A Approach to Understanding Cancer Perception: Contributions From Judgment and Decision-Making Research Ellen Peters, Ph.D. Decision Research, Eugene, Oregon and University of Oregon Kevin D. McCaul, Ph.D. North Dakota State University Michael Stefanek, Ph.D. and Wendy Nelson, Ph.D. National Cancer Institute

ABSTRACT uncertainty, patient understanding of risk is a critical part of the Background: The likelihood judgments that people make decision-making process (3–5). For example, a man who elects about their for cancer have important implications. At the to screen for prostate cancer by having a prostate-specific anti- individual level, risk estimates guide protective actions, such as gen test would need to weigh the potential positive and negative cancer screening. However, at the extremes, exaggerated risk outcomes of the test (e.g., cancer detection, unnecessary anxiety judgments can also lead to anxiety that degrades quality of life from a detected cancer whose best treatment is watchful wait- or to aggressive self-protective actions that are unwarranted ing) with the likelihood of each outcome. At a policy level, risk given the objective risks. At the policy level, risk judgments may judgments may reflect societal perceptions of cancer prevention serve as an indicator of societal perceptions of the “war” and control efforts. For example, news reports describing ele- against cancer. Using risk judgments, the public expresses its vated rates of cancer in Long Island, NY, or Marin County, CA belief about whether we are winning. Purpose: We present theo- (so-called cancer clusters) may be interpreted to mean that we retical perspectives from judgment and decision making, illus- are losing the “war on cancer” even though the objective exis- trate how they can explain some of the existing empirical find- tence of these cancer clusters remains an open question (6). ings in the cancer risk literature, and describe additional People do not always have a clear understanding of the risks predictions that have not yet been tested. Conclusions: Overall, of cancer, or of the likelihood of various outcomes of cancer we suggest that theories from the judgment and decision-making screening tests and treatments (7). For example, with genetic perspective offer a potentially powerful view for understanding testing, people tend to overestimate their risk of cancer and un- and improving risk judgments for cancer and other diseases. derestimate the risks associated with genetic testing itself, such as the stigma associated with testing positive for a cancer sus- (Ann Behav Med 2006, 31(1):45–52) ceptibility gene (8,9). In extreme cases, exaggerated risk judg- ments can lead to anxiety that degrades quality of life and causes INTRODUCTION excessive vigilance and self-protective behaviors as well as un- The judgments people make about their risk or statistical warranted, aggressive medical treatments (e.g., prophylactic probability of developing cancer (i.e., their risk perceptions) mastectomy in the absence of a family history or genetic vulner- have important implications for cancer prevention, screening, ability to breast cancer) (10,11). and treatment. At an individual level, risk perceptions can guide People estimate risk in a variety of ways. For example, they protective actions, such as screening for cancer or quitting may make risk judgments based on intuition, a recent experi- smoking (1,2). Risk perceptions are also relevant to medical de- ence, a vivid memory, or a sound bite on the news. Others may cisions that cancer patients make, especially when no consensus approach risk estimation more systematically and analytically, exists on the standard of care or best course of action to take. Be- basing their risk estimates on “hard data.” When facing uncer- cause cancer patients make many decisions under conditions of tain situations and complex decisions, people tend to rely on mental shortcuts to simplify the decision-making process and thus reduce its cognitive demands and psychological stress. This work was supported by grants from the National Science Founda- The purpose of this article is to summarize how the judg- tion (0111941 and 0241313) to Dr. Ellen Peters and from the National ment and decision-making literature can help us understand Cancer Institute (K05 CA92633) to Dr. Kevin McCaul. how and why people underestimate or overestimate cancer risk. Specifically, we examine heuristics, the mental shortcuts or We thank Paul Slovic and Stephanie Hess for helpful comments on the “rules of thumb” that decision makers consciously or uncon- article and Stephanie Hess for helping with the literature search. sciously employ to make judgments of uncertainty. We review Reprint Address: E. Peters, Ph.D., Decision Research, 1201 Oak Street, four heuristics—, representativeness, availability, and an- Suite 200, Eugene, OR 97401. E-mail: [email protected] choring and adjustment—and speculate how each may contrib- © 2006 by The Society of Behavioral Medicine. ute to risk judgments about cancer.

45 46 Peters et al. Annals of Behavioral Medicine

HEURISTICS, JUDGMENT, AND cision making; instead, they are critical for accuracy and effi- DECISION-MAKING RESEARCH ciency. According to the affect , all of the images in a The psychological study of judgment and decision making person’s mind are marked to varying degrees with affect, and it attempts to understand how decision makers process information is this “affect pool” that people consult when called on to make to form judgments (defined as beliefs or evaluations; e.g., a risk certain judgments (21). Relying on affective impressions can be perception) and make decisions (defined as a choice among two simpler and more efficient than using deliberative processes, or more options; e.g., a choice among treatments). Three general such as weighing the pros and cons of a situation or retrieving rel- themes guide research on judgment and decision making. The evant examples from memory, especially when the required judg- first is that people have limited resources to deal with the numer- ment or decision is complex or mental resources are limited. ous decisions and overwhelming quantity of information that The affective and experiential nature of risk perception is they face daily. People are limited by time, their cognitive and particularly relevant to health domains such as cancer, given the computational abilities at the moment of choice, and their envi- affect that cancer arouses. People who have personally experi- ronment. The second theme states that because people often do enced cancer are likely to have developed stronger and more ac- not know what they value or prefer, they tend to construct judg- cessible affective reactions to the disease compared with people ments “on the spot” when asked to make a particular judgment or who lack personal experience. According to the affect heuristic, decision.Consequently,valuesandpreferencesmaybelabileand individuals then will use these feelings as a cue to determining sensitive to the way in which choices are framed (12). The third their own risk, with the stronger feelings being associated with theme states that people process information using two distinct enhanced personal risk estimates. Although they did not assess modes of thinking: deliberative and experiential (13–15). feelings specifically, Fiandt, Pullen, and Walker (22) found evi- Whereasthedeliberativemodeisanalytical,reasonbased,verbal, dence consistent with the affect heuristic in that women who had and relatively slow, the experiential mode is automatic, associa- a family member or a friend with cancer used their experience as tive, and fast. The experiential mode of thought relies on affect a cue for increased risk perceptions. and categorical thinking (e.g., stereotypes) and functions to high- Experimental studies have shown that attention to salient light important information. Depending on the situation and na- affective cues can lead to neglect of probabilistic information ture of the decision, people may rely more heavily on one system. (23). These findings may explain why some women fail to be- For example, whereas a medical professional’s understanding of lieve accurate numerical risk information about breast cancer, risk as statistical probability may be more heavily influenced by overestimate their risk, and experience undue cancer distress the deliberative system, lay understanding may rely more on ex- (24). Because cancer is a dreaded disease, the strong negative af- periential ways of knowing (16). fect that it elicits may create insensitivity to its (often relatively The deliberative and experiential modes of thinking have low) objective risk. For example, Kraus, Malmfors, and Slovic important implications for how people judge and decide. Al- (25) found that, even though expert toxicologists were able to though people in Western cultures tend to believe that more de- accurately assess the risk of cancer posed by different levels of liberation will always produce better decisions, evidence sug- exposure to a toxic agent, the public tended to believe that any gests that, in some contexts, deliberation disrupts affective and level of exposure was risky. Because the possibility of cancer intuitive processes in decision making and reduces postchoice will likely remain after testing and monitoring, affective reac- satisfaction (e.g., 17). Research suggests that intuitive processes tions to it will also remain, and risk perceptions may not change such as affect may have greater influence when deliberative ca- much. A possible result is avoidance of genetic testing for can- pacity is diminished because of cognitive constraints, time pres- cer and prophylactic measures by patients who understand that sure, or age (18). uncertainty about getting cancer cannot be eliminated. Heuristic processing is central to the experiential system in The strength of affect conveyed about a risk also appears to that it operates intuitively and automatically. Although this is play a role in how people respond to risk communication. advantageous in certain decision-making contexts, at other Slovic, Monahan, and MacGregor (26) conducted a series of times reliance on heuristics results in suboptimal decision mak- studies in which experienced forensic psychologists and psychi- ing. Thus, an understanding of heuristics and the way in which atrists judged the likelihood that a mental patient would commit they are used may help us understand how people make risk an act of violence within 6 months after discharge. When clini- judgments. cians learned that “20 out of every 100 patients similar to Mr. Jones are estimated to commit an act of violence” (26), 41% of The Affect Heuristic the clinicians refused to discharge the patient. However, when Although judgment and decision-making research has tra- another group of clinicians learned that “patients similar to Mr. ditionally focused on reason-based explanations for how we Jones are estimated to have a 20% chance of committing an act judge and decide, in recent years research efforts have increas- of violence” (26), only 21% refused to discharge the patient. ingly focused on the role of affect in decision making. Affect re- Follow-up studies (21) showed that the percentage representa- fers to the specific feeling of “goodness” or “badness” evoked tions of risk (10% or 20%) led to relatively benign images of one by a stimulus. Affective responses occur rapidly and automati- person, unlikely to harm anyone; the “equivalent” frequency cally and accompany virtually all cognitions (19). Damasio (20) representations, however, created frightening images of violent suggested that affective responses are not merely helpful to de- patients (e.g., “Some guy going crazy and killing someone” Volume 31, Number 1, 2006 Cancer Risk Perception 47

[21]). Slovic, Finucane, Peters, and MacGregor (21) suggested personal vulnerability to the disease. Equally biased estimates that these affect-laden images likely induced greater risk per- may result when a stereotype does not seem to fit the individual. ceptions in response to the frequency frame. These findings sug- This logic may help explain why certain individuals underesti- gest that the way in which likelihood information about a dis- mate their risk of disease if they feel that they do not physically ease or treatment (e.g., the risk of endometrial cancer, the resemble someone with the disease (5). benefit of tamoxifen) is communicated will impact affective re- Research on the representativeness heuristic has shown that actions and risk perceptions in turn. reliance on stereotypes often results in base rate neglect. Be- The relative lack of affect associated with a percentage pre- cause of the strong epidemiologic link between smoking and sentation of risk could help explain why people are hesitant to lung cancer, people tend to equate being a smoker with develop- accept that their risk of cancer is lower than they thought (27). ing lung cancer. In fact, data suggest that both smokers and non- Although it is not surprising that people act defensively upon smokers overestimate how likely smokers are to contract lung learning that their risk is higher than expected, it is remarkable cancer (36). The smoker/lung cancer stereotype may lead peo- that people fail to adopt the positive feedback that their risk is ple to mistakenly conclude that smokers are more likely to die low (28,29). We know that decision makers generally do not up- from lung cancer than heart disease (37). A similar phenomenon date beliefs quickly enough based on new information. As Ed- is observed when people are asked to judge whether women are wards (30) argued, “It takes anywhere from two to five observa- more likely to die of breast cancer or heart disease. Because tions to do one observation’s worth of work in inducing [a heart disease is more stereotypic of men, women tend to view decision maker] to change his opinions” (p. 18). However, it is their risk of breast cancer as greater than the risk of heart dis- also conceivable that these findings depend on the lack of affect ease, despite objective evidence to the contrary (38). in the risk communications used. Specifically, if people base How stereotypes influence cancer risk perception and treat- their cancer risk judgments partly on the worry and dread asso- ment decisions has not been widely studied. For example, peo- ciated with this disease, then changing such beliefs may depend ple may have a stereotype of the “cancer survivor.” For some, on manipulations that change affect as opposed to those that this stereotype might include images of chemotherapy side ef- work solely on rational belief change. In social psychology, re- fects, such as hair loss and uncontrollable nausea and vomiting. searchers have consistently found that affectively based atti- For others, this stereotype might be an image of a successful, tudes are best modified with affective as opposed to cognitive long-term survivor. We do not know how such stereotypes influ- persuasion (31). ence decisions about cancer screening and treatment. For exam- The affect heuristic also may explain why risk perception ple, if cancer patients do not fit the stereotype of the debilitated differs on the basis of the affective state of the individual. A per- cancer patient, might they fail to receive the support they need to son’s assessment of risk may be congruent. That is, some- maintain and improve their state of health? one in a more negative mood (e.g., due to an upsetting event at Stereotypes may influence cancer risk perceptions in other the office) may be more likely to overestimate his or her risk of ways. For example, a teenager who recently started smoking disease compared with someone in a more positive mood state. may not regard herself as a smoker because she does not identify For example, Johnson and Tversky (32) demonstrated that per- with the stereotype of a smoker, which in her mind is an older ceptions of risks, including cancers, were increased among indi- person who has smoked for many years. Consequently, she may viduals exposed to a negative mood manipulation (reading a be resistant to arguments about the dangers of smoking (39). For tragic newspaper story). As a result, two people in different af- some, the smoker stereotype may include a chronic cough or re- fective states (e.g., a physician in a “cold” cognitive state and a spiratory infections. Smokers who perceive themselves as gen- patient in a “hot” affective state due to of cancer) may not be erally healthy, and therefore not “fitting” the stereotype, may er- able to comprehend the other’s point of view. These “hot–cold roneously conclude that they are less vulnerable to empathy gaps” may result in miscommunication and poorer de- smoking-related diseases. cisions (33).

The Representativeness Heuristic The People invoke the representativeness heuristic when they The availability heuristic allows us to judge probabilities by estimate a probability or frequency based on similarity with a retrieving examples from memory. The easier it is to retrieve ex- stereotype, schema, or other pre-existing knowledge structure. amples of an event, the higher the estimated likelihood of occur- In a classic demonstration of this heuristic, people were asked to rence. In a classic demonstration of this linkage, Tversky and judge whether a person was an engineer or a lawyer after read- Kahneman (40) asked people to estimate whether more English ing about traits the target person possessed. Rather than consid- words began with the letter k or had the letter k in the third posi- ering the actual base rate of the two occupations, people tended tion. Although the latter is much more frequent, people tended to focus on specific traits associated with their stereotypes of en- to believe that more words began with k, probably because it is gineers or lawyers (34). In similar fashion, Gerend, Aiken, West, easier to recall words that start with the letter k. Analogously, and Erchull (35) found evidence that women invoked the repre- people who undergo genetic testing for cancer susceptibility sentativeness heuristic and used stereotypes of “the typical may overestimate their risk of cancer and underestimate the woman who gets breast cancer” in making judgments of their risks involved in genetic testing because it is easier to recall ex- 48 Peters et al. Annals of Behavioral Medicine amples of cancer morbidity and mortality than it is to recall the cancer deaths (as for a person who suffers over a long time) risks associated with testing. could leave some individuals with a permanent change to their Availability can also be influenced by other aspects of a sit- assessment of personal risk. uation that impact retrieval from memory, such as the ease with The availability heuristic is linked to Support Theory (47), which one is able to visualize an event. For example, the myth which proposes that people estimate probability on the basis of exists that lung cancer can spread when exposed to air. The ease the perceived strength of evidence for a hypothesis (“I will die of visualizing this scenario may lead some patients to reject po- from cancer”) relative to the strength of evidence for an alterna- tentially life-saving surgery (41). In other cases, people who can tive hypothesis (“I will die from some other disease or acci- readily imagine themselves having cancer may overestimate dent”). Support Theory predicts that probability judgments will their cancer risk. differ depending on the number of salient alternatives that are The availability heuristic suggests that people who have presented. That is, if a person is asked to judge the probability of had a recent personal experience with cancer would overesti- a single event in the absence of other alternatives, a relatively mate their risk of cancer. In fact, people who have a family his- large probability will be elicited; if asked to judge the probabil- tory of cancer typically judge their cancer risk to be higher than ity of the same event in the context of other possible events (thus people without a family history (42,43). A tendency to overesti- making other events “available”), a smaller probability will be mate cancer risk has also been observed, however, among peo- elicited. One explanation for this phenomenon is that people ple who have friends with cancer (22). This would explain why, tend to discount the possibility of events that are not explicitly as one physician noted, “If a man develops prostate cancer, ev- mentioned (48). Smoking findings (49,50) demonstrated that ery friend he plays golf with will be in to be checked” (44, risk assessment for smoking-related diseases depended greatly p. B1). on the number of diseases presented. Specifically, participants The news media are another important source of cancer in- indicated that the likelihood of a smoker dying of lung cancer formation. Highly publicized events, particularly those associ- was far greater when lung cancer was inquired about by itself (M ated with celebrities, are likely to be salient and therefore more = 48%) than when it was presented in a list of nine other possible readily remembered. A case in point is the publicity surrounding causes of death (M = 30%) (50). This “unpacking” of other pos- the death of Katie Couric’s husband from colon cancer and her sible causes influences nonexperts as well as experts (51). The own televised colonoscopy, which may have increased aware- theory has implications for those studies reporting that women ness of colon cancer and generated a heightened risk of disease. greatly overestimate their numerical risks of breast cancer when In fact, risk comparisons among different cancers appear to have they are asked for numerical risk and breast cancer is one of only shifted following Couric’s reports such that people were more a few available causes of death. It is possible that this overesti- mation would disappear by increasing the availability of other likely to overestimate their risk of colon cancer (22). Better evi- causes of death. dence of the effects of media reporting was described by Cram et al. (45), who demonstrated that colonoscopy rates rose signif- icantly in the months following a colorectal cancer awareness The Anchoring and Adjustment Heuristic campaign on the Today television show in March 2000. It is Anchoring and adjustment refers to the tendency for deci- reasonable to speculate that the campaign may have increased sion makers to be systematically influenced by salient, but not risk perceptions, thus promoting increases in colon cancer necessarily relevant, numbers (the anchors) when making a nu- screening. merical estimate (40). In a classic demonstration of this heuris- On the basis of the availability heuristic, we predict that al- tic, researchers pose a comparative and an absolute question. though media effects may be powerful temporarily, their influ- For example, an experimenter might spin the number 16,000 on ence will fade over time without exposure to new examples be- a wheel of fortune and ask the participant the comparative ques- cause memories fade with time. We were unable to identify any tion, “Is the Mississippi River longer or shorter than 16,000 longitudinal studies of cancer risk judgments to test this propo- miles?” The participant then would be asked the absolute ques- sition. To the extent that policymakers examine risk judgments tion: to estimate the length of the Mississippi River. When given as an indication of the public’s perception of effectiveness in a low-anchor value, individuals tend to make lower absolute es- battling cancer, however, one would want to take into account timates than when they are given a high-anchor value. Accord- highly visible media communications. ing to Tversky and Kahneman (40), this is due to insuffi- Availability appears to be influenced by the degree to which cient adjustment from the anchor. information is emotionally compelling and vivid. For example, Anchoring effects are robust. They may occur even if the in decision making, vivid testimonials are given more weight anchor value is known to be random and uninformative (e.g., a than statistical summaries (46). One might expect that the vivid- number on a wheel of fortune) or implausibly extreme (e.g., Is ness of another person’s cancer experience may influence one’s Mahatma Gandhi more or less than 140 years old?) (52). An- own risk judgments. The more vivid an event, the easier it choring effects occur independent of participants’ motivation should be to recall, and the longer term effect it should have. In (e.g., awarding a prize for the best estimate) (53) and regardless other words, a friend dying of cancer should have a larger, lon- of participants’ expertise (e.g., real estate agents estimating the ger impact on risk judgments than her rapid recovery from sur- value of a house) (54). Explicit instructions to correct for the po- gery and adjunct treatment. The frightening salience of some tential influence of an anchor fail to mitigate the effect (53). Volume 31, Number 1, 2006 Cancer Risk Perception 49

What anchors might people use when asked to judge their those who overestimate their risk may worry excessively, risk of developing cancer? One could easily imagine how risk overdo protective behaviors, and burden the health care system estimates appearing in the media might serve as anchors. For ex- (58,59). Risk perceptions are also used to prioritize the many ample, women have probably heard the statement that “1 in 8” risks that people face in life (58). An understanding of heuristics women are likely to develop breast cancer. A woman without a may provide insight into how people estimate their risk of devel- family history of breast cancer might conclude that her own risk oping cancer. is 1 in 8, whereas someone with a family history of breast cancer Although people employ heuristics every day in ways that might adjust upward from that anchor and perhaps conclude that generally produce good decisions, they tend to do so with lim- shehasa1in5risk. Alternatively, women with a strong family ited awareness of when, how, or why they are using heuristics, history or personal experience with cancer may initially anchor thus leaving themselves vulnerable to bad decisions. Consider their risk at 100%, reasoning that developing cancer is a ques- the following hypothetical scenarios. tion of when, not if. Adjustments then might be made downward from 100%, but they would likely be insufficient, producing • A teenager ignores warnings about the dangers of smok- overestimates. ing because only old people get lung cancer (the repre- Other risk estimates may rely on different anchors. When sentativeness heuristic). using percentile estimates, people may find 50% to be a particu- • A healthy 19-year-old schedules a colonoscopy after larly accessible number because it indicates uncertainty (“I watching Katie Couric’s report on colon cancer on televi- don’t know. It may happen or it may not”) (55,56). Although sion (the affect heuristic, the availability heuristic). 50% becomes accessible and salient for nonnumeric reasons, • A physician recommends a particular cancer treatment people may use it as an anchor, even though it is irrelevant to the for a patient because his two previous patients received numerical estimate. Of course, 50% is a gross overestimate of benefit from that treatment (the availability heuristic, an- most cancer risks, except in rare circumstances where a choring and adjustment heuristic). well-defined genetic risk exists. The anchoring heuristic can be insidious, resulting in unin- tended consequences. Question order is often important, be- In each of these examples, the use of heuristics may pro- cause people may anchor a numerical estimate on a previous nu- duce suboptimal decision making: The teenager’s stereotype of merical response. For example, a patient who responds that he the elderly bronchitic smoker prevents her from recognizing the pays 100% of his bills on time might unconsciously use 100% as health risks of smoking, the extreme worry about developing co- an anchor when asked to estimate his risk of developing cancer. lon cancer generated by Katie Couric’s report overrides the fact One way to avoid this bias is to use best-case and worst-case that there is no indication for colonoscopy in an otherwise scenarios to illustrate risk. Dillard, McCaul, Kelso, and Klein healthy 19-year-old, and recent success with a particular cancer (27), for example, discovered that providing a social compari- treatment prevents a physician from accurately assessing that son of someone who was “worse off” to women who considered treatment’s likelihood of success for a patient who might be themselves at high risk for breast cancer decreased their risk helped more by a different treatment. perceptions to more appropriate levels. These social compari- Limitations exist as to the use of heuristics in cancer risk sons may have worked because they required women to con- perceptions. First, in the present article, we focus on a single di- sider other possibilities. mension of risk—risk as likelihood—similar to how risk is used An important research question is how people adjust from in some models of health beliefs (60). We recognize, however, an anchor. Peters, Slovic, Hibbard, and Tusler (57) proposed that risk is a multifaceted construct. For example, other impor- that adjustment depends on the strength of affective feelings. tant features include the seriousness of the risk (61) and whether They examined the influence of anchors and subjective worry people can easily avoid the risk (56). Risk has been defined as a on fatality estimates of various causes of death. Providing differ- negative outcome, a hazard, a feeling of dread, and as a multidi- ent anchors (i.e., either the number of deaths from appendicitis, mensional construct of beliefs (62). Second, it is also important 400, or kidney disease, 40,000) influenced death estimates by to recognize that multiple heuristics may be operating in any pulling them lower or higher consistent with the anchor. In addi- given situation. For example, overestimating one’s risk of can- tion, worry about each cause correlated highly with death esti- cer because a friend has cancer may result from the ease with mates (r = .87), regardless of anchor condition. These results which the friend’s situation is recalled (the availability heuristic) suggest that estimates of the likelihood of a particular cancer or from the strong affect associated with the friend’s cancer ex- will be influenced by salient numerical anchors (generated by perience (the affect heuristic). Finally, separating out the effects the self, a physician, or the media) and the individual’s feelings, of heuristics from simple provision of information can be diffi- including the extent of his or her worry about cancer. cult. In experimental studies, one can vary the levels of one vari- able (e.g., the provided anchor or the availability of information through repetition), but in real life health situations we can ex- DISCUSSION amine only correlations. Thus, the predictive value of heuristics Risk perception is critical to cancer prevention and control. in the complexity of real life is unknown. People who underestimate their risk of developing cancer may Nevertheless, we believe that heuristics can be employed to be less likely to engage in health-protective behaviors, whereas produce more informed risk estimates and health decisions. 50 Peters et al. Annals of Behavioral Medicine

First, however, one must consider whether the goal of risk com- (2) Weinstein ND: The precaution adoption process. Health Psy- munication is persuasion (e.g., persuading a person to take a chology. 1988, 7:355–386. proven screening test, such as colonoscopy) or informed deci- (3) Edwards A, Unigwe S, Elwyn G, Hood K: Personalised risk sion making (e.g., providing information about a controversial communication for informed decision making about entering medical procedure, such as prostate-specific antigen testing). screening programs. The Cochrane Database of Systematic Re- For persuasion, the precision of a risk judgment is less important views. 2003, Art No. CD001865. DOI: 10.1002/14651858. CD001865. than whether a person takes the healthy action. Thus, health pro- (4) Gattellari M, Butow P, Tattersal MH: Sharing decisions in can- fessionals might focus on making information more or less sa- cer care. Social Science & Medicine. 2001, 52:1865–1878. lient in order to lead to an appropriate action. For example, using (5) Parascandola M, Hawkins J, Danis M: Patient autonomy and the availability heuristic, we should be able to increase the use of the challenge of clinical uncertainty. Kennedy Institute of Ethics colorectal screening by making it easier to visualize a celebrity, Journal. 2001, 12:245–264. a friend, or someone similar to the patient who has developed (6) Trumbo CW: Public requests for cancer cluster investigations: this cancer. Although this strategy might produce exaggerated A survey of state health departments. American Journal of Pub- risk perceptions, the goal of persuasion would be met nonethe- lic Health. 2000, 90:1300–1302. less. (7) National Cancer Institute: How the Public Perceives, Processes, However, a host of clinical situations exist in which there is and Interprets Risk Information: Findings From Focus Group no “right” decision. If this is the case, then health care providers Research With the General Public. Bethesda, MD: Office of must provide easily available, accurate, and timely information Cancer Communications, 1998. about risk perceptions that patients are able to use (63). An ap- (8) Croyle RT, Lerman C: Risk communication in genetic testing for cancer susceptibility. Journal of the National Cancer Insti- preciation of heuristic processes may be able to improve risk tute Monographs. 1999, 25:59–66. judgments in these cases. For example, women at increased risk (9) Reyna VF, Lloyd FJ, Whalen P: Genetic testing and medical de- of developing breast cancer can choose risk reduction mastec- cision making. Archives of Internal Medicine. 2001, tomy or other options, such as more aggressive follow-up 161:2406–2408. screening (64). All of the options have both costs and benefits (10) Kash KM, Holland JD, Halper MS, Miller DG: Psychological that need to be weighed by the individual patient (65), so in- distress and surveillance behaviors of women with a family his- formed decision making is critical. Making available informa- tory of breast cancer. Journal of the National Cancer Institute. tion about former patients who have chosen and experienced the 1992, 84:24–30. outcomes of each of the options should help current patients un- (11) Erblich J, Bovbjerg DH, Valdimarsdottir HB: Psychological derstand each option more fully so that they can make a better distress, health beliefs, and frequency of breast self-examina- informed choice. In addition, the health care provider could use tion. Journal of Behavioral Medicine. 2000, 23:277–292. information presentation techniques that may engage the affect (12) Slovic P: The construction of preference. American Psycholo- heuristic. For example, the physician could explain each option gist. 1995, 50:364–371. with frequencies rather than percentages (e.g., more than 90 of (13) Epstein S: Integration of the cognitive and the psychodynamic unconscious. American Psychologist. 1994, 49:709–724. 100 women—vs. less than 90% of women—at increased risk of (14) Kahneman D: A perspective on judgment and choice: Mapping breast cancer who receive risk reduction mastectomy will not bounded rationality. American Psychologist. 2003, develop this disease) (66). Although these changes may seem 58:697–720. quite minor, heuristics have potentially important and still (15) Loewenstein GF, Weber EU, Hsee CK, Welch ES: Risk as feel- largely unexplored implications for how people make and use ings. Psychological Bulletin. 2001, 127:267–286. cancer risk judgments. (16) Reventlow S, Hvas AC, Tulinius C: “In really great danger … ”: Heuristics, of course, are likely relevant to a wide range of The concept of risk in general practice. Scandinavian Journal health domains, from heart disease to West Nile virus. The trend of Primary Health Care. 2001, 19:71–75. toward greater shared decision making between patients and (17) Wilson TD, Dunn DS, Kraft D, Lisle DJ: Introspection, attitude their caregivers across health domains demands that health care change, and attitude–behavior consistency: The disruptive ef- providers understand what information patients understand and, fects of explaining why we feel the way we do. Advances in Ex- importantly, what information they use in their health decisions. perimental Social Psychology. 1989, 22:287–343. If a mismatch exists between the information that is understood (18) Peters E, Finucane ML, MacGregor DG, Slovic P: The bearable lightness of aging: Judgment and decision processes in older and used and what information should be understood and used, adults. In National Research Council, Committee on Future Di- then health care providers, if properly educated about the heuris- rections for Cognitive Research on Aging, Stern PC, tic processes that operate in the human mind, can apply this Carstensen LL (eds), The Aging Mind: Opportunities in Cogni- knowledge to improve patient health. tive Research. Washington, DC: National Academy Press, 2000, 144–165. (19) Murphy ST, Zajonc RB: Affect, cognition, and awareness: Af- REFERENCES fective priming with optimal and suboptimal stimulus expo- (1) McCaul KD, Tulloch HE: Cancer screening decisions. Cancer sures. Journal of Personality and Social Psychology. 1993, risk communication: What we know and what we need to learn. 64:723–739. Journal of the National Cancer Institute Monographs. 1999, (20) Damasio AR: Descartes’Error: , Reason, and the Hu- 25:52–58. man Brain. New York: Avon, 1994. Volume 31, Number 1, 2006 Cancer Risk Perception 51

(21) Slovic P, Finucane ML, Peters E, MacGregor DG: Risk as anal- (40) Tversky A, Kahneman D: Judgment under uncertainty: ysis and risk as feelings: Some thoughts about affect, reason, Heuristics and . Science. 1974, 185:1124–1131. risk, and rationality. Risk Analysis. 2004, 24:1–12. (41) O’Connor A: Finding the fact: Myth about lung cancer can be (22) Fiandt K, Pullen CH, Walker SN: Actual and perceived risk for deadly. New York Times. 2003, October 7, p. F5. chronic illness in rural older women. Clinical Excellence for (42) Lerman C, Lustbader E, Rimer BK, et al.: Effects of individual- Nurse Practitioners. 1999, 3:105–115. ized breast cancer risk counseling: A randomized trial. Journal (23) Rottenstreich Y, Hsee CK: Money, kisses, and electric shocks: of the National Cancer Institute. 1995, 87:286–292. On the affective psychology of risk. Psychological Science. (43) McCaul KD, Branstetter AD, O’Donnell SM, Jacobson K, 2001, 12:185–190. Quinlan KB: A descriptive study of breast cancer worry. Jour- (24) Watson M, Lloyd S, Davidson J, et al.: The impact of genetic nal of Behavioral Medicine. 1998, 21:565–579. counseling on risk perception and mental health in women with (44) Christie T: The health files: Doctor’s office: Men’s last stop on a family history of breast cancer. British Journal of Cancer. ride of pain. The Register-Guard. 2003, December 15, p. B1. 1999, 79:868–874. (45) Cram P, Fendrick AM, Inadomi J, et al.: The impact of a celeb- (25) Kraus N, Malmfors T, Slovic P: Intuitive toxicology: Expert rity promotional campaign on the use of colon cancer screen- and lay judgments of chemical risks. Risk Analysis. 1992, ing. Archives of Internal Medicine. 2003, 163:1601–1605. 12:215–232. (46) Borgida E, Nisbett RE: The differential impact of abstract vs. (26) Slovic P, Monahan J, MacGregor DG: Violence risk assessment concrete information on decisions. Journal of Applied Social and risk communication: The effects of using actual cases, pro- Psychology. 1977, 7:258–271. viding instructions, and employing probability vs. frequency (47) Rottenstreich Y,Tversky A: Unpacking, repacking, and anchor- formats. Law and Human Behavior. 2000, 24:271–296. ing: Advances in support theory. Psychological Review. 1997, (27) Dillard AJ, McCaul KD, Kelso PD, Klein WMP: Resisting 104:406–415. good news: Reactions to breast cancer risk communication. (48) Fischhoff B, Slovic P, Lichtenstein S: Fault trees: Sensitivity of Health Communication (in press). estimated failure probabilities to problem representation. Jour- (28) Hopwood P: Breast cancer risk perception: What do we know nal of Experimental Psychology: Human Perception and Per- and understand? Breast Cancer Research. 2000, 2:387–391. formance. 1978, 4:330–344. (29) Lipkus IM, Biradavolu M, Fenn K, Keller P, Rimer BK: In- (49) Slovic P: Rejoinder: The perils of Viscusi’s analyses of smok- forming women about their breast cancer risks: Truth and con- ing risk perceptions. Journal of Behavioral Decision Making. sequences. Health Communication. 2001, 13:205–226. 2000, 13:273–276. (30) Edwards W: Conservatism in human information processing. In (50) Windschitl PD: Judging the accuracy of a likelihood judgment: Kleinmuntz B (ed), Formal Representation of Human Judg- The case of smoking risk. Journal of Behavioral Decision ment. New York: Wiley, 1968, 17–52. Making. 2002, 15:19–35. (31) Millar MG, Tesser A: The effects of affective–cognitive consis- (51) Reyna VF, Adam MB: Fuzzy-trace theory, risk communication, tency and thought on the attitude–behavior relation. Journal of and product labeling in sexually transmitted diseases. Risk Experimental Social Psychology. 1989, 25:189–202. Analysis. 2003, 23:325–341. (32) Johnson EJ, Tversky A: Representations of perceptions of risk. (52) Strack F, Mussweiler T: Explaining the enigmatic anchoring ef- Journal of Experimental Psychology: General. 1984, fect: Mechanisms of selective accessibility. Journal of Person- 113:55–70. ality and Social Psychology. 1997, 73:437–446. (33) Loewenstein GF, Schkade D: Wouldn’t it be nice? Predicting (53) Wilson TD, Houston C, Etling KM, Brekke N: A new look at future feelings. In Diener E, Schwartz N, Kahneman D (eds), anchoring effects: Basic anchoring and its antecedents. Journal Well-Being: The Foundations of Hedonic Psychology. New of Experimental Psychology: General. 1996, 125:387–402. York: Russell Sage Foundation, 1999, 85–105. (54) Northcraft GB, Neal MA: An anchoring-and-adjustment per- (34) Tversky A, Kahneman D: Extensional versus intuitive reason- spective on property pricing decisions. Organizational Behav- ing: The in probability judgment. Psycho- ior & Human Decision Processes. 1987, 39:84–97. logical Review. 1983, 90:293–315. (55) Fischhoff B, Bruine de Bruin W: Fifty-fifty = 50%? Journal of (35) Gerend MA, Aiken LS, West SG, Erchull MJ: Beyond medical Behavioral Decision Making. 1999, 12:149–163. risk: Investigating the psychological factors underlying (56) Weinstein ND: What does it mean to understand a risk? Evalu- women’s perceptions of susceptibility to breast cancer, heart ating risk comprehension. Journal of the National Cancer Insti- disease, and osteoporosis. Health Psychology. 2004, tute Monographs. 1999, 25:15–20. 23:247–258. (57) Peters E, Slovic P, Hibbard J, Tusler M: Why worry? Worry, (36) Viscusi WK: Smoking: Making the Risky Decision. New York: risk perceptions, and willingness to act to reduce errors. Health Oxford University Press, 1992. Psychology (in press). (37) Strecher VJ, Kreuter MW, Kobrin SC: Do cigarette smokers (58) Kreuter MW: Dealing with competing and conflicting risks in have unrealistic perceptions of their heart attack, cancer, and cancer communication. Journal of the National Cancer Insti- stroke risks? Journal of Behavioral Medicine. 1995, 18:45–54. tute Monographs. 1999, 25:27–35. (38) Woloshin S, Schwartz LM, Black WC, Welch HG: Women’s (59) Kreuter MW, Strecher VJ: Changing inaccurate perceptions of perceptions of breast cancer risk: How you ask matters. Medi- health risk: Results from a randomized trial. Health Psychol- cal Decision Making. 1999, 19:221–229. ogy. 1995, 14:56–63. (39) Slovic P: What does it mean to know a cumulative risk? Adoles- (60) Strecher VJ, Champion VL, Rosenstock IW: The health belief cents’perceptions of short-term and long-term consequences of model and health behavior. In Gochman DS (ed), Handbook of smoking. Journal of Behavioral Decision Making. 2000, Health Behavior Research I: Personal and Social Determi- 13:259–266. nants. New York: Plenum, 1997, 71–91. 52 Peters et al. Annals of Behavioral Medicine

(61) Lipkus IM, Green LG, Marcus A: Manipulating perceptions of (64) Stefanek M, Helzlsouer K, Wilcox P: Predictors of and satisfac- colorectal cancer threat: Implications for screening intentions tion with bilateral prophylactic mastectomy. Preventive Medi- and behavior. Journal of Health Communication. 2003, cine. 1995, 24:412–419. 8:213–228. (65) Stefanek M, Hartmann L, Nelson W: Risk-reduction mastec- (62) Slovic P: Perception of risk. Science. 1987, 236:280–285. tomy: Clinical issues and research needs. Journal of the Na- (63) Hibbard JH, Peters E: Supporting informed consumer health tional Cancer Institute. 2001, 93:1297–1306. care decisions: Data presentation approaches that facilitate the (66) Hartmann L, Schaid DJ, Woods JE, et al.: Efficacy of bilateral use of information in choice. Annual Review of Public Health. prophylactic mastectomy in women with a family history of 2003, 24:413–433. breast cancer. New England Journal of Medicine. 1999, 340:77–84.