ORIGINAL INVESTIGATION
Knowledge of Heart Attack Symptoms
in a Population Survey in the United States
The REACT Trial
David C. Goff, Jr, MD, PhD; Deborah E. Sellers, PhD; Paul G. McGovern, PhD; Hendrika Meischke, PhD;
Robert J. Goldberg, PhD; Vera Bittner, MD, MSPH; Jerris R. Hedges, MD, MS; P. Scott Allender, MD;
Milton Z. Nichaman, MD, ScD; for the REACT Study Group
Background: Greater use of thrombolysis for patients
mon. The median number of correct symptoms reported
with myocardial infarction has been limited by patient
was 3 (of 11). In a multivariable-adjusted model, signifi-
delay in seeking care for heart attack symptoms. Defi-
cantly higher mean numbers of correct symptoms were re-
ciencies in knowledge of symptoms may contribute to
ported by non-Hispanic whites than by other racial or eth-
delay and could be a target for intervention. We sought
nic groups, by middle-aged persons than by older and
to characterize symptom knowledge.
younger persons, by persons with higher socioeconomic
status than by those with lower, and by persons with pre-
Methods: Rapid Early Action for Coronary Treatment
vious experience with heart attack than by those without.
is a community trial designed to reduce this delay. At base-
line, a random-digit dialed survey was conducted among
Conclusions: Knowledge of chest pain as an important
1294 adult respondents in the 20 study communities. Two
heart attack symptom is high and relatively uniform; how-
open-ended questions were asked about heart attack
ever, knowledge of the complex constellation of heart at-
symptom knowledge.
tack symptoms is deficient in the US population, espe-
cially in low socioeconomic and racial or ethnic minority
Results: Chest pain or discomfort was reported as a symp-
groups. Efforts to reduce delay in seeking medical care
tom by 89.7% of respondents and was thought to be the
among persons with heart attack symptoms should ad-
most important symptom by 56.6%. Knowledge of arm pain
dress these deficiencies in knowledge.
or numbness (67.3%), shortness of breath (50.8%), sweat-
ing (21.3%), and other heart attack symptoms was less com-
Arch Intern Med. 1998;158:2329-2338
NUMEROUSSTUDIEShave symptomshasbeenidentifiedasamajor
documented the benefit
obstacle to more widespread use of throm-
of thrombolytic therapy
bolytic therapy.2,3
for patients with acute
The prehospital delay encompasses
myocardial infarction1;
the time required for recognizing the
thus, attention has shifted toward efforts
presence of symptoms, attributing the
to increase the use of this therapy.2,3 The
symptoms to a condition requiring medi-
frequency of use and the benefits of throm-
cal attention, deciding to seek care, ar-
bolytic therapy are greater in the first sev-
ranging transportation, and going to the
eral hours following the onset of heart at-
hospital. Several barriers to rapid action
tack symptoms than in the later stages.3
can arise in this process. At the earliest
Thrombolytic therapy is used much less
stages, delay in the recognition of symp-
frequently in patients who delay seeking
toms as being caused by a heart attack
care for at least 6 hours relative to pa-
may be due to inadequate knowledge of
tients who respond more rapidly.3-6 This
heart attack symptoms or misattribution
pattern of health care delivery appears to
of the symptoms to another, noncardiac,
be a rational response to results of clini-
and potentially less serious cause.7-12 The
cal trials that show diminishing benefit of
belief that symptoms were cardiac in ori-
thrombolytic therapy given later than 6
gin has been associated with shorter
The affiliations of the authors
hours following the onset of acute ische-
delay and more frequent and earlier use
appear in the acknowledgment
mic symptoms.3 Approximately 22% of pa-
of emergency medical services.10-12 Inad-
section at the end of the article.
tients with heart attack delay seeking care
equate knowledge of heart attack symp-
The members of the REACT
Study Group and contributors
for at least 6 hours following the onset of
toms can lead to erroneous symptom
to this work are listed on page
symptoms4; hence, patient delay in rec-
attribution. Individual knowledge of
2332.
ognition of and response to heart attack
heart attack symptoms may be based
ARCH INTERN MED/ VOL 158, NOV 23, 1998
2329
on August 14, 2009
www.archinternmed.com
Downloaded from
©1998 American Medical Association. All rights reserved.

---

PARTICIPANTS, MATERIALS,
on the proportion of households (listed and unlisted). Dis-
AND METHODS
proportionate stratified sampling was used to increase the
overall household rate. To adjust for the complex sam-
pling design, survey responses were weighted by the re-
ciprocal of the probability of selection. The sampling of
The design and rationale of the REACT community-
adults and the structured interviews were conducted by the
intervention randomized trial have been described in de-
Survey Research Center at the New England Research In-
tail previously.17,18 In brief, 20 communities were in-
stitutes, Watertown, Mass, the coordinating center for
cluded in a matched design yielding 10 pairs of communities
REACT. Since the REACT communities were not selected
in 5 regions throughout the United States. One commu-
at random from all communities in the United States, the
nity from each pair was randomized to receive an 18-
resulting sample formally represents only the 20 REACT
month educational program following a 4-month baseline
communities. The degree to which selected characteris-
data collection. Communities were chosen, in part, to pro-
tics of the survey sample approximated characteristics of
vide geographic diversity and to reflect the racial and eth-
the general US population was assessed qualitatively by com-
nic composition of the general population of the United
paring observed distributions of sociodemographic char-
States. The REACT communities are in Alabama, Louisi-
acteristics and medical attributes to published data on US
ana, Massachusetts, Minnesota, North Dakota, Oregon,
population samples.19-24
South Dakota, Texas, Washington, and Wisconsin. The
Information was collected regarding a variety of top-
project was approved by the institutional review boards of
ics of importance to the goals of REACT, including
all participating universities and hospitals.
knowledge of heart attack symptoms, coronary heart dis-
A random-digit dialed telephone survey was con-
ease risk factor status, and sociodemographic characteris-
ducted before the educational activities in all 20 commu-
tics. Sociodemographic characteristics included age, sex,
nities to establish a baseline for the assessment of the ex-
race or ethnicity, educational attainment, and household
posure to the intervention or similar programs in the study
income. Knowledge regarding heart attack symptoms was
communities and the impact of the educational program
assessed by asking the following 2 open-ended questions:
on the knowledge, attitudes, and beliefs of community mem-
“What would you say are the signs or symptoms that
bers. The survey was designed to collect data from 60 adults,
someone may be having a heart attack?” and “Of the heart
aged 18 years or older, in each of the 20 communities. Sam-
attack signs or symptoms you just mentioned, which one
pling for the survey was completed by random-digit dial-
would you say is the most important?” Participants were
ing. Each community’s designated geographical target area
encouraged to give multiple responses to the first ques-
was defined by a specified set of ZIP codes. A list of tele-
tion through the use of the repeated follow-up question,
phone exchanges and a count of the households with listed
“Anything else?” Survey Research Center staff coded
telephone numbers in each ZIP code area was obtained from
responses into predetermined categories or as other. Fol-
a commercial vendor. Counts of listed households were
lowing completion of data collection, reported symptoms
supplemented with estimates of unlisted households. Eli-
that were not coded into a predetermined category were
gible telephone exchanges were divided into 5 strata based
reviewed and categorized by 2 of us (D.C.G. and D.E.S.).
more on common media presentations of immediately
RESULTS
incapacitated heart attack victims than on the aggregate
experience of real patients with heart attack. Knowledge
A total of 5603 random telephone numbers was gener-
may differ by demographic attributes as a reflection of,
ated for this survey. The dispositions of these numbers
or independent of, level of education.13-16 Persons with
were as follows: 1584 were nonworking numbers; 635
risk factors for heart disease could be considered as
resulted in no contact after 5 calls (whether they were
having a greater need to know, and may have been edu-
households was not determined); 150 were not
cated about heart attack symptoms by health pro-
screened for eligibility due to refusal or language other
fessionals. Finally, persons who have had previous
than Spanish or English; 944 were not households; 55
experience with heart disease, whether personally or
were households with no one contacted after at least
in a family member or close friend, could also have
15 calls; 520 resulted in refusals; 285 were ineligible
greater knowledge of heart attack symptoms than other
due to nonstudy ZIP code or illness; 136 resulted in
persons.
incomplete interviews; and 1294 resulted in com-
Our purpose herein is to describe knowledge
pleted interviews. Overall, 36.9% of the numbers were
regarding heart attack symptoms as reported by par-
for ZIP code-eligible households. The overall in-
ticipants in a random-digit dialed survey conducted as
terview rate (completed interviews divided by poten-
part of the Rapid Early Action for Coronary Treatment
tially eligible households) was 64.5%. The range
(REACT) Trial, a multicenter community trial designed
across the 20 study communities was 48.0% to
to test an education program to reduce prehospital
77.3%. Participation rates were generally higher in the
delay among persons experiencing heart attack symp-
northern communities, intermediate in Alabama, and
toms. We also examined whether symptom knowledge
lowest in Texas and Louisiana. According to the 1990
was related to sociodemographic characteristics, coro-
census, telephone availability exceeded 80% in all
nary heart disease risk factors, and previous experience
communities and exceeded 90% in 18 of the 20 study
with heart disease.
communities.19
ARCH INTERN MED/ VOL 158, NOV 23, 1998
2330
on August 14, 2009
www.archinternmed.com
Downloaded from
©1998 American Medical Association. All rights reserved.

---

Correct symptoms were defined as those consistent with a
Data were analyzed using commercially available soft-
published list of heart attack symptoms, ie, chest pain or
ware (SAS version 6.11; SAS Institute, Cary, NC). The total
discomfort, arm or shoulder pain or discomfort, jaw pain,
and correct numbers of heart attack symptoms reported per
intrascapular (back) pain, shortness of breath, nausea or
individual, the percentage responding that chest pain or dis-
vomiting, sweating, lightheadedness, weakness, palpita-
comfort was the most important symptom, and the prevalence
tions, and a feeling of impending doom.1 All of these symp-
of specific symptom expectations were examined for the over-
toms except the last 2 were included in the REACT inter-
all group and in subgroups defined by sociodemographic char-
vention messages following a review of multiple sources
acteristics, coronary heart disease risk factors, and previous
of information regarding heart attack symptoms. Any other
experience with heart disease. Given the large number of com-
reported symptom not on the previous list was considered
parisons, statistical tests of association were not performed
incorrect for the purposes of this analysis. Chest pain or
at this stage; however, 95% confidence intervals were com-
discomfort was identified as the correct response to the sec-
puted to provide information regarding the precision of the
ond question posed, since chest pain or discomfort is the
point estimates. Two multivariable regression analyses were
most common symptom reported by patients with heart at-
performed to examine the independence of associations be-
tack. For our purposes, the symptom chest pain or dis-
tween knowledge of heart attack symptoms and sociodemo-
comfort is the combination of the 4 nonexclusive descrip-
graphic characteristics, risk factor status, and previous expe-
tive categories of chest pain or discomfort that were coded
rience with heart disease. The number of correct symptoms
in the survey. Respondents could report none, 1, or more
reported per individual was regressed on sociodemographic
than 1 of the following categories: pain, pressure, tight-
characteristics, risk factors, and previous experience with heart
ness, and discomfort. Other synonyms were coded into these
disease using mixed-model linear regression. In this analy-
categories.
sis, persons reporting 7 or more correct symptoms (n = 8) were
Presence of coronary heart disease risk factors (ie, dia-
reclassified with the group that originally reported 6 into a
betes, high blood pressure, high levels of serum choles-
new group to prevent those few cases with very high values
terol, and current cigarette smoking) and previous expe-
from exerting excessive influence on the results. Mixed-model
rience with heart disease were assessed by self-reports,
logistic regression was used to examine the independence of
including reports of physician diagnoses. Personal history
associations between a correct report of chest pain or discom-
of heart disease was defined as a positive response to ei-
fort as the most important heart attack symptom and socio-
ther of the following questions: “Have you ever had a heart
demographic characteristics, coronary heart disease risk fac-
attack?” and “Have you ever been told by a physician that
tor status, and previous experience with heart disease. In both
you have a heart condition?” Two other indicators of pre-
mixed models, community was a random effect nested within
vious experience with heart disease were assessed. His-
pair and region. Pair was a random effect nested within re-
tory of a heart attack in a spouse, parent, or sibling was as-
gion. Sociodemographic characteristics, risk factors, and pre-
sessed as an indication of an experience with high saliency.
vious experience with heart disease were fixed effects. In all
History of a heart attack in any other relative or close friend
analyses, survey responses were weighted by the reciprocal
was assessed as an indication of an experience with lesser,
of the probability of selection to adjust for the complex sam-
but still substantial, salience.
pling design within each community.
Sociodemographic characteristics of the survey re-
less knowledge and those with high levels of serum cho-
spondents and the prevalence of self-reported risk fac-
lesterol exhibited greater knowledge than respondents with-
tors for coronary heart disease are shown in Table 1.
out these conditions. Minimal differences in knowledge were
In aggregate, these findings support the contention that
observed between groups classified according to cigarette
the lowest socioeconomic group is underrepresented in
smoking and blood pressure status. Respondents with pre-
this sample relative to the total US population,19 as would
vious experience with heart disease exhibited greater knowl-
be expected for a random-digit dialed telephone survey.
edge of heart attack symptoms than those without such an
The total and correct mean numbers of heart attack
experience. Chest pain was reported to be the most im-
symptoms reported and the percentage of participants re-
portant symptom by 56.6%. Subgroup differences in re-
porting chest pain as the most important symptom of a heart
porting chest pain as the most important symptom were
attack are shown in Table 2 for the overall population and
observed. Older respondents and lower-income groups were
for subgroups defined by sociodemographic characteris-
more likely to report chest pain as the most important symp-
tics, coronary heart disease risk factor status, and previ-
tom. Persons with missing data for income responded simi-
ous experience with heart disease. The median number of
larly to low-income respondents. Persons with diabetes were
correct symptoms reported was 3. Only 40 (3.1%) of 1294
more likely to report chest pain as the most important heart
participants reported at least 6 correct symptoms. There
attack symptom than persons without this condition.
was evidence of a greater level of knowledge, as indi-
The prevalence of reported expectations of specific
cated by number of correct symptoms reported, in non-
symptoms is listed in Table 3, Table 4, and Table 5
Hispanic whites than in other race or ethnic groups, in the
for the overall population and in subgroups defined by
2 older age groups than in youngest group, and in the more
demographic characteristics (Table 3), socioeconomic
educated and higher-income groups than in the less edu-
characteristics (Table 4), and coronary heart disease risk
cated and lower-income groups. The presence of coro-
factor status and previous experience with heart disease
nary heart disease risk factors was not consistently asso-
(Table 5). Some form of chest pain or discomfort was re-
ciated with knowledge; respondents with diabetes exhibited
ported by 89.7% of respondents. Respondents reported
ARCH INTERN MED/ VOL 158, NOV 23, 1998
2331
on August 14, 2009
www.archinternmed.com
Downloaded from
©1998 American Medical Association. All rights reserved.

---

The REACT Study Group and Contributors to This Work
University of Alabama at Birmingham, School of Medicine: James Raczynski, PhD (principal investigator), Vera Bittner, MD
(coprincipal investigator and medical director), Carol Cornell, PhD (coprincipal investigator and project director), Diane
Caddell, Eli Capilouto, DMD, MPH, ScD, Diane Gibb, Janice Gilliland, MA, MSPH, Linda Goodson,RN, Liz Hester, Laura
Leviton, PhD, Janet Pribble, MD, Herman Taylor, MD, Judy Taylor, PhD, Sue Williams, Karen Nelson, Barbara Theobald,
Earle Overton, and Shelly Jordan; University of Massachusetts Medical School, Worcester: Robert Goldberg, PhD (prin-
cipal investigator), Jane Zapka, ScD (coprincipal investigator), Barbara Estabrook, MSPH (project director), Sharon Mel-
ville, Mark Robbins, Lucy Jenkins, Susan Demoga, Beth Ewy, MPH, Kathleen Lovell, MEd, and Lorraine Wallace, MPH; Uni-
versity of Minnesota School of Public Health, Minneapolis: Russell Luepker, MD (principal investigator), John Finnegan,
PhD (coprincipal investigator and project officer), Deborah Alexander, Neil Bracht, MSW, MPH, Richard Fouschee, PhD,
Bernadette Gloeb, Nancy Hall, Brian Laing, Paul McGovern, David Murray, PhD, Janelle Traut, Carol Grimm, and Michele
Hartshorn; University of Texas Health Science Center, School of Public Health, Houston: David Goff, MD, PhD (former
principal investigator), Milton Nichaman, MD, ScD (principal investigator), Alfred McAlister, PhD (coprincipal investiga-
tor), Angela Meshack, DrPH (project director), Ruby Benjamin-Garner, MA, Wenyaw Chan, PhD, Marcia Davis, Holly Har-
rison, Adriana Linares, MD, MPH, Dilip Pandey, MD, PhD, Paul Pepe, MD, MPH, Roy Reyna, MD, MPH, Richard Smalling,
MD, PhD, Janet Williams, MPH, Guillermo Chapa, Rafael Cardenas, PhD, Marilyn Fielding, MS, RN, Alfonso Gutierrez, MS,
and Suzanne Norwood, MA; King County Department of Emergency Medical Services and University of Washington Medi-
cal Center, Seattle, and Oregon Health Sciences University, Portland: Mickey Eisenberg, MD, PhD (principal investiga-
tor), Jerris Hedges, MD (coprincipal investigator), Hendrika Meischke, PhD (coprincipal investigator), Sheri Schaeffer (project
coordinator), Mohamed Daya, MD, Eric Dulberg, PhD, Dan Henwood, MA, Jon Jui, MD, Denise Macko, N. Clay Mann, PhD,
Colleen Gillespie, Gene Liddell, MA, Jean Lum, and Laura Philips, MSW; New England Research Institutes, Watertown,
Mass: Voula Osganian, MD (principal investigator), Henry Feldman, PhD (coprincipal investigator), Sarah McGraw, PhD
(coprincipal investigator), Deborah Sellers, PhD (project director), Kathy Schulman, MS (project director), Cheryl Caswell,
MBA, Norina Coppinger, Maggie Cotter, Vanessa Hosein, Paul Mitchell, MS, Kevin Smith, MA, Foss Tighe, and Dinh Tran;
and National Heart, Lung, and Blood Institute, Bethesda, Md: Denise Simons-Morton, MD, PhD (project officer), Lawton
Cooper, MD (deputy project officer), Scott Allender, MD, John Bradley, Mary Hand, MSPH, RN, Sandy Kamisar, Terry Long,
Michael Proschan, PhD, Elaine Stone, and Mario Stylianou, MS.
the specific nonexclusive subcategories as follows: pain,
disease were consistently more knowledgeable regard-
81.9%; pressure, 6.4%; tightness, 6.2%; and discomfort,
ing correct heart attack symptoms, but also tended to
6.2%. Approximately two thirds of respondents re-
report incorrect symptoms more often.
ported that arm pain and/or numbness was a heart at-
A weighted multivariable linear regression analysis pro-
tack symptom, and one half reported that shortness of
vided estimates of the adjusted subgroup differences in mean
breath was. The remaining 8 correct heart attack symp-
number of correct symptoms reported (Table 6). Statis-
toms were reported by fewer than one quarter of respon-
tically significant differences were observed between sev-
dents; 4 of the correct symptoms were reported by fewer
eral subgroups. These differences largely confirmed the re-
than 10.0% of respondents. None of the incorrect symp-
sults of the unadjusted analyses described previously. No
toms was reported by as many as 10.0% of respondents.
substantial difference was observed between women and
Subgroup differences in knowledge of symptoms
men. Hispanic and Asian groups demonstrated substan-
were observed (Tables 3-5). Differences between women
tially less knowledge than non-Hispanic whites. There was
and men were modest in magnitude and somewhat
a trend toward greater knowledge in the middle-aged group
inconsistent in direction. Differences between the racial
relative to younger and older respondents. The lower-
and ethnic groups were more obvious and more consis-
income groups showed less knowledge than the higher-
tent. Non-Hispanic whites showed consistently greater
income groups. After adjustment for income, educational
knowledge of correct symptoms. Furthermore, non-
attainment was not associated with heart attack symptom
Hispanic whites tended to report the incorrect symp-
knowledge. The presence of risk factors for heart disease
toms at a similar or lower frequency relative to other
did not identify subgroups with greater knowledge. In fact,
racial or ethnic groups. The youngest group (aged
persons with diabetes mellitus showed a trend toward less
18-34 years) showed generally less knowledge than the
knowledge than those without this condition. Previous ex-
older groups. The higher socioeconomic groups showed
perience with heart disease in the respondents, their fam-
greater knowledge of correct symptoms than the lower
ily, or their close friends was associated with greater knowl-
socioeconomic groups. The group with missing data on
edge of heart attack symptoms. In summary, this analysis
income most closely resembled the lowest-income
provided evidence of subgroup differences in knowledge
group. There were no large differences in the reporting
of the complex constellation of heart attack symptoms.
of incorrect symptoms across socioeconomic groups.
The results of a weighted, multivariable-adjusted
Differences in knowledge between individuals with and
logistic regression analysis of factors associated with the
without risk factors for coronary heart disease were less
reporting of chest pain as the most important heart
consistent. Of the risk factors examined, only persons
attack symptom are shown in Table 7. Only 1
with high levels of serum cholesterol showed consis-
adjusted odds ratio had a 95% confidence interval that
tently greater knowledge than those without this condi-
excluded 1.0; persons with household income of less
tion. Individuals with previous experience with heart
than $25 000 per year were approximately 50% more
ARCH INTERN MED/ VOL 158, NOV 23, 1998
2332
on August 14, 2009
www.archinternmed.com
Downloaded from
©1998 American Medical Association. All rights reserved.

---

ducted in 302 members of the general population in Dub-
Table 1. Prevalence of Sociodemographic Characteristics
lin, Ireland, 67.5% of respondents reported that chest pain
and Self-Reported Coronary Heart Disease Risk Factors*
was an indicator of heart attack.25 In a 1995 report of a
survey of 65 patients with heart attack in Chicago, Ill,
No. of
Prevalence, %
71% expected chest symptoms of an unspecified nature,
Respondents
Characteristic
(N = 1294)
REACT
US
and 75% expected pain of unspecified location. Only 43%
expected any other symptom in association with pain.12
Sex
Women
735
56.8
52.2
In a survey of 1315 Mexican Americans and non-
Men
547
42.3
47.8
Hispanic whites in the general population in San Anto-
Refused or missing
12
0.9
. . .
nio, Tex, from 1979 to 1981, knowledge of heart attack
Race or ethnicity
symptoms differed by ethnicity and socioeconomic sta-
Hispanic†
145
11.2
7.7
tus, ranging from 39% to 82% for chest pain, from 26%
Non-Hispanic black
106
8.2
10.8
to 51% for shortness of breath, and from 4% to 32% for
Non-Hispanic white
970
75.0
77.5
Native American
8
0.6
0.7
the combination of sweating and nausea.13 Compared with
Asian or Pacific Islander
31
2.4
2.8
these earlier surveys, the level of knowledge of heart at-
Other
34
2.6
0.3
tack symptoms represented in our survey is clearly bet-
Age, y
ter, with nearly complete recognition of chest pain and
18-34
464
35.9
37.7
substantial recognition of arm pain and/or numbness and
35-54
516
39.9
34.1
shortness of breath as heart attack symptoms. The greater
?55
289
22.3
28.2
knowledge shown in our survey compared with previ-
Refused or missing
25
1.9
. . .
Education
ous surveys may be related, in part, to the characteris-
Up to high school
473
36.6
54.6
tics of the populations studied or to other methodologi-
Beyond high school
805
62.2
45.4
cal differences. In contrast to previous surveys, the REACT
Refused or missing
16
1.2
. . .
population is fairly representative of the general US
Annual household income, $
population, with underrepresentation of the lowest
?25 000
382
29.5
41.9
socioeconomic group introduced by the choice of a
25 000-54 999
480
37.1
38.1
random-digit dialed telephone survey design. The
?55 000
287
22.2
20.0
Refused, do not know, or
populations surveyed in Dublin and Chicago were of
missing
145
11.2
. . .
unspecified social class; the population surveyed in San
Coronary heart disease risk factors
Antonio included substantial socioeconomic diversity
Diabetes
61
4.7
3.4
and was biased toward lower socioeconomic status. A
High blood pressure
277
21.4
24.2
temporal improvement in general knowledge of heart
High serum cholesterol level
270
20.9
20.0
Current cigarette smoking
311
24.0
25.6
attack symptoms is an alternative explanation for the
Previous experience with
greater knowledge seen in REACT than in these earlier
heart disease
studies.
Personal
166
12.8
7.3
Chest pain was identified as a heart attack symp-
Family or spouse
430
33.2
NA
tom by 89.7% of respondents and as the most important
Other relative or close friend
774
59.8
NA
heart attack symptom by more than half. Given this ap-
parently high level of knowledge regarding chest pain as
*REACT indicates Rapid Early Action for Coronary Treatment trial; NA, not
available. Prevalences have been rounded and may not sum 100. Ellipses
a heart attack symptom, why do individuals with chest
indicate not applicable.
pain or discomfort delay seeking care for a median of 2.5
†May be of any race.
to 3.5 hours?26 Our survey is limited by lack of detailed
information regarding the expected quality of the chest
likely than persons with household incomes of at least
pain; however, of the 4 categories coded, a nonspecific
$55 000 per year to report chest pain as the most
response of pain was much more common than more spe-
important heart attack symptom. This finding contrasts
cific descriptions of pressure, tightness, or heaviness. It
with the results of the first multivariable analysis that
may be that the quality of the chest discomfort experi-
showed less knowledge of the complex constellation of
enced by patients with heart attack fails to meet their ex-
heart attack symptoms in this same lower-income sub-
pectations, expectations that may often be based on dra-
group. In summary, this latter analysis provided little
matic portrayals of such persons in popular media. In
evidence of subgroup differences in knowledge of chest
addition, fewer than one quarter of respondents demon-
pain as an important heart attack symptom. These find-
strated knowledge of several other important symp-
ings suggest that the subgroup differences described in
toms, including sweating, nausea, and vomiting, thereby
Table 6 are related primarily to differences in knowl-
indicating that knowledge of the often complex constel-
edge of heart attack symptoms other than chest pain.
lation of heart attack symptoms is deficient in the US
population. Other possible explanations of prolonged de-
COMMENT
lay include denial, self-treatment, other maladaptive cop-
ing strategies, fear, concerns about costs, and embarrass-
The level of knowledge of heart attack symptoms exhib-
ment about being wrong despite awareness that the
ited in our survey compares favorably with that of the
symptoms might be cardiac in origin.7-10
few published reports from other studies. In a 1992 re-
Subgroup differences in knowledge of heart attack
port from a survey regarding heart attack recognition con-
symptoms have been reported once previously.13 Knowl-
ARCH INTERN MED/ VOL 158, NOV 23, 1998
2333
on August 14, 2009
www.archinternmed.com
Downloaded from
©1998 American Medical Association. All rights reserved.

---

Table 2. Total and Correct Symptoms Reported as Characteristic of a Heart Attack and Percentage Responding Chest Pain
as the Most Important Heart Attack Symptom by Sociodemographic and Risk Factor Status*
Mean (95% CI)
No. of
No. of Correct
Chest Pain as Most Important
Characteristic
Symptoms Reported
Symptoms Reported
Symptom, % (95% CI)
Total population
3.25 (3.16-3.34)
2.90 (2.74-3.07)
56.6 (53.9-59.3)
Sex
Women
3.30 (3.19-3.42)
2.95 (2.85-3.06)
57.6 (54.1-61.1)
Men
3.25 (3.11-3.39)
2.89 (2.77-3.01)
55.9 (51.6-60.2)
Race or ethnicity
Hispanic†
2.90 (2.57-3.23)
2.46 (2.20-2.72)
60.2 (51.8-68.6)
Non-Hispanic black
2.64 (2.34-2.94)
2.30 (2.07-2.53)
56.6 (46.6-66.6)
Non-Hispanic white
3.41 (3.31-3.51)
3.08 (2.99-3.17)
56.3 (53.2-59.4)
Asian or Pacific Islander
2.99 (2.26-3.72)
2.36 (1.79-2.91)
55.1 (36.3-73.9)
Age, y
18-34
2.95 (2.80-3.11)
2.62 (2.50-2.75)
55.8 (51.0-60.6)
35-54
3.55 (3.40-3.69)
3.17 (3.05-3.30)
53.7 (49.4-58.0)
?55
3.30 (3.13-3.46)
2.95 (2.79-3.11)
62.7 (57.5-67.9)
Education
Up to high school
3.00 (2.86-3.14)
2.72 (2.60-2.85)
58.4 (53.8-63.0)
Beyond high school
3.45 (3.34-3.56)
3.05 (2.94-3.15)
55.5 (52.1-58.9)
Annual household income, $
?25 000
2.95 (2.79-3.10)
2.55 (2.42-2.68)
61.8 (57.0-66.6)
25 000-54 999
3.31 (3.18-3.45)
3.01 (2.89-3.13)
56.1 (51.7-60.5)
?55 000
3.67 (3.45-3.88)
3.28 (3.09-3.47)
49.0 (42.9-55.1)
Not reported
3.06 (2.77-3.35)
2.75 (2.50-3.02)
59.8 (52.1-67.5)
Diabetes
Yes
2.84 (2.51-3.18)
2.52 (2.03-2.83)
69.2 (58.1-80.3)
No
3.30 (3.21-3.39)
2.94 (2.86-3.03)
56.0 (53.2-58.8)
High blood pressure
Yes
3.30 (3.11-3.48)
2.97 (2.80-3.14)
60.9 (55.4-66.4)
No
3.27 (3.17-3.37)
2.91 (2.82-3.00)
55.4 (52.3-58.5)
High serum cholesterol level
Yes
3.61 (3.41-3.80)
3.21 (3.04-3.39)
54.2 (48.5-59.9)
No
3.18 (3.08-3.29)
2.84 (2.75-2.93)
57.2 (54.1-60.3)
Current cigarette smoking
Yes
3.10 (2.93-3.26)
2.83 (2.68-2.97)
54.8 (49.3-60.3)
No
3.33 (3.22-3.43)
2.95 (2.86-3.05)
56.7 (53.5-59.9)
Personal history of heart disease
Yes
3.73 (3.51-3.96)
3.24 (3.03-3.45)
61.8 (55.1-68.5)
No
3.21 (3.11-3.31)
2.88 (2.79-2.96)
55.8 (52.8-58.8)
Heart attack in family or spouse
Yes
3.68 (3.53-3.83)
3.25 (3.11-3.38)
55.6 (51.0-60.2)
No
3.07 (2.96-3.18)
2.75 (2.66-2.85)
57.4 (54.0-60.8)
Heart attack in other relative or close friend
Yes
3.52 (3.40-3.63)
3.12 (3.02-3.23)
57.4 (53.9-60.9)
No
2.93 (2.79-3.06)
2.62 (2.50-2.74)
55.0 (50.6-59.4)
*CI indicates confidence interval.
†May be of any race.
edge of chest pain, nausea, and sweating as heart attack
Socioeconomic status also predicted knowledge, with less
symptoms was greater in non-Hispanic whites than in Mexi-
knowledge seen in the lower-income and lower educa-
can Americans, independent of socioeconomic status. No
tional attainment groups. In our study, income was more
gender-related differences in knowledge were seen; how-
strongly associated with heart attack symptom knowl-
ever, there was a strong, positive association between
edge than was educational attainment. Educational attain-
socioeconomic status and knowledge of heart attack symp-
ment was associated with knowledge in models that did
toms.13 Our findings with respect to gender, race or eth-
not include income (results not shown). The race and eth-
nicity, and socioeconomic status are similar and extend
nic findings probably represent the result of cultural in-
those findings to other race or ethnic groups. Knowledge
fluences and other barriers to access to information about
of heart attack symptoms was highest in non-Hispanic
health in general and heart attack symptoms in particular
whites. Hispanics, non-Hispanic blacks, and Asian Ameri-
that compound the effects of lower socioeconomic status
cans and Pacific Islanders showed less knowledge, al-
or that are not reflected in our measures of socioeconomic
though the difference for non-Hispanic blacks disap-
status. Together, these findings underscore the primary im-
peared following adjustment for socioeconomic status.
portance of designing interventions that can reach the so-
ARCH INTERN MED/ VOL 158, NOV 23, 1998
2334
on August 14, 2009
www.archinternmed.com
Downloaded from
©1998 American Medical Association. All rights reserved.

---

Table 3. Prevalence of Reported Signs or Symptoms of Heart Attack by Demographic Characteristics*
Demographic Characteristics of Respondents, %
Asian or
Age, y
Non-Hispanic Non-Hispanic Pacific
Total
Women
Men
Hispanics†
Blacks
Whites
Islander
18-34
35-54
?55
Symptom
(N = 1294) (n = 735)
(n = 547)
(n = 145)
(n = 970)
(n = 106)
(n = 31) (n = 464)
(n = 516)
(n = 289)
Correct
Chest pain or
89.7
90.2
90.4
87.0
89.5
91.4
76.6
90.9
90.7
88.6
discomfort
Arm pain or numbness
67.3
71.0
63.6
51.7
47.0
73.2
32.4
55.4
77.9
69.8
Shortness of breath
50.8
49.0
54.3
47.7
45.1
52.4
51.8
52.9
53.5
43.9
Sweating
21.3
18.9
25.0
10.6
12.3
24.9
8.6
16.0
24.9
24.1
Nausea or vomiting
14.5
17.0
11.5
2.5
5.6
17.6
12.4
7.5
18.9
17.9
Dizziness or
12.1
10.2
14.8
7.2
4.3
12.5
36.2
13.9
12.6
9.3
light headedness
Weakness, fatigue,
10.8
10.3
11.7
11.7
7.5
11.3
12.8
11.7
11.4
9.0
or malaise
Jaw or neck pain
9.9
13.4
5.3
11.1
6.7
10.7
0.0
3.9
11.6
16.3
Palpitations
7.1
7.4
6.7
8.4
11.5
6.6
5.0
6.4
8.2
5.8
Back pain
6.3
6.7
5.9
6.4
0.7
7.1
0.0
3.4
6.5
10.1
Sense of impending
0.6
1.0
0.0
1.3
0.0
0.6
0.0
0.3
1.1
0.0
doom
Other
Heartburn
8.0
10.4
4.9
9.9
6.1
8.4
0.0
4.9
9.1
11.7
Loss of consciousness
6.4
7.1
5.5
3.6
3.2
6.6
24.7
6.1
7.2
5.8
Change in color
5.5
3.9
7.6
0.9
1.5
6.0
17.4
6.2
5.1
4.4
Other
14.7
13.2
16.9
28.6
22.6
11.5
21.2
15.3
15.6
11.7
*Data are reported in percentages. For some subgroup comparisons, the sum of the groups dose not equal the total sample size because of missing data
†May be of any race.
Table 4. Prevalence of Reported Signs or Symptoms of Heart Attack by Socioeconomic Status*
Education
Annual Household Income, $
Up to High School
Beyond High School
?25 000
25 000-54 999
?55 000
Not Reported
Symptom
(n = 473)
(n = 805)
(n = 382)
(n = 480)
(n = 287)
(n = 145)
Correct
Chest pain or discomfort
89.6
90.8
88.3
92.9
88.7
84.6
Arm pain or numbness
65.2
69.5
51.0
76.4
76.3
62.8
Shortness of breath
52.3
50.8
48.4
52.6
55.6
41.6
Sweating
15.0
25.1
12.8
20.5
32.2
24.6
Nausea or vomiting
10.8
17.0
9.3
15.4
20.9
12.4
Dizziness or light headedness
8.1
14.5
11.3
12.7
13.5
8.9
Weakness, fatigue, or malaise
9.2
12.0
8.7
9.2
16.0
11.7
Jaw or neck pain
9.0
10.4
10.1
9.1
10.4
11.0
Palpitations
6.6
7.4
9.8
4.7
7.6
6.5
Back pain
6.1
6.5
4.1
7.1
6.5
8.9
Sense of impending doom
0.4
0.6
1.2
0.4
0.2
0.0
Other
Heartburn
7.0
8.8
7.3
8.3
7.4
10.2
Loss of consciousness
5.7
6.9
7.9
6.4
6.3
2.4
Change in color
3.2
6.9
5.9
5.3
5.3
5.1
Other
10.8
17.2
17.5
10.1
18.7
14.3
*Data are reported in percentages.
cioeconomically disadvantaged and racial and ethnic mi-
persons. The observation of greater knowledge in the
nority groups with health and urgent care information in
middle-aged than the youngest group was consistent with
attempts to improve the public’s health and response to car-
our expectation. However, the observation of less knowl-
diac emergencies.
edge in the oldest group relative to the middle-aged group
The pattern of age-group differences in knowledge
was not. The basis for this finding is not obvious, but may
of heart attack symptoms that we observed was some-
include differences in education and access to informa-
what surprising. We had anticipated greater knowledge
tion about health that are not reflected in our unrefined
in older than younger individuals related to the greater
measures of socioeconomic status. These issues deserve
importance of heart disease as a health problem in older
further attention.
ARCH INTERN MED/ VOL 158, NOV 23, 1998
2335
on August 14, 2009
www.archinternmed.com
Downloaded from
©1998 American Medical Association. All&nbs