PUBLIC HEALTH SURVEILLANCE IN THE UNITED STATES In 1963, Alexander D. Langmuir defined disease surveillance as "the continu

Page 1

EPIDEMIOLOGIC REVIEWS

Vol. 10, 1988

Printed in U.S.A.

PUBLIC HEALTH SURVEILLANCE IN THE UNITED STATES

STEPHEN B. THACKER1 AND RUTH L. BERKELMAN2

In 1963, Alexander D. Langmuir defined

disease surveillance as "the continued

watchfulness over the distribution and

trends of incidence through the systematic

collection, consolidation and evaluation of

morbidity and mortality reports and other

relevant data" and the regular dissemina-

tion of data to "all who need to know" (1,

pp. 182-183). Langmuir was careful to dis-

tinguish surveillance both from direct re-

sponsibility for control activities and from

epidemiologic research, although he recog-

nized the important interplay among epi-

demiologic studies, surveillance, and con-

trol activities. In 1968, the 21st World

Health Assembly held technical discussions

on the National and Global Surveillance of

Communicable Disease and identified these

main features of surveillance: 1) the sys-

tematic collection of pertinent data; 2) the

orderly consolidation and evaluation of

these data; and 3) the prompt dissemina-

tion of the results to those who need to

know, particularly those who are in a posi-

tion to take action (2).

Subsequently, the applications of sur-

veillance concepts have broadened to in-

clude a wider range of health data—risk

factors, disability, and health practices—as

well as disease. This is reflected in the 1986

Centers for Disease Control (CDC) defini-

tion of epidemiologic surveillance:

Abbreviations: AIDS, acquired immunodeficiency

syndrome; CDC, Centers for Disease Control; NCHS,

National Center for Health Statistics; NIOSH, Na-

tional Institute for Occupational Safety and Health;

WHO, World Health Organization.

1 Center for Environmental Health and Injury Con-

trol, Centers for Disease Control, Atlanta, GA 30333.

(Reprint requests to Dr. Stephen B. Thacker.)

2 Epidemiology Program Office, Centers for Disease

Control, Atlanta, GA.

The authors thank Drs. Philip S. Brachman, Mi-

chael B. Gregg, Alexander D. Langmuir, and R. Gibson

Parrish for their contributions to the manuscript.

Epidemiologic surveillance is the ongoing sys-

tematic collection, analysis, and interpretation

of health data essential to the planning, imple-

mentation, and evaluation of public health prac-

tice, closely integrated with the timely dissemi-

nation of these data to those who need to know.

The final link in the surveillance chain is the

application of these data to prevention and con-

trol. A surveillance system includes a functional

capacity for data collection, analysis, and dis-

semination linked to public health programs (3,

p. ii).

A critical word in this definition is "on-

going"; one-time surveys or sporadic stud-

ies do not constitute surveillance. An on-

going system of data collection and colla-

tion is also not sufficient to constitute

public health surveillance, because to be

useful the data must be integrated into the

conduct and evaluation of specific public

health programs, which may include epi-

demiologic research leading to prevention.

The purpose of this review is to describe

the historical and current practice of public

health surveillance, to discuss new direc-

tions for surveillance both in terms of new

public health priorities and new methodo-

logical tools, and to assess the limitations

of surveillance.

HISTORICAL OVERVIEW

Current concepts of public health sur-

veillance have evolved from public health

activities developed to control and prevent

disease in the community. In the late Mid-

dle Ages, governments in Western Europe

assumed responsibility for both health pro-

tection and health care of the population of

their towns and cities (4). A rudimentary

system of monitoring illness led to regula-

tions against polluting streets and public

water, instructions for burial and food

handling, and the provision of some types

of care. In the 17th century, John Graunt

used the Bills of Mortality to monitor dis-

164

at Pennsylvania State University on March 5, 2016

http://epirev.oxfordjournals.org/

Downloaded from

Page 2

US PUBLIC HEALTH SURVEILLANCE

165

ease in London (5). In 1766, Johann Peter

Frank advocated a more comprehensive

form of public health surveillance with his

system of police medicine in Germany,

which covered school health, injury preven-

tion, maternal and child health, and public

water and sewage (4). In addition, the gov-

ernmental measures to protect the public

health were delineated (4).

William Farr (1807-1883) is recognized

as the founder of the modern concepts of

surveillance (6). As Superintendent of the

Statistical Department of the Registrar

General's Office of England and Wales

from 1839 to 1879, Farr concentrated his

efforts on collecting vital statistics, on as-

sembling and evaluating those data, and on

reporting them to both responsible health

authorities and to the general public.

In the United States, public health sur-

veillance has focused primarily on infec-

tious diseases. Basic elements of surveil-

lance were found in Rhode Island in 1741

when the colony passed an act requiring

tavern keepers to report contagious disease

among their patrons. Two years later, the

colony passed a law requiring reporting of

smallpox, yellow fever, and cholera (7).

National disease monitoring activities

did not begin until 1850 when mortality

statistics based on the decennial census of

that year were first published by the federal

government for the entire United States

(8). In 1878, Congress authorized the fore-

runner of the Public Health Service (PHS)

to collect morbidity reports for use with

quarantine measures against pestilential

diseases such as cholera, smallpox, plague,

and yellow fever (9). In 1893, an act pro-

vided for the collection of information each

week from state and municipal authorities

throughout the United States. By 1901, all

state and municipal laws required notifi-

cation (i.e., reporting) of selected commu-

nicable disease to local authorities such as

smallpox, tuberculosis, and cholera (10). In

1914, PHS personnel were appointed as

collaborating epidemiologists to serve in

state health departments to telegraph re-

ports weekly to the Public Health Service.

It was not until 1925, however, following

markedly increased reporting associated

with the severe poliomyelitis epidemic in

1916 and the influenza pandemic of 1918-

1919, that all states were participating in

national morbidity reporting (11). After a

1948 PHS study led to the revision of mor-

bidity reporting procedures, the National

Office of Vital Statistics assumed the re-

sponsibility for morbidity reporting. In

1949, weekly morbidity statistics that had

appeared for several years in Public Health

Reports were published by the National

Office of Vital Statistics. In 1952, mortality

data were added to what is now known as

the Morbidity and Mortality Weekly Report.

Since 1961, this publication has been the

responsibility of CDC.

The Malaria Eradication Program was

undertaken by CDC and state health de-

partments in 1946 to address endemic ma-

laria in the United States at a time when

World War II veterans were returning from

Africa and from the Mediterranean and

Pacific theaters and introducing Plasmo-

dium vivax to the population (1). Spraying

of dichlorodiphenyltrichloroethane (DDT)

had begun before surveillance was initiated.

By 1947, it was clear that earlier reports of

morbidity and mortality had been erro-

neous. Mississippi, South Carolina, and

Texas had the highest reported incidences

of malaria, but because there was no diag-

nostic verification, the reported occurrence

was exaggerated. A change in reporting re-

quirements that included case reports with

diagnostic verification was illuminating. In

Mississippi, for example, the reported in-

cidence of provisional cases dropped from

17,764 to 914 in the first year, only a very

few of which could be confirmed. Such new

criteria revealed that malaria had disap-

peared as an endemic disease from the

South. The malaria experience was a major

factor emphasizing the necessity of a more

current and comprehensive system of sur-

veillance.

The critical demonstration in the United

States of the importance of surveillance

was made following the Francis Field Trial

at Pennsylvania State University on March 5, 2016

http://epirev.oxfordjournals.org/

Downloaded from

Page 3

166

THACKER AND BERKELMAN

of poliomyelitis vaccine in 1955 (12, 13).

Within two weeks of the announcement of

the results of the Field Trial and initiation

of a nationwide vaccination program, six

cases of paralytic poliomyelitis were re-

ported through the notifiable disease re-

porting system to state and local health

departments; case investigations revealed

that these children had received vaccine

produced by a single manufacturer. The

Surgeon General requested the manufac-

turer to recall all outstanding lots of vac-

cine and directed that a national poliomy-

elitis surveillance program be established

at CDC. Intensive surveillance and appro-

priate epidemiologic investigations by fed-

eral, state, and local health departments

found 141 vaccine-associated cases of par-

alytic disease, 80 of which were found in

family contacts. Daily surveillance reports

were distributed by CDC to all persons

involved in these investigations. This na-

tional common-source epidemic was ulti-

mately related to a particular brand of vac-

cine that had been contaminated with live

virus. Had the surveillance program not

been in existence, many and perhaps all

vaccine manufacturers would have ceased

production.

Surveillance was critical to the contain-

ment strategy for global eradication of

smallpox, and the success of the program

demonstrated to the international commu-

nity the practical value of surveillance (14).

To facilitate early outbreak detection, sur-

veillance teams actively investigated re-

ported cases, sought nearby cases, and ini-

tiated rapid containment measures. Sur-

veillance was intensified in areas in which

cases were confirmed. When outbreaks de-

creased, these teams continued to search

high-risk areas for cases until independent

assessment confirmed that transmission

had been interrupted. Routine reporting of

cases and the work of the surveillance

teams were further supplemented in some

settings by one-week, village-level, inten-

sive case identification. These surveillance

activities were clearly linked to contain-

ment measures that included isolating pa-

tients at home and rapidly vaccinating per-

sons in surrounding houses. Key population

contacts for surveillance included not only

government officials and religious leaders

but also school children, tea shop owners,

people in markets, nomads, and refugees.

The Conference (now Council) of State

and Territorial Epidemiologists was au-

thorized in 1951 by its parent body, The

Association of State and Territorial Health

Officials, to determine what diseases should

be reported by states to the Public Health

Service and to develop reporting proce-

dures. The Council currently meets an-

nually and, in collaboration with CDC, rec-

ommends to its constituent members ap-

propriate changes in morbidity reporting

and surveillance, including what diseases

should be reported to CDC and published

in the Morbidity and Mortality Weekly Re-

port.

Until 1950, the term surveillance was

restricted in public health practice to

watching contacts of serious communicable

diseases, such as smallpox, to detect early

symptoms so that prompt isolation could

be instituted (15). Langmuir has been cred-

ited with broadening the application of sur-

veillance to populations (1), and in 1968,

the 21st World Health Assembly focused

on national and global surveillance of com-

municable diseases, applying the term to

diseases rather than to the monitoring of

individuals with selected communicable

diseases (2). Over the intervening years, a

wide variety of health events, such as child-

hood lead poisoning, leukemia, congenital

malformations, abortions, injuries, and be-

havioral risk factors have been brought un-

der surveillance. In 1976, recognition of the

breadth of surveillance activities through-

out the world was made evident by a special

issue of the International Journal of Epi-

demiology devoted to papers specially com-

missioned to examine health surveillance

(16).

In 1986, CDC, in collaboration with the

Council of State and Territorial Epide-

at Pennsylvania State University on March 5, 2016

http://epirev.oxfordjournals.org/

Downloaded from

Page 4

US PUBLIC HEALTH SURVEILLANCE

167

miologists, published its first Comprehen-

sive Plan for Epidemiologic Surveillance

(3). In this document, CDC explicitly delin-

eated its policies and goals in surveillance,

specified plans to establish and evaluate

surveillance systems, and described rele-

vant activities in research and training.

Since the term surveillance was first ap-

plied to a disease rather than to an individ-

ual in 1950, it has assumed major signifi-

cance in disease control and prevention. Its

specific connotations, however, have not

been universally understood. In 1963,

Langmuir clearly limited surveillance to

the collection, analysis, and dissemination

of data (1). The term did not encompass

direct responsibility for control activities.

In 1965, the Director General of the World

Health Organization (WHO) established

an Epidemiological Surveillance Unit in

the Division of Communicable Diseases at

WHO (17). The Division Director, Karel

Raska, defined surveillance much more

broadly than Langmuir and included in it

"the epidemiological study of disease as a

dynamic process." In the case of malaria,

he saw epidemiologic surveillance as en-

compassing control and prevention activi-

ties. Indeed, the WHO definition of malaria

surveillance included not only case detec-

tion but also taking of blood films, drug

treatment, epidemiologic investigation, and

follow-up (18).

The 1968 World Health Assembly dis-

cussions reflected the broadened concept of

epidemiologic surveillance and addressed

the application of the concept to public

health problems other than communicable

diseases (2). In addition, epidemiologic sur-

veillance was said to imply "the responsi-

bility of following up to see that effective

action has been taken" (2, p. 9).

The use of epidemiologic to describe sur-

veillance first appeared in the mid-1960s

and was associated with the establishment

of the WHO unit of that name. This was

done both to distinguish this activity from

other forms of surveillance, such as for

military intelligence, and to reflect its

broadened applications. The use of the

term epidemiologic, however, also engen-

dered both confusion and controversy. In

1971, Langmuir noted that some epide-

miologists tend to equate surveillance with

epidemiology in its broadest sense, includ-

ing epidemiologic investigations and re-

search (15, p. 12). He found this "both

etymologically unsound and administra-

tively unwise," favoring a definition of sur-

veillance as "epidemiological intelligence."

Surveillance activities, however, have

frequently led to epidemiologic investiga-

tions of etiology. After the initiation of the

National Influenza Immunization Program

in October 1976, cases of Guillain-Barre

syndrome were reported to CDC through a

nationwide surveillance system established

to monitor illnesses occurring after influ-

enza vaccination (19). Subsequent epide-

miologic studies demonstrated a relation of

Guillain-Barre syndrome to the swine in-

fluenza vaccine that was in use, which re-

sulted in the cessation of the vaccination

program for the year (20). To test whether

the syndrome could result from use of other

influenza vaccines, a special surveillance

system was established in 1978 which used

1,813 neurologists as reporters (21). The

data collected for that and several subse-

quent years showed no association between

influenza vaccines and Guillain-Barre syn-

drome.

In addition, public health surveillance

systems are often the source of cases for

case-control studies. For example, in re-

sponse to concerns expressed by Vietnam

veterans about the possibility of increased

risk for fathering children born with birth

defects, CDC conducted a case-control

study using as cases with serious structural

birth defects infants identified by the Met-

ropolitan Atlanta Congenital Defects Pro-

gram (22). This surveillance system at-

tempted to ascertain all infants with de-

fects diagnosed during the first year of life

born to mothers who resided in the Atlanta

area. Cases and controls were selected from

infants born alive in the Atlanta area dur-

at Pennsylvania State University on March 5, 2016

http://epirev.oxfordjournals.org/

Downloaded from

Page 5

168

THACKER AND BERKELMAN

ing the years 1968 through 1980. This study

found that Vietnam veterans did not have

an increased risk of fathering children with

defects. Other examples of epidemiologic

research facilitated by case ascertainment

through surveillance include the demon-

stration of the association of tampon use

with the development of toxic shock syn-

drome (23), the relation between salicylate

use and Reye's syndrome (24), the risk of

breast cancer associated with long-term

oral contraceptive use (25), and quantifi-

cation of the risk of acquired immunodefi-

ciency syndrome (AIDS) from certain sex-

ual practices (26).

We believe that there are two issues to

be addressed in this discussion. First, what

are the boundaries of surveillance practice?

Second, is epidemiologic an appropriate

modifier of surveillance as it is used in

public health practice? To address these

questions, we must first examine the struc-

ture of public health practice. One can di-

vide public health activities into surveil-

lance, epidemiologic and laboratory re-

search, service (including program

evaluation), and training. Surveillance data

should be used to identify areas needing

research and service which, in turn, help to

define training needs. Unless data are pro-

vided to those who set policy and imple-

ment programs, their use is limited to ar-

chives and academic pursuits and are ap-

propriately considered to be health

information rather than surveillance data.

Surveillance, however, does not encompass

research or service. These are related but

independent public health activities and

may be based on surveillance. Hence, the

boundary of surveillance practice is drawn

before the actual conduct of research and

the implementation of delivery programs.

Given this context, the use of the term

epidemiologic to modify surveillance is mis-

leading. Epidemiology is a broad discipline

that incorporates research and training

that is distinct from a public health process

that we call surveillance (table 1). Because

of the much broader content of epidemiol-

ogy, the use of epidemiologic confuses the

meaning of surveillance in the public health

setting, having led in the past to the inap-

propriate incorporation of research into the

definition of surveillance (18). For this rea-

son, in this paper, we will not adhere to the

current practice of using the term epide-

miologic to modify surveillance. We pro-

pose that a more appropriate term is public

health surveillance, because its use retains

the original benefits of the term epidemio-

logic cited previously and removes some of

the confusion surrounding current practice.

Surveillance is more correctly an element

of public health, and persons encountering

the term should understand this.

SURVEILLANCE PRACTICE

Data collection

Surveillance data are collected from mul-

tiple sources. Physicians, laboratories, and

other health care providers are required to

report all cases of those diseases or health

conditions specified by state law to be no-

tifiable (or reportable); most of these con-

ditions are of infectious origin. Typically, a

case report form is completed for each case

by the health care provider or laboratory

and mailed to the local or state health

department. In some states, the authority

to change the list of notifiable diseases is

granted to the state health authorities; in

other states, each change must be newly

legislated. Penalties for failure to report a

notifiable condition may include suspen-

sion of a physician's license (27), but in

practice such penalties are rarely enforced.

Physician reporting is influenced by disease

severity, availability of public health mea-

sures, public concern, ease of reporting, and

physician appreciation of public health

practice in the community.

A disease traditionally is notifiable only

when there is a clear link between a case

report and a public health action. For many

of these diseases, case investigations are

performed by the state or local health de-

partment. Individual names and other per-

sonal identifiers are often required for pur-

poses of contact identification or treat-

at Pennsylvania State University on March 5, 2016

http://epirev.oxfordjournals.org/

Downloaded from

Page 6

US PUBLIC HEALTH SURVEILLANCE

169

TABLE 1

Distinctions between public health surveillance and epidemiologic research

Public health surveillance

Epidemiologic research

Reason for initiating data

collection

Frequency of data collec-

tion

Method of data collection

Amount of data collected

per case

Completeness of data col-

lected

Analysis of data

Dissemination of data

Use of data

Problem detection

Problem description

Identify cases for epidemiologic studies

May be legally required

Monitor geographic and temporal trends

in disease occurrence

Ongoing

Established systems or procedures

Many persons involved

Traditionally depends on voluntary par-

ticipation

Usually minimal

Often incomplete

Traditionally simple

Primarily to detect change in incidence

Usually historical comparison groups

Timely

Regular

Review in public health agency

Targeted to public health and clinical au-

dience

Identifies a problem

Triggers intervention

Suggests hypotheses

Commonly used to evaluate programs

Estimates magnitude of a problem

Hypothesis testing

Problem description

Usually time-limited

Special procedures tailored to hypotheses or

questions of interest

Fewer persons involved

Depends on paid, supervised employees

Can be considerable and usually detailed

Usually complete

Can be complex

Hypothesis testing often requires statistical

methods

Concurrent controls

Not timely

Sporadic

External review

Targeted to academic as well as public health

and clinical audience

Describes a problem in detail

Provides etiologic information

Tests hypotheses, suggests additional hy-

potheses

Less often used to evaluate programs

ment. In addition, collecting names aids in

identifying duplicate reports. Because of

the need to identify individuals, however,

concerns about confidentiality affect noti-

fiable disease reporting, and individual

identifiers are not usually collected at the

national level. These concerns have been

heightened by the epidemic of AIDS (28).

The Council of State and Territorial Ep-

idemiologists determines which notifiable

diseases should be reported from the state

health department to CDC. In addition, the

quarantinable diseases—yellow fever, chol-

era, and plague—are reportable by inter-

national regulation. To obtain information

(case reports) on specific topics such as

birth defects, influenza, low birth weight,

and nosocomial infections, CDC has collab-

orated with state and local health depart-

ments to establish specialized disease re-

porting systems. Other federal agencies are

involved in the collection of surveillance

data; for example, the Food and Drug Ad-

ministration (FDA) conducts postmarket-

ing surveillance of adverse reactions to

drugs (29), and the Consumer Product

Safety Commission conducts surveillance

of product-related injuries (30).

For many health events, national sur-

veillance systems rely on data collection

efforts by the National Center for Health

Statistics (NCHS) of CDC, including the

National Health Interview Survey (31), the

National Hospital Discharge Survey (32),

and the National Health and Nutrition Ex-

amination Survey (33) (table 2). Although

such surveys do not constitute public health

surveillance systems, data obtained in these

at Pennsylvania State University on March 5, 2016

http://epirev.oxfordjournals.org/

Downloaded from

Page 7

TABLE 2

Selecteiyiational data sources that support public health surueillance, United States

-3

Title

Scope

Responsible organization

Sources of data

Dates

Ambulatory Sentinel Practice

Network for North America

Boating Accident Reporting

System

Fatal Accident Reporting Sys-

tem

Hazardous Materials Informa-

tion System

McAuto

National Accident Sampling

System

National Ambulatory Medical

Care Survey

National Burn Registry

National Disease and Thera-

peutic Index

National

(unrepresentative)

National

(unrepresentative)

National

(unrepresentative)

National

(unrepresentative)

National

Ambulatory Sentinel Practice

Network

Coast Guard

Department of Transportation

McDonnell-Douglas Corpora-

tion

Department of Transportation

NCHS (CDC)*

National Institute of Burn

Medicine

IMS, Inc.

Family physicians

1981-present

Boat operators

1961-present

Police records, vital records, medical exam-

1975-present

iners, coroners, hospital records

Highway patrol

1971-present

Hospital discharge abstracts

1982-present

Police, hospitals

Office-based, medical practices

1973-1981,1986

Burn centers

1964-present

Office-based, medical practices

1960-present

at Pennsylvania State University on March 5, 2016

http://epirev.oxfordjournals.org/

Downloaded from

Page 8

US PUBLIC HEALTH SURVEILLANCE

171

2-2

1 5

^~-

§ 6

£ E

III

S oco

—•

a 3

1 1

> K

2 5

tn •—

a- <

tice

c c

1°

a £

-—.

o 5

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

Activities (34), provide information on

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-

lic domain (35). Again, such data collection

activities do not constitute surveillance sys-

tems, but may provide useful data for sur-

veillance.

There are relatively few national data

sets in the area of ambulatory care, and

these are used only rarely for surveillance

purposes (36-38). National data on diag-

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).

National influenza surveillance efforts

have been complemented by a convenience

sample of family practitioners that provides

CDC with demographic data and culture

specimens for all cases of influenza-like

illness seen in their offices during influenza

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

Abuse (40).

Registries are also useful sources of in-

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-

veillance systems, but data from registries

can be used for public health surveillance.

CDC's Metropolitan Atlanta Congenital

Defects Program is an example of a registry

that has been developed into a system of

public health surveillance (42). The best

at Pennsylvania State University on March 5, 2016

http://epirev.oxfordjournals.org/

Downloaded from

Page 9

TABLE 3

Selected national public health surveillance systems, United States

Title

Scope

Responsible organization

Types of data

Dates

Adverse drug reaction program

Behavioral Risk Factor Survey

Birth Defects Monitoring Pro-

gram

Disease- and condition-specific

surveillance systemst

Fatal Accident Circumstances

and Epidemiology

National Electronic Injury Sur-

veillance System

National Notifiable Disease

Surveillance System

Nutrition (pediatrics and preg-

nancy)

Surveillance, Epidemiology, and

End Results

National

(unrepresentative)

National

(unrepresentative)

National

(unrepresentative)

National

(unrepresentative)

Food and Drug Administration

CDC* via state health departments

Commission of Professional and Hospital

Activities

CDC via state health departments

National Institute for Occupational Safety

and Health (CDC)

Consumer Product Safety Commission

CDC via state health departments

CDC

National Cancer Institute

Pharmaceutical manufacturers,

1969-present

physicians

Household telephone survey

1981-present

Hospital discharge abstracts

Physicians, hospitals, laboratories

Medical examiners

Emergency rooms

Physicians hospitals, laboratories

Public clinics

Cancer registry

1970-present

Various, by con-

dition (since

1954)

1980-present

1972-present

1920-present

1975-present

1972-present

:KE

KELM

• CDC, Centers for Disease Control.

t This includes approximately 80 independent surveillance systems for specific diseases or conditions, mostly infectious.

at Pennsylvania State University on March 5, 2016

http://epirev.oxfordjournals.org/

Downloaded from

Page 10

US PUBLIC HEALTH SURVEILLANCE

173

known and most widely used registries are

those for cancer. There are population-

based cancer registries in 38 states; 11 of

these are part of the National Cancer In-

stitute's Surveillance, Epidemiology, and

End Results Program (43).

Data collation and analysis

Data on infectious diseases are collated

and analyzed in local and state health de-

partments as well as at CDC. Descriptive

statistics, including sex, age, race, and

county of occurrence, have been the most

useful for analyzing infectious disease data

with emphasis on total number of cases for

a defined time period (e.g., weekly for no-

tifiable diseases). Additional analyses for

trends over time and summary statistics on

demographic information on cases may be

performed, and rates of disease may be

calculated. An exception to these limited

analyses has been the application of regres-

sion and time series analyses to mortality

data for the surveillance of influenza (44-

46).

Public health surveillance of noninfec-

tious conditions emphasizes population-

based rates of disease. Linkage of data

sources has facilitated calculation of rates

and improved reporting (e.g., birth-weight-

specific death rates linked by birth and

death certificates) (47). Typically, health

department statistical staff calculate dis-

ease rates by sex, race, and age. In addition,

trends over time (by year for most chronic

conditions) are determined. In the past,

only national and regional data have been

available for estimates of morbidity related

to many noninfectious conditions, and few

small-area comparisons have been made.

As increasingly large morbidity data sets

are being used (e.g., hospital discharge ab-

stracts), the number and variety of appli-

cations are increasing (35).

Dissemination of data

An important purpose of data analysis

and dissemination is to provide easily

understood information in tabular or

graphic formats (in contrast to raw data)

to those who implement or influence public

health practice. Public health surveillance

data can be used to inform policymakers

and the public about the nature and extent

of health problems and to persuade these

audiences to address particular issues. In

this way, a health agency can develop a

constituency to support public health pro-

grams and to justify the expenditure of

public funds.

More than half of all state health depart-

ments and 40 per cent of county health

departments publish surveillance data in a

routine bulletin or newsletter for the local

medical and public health community (48).

State-specific notifiable disease data are

presented weekly in tabular format in

CDC's Morbidity and Mortality Weekly Re-

port. Infectious disease data are also pub-

lished annually in the CDC's Summary of

Notifiable Diseases and in similar publica-

tions by state health departments. In 1982,

CDC began publishing the CDC Surveil-

lance Summaries, which contains surveil-

lance reports on specific health events for

which CDC has program responsibilities.

Surveillance data from other agencies may

be published in special publications (e.g.,

the FDA Drug Bulletin). State and local

health department reports, federal publi-

cations like the Morbidity and Mortality

Weekly Report, and peer-reviewed public

health and clinical journals, however, are

the most common forms of disseminating

surveillance data. The "Rainbow Series" of

NCHS publications reflects data collected

and analyzed from vital statistics and na-

tional surveys (49). Although NCHS data

sets are not specifically linked to public

health programs, they are frequently used

to establish policy and monitor the effect

of national intervention programs (50).

Application to program

The concept of public health surveillance

has evolved from primarily an archival

function prior to 1950 to one in which there

is timely analysis of the data with an ap-

propriate response. Because surveillance is

part of public health practice, it should be

at Pennsylvania State University on March 5, 2016

http://epirev.oxfordjournals.org/

Downloaded from

Page 11

174

THACKER AND BERKELMAN

used to guide control and/or prevention

measures (or relevant research). No public

health surveillance system is complete

without being linked to action. The uses of

surveillance include detecting new health

problems (e.g., antibiotic-resistant strains

of bacteria), detecting epidemics, docu-

menting the spread of disease, providing

quantitative estimates of the magnitude of

morbidity and mortality, describing the

clinical course of disease, identifying poten-

tial factors involved in disease occurrence,

facilitating epidemiologic and laboratory

research, and assessing control and preven-

tion activities (51).

Surveillance has been vital to developing

hypotheses and stimulating epidemiologic

research. Historically, acute infectious dis-

ease problems have almost always been de-

fined by epidemiologists in terms of their

temporal and geographic patterns. The

need to define chronic as well as acute

diseases in terms of temporal and geo-

graphic trends is being increasingly recog-

nized (52).

Public health surveillance efforts have

often been intensified when the means for

primary prevention of most or all cases is

at hand (e.g., vaccine for measles or small-

pox) or when the disease is severe and

newly emerging, with major efforts being

made to develop control and prevention

measures (e.g., toxic shock syndrome). Ad-

ditionally, public health surveillance has

served as the means for identifying persons

with a health problem who can participate

in epidemiologic studies for developing pre-

vention strategies (53). Even before AIDS

was documented to have a viral etiology,

for example, measures to lower a person's

risk of disease were suggested by studies of

cases detected through public health sur-

veillance (26).

Evaluation of surveillance programs

Established surveillance systems require

regular review and modification based on

explicit criteria of usefulness, cost, and

quality (51). Most published evaluations of

surveillance systems have been limited to

infectious diseases (54-57), although there

have been some efforts to assess the appro-

priateness of various data sources for the

surveillance of other kinds of health prob-

lems (58-60).

A surveillance system is useful if it can

be applied to a public health program to

control and prevent adverse health events

or to better understand the process leading

to an adverse outcome. The simplest way

to assess usefulness is to ask those involved

in public health practice by means of inter-

views or surveys (48, 61, 62). A more rig-

orous approach to defining usefulness is

through the assessment of the impact of

surveillance data on policies and interven-

tions (63), but there are no published stud-

ies of this kind. Decisions affecting public

health surveillance programs are more

often based on changes in more general

program directions than on detailed analy-

sis of a particular system (e.g., directing

resources away from routine contact trac-

ing for gonorrhea control to programs for

preventing AIDS).

The economic analysis of surveillance

systems has received little systematic at-

tention apart from the accounting of direct

costs to health agencies. In a 1983 report

from Vermont, the authors reasoned that

costs were too high to justify active, health-

department-initiated surveillance of se-

lected acute infectious diseases unless un-

quantified subjective benefits, such as im-

proved relations with practicing physicians,

were great (56). In a 1985 report from Ken-

tucky, on the other hand, the benefits as-

sociated with health-department-initiated

surveillance of hepatitis A were found to

outweigh the costs (64).

CDC has proposed a systematic method

to evaluate surveillance systems on the ba-

sis of usefulness and cost as well as seven

attributes of quality: sensitivity, specificity

and predictive value positive, representa-

tiveness, timeliness, simplicity, flexibility,

and acceptability (51, 65). These attributes

of a surveillance system are interdepen-

dent, and the improvement of one may

improve or compromise another. Increasing

at Pennsylvania State University on March 5, 2016

http://epirev.oxfordjournals.org/

Downloaded from

Page 12

US PUBLIC HEALTH SURVEILLANCE

175

the sensitivity of a system to detect a

greater proportion of a given health event

in a population may also improve repre-

sentativeness and usefulness of the system,

yet may lead to greater cost, lower specific-

ity, and more false positive events. This

method of evaluation is currently being

used to assess all surveillance systems at

CDC at least once every three years. Such

an approach for evaluating surveillance

systems should enable public health prac-

titioners to efficiently assess their surveil-

lance practices and thus improve the deliv-

ery of public health services.

NEW PUBLIC HEALTH PRIORITIES

Chronic diseases

Better data are essential for progress in

chronic disease prevention and control,

particularly incidence data to establish

priorities and to evaluate programs (66).

These data should describe the burden and

the determinants of disease and help to

evaluate programs.

Three aspects of chronic diseases make

surveillance difficult. First, for some dis-

eases (e.g., mesothelioma following asbes-

tos exposure), the latency between a precip-

itating event or exposure and the eventual

chronic disease not only hinders linkage

between exposure and outcome but also

complicates development and evaluation of

prevention programs. Second, the multifac-

torial etiology of many chronic conditions

often prevents accurate linkage between

exposures, risk factors, or interventions

and outcomes. Third, because the public

health community is often interested in

arresting or reversing the progression of a

chronic condition, surveillance of various

stages of disease is important.

There has been extensive experience in

data collection and analysis of chronic dis-

ease occurrence. Indeed, at the community

level, there have been many examples of

monitoring of heart disease, stroke, and

cancer, although these programs are typi-

cally not ongoing, are usually limited to

data collection and analysis, and are rarely

directly linked to public health prevention

programs. For example, since 1945,

population-based community studies on

the natural history of stroke have been

conducted on data collected from medical

records and death certificates in Rochester,

Minnesota (67). Similar community studies

have been conducted on cardiovascular dis-

ease (68, 69). For cancer, the most success-

ful approach to community-based surveil-

lance has been the use of registries (43, 70),

an approach that has also been used for

stroke (71) and hypertension (72).

For various chronic conditions, efforts

have been made to obtain comprehensive

data not only from medical records and

death certificates but also from special sur-

veys (73, 74). Cancer has become a notifi-

able disease in at least 36 states in an effort

to broaden the scope of data collection for

that condition (T. Aldrich, Oakridge Na-

tional Laboratories, personal communica-

tion, 1988).

At the state and national levels, large

data sets are available for application to

the surveillance of chronic diseases (table

1). For example, national stroke mortality

(75) and cancer deaths (76) have been mon-

itored using death certificate data available

from the NCHS. An alternative approach

to the use of such national data bases is the

pooling of information from state and local

sources to monitor national trends, as has

been done with nutrition data. Data on

height and weight obtained from publicly

supported health and nutrition programs

demonstrated high prevalences of growth

stunting in Native American and Hispanic

children (77). This finding, together with

high weight-for-height in these popula-

tions, suggests that the diet of these chil-

dren is adequate in quantity but inadequate

in quality of nutrient intake. If confirmed,

such findings call for nutritional programs

focused on high quality protein and in-

creased essential vitamins rather than sim-

ply increased calories. Although pooling of

state data is less efficient than conducting

national samples, the close linkage at the

state level of data collection and analysis

at Pennsylvania State University on March 5, 2016

http://epirev.oxfordjournals.org/

Downloaded from

Page 13

176

THACKER AND BERKELMAN

to program intervention is an important

consideration. In addition, there have been

efforts to bring together national data from

various sources for chronic health prob-

lems, such as esophageal cancer (78) and

alcohol abuse and alcoholism (79).

The Surveillance, Epidemiology, and

End Results Program of the National Can-

cer Institute is an elaborate registry of in-

cident cancers in 11 geographic areas in the

United States which provides detailed

population-based information on mortality

due to malignant neoplasms (43). It is the

principal source of national estimates of

site-specific cancer incidence and trends,

documenting increases in cancer of the lung

and bronchus and declines in the incidence

of gastric cancer (80). This program, which

costs about $5,800,000 annually, is partic-

ularly useful for national trend estimates

and epidemiologic research. Its uses for lo-

cal prevention and control activities, how-

ever, have been limited. Less elaborate

state cancer registries may be more closely

linked to state cancer control efforts (70).

None of these data collection activities

constitute public health surveillance sys-

tems. The usefulness of existing data sets

for chronic disease surveillance has not

been proven. Such data may, nonetheless,

prove useful for public health and are es-

sential to assess the completeness and ac-

curacy of existing chronic disease data and

their appropriateness for this purpose. To

date, there has been a limited effort to

apply the principles of public health sur-

veillance to specific chronic conditions (81)

or to assess alternative approaches to col-

lecting chronic disease data for public

health surveillance (82-84).

Occupational safety and health

In 1984, J. Donald Millar, the Director

of the National Institute for Occupational

Safety and Health (NIOSH), told Congress

that federal surveillance of occupational

illness was "70 years behind that of com-

municable disease surveillance" (85, p. 11).

Before the enactment of the 1970 Occupa-

tional Safety and Health Act, no compre-

hensive national data base on workplace

hazards existed (86).

Major efforts are currently under way to

perform surveillance of occupational dis-

eases both at the national and state levels.

Although past efforts have focused on data

gathering and analysis, current efforts are

motivated by attempts to collect data in a

way that will lead directly to intervention.

NIOSH first developed a list of the "Ten

Leading Work-related Diseases and Inju-

ries" and is now identifying those occupa-

tions and industries at high risk for adverse

health events (87).

A survey of states conducted by the Iowa

Department of Health in 1985-1986

showed that at least 30 (60 per cent) states

had either voluntary or mandatory report-

ing programs for selected occupational

health conditions (88). The states have not

been uniform, however, concerning the

conditions for which they require reporting,

although lead poisoning, silicosis, and as-

bestosis are frequently included on the re-

portable disease list. Also, the reporting

criteria for these occupational health

events are not uniform across states (e.g.,

Texas requires reporting of blood lead lev-

els >40 mg/ml, whereas New York requires

reporting of all blood lead levels >25 mg/

ml) (P. Honchar, CDC, personal commu-

nication, 1987).

Although many state health departments

have reporting laws, few have maintained

a professional staff that could respond to

the incoming reports. Fortunately, this gap

in surveillance is being addressed. For ex-

ample, the Texas Department of Health

performs case investigations routinely in

response to reported cases of occupation-

ally acquired lead poisoning (P. Honchar,

CDC, personal communication, 1987). Case

investigation includes 1) assuring proper

clinical management of the affected person

and 2) offering an evaluation of the work-

site to detect factors potentially responsible

for the case. This evaluation is accom-

panied by recommendations for preventing

further cases. Screening for elevated blood

lead levels in coworkers may be conducted.

at Pennsylvania State University on March 5, 2016

http://epirev.oxfordjournals.org/

Downloaded from

Page 14

US PUBLIC HEALTH SURVEILLANCE

177

As another example, NIOSH is conducting

the Fatal Accident Circumstances and Ep-

idemiology Project, which focuses upon se-

lected electrical-related and confined

space-related fatalities (89). The purpose

of the program is to identify factors influ-

encing the risk of fatal injuries in the work-

place.

In addition to data gained from case re-

ports, 30 states now include occupational

information on death certificates; only 18

states collected such information in 1981

(90). Fourteen health departments include

parents' occupations on the birth certifi-

cate. Reporting to health departments will

be expanded through the Sentinel Event

Notification System for Occupational

Risks, a NIOSH-sponsored health event

reporting system based on reporting se-

lected occupational disease and injury out-

comes amenable to control and prevention

(E. Baker, CDC, personal communication,

1987). Other data used by state health de-

partments for surveillance include hospital

discharge data, workmen's compensation

data, and cancer registries (11 states report

occupational history on all cancer cases).

At the national level, questions on health

outcomes and health risks of relevance to

occupational health have been incorporated

into the National Health and Nutrition

Examination Survey and the National

Health Interview Survey.

Health effects of environmental toxic

exposures

Public health surveillance in environ-

mental health includes both hazard (expo-

sure) and health effects monitoring. An

example of an ongoing national system for

collecting data on potential exposures is the

Hazardous Materials Information System

of the Department of Transportation,

which was established in 1971 by a federal

law that seeks voluntary reporting of spills

occurring during interstate commerce (91).

Comparisons of these reports with inde-

pendently collected data from the state of

Washington, however, indicated that the

federal system missed over 80 per cent of

spills and had inadequate data on injury,

death, and cost (91). The state of California

compared data from the Hazardous Mate-

rials Information System with similar in-

formation collected by the California High-

way Patrol related to hazardous material

spills to determine number and nature of

incidents (59). Of 941 incidents involving

highway transport of hazardous materials

and related exposures and injuries, only 18

were reported in both systems. Despite such

limitations, the Hazardous Materials Infor-

mation System could be integrated into a

system of public health surveillance be-

cause it offers useful data on place, cause,

and mode of spill.

The most extensive public health sur-

veillance system developed for outcomes

related to an environmental hazard evolved

from 62 childhood lead-poisoning preven-

tion programs (92). Over a 10-year period

ending in 1981, 247,000 children with lead

toxicity were identified among nearly 4 mil-

lion screened. The data were disseminated

at both the local and national levels and

were applied to program planning and im-

plementation. This routine reporting sys-

tem was complemented with data from the

Second National Health Assessment and

Nutrition Examination Survey (92). These

data sources documented the decrease in

blood lead levels associated with the reduc-

tion of lead used in gasoline. When federal

funding was discontinued in 1981, the na-

tional program stopped, and most local ac-

tivities were curtailed or eliminated.

More typically, public health surveillance

of specific environmental health outcomes

is established, often in the form of regis-

tries, and then attempts are made to relate

these outcomes to particular exposures or

etiologies. Important examples are the sys-

tems established for the surveillance of con-

genital malformations. Widespread interest

in birth defects followed the epidemic of

limb reduction deformities which was as-

sociated with women taking thalidomide

during early pregnancy. This event, coupled

with epidemiologic patterns for several

malformations indicative of an environ-

at Pennsylvania State University on March 5, 2016

http://epirev.oxfordjournals.org/

Downloaded from

Page 15

178

THACKER AND BERKELMAN

mental etiology, led to the establishment of

the Metropolitan Atlanta Congenital De-

fects Program and the nationwide Birth

Defects Monitoring Program in 1967 and

1974, respectively (42). These two systems

are used to monitor trends in specific birth

defects or combinations of defects and to

stimulate epidemiologic investigations

when increases are identified. The data

have been used to demonstrate the lack of

teratogenicity of exposures of serious public

concern such as spray adhesives (93), vinyl

chloride (34), airport noise (94), and mili-

tary service in Vietnam (22). Similarly,

cancer registries have been used to study

possible relations between specific cancers

and environmental exposures (43).

The first challenge in the public health

surveillance of environmental hazards is to

determine which hazards warrant ongoing

programs of surveillance. The major con-

straint of the outcome approach is the lim-

ited knowledge of the health effects of spe-

cific toxins (i.e., natural) and toxicants (i.e.,

man-made), which inhibits our ability to

detect unexpected associations between

disease and exposure. Humans have re-

leased thousands of toxins and toxicants

into the environment, but both the health

impact and the exposure potential of most

of these substances are unknown or, at best,

established only in laboratory animals. The

Agency for Toxic Substances and Disease

Registry has been given the responsibility

of ranking the leading priority chemicals in

terms of risk to human health (95). Yet,

even when this task has been accomplished,

policies for establishing systems of public

health surveillance will need to be formu-

lated by local, state, and federal agencies.

Once priorities are established, data must

be identified for both exposures and out-

comes. Fortunately, many data sets (e.g.,

the Birth Defects Monitoring Program) al-

ready exist for both and need only to be

integrated into public health programs (42).

It is simpler and less costly to use existing

data and data systems than to establish

new ones. Additionally, historical data en-

able one to analyze long-term trends.

Public health surveillance in a disaster

setting is critical to the optimal allocation

of scarce and often poorly organized re-

sources. Surveillance systems were estab-

lished, for example, to monitor exposure to

radiation following the incident at the nu-

clear reactor at Three Mile Island (96).

Surveillance systems were also developed

to monitor the health effects of the volcanic

ash plume created by the eruption of the

Mount St. Helens volcano in 1980 (97) and

the health effects of exposure to toxic waste

at Love Canal (98). Similar short-term, lo-

cal environmental monitoring systems have

been established in response to chemical

spills (99). Although these are examples of

ad hoc surveillance established in an acute

situation, there are few examples of ongo-

ing systems of public health surveillance

linked to public health programs of control

and prevention. Occasionally, emergency

preparedness plans include surveillance,

such as during the 1984 Olympics when the

potential for terrorist activities was consid-

ered high (100). Currently, CDC is working

with the American Red Cross to organize

disaster surveillance and to establish an

international activity in this area (P. Du-

clos, CDC, personal communication, 1987).

Finally, there have been efforts to com-

bine environmental monitoring data with

health outcome information. After the se-

vere heat wave of 1980, for example, CDC,

in collaboration with medical examiners,

state and local health departments, and the

National Weather Service, developed a sys-

tem of surveillance of mortality related to

summer heat waves (101).

Injuries

The recognition of both intentional (e.g.,

homicide) and unintentional (e.g., falls) in-

juries as major public health problems has

led to the need for developing systems of

public health surveillance (102-104). Be-

cause of the acute nature of injury events,

surveillance principles learned from expe-

rience with acute infectious diseases are

often readily adaptable to injuries (105).

The current approach to establishing public

at Pennsylvania State University on March 5, 2016

http://epirev.oxfordjournals.org/

Downloaded from

Page 16

US PUBLIC HEALTH SURVEILLANCE

179

health data bases for injury has been to

adapt data, such as medical examiner re-

ports and vital statistics, to public health

needs (103, 105). This approach has been

used most widely at the state level where

vital statistics, hospital discharge data,

emergency room data, and household sur-

veys have been used to measure the extent

and nature of the unintentional injury

problem in particular populations, as well

as to assess the impact of prevention pro-

grams (102, 106-108). Medical examiner

data have also been used in the surveillance

of injuries and associated risk factors such

as alcohol and drug use (109).

Approaches to injury surveillance vary at

the state and local levels. During one year,

the Statewide Childhood Injury Prevention

Program in Massachusetts detected 5,953

fatal and nonfatal injuries in 87,022 chil-

dren and adolescents through a public

health surveillance system based on hospi-

tal and emergency room records from 23

hospitals in 14 communities (102). Using

these data, program personnel focused pre-

vention resources on the injury problems

of highest incidence in particular commu-

nities. In 1987, the Council of State and

Territorial Epidemiologists adopted a res-

olution to recommend that spinal cord in-

jury be made reportable in all states (168).

North Dakota has already made notifiable

all injuries resulting in at least one day of

disability (J. Pearson, North Dakota State

Department of Health, presented at the

annual Council of State and Territorial Ep-

idemiologists meeting, May 1987). Trauma

registries can also be adapted for surveil-

lance (110).

Several national data sets are available

for the surveillance of unintentional inju-

ries (table 2). NCHS compiles and analyzes

mortality statistics, hospital discharge

data, office-based physician utilization

data, and data collected in an ongoing

health interview survey of the general pop-

ulation. Other sources for national data

include the National Electronic Injury Sur-

veillance System maintained by the Con-

sumer Product Safety Commission (39), the

Fatal Accident Reporting System (111) and

the National Accident Sampling System

(112) maintained by the National Highway

Traffic Safety Administration, the Na-

tional Burn Registry initiated by the Na-

tional Institute of Burn Medicine (113), the

National Fire Incident Reporting System

established by the Federal Emergency

Management Agency (114), the US Coast

Guard investigations of boating incidents

(115), and the National Spinal Cord Injury

Network (116). As with chronic diseases,

the usefulness of many of these data

sources for public health surveillance re-

mains to be assessed.

The challenge of surveillance of inten-

tional injuries is even more complex. Data

are available from vital records and medical

examiners, but information on the circum-

stances of homicide and suicide is often

absent or limited in these data sets. Public

health surveillance of intentional injuries

will require the collaboration of the public

health community with a new array of ex-

perts, especially in the fields of law enforce-

ment and sociology (113). At the state and

local levels, data from criminal justice agen-

cies, medical examiners and coroners, and

medical and social service agencies are

being explored for use in the surveillance

of intentional injury (117). Illinois insti-

tuted mandatory uniform crime reporting

in 1972; the state maintains the data on

computer and publishes a report each year

(118). Few data exist on morbidity related

to assault or child abuse, and only rarely

have epidemiologic studies been conducted

in this area (119). Efforts are under way to

assess the feasibility of alternative ap-

proaches to the surveillance of domestic

violence (120).

On the basis of data from the national

mortality files of NCHS and population

estimates of the US Bureau of the Census,

a 40 per cent increase in youth suicide was

documented in the decade ending in 1980

(121). This increase was found primarily in

white males 15-24 years of age. These sur-

veillance data documented the dramatic

change of suicide as a problem of the elderly

at Pennsylvania State University on March 5, 2016

http://epirev.oxfordjournals.org/

Downloaded from

Page 17

180

THACKER AND BERKELMAN

to a problem of the young. Current efforts

in suicide surveillance have demonstrated

the importance and difficulty of arriving at

uniform definitions, a problem complicated

by the interdisciplinary nature of this en-

deavor. Uniform definitions of child and

spouse abuse, problems for which incidence

data are sparse, are also needed. Yet, as

such data bases are developed, surveillance

will play a crucial role in public health

programs aimed at controlling and pre-

venting these and other injuries. Other na-

tional data sources, such as the Uniform

Crime Reports of the Federal Bureau of

Investigation and the annual National

Crime Survey of the US Department of

Justice, have proven to be useful (121-125).

Personal health practices

At a national level, the Health Interview

Survey conducted by NCHS has provided

the most information on personal health

practices such as alcohol use and smoking.

The prevention supplements to the 1982

and 1985 surveys have provided more de-

tailed information in this area (126). As the

role of personal health behavior in the de-

velopment of chronic diseases and injuries

has become more fully recognized, state-

based programs to reduce the prevalences

of unhealthy behaviors have been estab-

lished. In turn, interest in providing a sys-

tematic means of collecting population-

based prevalence data on a state-specific

basis resulted in the initiation of the Be-

havioral Risk Factor Surveys in 1981 (127).

National estimates can be obtained more

efficiently, but local programs benefit from

involvement in data collection as well as

from the ability to adapt the collection

process to their particular needs. As of

1987, 35 state health departments are con-

ducting ongoing surveys of behavioral risk

factors in persons aged 18 years or older.

Each state uses a standardized question-

naire to determine the prevalence of a va-

riety of personal health practices including

cigarette smoking, smokeless tobacco use,

alcohol consumption, exercise, seat belt

use, dieting, and hypertension control

(127). The sample for each survey, con-

ducted by telephone, is selected generally

with a multistage cluster design based on

the Waksberg method (128).

Interview surveys can obtain personal,

health-related information with only minor

differences in the prevalence of various

health conditions when conducted by tele-

phone or in person (127). Telephone inter-

views have the advantages of lower cost

(about one-third to one-half the cost of

personal interviews) and the ease of super-

vising interviewers. Although there are

problems of bias related to omitting those

households without telephones, telephone

coverage exceeds 93 per cent in the United

States (127).

Results of the surveys are published by

both CDC and state health departments

(129, 130). They are also distributed to the

press and to a variety of local and state

organizations, including voluntary health

agencies, hospitals, health maintenance or-

ganizations, and state legislators. In 1986,

the 43 states that had conducted these sur-

veys reported that these data were fre-

quently used by the health department to

prepare state planning documents and to

establish state-level health objectives (62).

Sixty-five per cent of these states reported

using the data to support legislative initia-

tives, especially seat belt and anti-smoking

legislation (62). Limitations, however, exist

when these data are used; many states cited

a circumscribed authority to disseminate

the findings. In addition, because the sur-

veys only recently have been initiated at

the state level, not enough time has elapsed

to adequately analyze trends.

Preventive health technologies

Health technology includes the drugs, de-

vices, and medical and surgical procedures

used in health care, and the organizational

and supportive systems within which such

care is provided (131). The implementation

of new technologies is a prominent growth

industry in health. Dramatic examples,

such as carotid endarterectomy, artificial

hearts, osteoporosis screening, and AIDS

at Pennsylvania State University on March 5, 2016

http://epirev.oxfordjournals.org/

Downloaded from

Page 18

US PUBLIC HEALTH SURVEILLANCE

181

testing, are very much in the public eye.

Concerns regarding premature diffusion or

misapplication of health technologies have

highlighted the need for routine surveil-

lance of the application of the technologies,

particularly as these new technologies are

used in healthy or asymptomatic popula-

tions to prevent disease (132). Currently,

efforts are under way in several state health

departments to assess the effectiveness of

both cervical cancer and breast cancer

screening programs. Systems of public

health surveillance are an integral part of

these assessments.

There are, however, few examples of sur-

veillance of health technologies despite this

widespread diffusion of new devices and

practices. Immunization against selected

infectious diseases is probably the most

effective and well-known technology used

in public health. More recently, public

health surveillance of selected medical

technologies has been developed by CDC in

response to concerns in the public health

community. For example, in response to

state health officials during a perceived cri-

sis in 1982 concerning the use of insulin

pumps, CDC established a short-term sur-

veillance system to determine the fre-

quency and severity of complications asso-

ciated with these devices (133). Using phy-

sician reporting, the investigators iden-

tified previously unrecognized adverse

events associated with pump use as well as

35 deaths among pump users. The data

were used to assist the American Diabetes

Association in developing a policy state-

ment for clinicians that included new cri-

teria for initiating pump use (134).

Public health surveillance of technology

use provides a mechanism for monitoring

the use of a practice or device and, together

with data on morbidity and mortality, pro-

vides an ongoing measure of its effective-

ness and safety in the populations being

monitored. Surveillance will also indicate

whether an effective technology is being

applied to the population that is likely to

benefit from such technology. It is not

known, for example, whether the women

undergoing mammography are those most

likely to benefit from screening.

The need for surveillance of technology

use is evident, but the process of gathering

the primary data is not established cur-

rently for most technologies other than

drugs—the latter being a responsibility of

the Food and Drug Administration. As il-

lustrated by the surveillance of tubal steri-

lization, some hospital data sets can be

helpful in tracking inpatient procedures.

There is a lack of state and national sur-

veillance information, however, to track

diffusion of technologies in the outpatient

setting, where complex and expensive tech-

nologies are being used increasingly (135).

Although surveillance is usually under-

taken by public health agencies at the local,

state, and federal levels in collaboration

with the medical community, efforts to es-

tablish surveillance systems at all levels

have faltered in recent years. In its lead

federal role in health care technology, the

National Center for Health Services Re-

search and Health Care Technology As-

sessment should be encouraged to develop

priority-setting criteria for bringing tech-

nologies under surveillance and subse-

quently for analyzing the impact of tech-

nologies in terms of their effect on morbid-

ity, mortality, disability, and cost.

NEW TOOLS FOR PUBLIC HEALTH

SURVEILLANCE

Computers

The introduction of computer hardware

and software has provided public health

professionals with the capability to perform

surveillance more efficiently on common

conditions. Large data bases may be better

managed and analyzed, and in some in-

stances may be linked. In addition, the

microcomputer has empowered the public

health professional with an increased abil-

ity to organize, communicate, tabulate, and

analyze data. Use of the computer has in-

creased the timeliness of both data collec-

tion and analysis and has decreased the

epidemiologist's reliance on programmers

at Pennsylvania State University on March 5, 2016

http://epirev.oxfordjournals.org/

Downloaded from

Page 19

182

THACKER AND BERKELMAN

and biostatisticians for data analysis and

interpretation.

The Public Health Foundation initiated

an electronic mail system in 1983. Several

federal health agencies, including CDC, and

44 state health departments are now on-

line. In addition, three states have enrolled

their local health departments and can

telecommunicate with them. In 1984, this

network was used in six states to pilot-test

the transfer of notifiable infectious disease

data weekly to CDC (136). By early 1988,

37 reporting areas were transferring indi-

vidual case data on over 40 notifiable dis-

eases to CDC each week. The ability to

transfer binary file will allow the telecom-

munication of graphics, which facilitates

review of aggregate data. Surveillance at

the state level has been hampered by a lack

of microcomputer software for managing

and analyzing large numbers of disease rec-

ords. Currently, software developed for use

in epidemic investigations has been

adapted for surveillance and used in 20

states (A. Dean, CDC, personal communi-

cation, 1987). Use of such software in Geor-

gia has enabled early detection of an epi-

demic of illness due to Salmonella havana,

facilitating efforts to identify the environ-

mental source of the organism (137).

Programs at CDC for vaccine-

preventable diseases, tuberculosis, AIDS,

and diabetes have also developed computer

networks with state health departments to

enhance their surveillance capabilities. In

addition, state health departments have in-

itiated computer linkage with selected local

health departments for disease reporting

(138). In Wisconsin, for example, case data

from sexually transmitted diseases clinics

are telecommunicated to each other and to

local and state health departments, improv-

ing the efficiency of follow-up of patients

(A. Dean, CDC, personal communication,

1987).

Use of microcomputers has also ex-

panded surveillance activities to nontradi-

tional reporting sources. Computers in

medical examiners' offices will aid in injury

surveillance (139); microcomputers in a na-

tional sample of hospitals currently aid in

collecting information on nosocomial infec-

tions (140).

Statistical methods

The increased sophistication of statisti-

cal methods, the availability of computers,

and the development of statistical software

for analysis have broadened the potential

of statistical analysis in day-to-day public

health practice and have led to the inves-

tigation of new methods of data analysis.

The usefulness of time series analysis (45),

of detecting clusters of adverse health

events in time and place (141-144), and of

mathematical models to forecast epidemics

based on surveillance data (145) remains to

be fully assessed.

Although detecting temporal and spatial

clusters of disease has always been a goal

of public health surveillance, formal statis-

tical testing for clusters has rarely been

applied to routinely collected surveillance

data. The statistical problems associated

with determining whether an "outbreak"

has occurred were addressed in depth in the

1960s, and a variety of alternative analyses

were proposed. Two commonly used meth-

ods for space-time clustering, proposed by

Knox and Lancashire (143) and by Ederer

et al. (146), are based on the number of

"close" pairs of cases and the sum, over all

space divisions, of the maximum number

of cases in any time unit within a space

division. For example, Ederer et al. em-

ployed a summary statistic to detect both

clusters of leukemia over time and out-

breaks of polio and hepatitis.

The SCAN statistic, based on such sum-

mary statistics as those used by Mantel

(144) and Ederer et al. (146), was proposed

by Naus (147) and has recently been ap-

plied to a cluster of trisomies in three New

York City hospitals (148). This statistic is

computed by plotting points over time, tak-

ing a "moving window" of a fixed length of

time, and then finding the maximum num-

ber of observations revealed through the

window as it scans or slides over the entire

time period. The statistic is based on the

at Pennsylvania State University on March 5, 2016

http://epirev.oxfordjournals.org/

Downloaded from

Page 20

US PUBLIC HEALTH SURVEILLANCE

183

assumption that the size of the population

at risk remains fairly constant and that the

condition shows no seasonal or cylical pat-

tern over the time period plotted. Other

analytic methods have been suggested for

using environmental data to predict the

occurrence of Rocky Mountain spotted fe-

ver, but such methods have not been used

routinely (142).

The Chandra Sekar-Deming method de-

veloped by demographers has been used to

estimate completeness of reporting by com-

parison of two independent surveillance

systems with individual identifiers so that

the data may be linked (149). This method

has recently been applied to AIDS data

reported through the notifiable disease sys-

tem and through death certificate registra-

tion (150). It has also been applied to esti-

mate the sensitivity of two systems for de-

tecting vaccine-preventable diseases (151).

Surveillance systems are subject to both

selection and information biases. Notifi-

able disease reports, for example, are likely

to come from a nonrepresentative sample

of practicing physicians who may report

specific diseases because of personal inter-

est. Private practitioners, for example, may

be less likely than physicians at public

health clinics to report certain conditions

(e.g., sexually transmitted diseases). At the

same time, certain kinds of data are less

likely to be reported than others because of

ease of ascertainment (e.g., age or sex vs.

pathologic diagnoses). Analytic models are

required to measure the impact of bias on

surveillance data. Other important re-

search issues on the statistics of surveil-

lance include the development of methods

to handle incomplete or missing data, the

use of multiple subset sampling, modeling

of timeliness, and the combination of data

from independently collected data sets.

Graphic methods for data analysis and

display

Graphics have the potential to serve as

powerful tools for displaying data both for

analysis and for communication. Tukey

(152) has clearly demonstrated the impor-

tant role graphs can play in visual decoding

of large quantities of data. Although Tu-

key's methods have not yet been widely

applied to surveillance data, his pioneering

work together with the introduction of

computer graphics has laid the foundation

for graphic analysis of surveillance data

(153). Microcomputer graphics, in particu-

lar, have also made the results of data

analysis far more useful to private and pub-

lic policymakers in their planning and man-

agement of health care resources (35). Al-

though simple data still are incorporated

best into textual material or a tabular for-

mat, a graphic display can give the reader

an understanding of large and complex data

sets that cannot be conveyed easily in other

ways (154).

The interest in computer mapping in

public health is strong. A 1976 workshop