Tài liệu tâm lý học sức khỏe: Các hành vi nguy cơ sức khỏe

44
Subjective Risk and Health Protective
Behavior:
Cancer Screening and Cancer Prevention
Leona S. Aiken
Mary A. Gerend
Arizona State University
Kristina M. Jackson
University of Missouri
This chapter explores the role of perceived risk in health protective behavior.
Cancer serves as the context of the presentation; the discussion employs
the literature on cancer screening and prevention to highlight theory and
findings on the perception of risk in relation to health behavior. The origins
of perceived risk, its role in health behavior models, and the linkages
between perceived risk and behavior are explored. In models of health
behavior, perceived risk for disease is the motivational engine for health
protective action. This chapter is intended to serve two purposes: to provide
both a broad picture of the literature on risk perception in health psychology
and to characterize research on perceived risk for cancer as a putative
determinant of cancer screening and preventive behavior.

CHAPTER OVERVIEW
The chapter first addresses the role of perceived susceptibility in models of
health behavior. It then turns to perceived risk as a construct, its
measurement, its observed relation to objective risk for cancer, and its
determinants. It next explores the relation of perceived susceptibility and
cancer distress to cancer screening and cancer preventive behavior. Here it
considers not only the susceptibility-behavior link, but also explores other
variables that may moderate or even mitigate the impact of perceived
susceptibility on specific cancer protective behaviors. The chapter then
considers interventions to increase screening and preventive behavior that

involve the perceived susceptibility construct. The emphasis is on the use of
mediational analysis to assess the direct and indirect impact of perceived
susceptibility on screening and preventive behavior. Finally, it explores a
number of issues that arise in consideration of how perceived susceptibility
impacts health protective behavior.

CANCER PREVALENCE
Cancer is a feared disease of high prevalence. By age 59, over 8% of men
and 9% of women will have developed an invasive cancer; from birth to
death, these percentages rise to 47% of men and 38% of women (Landis,
Murray, Bolden, & Wingo, 1998). Cancer is the second leading cause of
death (23% of all deaths) behind heart disease (32%) among adults in the
United States. In all, 1.23 million new cases of cancer and over 564,800
cancer deaths are expected in the United States in 1998 (Landis et al.,
1998).


SCREENING AND PREVENTIVE
RECOMMENDATIONS
The public is inundated with information about cancer and with
recommendations for cancer screening and prevention.
-727As of 1998, the American Cancer Society recommended extensive cancer
screening (American Cancer Society, 1998). Screening recommendations
include an annual mammogram for women age 40 and older; colon and
rectal screening with fecal occult blood test (FOBT) annually, plus flexible
sigmoidoscopy every 5 years for men and women over age 50; annual
prostate-specific antigen (PSA) blood test and digital rectal examination
(DRE) for men age 50 and older; annual pelvic examinations for all women
age 18 and older, with annual Pap tests until at least three negative Pap
tests have been achieved, and then less frequent Pap tests; and endometrial

screening for women at high risk for uterine cancer (American Cancer
Society, 1998). Regular self-examinations are also recommended, including
skin self-examination (American Academy of Dermatology, 1994; National
Cancer Institute, 1995), breast self-examination (BSE) for all women
beginning at age 20 (American Cancer Society, 1998), and testicular selfexamination for men (National Cancer Institute, 1992). (See also the
screening recommendations of the U.S. Preventive Services Task Force,
1996.) Beyond screening are recommendations for cancer prevention
through lifestyle modification, including skin protection (American Cancer
Society, 1997b) and diet (American Cancer Society, 1996).

SCREENING UTILIZATION IN THE
UNITED STATES
According to the National Health Interview Surveys of 1987, 1992, and 1994,
a population-based national survey of 40,000 households (American Cancer
Society, 1997a; National Center for Health Statistics, 1996), the percentage
of women age 50 and over who have had a mammogram in the past 2 years
rose from 25% to 56% between 1987 and 1994. These rates were similar for
Black and Hispanic women (56% and 50%, respectively, in 1994, up from
19% and 18%, respectively, in 1987), although rates lagged for low income
women (38% in 1994) and those with less than a high school education
(42%). Rates for Pap test utilization (within the past 3 years) achieved 77%
in 1994 (74% for Hispanic women), again lagging behind for low education
women (62%). As of 1992, about a third of the population had had one of
three colorectal screening tests, DRE within the past year, FOBT within the
past 2 years, or a sigmoidoscopy at least once.

OPENLY AIRED CANCER DEBATES
Epidemiological findings make the news, and the public hears a relentless
array, often contradictory, of associations between behaviors and cancer
(Taubes, 1995). Medical debates about the efficacy of screening for mortality

reduction are publicly aired (Aiken, Jackson, & Lapin, 1998). The debate on
the efficacy of mammography screening for women under age 50 raged in
the public media for most of this decade (Aiken et al., 1998). Prostate
screening currently is occasioning considerable debate, along the lines of
the previous mammography debate (Albertsen, 1996; E. S. Wolfe & W. W.
Wolfe, 1997). Increasingly, laypersons are asked by their physicians to


decide whether they wish to be screened, with the argument that patient
choice must be preserved (Woolf & Lawrence, 1997). Issues concerning
appropriate screening are aired against a backdrop of economic constraints
posed by the health care industry.

CLASSES OF CANCER PROTECTIVE
BEHAVIORS
For the exploration of perceived risk and behavior, cancer protective
behaviors must be divided into two broad categories- screening for early
detection versus prevention. Screening behaviors may further be divided
into those that are medically based (e.g., mammography) versus those that
involve self- examination (e.g., BSE). This distinction is important because
the barriers to screening are expected to be very different for the two
categories. These barriers may interact with or obscure the role of perceived
risk. A similar argument can be made for specific preventive behaviors.
Although the discussion draws on literature on a variety of cancers,
medically based screening is exemplified with mammography, self-screening
with BSE, and preventive behavior with skin protection.

PERCEIVED RISK IN MODELS OF HEALTH
BEHAVIOR
This section is devoted to an overview of the role of perceived risk in models

of health protective behavior. Health psychology is rich in models of the
putative determinants of health protective behavior. At the core of
essentially all these models is the concept of perceived risk-the extent to
which individuals believe that they are subject to a health threat (Becker,
1990; Gerrard, Gibbons, & Bushman, 1996; Kowalewski, Henson, &
Longshore, 1997; van der Pligt, 1998; Weinstein, 1993).
Health psychology draws on a theoretically based literature in risk
perception and its determinants (Kahneman & Tversky, 1973; Kasperson et
al., 1988; Slavic, 1987; Tversky & Kahneman, 1974). Formal models of risk
(Kasperson et al., 1988) postulate that risk is a joint function of the
probability of occurrence of a negative event, and the magnitude of its
consequences; risk is the product of these factors.
Literature applying perceptions of risk to health behavior is less precise. The
term perceived risk, as well as the terms perceived susceptibility and
perceived vulnerability are used interchangeably for measures of the
subjective likelihood of contracting a disease, absent any consideration of
severity. Consistent with application in health, the terms perceived risk, used
here to refer to subjective estimates of the likelihood of personally
contracting a disease, and not the combination of likelihood and
consequences. Perceived severity is used here to refer to perceptions of
disease consequences independent of likelihood.
According to Weinstein (1993), models of health behavior assume that the
motivation for health protective behavior stems from anticipation of some
negative health outcome coupled with hope of avoiding the outcome.
Anticipation of a negative outcome involves foremost the perception that
one is


-728personally susceptible to some disease; for strong health motivation to be
achieved, this perception must be coupled with the anticipation that the

disease consequences are severe (Weinstein, 1993).
Our particular interest in this chapter is the linkage of perceived
susceptibility to health protective behavior. A theoretical context for this
linkage is provided by consideration of the way in which perceived
susceptibility is used in models of health protective behavior. Three widely
applicable models of health behavior, the health belief model (HBM; Becker
$I Maiman, 1975; Rosenstock, 1966; 1974a, 1974b, 1990), protection
motivation theory (PMT; Prentice-Dunn & Rogers, 1986; Rogers, 1975,1983),
and the precaution adoption process model (PAPM; Weinstein, 1988) employ
the perceived susceptibility construct as a driving force in health protective
behavior. Perceived risk appears as well in the transtheoretical model of
change (TTM; Prochaska, DiClemente, & Norcross, 1992), and the recently
proposed cognitive-social health information processing (C-SHIP) model (S.
M. Miller, Shoda, & Hurley, 1996). Perceived risk is also implicit in the theory
of reasoned action (TRA; Ajzen & Fishbein, 1980; Fishbein & Ajzen, 1975) and
its extension, the theory of planned behavior (Ajzen, 1991), as well as
subjective expected utility theory (Ronis, 1992; Weinstein, 1993), as they are
applied to health behavior.
Although perceived susceptibility is consistently cast as the motivating
engine for health protective behavior, the specific role of perceived
susceptibility and assumptions about how it combines with other constructs
vary in informative ways across models. A brief characterization of the role
of the perceived susceptibility construct in several well-established models
of health behavior and the newer C-SHIP model is provided. A
characterization of the complete health models is beyond the scope of this
chapter; Conner and Norman (1996); Glanz, Lewis, and Rimer (1990);
Weinstein (1993); and Weinstein, Rothman, and Sutton (1998) provided
explications of these and other models; Conner and Norman (1996) provided
extensive reviews of literature employing these models as well. Curry and
Emmons (1994) provided a thorough summary of applications of the HBM,

TRA, and the TTM to breast cancer screening.

Perceived Susceptibility as
Motivator: Health Belief Model
The health belief model (HBM) traces its origins to problems encountered in
the U.S. Public Health Service nearly half a century ago-problems of failures
of usymptomatic individuals to undergo screening tests or to engage in
preventive health behaviors (Rosenstock, 1966, 1974a; 1990). Ironically, the
health belief model is still being applied to the same issues, which are
abundant in the area of cancer prevention and early detection. The HBM
states that individuals will undertake a health action to the extent that they
believe themselves to be susceptible to a health threat
(perceivedsusceptibilio), believe that the consequences of the disease are
serious berceived severity or seriousness), believe that the proposed health
action will offer protection against the health threat (perceived benefits),
and believe that barriers to performing the health action can be overcome
(perceived barriers). Finally, individuals must receive some trigger, or cue, in
order to act (cue to action). Interestingly, in current interventions to increase
cancer screening, a reminder letter (a cue to action) is a common
component (e.g., Bastani, Marcus, Maxwell, Das, & Y an, 1994). Physicians'
recommendations for screening have been conceptualized as a cue to action


as well (Fox, Siu, & Stein, 1994).
Perceived susceptibility is, in a sense, the centerpiece of the HBM. There are
two aspects to perceived susceptibility: the individuals' belief that
contracting a disease is a realistic possibility for themselves, and the
individuals' belief that they may have the disease in the complete absence
of symptoms (Rosenstock, 1990). Failure to utilize cancer screening tests
may be attributed to a lack of belief that pathology can exist in the absence

of symptoms (Rosenstock, 1990). Perceived susceptibility and perceived
severity combine to form perceived threat, a determinant of the likelihood of
adopting a health action; this combination closely reflects the formal
definition of risk (Kasperson et al., 1988) provided earlier. The HBM is silent
on the nature of the combinatorial rules for the constructs; in most
applications of the health belief model, simple additive effects of the
constructs have been explored., The interplay of perceived threat with
perceived benefits is important for cancer screening, in that high risk
individuals, although they perceive heightened vulnerability, may avoid
seeking screening if they believe that cancer treatment cannot save them
(e.g., Lerman & M. D. Schwartz, 1993, for breast cancer; M. D. Schwartz,
Lerman, Daly, et al., 1995, for ovarian cancer). Ronis (1992) suggested a
combinatorial rule for HBM constructs in which perceived susceptibility and
severity are necessary precursors to the perception of benefits of health
action, a characterization on which we draw in our later discussion of
interventions.
The health belief model has made sustained contributions as a heuristic for
the study of psychosocial correlates of preventive health behavior. (See
reviews by Harrison, Mullen, & Green, 1992; Janz & Becker, 1984; and
Sheeran & Abraham, 1996.) Typically, the perceived barriers construct has
been the strongest correlate of lack of protective behavior, whereas
perceived susceptibility has typically exhibited low to moderate positive
correlations with protective behavior.
Whereas perceived susceptibility is expected to combine with perceived
severity to motivate health protective behavior, perceived severity by itself
rarely correlates with preventive behavior or screening behavior (Janz 8z
Becker, 1984; Harrison et al., 1992). This is certainly true for cancer
research: Perceived severity has failed to show predictive utility and has not
been amenable to change via intervention, as cancer apparently is seen as
uniformly serious (Champion, 1994; Curry & Emmons, 1994; Rimer, 1990;

but see Ronis & Harel, 1989, for an exception). Researchers often forgo the
measurement of perceived severity in characterizing the HBM for cancerrelated behavior (e.g., Hyman, Baker, Ephraim, Moadel, & Philip, 1994;
Vernon, Myers, & Tilley, 1997). Thus, perceived susceptibility by itself, rather
than the combination of susceptibility
-729and severity, is de facto characterized as the motivating force for cancer
protective behavior.

Fear Amusing Comnmn;cat;on,
Percehed Susceptibiky, and
Behavior: Protection Motivation
Theory


Perceived susceptibility also plays a central role as a motivator of health
protective behavior in PMT, a model that arose from consideration of the
impact of fear arousing communication on the adoption of health protective
behavior (Beck & Frankel, 198 1; Rogers, 1975). As in the HBM, perceptions
of susceptibility and severity that resulted from fear communications were
expected to combine with perceptions of the existence of an effective health
protective behavior to arouse protection mot&- tion, which in turn led to
intentions to adopt the protective health behavior (Rogers, 1975). In the
revised form of PMT (Rogers, 1983; Prentice-Dunn & Rogers, 1986), a special
motivating role for perceived susceptibility coupled with perceived severity
was provided, that of lowering the probability of a maladaptive response
(e.g., delay in seeking treatment for suspected cancer symptoms,
persistence in behaviors that put one at increased cancer risk). Rippetoe and
Rogers (1987) applied PMT to an experimental investigation of breast selfexamination.

How Perceptions of Susceptibility
Accrue: The Precaution Adoption

Process Model
Models of health behavior assume that in order for perceived susceptibility
to act as a motivational force, perceptions of susceptibility must be personal
(i.e., individuals must feel that they, themselves, are vulnerable). Weinstein
(1988) proposed the precaution adoption process (PAPM) as a stage model
of the adoption of health behavior. In general, stage models (Weinstein et
al., 1998) characterize individuals as falling into a series of ordered
categories with regard to adoption of a health behavior. In PAPM, these
stages move from lack of awareness of the health issue (Stage I) through
health behavior maintenance (Stage 7). Consistent with this stage structure,
beliefs about perceived susceptibility are assumed to develop in a series of
cumulative stages. First, individuals are assumed to become aware of a
health hazard (awareness), then to believe in the likelihood of the hazard for
others (general susceptibility), and finally to acknowledge their own personal
vulnerability (personal susceptibility). Personal susceptibility is assumed to
be critical in the decision to take precautionary action (Weinstein et al.,
1998). Assessment of discrepancies between general susceptibility versus
personal susceptibility has uncovered optimistic biases (Weinstein, 1980)
about people's vulnerability; these biases are discussed later. The PAPM has
been applied to home testing for radon gas, which is an environmental
cancer threat (Weinstein & Sandman, 1992).

The Growth of Perceived Suscept&&y
and the Process of Adopting Health
Behaviors: Transtheoretical Model of
Change
The TTM, as the PAPM, is a stage model of health behavior adoption
(Prochaska et al., 1992). Progress through the first two stages is
hypothesized to be driven by the growth of awareness of perceived
susceptibility from precontemplation, of vulnerability to a health threat, to

contemplation, in which there is an awareness of one's own vulnerability to a
health threat but no commitment to health action. Although the
transtheoretical model was initially applied to smoking cessation as a cancer


preventive action, the model has now been applied to cancer screening as
well. For example, Lipkus, Rimer, and Strigo (1996), Rakowski et al. (1992),
and Siegler, Feaganes, and Rimer (1995) applied this model to
mammography screening.

Cognitive-Social Health Information
Processing
The C-SHIP model (S. M. Miller et al., 1996) is a comprehensive model of the
genesis and maintenance of health-protective behavior, initially expounded
in the context of the complex sustained behavior of breast self-examination
(BSE). The model considers five classes of determinants of health behavior
that incorporate both cognitions and affect. Among these, two classes
address issues of perceived susceptibility: health-relevant encodings,
including health risks and vulnerabilities, plus attentional strategies for
gathering versus avoiding health relevant information; and health beliefs
and expectancies, including how vulnerabilities, such as genetic
predisposition, impact subjective likelihood of disease development. The
model specifies how information about objective risk and resultant
perceptions of susceptibility interact with emotions associated with receiving
health information, with health goals, and with self-regulation in producing
health behaviors. That the model addresses the interplay of emotion with
cognitions about people's vulnerability is important for an understanding of
cancer screening behavior among high risk individuals; this issue is
discussed later.

Perceived Susceptibility as a
Predisposing Factor in Complex Hybrid
Models of Health Behavior Adoption
Within this decade, a number of authors have proposed extensive
integrative frameworks of the putative determinants of health protective
behavior, which have been employed in the design of interventions to
increase health behavior. Four such frameworks are summarized in Curry
and Emmons (1994). Each framework specifies a complex causal chain of
variables that ultimately leads to health behavior. Most important for our
consideration is the fact that perceived susceptibility is included as a
predisposing factor for health behavior adoption early in the causal chain, a
factor that may facilitate overcoming barriers to the health protective
behavior (McBride, Curry, Taplin, Anderman, & Grothaus, 1993) and lead to
receptiveness to health promotion interventions (the PRECEDE-PROCEED
model of Green & Kreuter, 199 1).
-730-

Summary: Perceived Susceptibility as a
Predisposing Factor for Health Behavior
Adoption
Not surprisingly, models of health behavior have matured and increased in complexity. Early
models have been augmented with new variables, for example, the addition of self-efficacy for
health behavior to both the HBM (Rosenstock, Strecher, & Becker, 1988) and PMT (Rogers,


1983). New stage models have viewed health behavior adoption as dynamic, in part driven by
perceived susceptibility. The interplay of susceptibility cognitions with emotion has been
elucidated. Hybrid models have incorporated a complex network of environmental and medical
system variables along with individual cognitions, including perceived susceptibility. The
evolution of these models has clarified the role of perceived susceptibility as a potentially

powerful predisposing factor at the outset of the process of adoption of health behaviors, a
factor that motivates this process of adoption. Drawing on this evolution of health behavior
models, we conceptualize perceived susceptibility to disease as a distal construct in a
mediational chain of constructs that eventuates in protective health behavior.

PERCEIVED SUSCEPTIBILITY: MEASUREMENT AND
DETERMINANTS
This section first considers approaches to the measurement of perceived susceptibility. It then
reviews comparisons of objective risk for cancer with subjective risk, raising the question of
whether individuals overestimate or underestimate their cancer risk relative to objective risk.
Finally, it explores the putative determinants of perceived susceptibility, drawing on both a
broad literature on determinants of risk, and a cancer-specific literature.

Measurement of Perceived Susceptibility
Alternative approaches to the measurement of perceived susceptibility lead to varying pictures
of personal perceptions of risk for developing cancer. Two broad classes of measures are
absolute measures, in which personal ratings are made without reference to any outside
group, and comparative measures, in which personal perceived susceptibility is compared to
susceptibility in some normative group (Weinstein 8z Klein, 1996).
Absolute Measures
Rating Scales. Among absolute measures, typical rating scales ask individuals for Likert scale
judgments of their likelihood of developing cancer-for example, “What do you think are the
chances that you personally will get breast cancer someday” (5-point scale; Bastani, Marcus, &
Hollatz-Brown, 1991). These are the most commonly used measures of perceived
susceptibility, employed both in studies of the psychosocial correlates of health protective
behavior and in evaluations of interventions.
Numerical Estimates. Numerical estimates of the chance of contracting cancer are also taken
as absolute indicators of perceived susceptibility-for example, “Risk of developing breast
cancer in the next 10 years” (c I%, l–5%, 640%, 1 l–20% or > 20%; Dolan, Lee, & McDermott,
1997). Perceived risk has also been measured with rate judgments -for example, “the number

of women out of 1000 whom you think would develop breast cancer in the next 10 years”
(Black, Nease, 8z Tosteson, 1995; see also L. M. Schwartz, Woloshin, Black, & Welch, 1997).
Such measures have enjoyed relatively limited application, most often in studies comparing
perceived to objective risk.
Comparative Risk
Direct cumparative risk is measured with some form of the following question: “What do you
believe are your chances of getting (disease) compared to other (men/women) your own age?',
with typical responses of “a lot lower, somewhat lower, about the same, somewhat higher, and
a lot higher.” This measure has been applied to cancer in general (Kreuter & Strecher, 1995);
lung cancer, skin cancer, and cancer in general (Weinstein, 1987); breast cancer (e.g., Aiken,
Fenaughty, West, Johnson, & Luckett, 1995); and colorectal cancer (e.g., Blalock, DeVellis,
comparative risk is assessed by having individuals rate the perceived likelihood of developing
the disease for themselves and for others on separate scales; the difference between these
two ratings reflects comparative risk (Weinstein & Klein, 1996).


Measure of comparative risk are sometimes used in combination with absolute rating scales in
the formation of multi-item susceptibility measures. However, the two most common
applications of comparative risk items have been in research on optimistic bias (e.g.,
Weinstein, 1980) and in studies of individual attributions of risk (e.g., Aiken et al., 1995). Direct
comparative risk items provide risk estimates only in relation to others; thus, the specification
of the comparison group is critical. An individual who felt quite vulnerable to a disease, when
measured on an absolute rating scale, might nonetheless feel less at risk than more
unfortunate others (Klein & Weinstein, 1997), yielding a comparative rating of relatively low
comparative risk.
Perceived Susceptibility Versus Cancer Worry and Cancer Distress
Perceived susceptibility has been distinguished from more emotional aspects of vulnerability in
studies of cancer-related health behaviors, consistent with the C-SHIP model. Items such as
“Thinking about breast cancer makes me feel upset and frightened” (McCaul, Schroeder, &
Reid, 1996) have been used to characterize cancer worry, as distinct from perceived

susceptibility. Sjiiberg versus perceived risk reflect emotional versus cognitive reactions to
threat, respectively. The two variables are weakly correlated (e.g., r =.20; McCaul et al., 1996)
and form independent factors in the measurement of predictors of colorectal cancer screening
adherence (Vernon et al., 1997). In addition
-731to worry, fear of cancer and cancer treatment (Berman & Wandersman,
1992; Salazar & de Moor, 1995), cancer anxiety (Gram & Slenker, 1992), and
morbid concern about breast cancer (Irwig et al., 1991) also have been
included in research. A growing literature on breast and ovarian cancer
screening among high risk women (e.g., Audrain et al., 1998; M. D.
Schwartz, Taylor, et al., 1999) has employed such measures of cancerspecific distress.

Perceived Vulnerability to
Cancer and Objective Risk
Comparisons of Objective and Subjective Risk

A number of approaches have been taken to the comparison of objective
with subjective risk for cancer. Objective risk measures are of two types.
First are rates of risk in thepopulcation (e.g., the percent of women in the
population ever contracting breast cancer). Second are risk estimates
derived for specific individuals, based on their particular status on known
risk factors; derivation of these latter estimates employs epidemiological
models of risk. Subjective measures are numerical estimates of risk, or
rating scale measures of absolute or comparative risk.
Population Estimates. When determined from actual versus estimated
population rates, community samples overestimate their probability of
developing and dying of cancer (e.g., Helzlsouer, Ford, Hayward, Midzenski,
& Perry, 1994, for cancer in general; Ward, Hughes, Hirst, & Winchester,
1997, for prostate cancer).
Epidemiological Estimates and Individual Assessments. Individuals' risk
estimates based on epidemiological models have been compared with their

own subjective numerical risk estimates. The Gail model (Gail et al., 1989), a
five-factor epidemiological model of breast cancer risk among Caucasian
women, predicts risk for breast cancer among women who obtain annual
mammograms. Risk factors include number of first-degree relatives (mother
or sisters) with breast cancer, age at menarche, age at first live birth,


number of previous biopsies and chronological age. Using this model, Black
et al. (1995) and Dolan et al. (1997) found that women grossly overestimate
their chances of developing breast cancer, for example, by a factor of 20 in
women under age 50 (Black et al., 1995). Subjective and objective estimates
correlate moderately (for example, r =.46 for breast cancer in a sample of
women under age 50; Siegler et al., 1995).
Those at heightened cancer risk due to their being first- degree relatives
(FDRs) of individuals with cancer also overestimate their personal risk. More
than 60% of FDRs greatly overestimated their lifetime risk of breast cancer
as compared to Gail estimates (Lerman et al., 1995). Among FIN&, specific
Gail model risk components were found to be unrelated to numerical ratings
(0%100%) of the chance of getting breast cancer someday (Daly et al.,
1996).
Comparative Risk Ratings Versus Epidemiological Estimates. A number of
investigations have compared an epidemiological estimate of risk with a
direct measure of comparative risk. Such studies provide a very different
picture of the relation of objective to subjective risk, depending on the
disease. Individuals are optimistic (i.e., underestimate their risk) for heart
attack within the next 10 years (Avis, Smith, & McKinlay, 1989; Kreuter &
Strecher, 1995), but overestimate their risk of cancer in this time frame, with
almost half of individuals showing pessimistic bias (Kreuter & Strecher,
1995).

Comparative Risk and Unrealistic
Optimism
While people overestimate their absolute risk of contracting cancer, studies
of comparative risk suggest that individuals exhibit unrealistic optimism or
optimistic bias (Weinstein, 1980, 1987; Weinstein 8z Klein, 1996); that is,
they believe they are less likely to contract specific cancers than are others
their own age. This bias has been demonstrated for breast cancer (Aiken et
al., 1995), skin cancer (A. J. Miller, Ashton, McHoskey, & Gimbel, 1990), and
colorectal cancer (Blalock et al., 1990; Lipkus, Rimer, Lyna, et al., 1996), as
well as brain cancer, leukaemia, and lung cancer (Lek & Bishop, 1995). In
contrast, comparative judgments of “cancer in general” do not yield
optimistic bias (Weinstein, 1980, 1984), even when “cancer in general” and
specific cancers are rated by the same sample (Weinstein, 1982, 1987).
FDRs of women with breast cancer accurately estimate their comparative
risk as high when asked to compare their risk to women without a family
history of breast cancer (Audrain et al., 1995; Lerman, Kash, & Stefanek,
1994). However, when asked to compare their risk to others their own age,
with family history unspecified, a substantial portion of FDRs incorrectly rate
their risk as lower than average (Aiken et al., 1995; Blalock et al., 1990).
Whereas much effort has been made to untangle the sources of optimistic
bias and risk perceptions in general, the behavioral implications of the
optimistic bias for protective behavior are unclear (van der Pligt, 1998;
Weinstein & Klein, 1996), and have not been addressed in the context of
cancer. Van der Pligt (1996) argued that comparative risk appraisal may not
be a determinant of health behavior and does not contribute to the
prediction of health behavior beyond perceived vulnerability.

Determinants of Perceived Risk for

Cancer
Two literatures inform the question of the determinants of perceived risk for
cancer. The first, cancer-specific literature examines individuals' rationales
for their ratings of their own risk of cancer relative to others their own age,
following a methodology first employed by Weinstein (1984). The second is a
broader literature on the determinants of risk.
Determinants of Comparative Risk for Cancer
Weinstein (1984) coded the reasons generated by individuals for their
comparative risk judgments into five categories:
-732actions and behavior patterns, heredity, physiology or physical attributes,
environment, and psychological attributes. For cancer in general, breast
cancer (Aiken et al., 1995; Lipkus, Rimer, & Strigo, 1996; Salazar, 1994), and
colorectal cancer (Blalock et al., 1990; Lipkus, Rimer, Lyna, et al., 1996),
personal lifestyle-related actions were seen as decreasing risk (e.g., proper
diet, exercise). For lung cancer, personal actions (smoking) were seen as
increasing risk (Lek & Bishop, 1995). Across cancers, attributions for heredity
were that the absence of disease in the family reduced risk below average.
In contrast, women who believed their risk to be above average for breast
cancer mentioned heredity most often as the determining factor (Aiken et
al., 1995; McCaul & O'Donnell, 1998; Savage & Clarke, 1996). Of interest is
that FDRs of colorectal cancer patients rarely mentioned heredity as
increasing their risk, even after they had been informed they were at
increased risk due to their sibling's cancer (Blalock et al., 1990). These
results taken together suggest that individuals believe they have some
control over whether they get cancer through their own actions. However,
there is lack of understanding of the role of heredity in cancer. Absence of
family history is viewed as highly protective, even though most cancers are
not associated with family history. At the same time, a family history of
cancer may not lead to perceptions of increased risk.
General Determinants of Perceived Risk

Van der Pligt (1996,1998) summarized an extensive literature on the
determinants of perceived risk. Classes of determinants include cognitive
heuristics, disease characteristics, personal motivations, and personality and
information- processing strategies (Gerend, 1998). (See Fischhoff, Bostrom,
& Quadrel, 1993, for a discussion of risk perception and communication.)
Cognitive Heuristics. Individuals rely on cognitive heuristics in estimating
uncertain events (Kahneman & Tversky, 1973; Tversky & Kahneman,
1973,1974), and these heuristics may underlie inaccurate perceptions of risk
in the health domain. The availability heuristic (Tversky & Kahneman, 1973)
indicates that individuals base frequency estimates on the salience of the
event in question, or the ease with which the event comes to mind. Personal
experience with other individuals who have cancer (Wardle, 1995), coupled
with the extensive media coverage of cancer, may contribute to the
observed overestimates of cancer risk (Slavic, Fischhoff, & Lichtenstein,
1979; van der Pligt, 1998). The representativeness heuristic (Kahneman &
Tversky, 1973) indicates that individuals base likelihood estimates for a
hypothetical event (e.g., a personal diagnosis of breast cancer) on their
similarity to events with comparable characteristics (e.g., the individual's
similarity to others diagnosed with breast cancer).


Characteristics of the Health Threat. Bias in perceptions of comparative risk
has been hypothesized to depend on disease characteristics (Weinstein,
1984,1987; Weinstein & Klein, 1996). Harris (1996) and Weinstein (1987)
provided support for a direct relation between the perceived controllability
or preventability of a disease and optimistic bias concerning risk; that is, the
more controllable or preventable a disease was perceived to be, the greater
the optimistic bias. Evidence of a relation of optimistic bias with disease
heritability is lacking (Weinstein, 1982). The “absentlexempt” principle (e.g.,
“If I haven't gotten the disease by now, I won't get it”; Weinstein, 1987) is

associated with lower perceived risk with increasing age (e.g., Aiken, West,
Woodward, & Reno, 1994, for breast cancer), although cancer incidence
increases with age, Maintenance of Self-Esteem. Optimistic biases for
perceived personal risk have in part been attributed to a motivation to
protect oneself from feelings of distress or anxiety about future negative
events (e.g., Perloff, 1983). This protection may accrue from downward
social comparisons, that is, comparisons of one's own risk with the risk of
others who are actually more vulnerable (Klein, 1996; Klein & Weinstein,
1997; Perloff & Fetzer, 1986).
Personality Characteristics and Modes of Information Processing. A variety of
personality dimensions have been associated with perceived risk. Among
them are monitoring blunting (M. D. Schwartz, Lerman, S. Miller, Daly, &
Masny, 1995), psychological defense (Dziokonski & Weber, 1977; Paulhus,
Fridhandler, & Hayes, 1997), anxiety (MacLeod, Williams, & Berekian, 1991),
and neuroticism (Darvill & Johnson, 1991). This may explain linkages noted
between personality factors and breast screening behavior (Siegler et al.,
1995) reviewed by Siegler and Costa (1994).

PERCEIVED SUSCEPTIBILITY
AND CANCER RELATED
BEHAVIOR
This section considers the relations of perceived susceptibility to both
screening for early detection of cancer and cancer preventive behavior. A
critical issue for health psychology is the implication of perceptions of
susceptibility for protective behavior. As we have already indicated, we
conceptualize perceived susceptibility to disease as a distal construct in a
mediational chain of constructs that eventuates in health behavior. Relations
of perceived susceptibility to behavior are likely to be complex, to be
mediated, moderated, or nullified by other determinants of the particular
behavior in question, determinants that are explored in the discussion of

perceived susceptibility and protective behavior. Given space limitations,
this chapter does not provide a comprehensive review, but it does reference
and summarize existing reviews and highlight important themes (see RoyakSchaler, Stanton, & Danoff-Burg, 1997, for related work).

Perceived Susceptibility, Distress, and
Screening
Accuracy of Self-Report of Screening Behavior
Studies of screening behavior often rely on self-report of screening. Several
reports suggest approximately 95% accuracy


-733for self-reports of having had a mammogram when compared to clinic
records (Aiken, West, Woodward, Reno, & Reynolds, 1994; Degnan et al.,
1992; Etzi, Lane, & Grimson, 1994; King, Rimer, Track, Balshem, & Engstrom,
1990; Rimer et al., 1992). Correct recollection of whether a mammogram
occurred within the past year or 2 years appears somewhat lower (73%
accuracy; Degnan et al., 1992). Self-report accuracy is lower for screening
tests that occur during the course of physician examination, for example,
61% verification of Pap smears against laboratory records (Bowman, SansonFisher, & Redman, 1997), and very low verification rates for digital rectal
examination (DRE) and fecal occult blood test (FOBT) against medical charts
(Lipkus, Rimer, Lyna et al., 1996). Finally, self-reports of breast selfexamination (BSE) may overestimate actual performance (Alagna, Morokoff,
Bevett, & Reddy, 1987).
Perceived Susceptibility and Mammography Screening
McCaul, Branstetter, Schroeder, and Glasgow (1996) provided an extensive
meta-analysis of the relation between breast cancer risk and mammography
screening. The weighted average correlation between family history and
screening was r =.27, with only one article reporting a nonsignificant
negative correlation. For perceived vulnerability and screening, the average
weighted correlation was a somewhat lower, r =.16, with a stronger
relationship evidenced in cross-sectional (r =.19) than in prospective designs

(I =.lO). Higher screening likelihood was noted among women who had
breast problems, r =.30. Worry about breast cancer was positively
associated with screening, r =.14. The positive relation of perceived
susceptibility to screening has been confirmed in more recent studies (Cole,
Bryant, McDermott, Sorrell, & Flynn, 1997; Lauver, Nabholz, Scott, & Tak,
1997; Lipkus, Rimer, & Strigo, 1996). Perceived susceptibility is not a proxy
for family history, and predicts screening compliance above and beyond
family history (Aiken, West, Woodward, & Reno, 1994).
As we have argued, the relation of perceived susceptibility to screening has
been found to be moderated by other psychosocial variables. Aiken, West,
Woodward, and Reno (1994) found that susceptibility related to compliance
with mammography screening only when perceived barriers to screening
were low; under high perceived barriers, no such relation was observed.
Medical System and Demographic Determinants of Screening. ***domain
provides documentation of medical system determinants of the use of
medically based cancer screening tests. The impact of health care coverage
(e.g., Potosky, Breen, Graubard, & Parsons, 1998) and, moreover, continuity
of care, (e.g., O'Malley, Mandelblatt, Gold, Cagney, & Kerner, 1997), have
been documented. This literature further reflects the impact of demographic
variables on screening utilization, among them race (e.g., Frazier, Jiles, &
Mayberry, 1996; Paskett, Rushing, D'Agostino, & Tatum, 1997; Pearlman,
Rakowski, Ehrich, & Clark, 1996), acculturation among minority women
(Kaplan et al., 1996), and age (Caplan & Haynes, 1996; M. E. Costanza,
1992) in interaction with race (Fox & Roetzheim, 1994). These variables set
limits on the impact of psychosocial variables on screening utilization.
Perceived Susceptibility and Breast Self-Exumination (BSE)
Evaluation of the relation of perceived vulnerability to BSE performance
takes into account not only the frequency of BSE performance relative to the
recommended monthly schedule (American Cancer Society, 1998), but also
the adequacy of BSE performance (see review by Zapka & Mamon, 1986). S.

M. Miller et al. (1996), Savage and Clarke (1996), and Aiken et al. (1995) all


pointed out the mixture of positive and null results for the relation of
perceived vulnerability to BSE frequency. When relations of vulnerability to
BSE frequency are found, they are modest, ranging from. 14 to.25 (S. M.
Miller et al., 1996). The balanced mix of positive and null results yields a
lower average correlation across studies. Interestingly, perceived
susceptibility relates to thoroughness and accuracy of BSE performance
(Fletcher, Morgan, O'Malley, Earp, & Degnan, 1989; Wyper, 1990). However,
across studies, the perceived barriers construct (including such factors as
large breast size, difficulty of performing BSE, lack of expertise in BSE;
Salazar, 1994) dominates as the strongest predictor of BSE frequency within
the HBM framework, with correlations approaching -.5 (Wyper, 1990). Strong
barriers may override perceptions of susceptibility in influencing
performance versus nonperformance of BSE.
Self-Eficucy and Screening. As would be expected for a self-screening
behavior, self- efficacy or self-confidence in the ability to adequately perform
BSE is correlated strongly with BSE frequency (e.g., Alagna et al., 1987;
Champion,' 1991; Rutledge & Davis, 1988; Sortet & Banks, 1997; see
reviews by S. M. Miller et al., 1996, and Salazar, 1994). This relation has
been found both retrospectively and prospectively. Similarly, the importance
of self-efficacy has also been demonstrated for testicular self-examination
(Brubaker & Wickersham, 1990).
Few, Worry, Cancer Distress, and Screening Behavior
In both the general population and in FDRs of individuals with cancer, fear of
cancer, worry about cancer, and cancer distress have been associated with
both insufficient and excessive screening, thus providing a plethora of
conflicting results across studies.
General Population. McCaul and colleagues (McCaul, Reid, Rathge, &

Martinson, 1996; McCaul, Schroeder, & Reid, 1996) found a positive relation
of breast cancer worry to mammography screening in the general
population, as did Ward et al. (1997) for prostate cancer. However,
-734in an inner-city population, an inverted U-shaped relation was observed:
Moderate worry about breast cancer was associated with greater attendance
at a first mammography screening than was either extreme (Sutton, Bickler,
Sancho-Aldridge, & Saidi, 1994). The same inverted U-shaped relation was
observed between BSE frequency and breast cancer worries (Lerman et al.,
1991). Among older low income Mexican American women, fear of and
fatalism about cancer were associated with lower Pap smear rates (Suarez,
Roche, Nichols, & Simpson, 1997). Worry appears to serve as a barrier to
mammography among African American women (Friedmanet al., 1995).
Again in a sample with a substantial inner-city component, Bastani et al.
(1994) reported a strong negative association between fear of finding breast
cancer and screening. This brief sampling of articles suggests possible
demographic differences in the relation of emotional aspects of cancer
threat on screening, with cancer worry adversely affecting screening among
inner-city, low income, and minority individuals; these findings, however, are
not universal.
High Risk Individuals. A conflicting pattern of results is also observed for
high risk individuals, FDRs of individuals with cancer. Ovarian cancer worries
among FDRs have been positively associated with screening (M. D.
Schwartz, Lerman, Daly, et al., 1995). In contrast, high breast cancer


distress (i.e., extreme worry, intrusive thoughts about breast cancer) among
FDRs is associated with reduced screening (Lerman et al., 1993; see also
Kash, Holland, Halper, & D. G. Miller, 1992; Lerman et al., 1994), although
the opposite has also been found (Stefanek & Wilcox, 1991). Interestingly,
distress has been associated with either excessive or insufficient BSE

(Epstein et al., 1997; Lerman et al., 1994). Cancer distress among FDRs of
women with breast and ovarian cancer is associated with high perceived risk
of cancer and low perceived control over cancer development (Audrain et
al., in press).
Conflicting Findings and the Elusive Inverted U-Shaped Function
In the now classic fear communication literature, Janis and Feshbach (1953)
argued that fear served as a positive motivator for protective behavior up to
some critical level of fear. Above that critical fear level, avoidance of the
threat was expected to replace protective behavior, yielding an inverted Ushaped relation between level of fear and behavior.
Resolving Conflicting Findings. Lerman and M. D. Schwartz (1993) used the
notion of an inverted U-shaped relation to highlight an important issue in
resolving conflicting literature on the relations of worry and distress to
cancer screening- the range and level of distress represented among
participants in any individual study. If it is assumed for a moment that an
inverted U-shaped relation of distress to screening exists, then all relations
(positive, inverted U, negative, or no relation) are possible as segments of
the distress continuum are sampled. The resolution of conflicting study out
comes may lie in the segment of the distress continuum represented in any
study. The McCaul, Reid, et al. (1996) meta-analysis showed only monotonic
increasing relations of both susceptibility and cancer worry to behavior.
However, the meta-analysis did not include articles in which avoidance of
screening by FDRs of breast cancer victims was associated with high cancer
distress, discussed further later (e.g., Kash et al., 1992; Lerman et al., 1994).
In samples from the general population- samples such as those of McCaul,
Reid, et al. (1996)-it is possible that there are insufficient very high distress
cases for a curvilinear relation to be manifested and/or detected statistically.
Distress Versus Perceived Susceptibility. Support for the inverted U-shaped
relation is found when emotional distress and not the more cognitive
assessment of perceived susceptibility serves as the predictor of cancer
screening behavior. Perhaps the inverted U-shaped relation of risk to

behavior has been sought in the wrong variable, that is, in perceived
susceptibility rather than cancer distress. (See Hailey, 199 1, for
consideration of an inverted U-shaped relation of distress to screening
among FDRs of women with breast cancer.)
Modifying Perceived Susceptibility and Cancer Distress Through
Training
High Risk Women. As already described, FDRs of women with cancer
typically exhibit excessive perceived risk and associated high cancer
distress, apparently leading to failure to follow screening recommendations
(i.e., either excessive or insufficient screening) and even to requests for
prophylactic surgery (Lerman et al., 1995). Interventions to reduce perceived
susceptibility among FDRs have sometimes been successful (Alexander,
Ross, Sumner, Nease, & Littenberg, 1995). However, women with high
cancer distress benefit less from such susceptibility focused interventions,
suggesting that both cancer distress and inaccurate perceptions of risk must


be simultaneously addressed (Lerman et al., 1995). Reductions in cancer
distress have been achieved through individual counseling (Lerman et al.,
1996; Schwartz, Lerman, et al., 1998). An important issue is whether
clarifying that perceived risk is overestimated will lead to underutilization of
mammography screening (M. D. Schwartz, Rimer, Daly, Sands, & Lerman,
1998).
General Population. The extent to which subjective risk estimates can be
made more accurate through intervention has also been explored in the
general population (Weinstein & Klein, 1995), where perceived risk typically
exceeds objective risk. Kreuter and Strecher (1995) reported increased
accuracy in perceived risk (i.e., decreased perceived risk) for cancer in the
general population following an educational intervention. Lipkus, Biradavolu,
Fenn, Keller, and Rimer (1998) explored strategies for increasing accuracy of

risk perceptions for cancer.
-735Repeated Screening
An important question is whether screening behaviors are sustained over
time among asymptomatic women. Ronis, Yates, and Kirscht (1989) argued
that the factors that lead to initiation of a behavior do not sustain the
behavior, and that habits, rather than attitudinal variables, determine
repeated behavior. Similar arguments were made by S. M. Miller et al. (1996)
in the context of BSE performance.
Correlations of perceived susceptibility with repeated mammography
screening are not reliably observed, with positive associations noted by
Lerman, Rimer, Track, Balshem, and Engstrom (1990) and Fenaughty, Aiken,
and West (1993), but no association noted by Marshall (1994), Cockbum,
Schofield, White, Hill, and Russell (1997), and Orton et al. (199 1). Anxiety
about mammography appears to be negatively related to repeated
screening (Lerman et al., 1990).
Medical System and Demographic Determinants of Repeated Screening. A
host of demographic and medical system variables relate to repeated
screening, just as with mammography compliance taken at any single point
in time. Younger age, physician recommendation, having had regular clinical
breast examinations by a physician, and family breast cancer history are
positively associated with repeated mammography screening (Hitchcock,
Steckevicz, & Thompson, 1995; Lerman et al., 1990; Zapka, Stoddard, Maul,
& Costanza, 199 1). Failure of asymptomatic women to return for a second
mammogram at a regular interval is associated with negative experiences
with the initial mammogram, among them pain, embarrassment, and
unpleasant interaction with clinic staff (Marshall, 1994; Orton et al., 1991).
Again, it appears that many variables operate on repeated screening that
weaken the potential impact of perceptions of susceptibility.
Sequelue of Abnormal Screening Tests and Discovery of Symptoms
Two related literatures highlight the reciprocal nature between screening and

perceived vulnerability and cancer distress. The first literature, which
addresses the impact of abnormal screening tests on psychological
functioning, was reviewed by Paskett and Rimer (1995). This literature shows
clear negative psychological effects of abnormal Pap smear and
mammography test results, including heightened cancer distress, with


varying levels of follow-up screening (from 20% to 95% across studies). The
second literature, on delay in seeking treatment following the selfidentification of a possible cancer symptom (e.g., a breast lump), was
reviewed by Facione (1993). This literature characterizes the myriad fears
engendered by discovery of cancer symptoms and their association with
delay in seeking treatment.

Preventive Behavior: Sun
Protection
Although the focus is primarily on screening, this section touches on cancer
prevention, with a consideration of perceived vulnerability as a correlate of
sun protection. The incidence of deadly melanoma has risen 4% per year
since 1973. Skin protection through limiting sun exposure and sunscreen use
is recommended (American Cancer Society, 1998). (However, there is now
significant controversy as to the efficacy of sunscreen for protection against
melanoma; Facelmann & Wu, 1998).
Sun protection against skin cancer poses four related issues. First, because
intensive sun exposure between age 10 and 24 is associated with melanoma
development (Holman et al., 1986), adolescents must adopt sun protection.
Second, sun exposure early in life is associated with much later
development of skin cancer, thus raising the issue for health psychology of
how to induce behavior change against distal risk. Third, normative
influences play heavily in tanning: A suntan is perceived as healthy (e.g., Hill
et al., 1992; Mermelstein & Riesenberg, 1992) and attractive (A. J. Miller et

al., 1990). Fourth, parents must play an active role in their children's skin
protection (Rodrigue, 1996).
Objective risk based on skin type (Fitzpatrick, 1988) is associated with
perceived susceptibility to skin cancer (Clarke, Williams, & Arthey, 1997;
Jackson, 1997; Webb, Friedman, Lute, Weinberg, & Cooper, 1996). Arthey
and Clarke (1995) provided a review of the psychological literature on
suntanning and sun protection. Positive associations between perceived
susceptibility and sun protection have been noted among high school
students (Mermelstein & Riesenberg, 1992; Wichstrom, 1994), university
students (Cody & Lee, 1990), the general U.S. population (Hall, May, Lew,
Koh, & Nadel, 1997), and parents protecting their children (Lescano &
Rodrigue, 1997), but such relations are not uniformly observed. In fact, a
negative relation has been found between sun protection and perceived risk
among individuals with chronically high sun exposure (Carmel, Shani, &
Rosenberg, 1996). Elevated perceptions of susceptibility in this case appear
to result from past high risk behavior. A similar relation has been observed in
the HIV/AIDS literature; those who have engaged in high risk sexual behavior
subsequently report high perceived vulnerability to HIV/AIDS (Gerrard et al.,
1996). Manipulations of perceived susceptibility in interventions have
resulted in increased intentions for sun protection (e.g., Cody & Lee, 1990;
Mahler, Fitzpatrick, Parker, & Lapin, 1997).
Normative Influences. Normative influences (Pratt & Borland, 1994;
Wichstrom, 1994), particularly for appearance (A. J. Miller et al., 1990), have
shown reliable relations with sun tanning versus sun protective behavior,
particularly among adolescents. Self-presentation (impression management)
may well lead to health risks (Leary, Tchividjian, & Kraxberger, 1994);
suntanning exemplifies this phenomenon. Recent interventions (e.g., Jones &


Leary, 1994; Prentice-Dunn, Jones, & Floyd, 1997) also highlight the impact

of appearance concerns. These powerful normative influences, which are
much less often considered in relation to screening behaviors such as
mammography (but see Montano & Taplin, 1991), highlight the unique
forces, in addition to perceived vulnerability, that influence particular
cancer- specific behaviors.
-736-

INTERVENTIONS TO INCREASE
SCREENING
This section addresses interventions to increase cancer screening, and, more
specifically, attempts to link manipulations of perceived susceptibility to
increased screening. Experimental interventions provide the vehicle for
untangling the causal impact of putative determinants such as perceived
vulnerability on cancer protective behavior. The use of mediational analysis
to assess the extent of direct and indirect impact of manipulations of
perceived susceptibility on screening outcomes is also highlighed.
Comprehensive summaries of interventions to increase cancer screening
have been provided by Rimer (1994) for mammography screening and by
Snell and Buck (1996) for breast, cervical, and colorectal cancer.
From the perspective of health psychology, theory-based interventions that
employ models such as the HBM to design program components are most of
interest, because they permit the linking of changes in constructs in the
model (e.g., perceived susceptibility) to changes in screening behaviors. A
number of mammography screening interventions have included
components designed to increase perceived vulnerability to breast cancer
(Aiken, West, Woodward, Reno, & Reynolds, 1994; Champion, 1994; Curry,
Taplin, Anderman, Barlow, & McBride, 1993; Rimer et al., 1992; Skinner,
Strecher, & Hospers, 1994; Zapka et al., 1993). In some studies, the
perceived susceptibility component was only one small part of a large
complex intervention, and no attempt was made to establish a direct linkage

from this component to behavioral outcomes (Champion, 1994; Rimer et al.,
1992; Zapka et al., 1993). In contrast, Curry et al. (1993) showed that
providing tailored personal objective risk information to FDRs of breast
cancer victims increased screening; Skinner et al. (1994) showed a similar
impact of tailored messages in a community sample.

Mediational Analysis of Intervention
Impacts
Aiken, West, Woodward, Reno, and Reynolds (1994) implemented an HBMbased mammography intervention, with individual program components that
targeted each of the four HBM constructs: perceived susceptibility, severity,
benefits, and barriers. Mediational analysis, a statistical procedure that
establishes linkages among chains of variables, was used to test the
linkages from an intervention through intermediate mediators (the HBM
components) to mammography compliance (West & Aiken, 1997). This
mediational analysis is presented here because of our strong conviction
(West & Aiken, 1997) that mediational analysis provides important insights
into the way in which theoretical constructs influence behavior. To date,
mediational analysis has been used productively in both mental health and
substance abuse research, as well as in several areas of basic psychological


research.
Requirements for Mediational Analysis. In order to test the theory of an
intervention through mediational analysis, the following are required: a
specified theoretical model on which the program will be built, a
measurement instrument that provides distinct measures of each construct
in the model that will serve as a mediator, a translation of each construct of
the model into a distinct component of the intervention, assessment of
postintervention levels on each of the constructs targeted in the model in an
experimental versus control group (with adequate statistical control of

pretest levels), and measurement of the outcome. West and Aiken (1997)
summarized the conditions that must be met in order to demonstrate that a
putative mediator (here, perceived susceptibility) produced change in the
outcome (here, mammography screening), as specified by Judd and Kenny
(1981), Baron and Kenny (1986), and MacKinnon (1994).

The Mediational Role of Susceptibility in
Intervention
In the intervention, the HBM was amended by assessing intentions for
screening at immediate posttest as well as actual compliance 3 months
following the intervention. Mediational paths were established from
perceived susceptibility and perceived benefits to intentions, as was a strong
link from intentions to subsequent screening. The role of perceived
susceptibility in the causal chain from intervention through compliance is of
interest here. The model of the impact of HBM constructs on outcomes is
illustrated in Fig. 44.1. It differs from typical characterizations of the HBM in
that the four HBM constructs are not treated as coequal predictors of
outcome. Rather, following Ronis (1992), a model was specified in which
perceived susceptibility and perceived severity were antecedents of
perceived benefits, under the assumption that a woman would not perceive
the benefits of mammography screening unless she felt threatened
(perceived susceptibility plus severity) by breast cancer. Again following
Ronis (1992), it was specified that the effect of perceived susceptibility on
outcome would be mediated through perceived benefits, that is, that the
effect of susceptibility would be an indirect effect through benefits, in the
following causal sequence:
Intervention → Susceptibility → Benefits → Intentions.
This mediational chain was confirmed. In addition, a direct path from
susceptibility to intentions was confirmed, that is,
Intervention → Susceptibility → Intentions.

The size of the indirect effect of susceptibility, over and above the direct
effect, was substantial.
The full details of the mediational analysis, including explorations of possible
roles for perceived susceptibility, are provided in West and Aiken (1997).
What is critical here is a conception of perceived susceptibility at the outset
of a causal chain that flows through other constructs. Examining only the
direct effects of susceptibility on intentions or behavior may obscure the role
of perceived susceptibility in the behavioral compliance process, potentially
leading to underestimates of the total effect of perceived susceptibility on


behavior.
Some research on screening and preventive behavior omits considerations
of perceived susceptibility and examines variables that are conceptually
downstream of perceived susceptibility
-737FIG. 44.1. Mediational analysis of the impact of a health belief model (HBM)
based intervention on compliance with mammography screening
recommendations. The indirect mediational path from intervention to
perceived susceptibility through perceived benefits to intentions for
screening illustrates how perceived susceptibility serves as an apparent
precursor to benefits in the HBM. For paths, *p <.OS, **p <.Ol, ***p <.OOl.
From “Increasing Screening Mammography in Asymptomatic Women:
Evaluation of a Second-Generation, Theory-Based Program” by L. S. Aiken, S.
G. West, C. K. Woodward, R. R. Reno, and K. D. Reynolds, 1994, Health
Psychology, 13, p. 534. Copyright 0 1994 by the American Psychological
Association. Reprinted with permission.
in the causal chain from perceived susceptibility to behavior. For example,
Rakowski (Rakowski et al., 1992, 1997; Rakowski, Fulton, &Feldman, 1993)
focused on the impact of decisioPzal balance, or the balance between
perceived pros and cons of mammography on screening uptake. Such a

focus on downstream variables (here analogous to benefits of and barriers to
screening) does not diminish the role of perceived susceptibility earlier in
the causal process.

ISSUES IN CHARACTERIZING THE
ROLE OF PERCEIVED
VULNERABILITY IN CANCERRELATED BEHAVIOR
A number of issues have arisen in the course of this consideration of the
impact of perceived vulnerability on cancer protective behavior. These have
included the impact of perceived vulnerability versus cancer fear, worry, and
distress on screening behavior; the existence of an inverted U-shaped
relation between cancer distress and protective behavior; the difference in
determinants of medically based versus self-implemented screening; and
the place of perceived vulnerability in mediational chains of putative
determinants of cancer- related behavior.
To these issues the following are added: the level of numeracy in the lay
public, defined as “facility with basic probability and numerical concepts” (L.
M. Schwartz et al., 1997), and the ability of lay individuals to understand and
produce risk estimates; public understanding of screening tests and the
impact of confusion between early detection and prevention; the difference
in relation of perceived vulnerability to screening versus preventive health
action over time; the impact of the complementary relation among
screening behaviors (e.g., BSE versus mammography) on the observed link
between vulnerability and screening; the role of conditional versus
unconditional threat in accounting for cancer-related behavior; the impact of


message framing on the link from risk to behavior; the existence of
moderators of the relation of perceived vulnerability to cancer related
behavior, including demographics, level of risk, personality, and barriers to

behavior; and the need for experimental intervention as well as
-738psychosocial research, coupled with mediational analysis, to understand the
role of individual constructs in cancer protective behavior.

Numeracy
One critical issue that has only very recently been broached in the context of
cancer screening is the numeracy of the lay public, a parallel in the
quantitative domain to literacy. Although individuals' numerical estimates of
perceived risk for cancer are interpreted as meaningfully reflecting
perceived vulnerability, the competency of the lay public in understanding
and using probability estimates or rates has not been considered. More
numerate women exhibit smaller overestimates of the probability of
developing and dying of breast cancer, as well as of the absolute mortality
risk reduction attributable to mammography (Black et al., 1995), and are
better able to apprehend information about mortality reduction attributable
to mammography (L. M. Schwartz et al., 1997). Consideration of numeracy is
critical in risk-related research (Fischhoff et al., 1993), as well as in medical
settings, where individuals are asked to make their own medical decisions
with regard to screening and treatment based on probability and rate
information (L.M. Schwartz et al., 1997).
Humans as Intuitive Statisticians. ***decades ago, a literature developed in
applications of Bayesian decision theory in psychology (Edwards, Lindman, &
Savage, 1963), which examined the ways in which individuals estimate the
probability of events and revise their probability estimates based on new
information. Two principles emerged that may help to explain biases in the
lay public's understanding of risk for cancer. First, people appeared to
overestimate low probabilities and underestimate high probabilities (Mueller
& Edmonds, 1967; see also Lichtenstein, Slavic, Fischhoff, Layman, &
Combs, 1978). This may partially explain the overestimates of rates of
specific cancers, which are low percentage wise, on a cancer by cancer basis

(American Cancer Society, 1998). Second, people are conservative in
revising their estimates of probabilities in the face of new information
(Phillips & Edwards, 1966), which may partially account for failures of
training to eradicate biases in perceived risk.

Public Understanding of Screening Tests,
and Detection Versus Prevention
The public historically has exhibited a lack of understanding of screening
tests for asymptomatic individuals. Both Rimer, Keintz, Kessler, Engstrom,
and Rosan (1989) and Zapka, Stoddard, M. D. Costanza, and Greene (1989)
reported that women believed mammograms were unnecessary in the
absence of breast symptoms (see review by Vernon, Laville, & Jackson,
1990). Recall that in the original HBM formulation, perceived vulnerability in
the absence of symptoms was seen as requisite for health behavior
(Rosenstock, 1990). Public belief that screening tests are unnecessary for
asymptomatic individuals persists (Cockburn, Redman, Hill, & Henry, 1995).


Aiken et al. (1995) and Barnard and Nicholson (1997) also found confusion
between screening for cancer detection versus cancer prevention. Cancer
detection aims at finding existing cancers in an early, treatable state,
whereas prevention aims to avoid cancer development. Women, failing to
distinguish early detection from prevention, mistakenly attributed their lower
perceived personal risk for contracting breast cancer to the fact that they
receive regular mammograms. From an intervention perspective, correcting
this misconception ironically may result in decreased screening rates. Even
with these misconceptions, there is apparently understanding of and
acceptance of the fact that screening tests may result in both false negative
and false positive outcomes (Aiken et al., 1998; Cockburn et al., 1995).

Temporal Factors in the Perceived RisikBehaviour Linkage
For both prevention and detection, the cross-sectional relation of perceived
risk to behavior changes in complex ways as health innovations diffuse over
time (Weinstein & Nicolich, 1993). When a health protective behavior is fast
introduced, those who perceive themselves at the highest risk for the health
threat may self-select the behavior, occasioning a strong positive correlation
between perceived risk and behavior. If the behavior is screening, then
perceived vulnerability to the occurrence of the disease should not diminish
as the behavior is adopted, because screening is not, of course, preventive.
In fact, perceived severity may diminish if people come to believe in the
benefits of early detection for cancer survival. However, as a screening
innovation is adopted broadly by the medical profession and increasing
numbers of individuals are screened, the pool of screened individuals will
contain individuals at lower perceived risk, thus diluting the correlation of
perceived risk and screening. For preventive behavior, the initial selfselection of high risk individuals may again result in substantial positive
correlations between perceived risk and behavior. However, should the
disease risk be mitigated or substantially lessened by the preventive
behavior, then, in time, a negative correlation may be observed between
perceived risk and behavior-those who reliably engage in the behavior may
correctly perceive themselves to be at lower risk. (See Aiken et al., 1995, for
a consideration of temporal factors in perceived and objective risk as related
to mammography screening, and Gerrard et al., 1996, for a critical
discussion of these relations in the HIV/AIDS context.)

Complementary Screening Tests
Multiple screening tests for the same cancer (e.g., BSE and mammography)
may be considered by some individuals to be complementary (i.e., a woman
might opt for regular BSE and forego mammography). If so, this would
weaken the relation of psychosocial variables to each behavior taken
separately. BSE frequency and mammography compliance appear to be

uncorrelated (Aiken et al., 1995). Perceived vulnerability might better predict
a breast screening index crediting either mammography or BSE, but this
possibility has not been explored empirically. Similar arguments can be
made for FOBT, sigmoidoscopy, and DRE for colorectal cancer, and PSA and
DRE for prostate screening.
-739-

Conditional Health Threat


Perceptions of susceptibility to cancer or its severity are in part governed by
beliefs about personal actions that mitigate or increase cancer risk. Ronis
(1992) characterized conditional health threat as the perception of threat
under some behavior specification, that is, if the individual were to take a
specific protective health action versus to take no health action (see also
Rogers, 1983). Ronis (1992) and van der Pligt (1998) argued that the
measurement of conditional health threat would provide better
understanding of protective health behavior because such conditional
measures untangle the influence of current protective behavior on perceived
vulnerability. Weinstein and Nicolich (1993) theorized that the discrepancy in
level of perceived risk associated with participation versus nonparticipation
in a health protective behavior reflected perceived effectiveness of the
health precaution.
It is proposed that the components of conditional health threat-that is,
conditional susceptibility versus conditional severity-will have differential
associations with preventive versus screening behavior. For preventive
behavior, high perceived susceptibility given inaction coupled with high
perceived benefits of the health action is expected to produce preventive
behavior, with a subsequent reduction in perceived susceptibility. For
screening, the matter is different because susceptibility is not reduced by

screening; rather, the argument for screening is that consequences
(severity) of cancer will be reduced with early detection, so that the
appropriate conditional characterizations of perceived severity are “severity
if treated early” versus “severity if treated late” (Ronis & Hare& 1989). Ronis
and Hare1 (1989) applied this dual conception of perceived severity to BSE
performance and showed a link of these severity measures, but not
conditional susceptibility, to BSE, yielding new insight into the potential role
of perceived severity in screening. Jackson (1997) applied conditional
perceived susceptibility and severity to skin cancer preventive behaviors
and found the opposite effect-that conditional measures of perceived
susceptibility, but not severity, predicted skin protection. Measures of
conditional threat may provide help to clarify the links of perceived
susceptibility and severity to cancer protective behaviors.

Message Framing, the Understanding of
Risk, and Health Protective Behavior
The impact of perceptions of risk on decisions concerning health behavior is
strongly affected by the manner in which risks are framed. As specified in
prospect theory (Kahneman & Tversky, 1979), individuals respond differently
to information presented as gains (e.g., the number of breast cancer deaths
averted by regular mammography screening) versus as losses (e.g., the
number of breast cancer deaths associated with failures to be screened).
Thus considerations of perceived risk in relation to cancer-related behavior
must take into account message framing as a moderator of the perceived
risk-behavior link. Rothman and Salovey (199'7) provided an extensive
review of the impact of message framing on health behavior. Meyerowitz
and Chaiken (1987), Rothman, Salovey, Antone, Keough, and Martin (1993),
and Banks et al. (1995) provided examples of the impact of message
framing on BSE, skin protection, and mammography utilization, respectively.

Moderators of the Role of Perceived
Susceptibility


Variables like demographics, personality, and barriers to health action may
moderate (change) the relation of perceived vulnerability to protective
health behavior. High perceived barriers nullified the effect of susceptibility
on mammography compliance (Aiken, West, Woodward, & Reno, 1994).
Conscientiousness as a personality trait moderated the relation of cancer
distress to mammography screening among FDRs of breast cancer victims
(Schwartz, Taylor, et al., 1999). Moderation of the perceived susceptibilitybehavior link by demographic, medical system, personality, and other
psychosocial variables should be explored.

Mediational Chains and Experimental
Intervention Researc h
Our conception of perceived susceptibility is that it stands at the outset of a
causal chain that flows through other constructs to health behavior.
Consideration of mediational chains from perceived susceptibility through
other variables to behavior is critical for advancing an understanding of the
way health behaviors accrue. To reiterate, examination of both the direct
effect and the indirect effects of susceptibility through other variables on
behavior is required to estimate accurately the total effect of perceived
susceptibility on health protective behavior.
The understanding of whether, how, and to what extent individual variables
such as perceived susceptibility operate in determining health protective
behavior is best advanced through the evaluation of model-based
interventions, with research structured so that mediational analysis of the
effects of putative determinants of behavior can be accomplished (West &
Aiken, 1997). A distinguishing feature of psychology as a discipline is the
strength in theory and experimentation. Thus, health psychologists may

entertain a special role in health behavior research, providing careful theory
testing in controlled settings, and the refinement of models of health
behavior on a strong empirical base.