Biostatistics, Vol 19).
uous subcutaneous insulin-infusion pumps. N 152. Tukey JW. Exploratory data analysis. Reading,
Engl J Med 1984;310:361-8. MA: Addison-Wesley, 1977.
134. American Diabetes Association. Continuous sub- 153. Cleveland WS. The elements of graphing data.
cutaneous insulin infusion. Diabetes Care 1985; Murray Hill, NJ: Bell Telephone Laboratories,
8:516-17. 1985.

Downloaded from http://epirev.oxfordjournals.org/ at Pennsylvania State University on March 5, 2016

135. Marwick C. Legislation expands federal role in 154. Tufte ER. The visual display of quantitative
medical technology assessment. JAMA 1984; information. Cheshire, CT: Graphics Press, 1983.
252:3235-7. 155. National Center for Health Statistics. Proceed-
136. Graitcer PL, Burton AH. The epidemiologic sur- ings of the 1976 workshop on automated cartog-
veillance project: a computer-based system for raphy and epidemiology. Hyattsville, MD: Na-
disease surveillance. Am J Prev Med 1987;3:123- tional Center for Health Statistics, 1979.
7. (DHEW publication no. (PHS)79-1254).
137. Pavia A, Shipman L. Salmonella Havana, Geor- 156. Riggan WB. US Cancer mortality rates and
gia, 1987. Georgia Epidemiology Report 1987; trends, 1950-1979. Environmental Protection
3(7):4. Agency. Vol 4. Research Triangle Park, NC:
138. Graitcer PL, Thacker SB. The French connec- House of Facts Research Laboratory (in press).
tion. Am J Public Health 1986;76:1285-6. 157. Baker SP, Whitfield RA, O'Neal B. Geographic
139. Centers for Disease Control. Medical examiner variations in mortality from motor vehicle
and coroner systems in the United States. crashes. N Engl J Med 1987;316:1384-7.
MMWR (in press). 158. Olson JM. Noncontiguous area cartograms. Prof
140. Centers for Disease Control. Nosocomial infec- Geographer 1976;28:371-80.
tion surveillance, 1984. In: CDC Surveillance 159. Flatman GT, Brown KW, Mullins JW. Proba-
Summaries (MMWR) 1986;35(no. 1SS):17SS- bilistic spatial contouring of the plume around a
29SS. lead smelter. Silver Springs, MD: Superfund '85
141. Smith PG. Spatial and temporal clustering. In: Hazardous Materials Control Research Institute,
Shottenfeld D, Fraumeni JF, eds. Cancer epide- 1986.
miology and control. Philadelphia: WB Saun- 160. Health surveillance program. Congressional Rec-
ders, 1982. ord—House, H10867; Dec. 4,1985.
142. Newhouse VF, Choi K, D'Angelo LJ, et al. Analy- 161. Brachman PS. Surveillance. In: Evans AS, Feld-
sis of social and environmental factors affecting man HH, ed. Bacterial infections of humans.
the occurrence of Rocky Mountain spotted fever New York: Plenum Medical, 1982:49-61.
in Georgia, 1961-1975. Public Health Rep 162. Bennett JV, Brachman PS, eds. Hospital infec-
1986;101:419-28. tions. Boston, MA: Little, Brown & Co, 1986.
143. Knox G, Lancashire R. Detection of minimal 163. Rosenberg ML. Shigella surveillance in the
epidemics. Stat Med 1982;l:186-9. United States, 1975. J Infect Dis 1977;136:458-
144. Mantel N. Re: "Clustering of disease in popula- 9.
tion units: an exact test and its asymptotic ver- 164. Thacker SB, Choi K, Brachman PS. The sur-
sion." Am J Epidemiol 1983; 118:628-9. veillance of infectious diseases. JAMA 1983;
145. Longini IM Jr, Fine PEM, Thacker SB. Predict- 249:1181-5.
ing the global spread of new infectious agents. 165. Davis JP, Vergeront JM. The effect of publicity
Am J Epidemiol 1986;123:383-91. on the reporting of toxic-shock syndrome in Wis-
146. Ederer F, Meyers M, Mantel N. A statistical consin. J Infect Dis 1982; 145:449-57.
problem in space and time: do leukemia cases 166. Hinman AR, Eddins DL, Kirby CD, et al. Prog-
come in clusters? Biometrics 1964;20:626-38. ress in measles elimination. JAMA 1982;
147. Naus J. The distribution of the size of the max- 247:1592-5.
imum cluster of points on a line. J Am Stat Assoc 167. Brachott D, Mosley JW. Viral hepatitis in Israel:
1965:60:532-8. the effect of canvassing physicians on notifica-
148. Wallenstein S. A test for detection of clustering tions and the apparent epidemiological pattern.
over time. Am J Epidemiol 1980;lll:367-72. Bull WHO 1972;46:457-64.
149. Chandra Sekar C, Deming WE. On a method for 168. Centers for Disease Control. Acute traumatic
estimating birth and death rates and the extent spinal cord injury surveillance—United States,
of registration. J Am Stat Assoc 1949;44:101-15. 1987. MMWR 1988;37:285-6.
150. Hardy AM, Starcher ET, Morgan WM, et al.