Background: Aversion to “ambiguity”—uncertainty about the reliability, credibility, or adequacy of information—about medical tests and treatments is an important psychological response that varies among individuals, but little is known about its nature and extent. The purpose of this study was to examine how individual-level ambiguity aversion relates to important health cognitions related to different cancer screening tests.

Methods: A survey of 1,074 adults, ages 40 to 70 years, was conducted in four integrated U.S. healthcare systems. The Ambiguity Aversion in Medicine (AA-Med) scale, a measure of individual differences in aversion to ambiguity (AA) about medical tests and treatments, was administered along with measures of several cancer screening-related cognitions: perceived benefits and harms of colonoscopy, mammography, and PSA screening, and ambivalence and future intentions regarding these tests. Multivariable analyses were conducted to assess the associations between AA-Med scores and cancer screening cognitions.

Results: Individual-level AA as assessed by the AA-Med scale was significantly associated (P < 0.05) with lower perceived benefits, greater perceived harms, and greater ambivalence about all three screening tests, and lower intentions for colonoscopy but not mammography or PSA screening.

Conclusion: Individual-level AA is broadly and simultaneously associated with various pessimistic cognitive appraisals of multiple cancer screening tests. The breadth of these associations suggests that the influence of individual-level AA is insensitive to the degree and nonspecific with respect to the causes of ambiguity.

Impact: Individual-level AA constitutes a measurable, wide-ranging cognitive bias against medical intervention, and more research is needed to elucidate its mechanisms and effects. Cancer Epidemiol Biomarkers Prev; 23(12); 2916–23. ©2014 AACR.

Uncertainty in health care is a problem that originates from several sources (1). One major source is what decision theorists since Ellsberg have termed “ambiguity”—uncertainty arising from limitations in the reliability, credibility, or adequacy of available information (2, 3). Ambiguity is a problem of growing significance in health care due to several recent trends: the increasingly rapid emergence and diffusion of new technologies such as genomic testing; heightened awareness—stimulated by the evidence-based medicine movement—of limitations in medical knowledge; expanded efforts to help patients make informed medical decisions based on an understanding of the benefits, harms, and uncertainties associated with medical tests and treatments; and widespread mass media coverage of medical controversies.

Ambiguity is also a significant problem in health care because it can cause people to appraise choice options pessimistically and to avoid decision making. This cognitive bias, known as “ambiguity aversion” (AA), has been demonstrated in numerous decision-making domains. It has cognitive, affective, and behavioral manifestations (4), and is distinguishable from other related but more general constructs including “pessimism” (a cognitive tendency toward negative expectations about future outcomes; ref. 5), lack of “trust” in physicians and medical institutions (a multidimensional construct signifying the optimistic acceptance of a vulnerable situation in which the truster believes a trustee will care for the truster's interests; ref. 6), and “preference for self-treatment” (a general propensity to avoid seeing physicians and using prescription drugs; ref. 7). These other constructs point to phenomena that may result in similar outcomes (e.g., avoidance of decision making), but are nonspecific in terms of their causes and are not produced by ambiguity per se.

In the domain of health care, AA poses potentially significant problems, to the extent that it may impede decision-making and lead people to forego potentially beneficial interventions in an unreflective, non-deliberative manner. Yet much remains unknown about the causes and effects of AA in health care. In non-health decision-making domains, for example, a substantial minority of people have been shown to be indifferent to ambiguity (3), suggesting the existence of individual differences in AA. This finding motivated development of the Ambiguity Aversion in Medicine (AA-Med) scale, a brief measure of individual differences in aversion to ambiguity regarding medical tests and treatments (4). As assessed by the AA-Med scale, greater individual-level AA (propensity toward aversion to ambiguity regarding medical tests and treatments) was found to be associated with lower interest in an ambiguous hypothetical cancer screening intervention. These findings corroborate the existence of individual differences in aversion to ambiguity about medical tests and treatments.

However, several important questions remain. The effects of individual differences in AA and the mechanisms of these effects have yet to be elucidated. The phenomenon of ambiguity aversion is characterized by pessimistic cognitive appraisals—for example, higher perceived risks, lower perceived benefits—of ambiguous choice options (8–10). In theory, individual-level AA should moderate this phenomenon; people with higher individual-level AA should demonstrate more pessimistic appraisals of ambiguous choice options than people with lower individual-level AA. Furthermore, these pessimistic appraisals should be specific to ambiguous as opposed to nonambiguous interventions. However, we currently lack empirical evidence on the nature and specificity of the effects of individual-level aversion to ambiguity about medical tests and treatments.

The overall objective of the current study was to address these knowledge gaps. Its specific aim was to determine how individual differences in aversion to ambiguity about medical tests and treatments relate to cognitive appraisals of medical tests that vary in associated ambiguity. The current study focused on cancer screening tests and the relationships between individual-level AA (measured by the AA-Med scale) and key health cognitions consistent with the phenomenon of ambiguity aversion (3, 10): pessimistic perceptions of the benefits and harms of screening, ambivalence about screening, and (lower) intentions to be screened. By examining these relationships, we sought to generate new insights about both the nature and specificity of the effects of individual-level ambiguity aversion on cognitive appraisals of medical interventions.

Data source and study population

This study was conducted as part of a larger National Cancer Institute (Bethesda, MD)-supported project, Cancer Research Network Across Health Care Systems (Grant Number U19 CA079689), conducted within the HMO Cancer Research Network (CRN), a consortium of research organizations affiliated with 14 nonprofit integrated healthcare delivery systems with nearly 11 million enrollees across the United States. The current study was part of a project to develop new measures to assess comprehension of spoken health messages related to cancer prevention and screening (11, 12). Four CRN sites participated in recruiting participants for the study: Kaiser Permanente Georgia, Kaiser Permanente Hawaii, Kaiser Permanente Colorado, and Fallon Community Health Plan in Massachusetts. The study protocol was approved by the Institutional Review Boards at each site.

Eligible participants were adults ages 40 to 70 years at the time of recruitment, who had been enrolled with one of the participating health plans for at least 5 years, were able to speak and understand English, and had no physical or psychological limitations that would interfere with participating in data collection efforts including interviews and surveys. To recruit a diverse study population, sampling at three sites (Fallon Community Health Plan, Kaiser Permanente Georgia, and Kaiser Permanente Hawaii) was stratified by U.S. Census–based estimates of educational level. At Kaiser Permanente Georgia, sampling was further stratified to ensure that African American and White members were invited in equal numbers within educational strata. At Kaiser Permanente Colorado, only Hispanic adults were recruited; potential participants were identified using health care system data on race/ethnicity and language preference to identify members who self-identified as Latino and had English as their preferred language. Recruitment methods included direct mailings, telephone follow-up, and offering multiple study session locations. Participants provided written informed consent and received $50 incentives for participation. Across the four sites, a total of 1,074 individuals participated and are included in the present study. Further details on participant recruitment and data collection procedures are reported elsewhere (11–13).

Data collection and measures

The study measures were part of a broad-ranging survey assessing numerous health cognitions and behaviors. Trained research assistants at each site administered all study measures in-person and orally during an approximately 2-hour session.

Ambiguity aversion was assessed using the AA-Med scale (Table 1), a 6-item measure assessing cognitive, affective, and behavioral manifestations of aversion to ambiguity about medical tests and treatments (4). Ambiguity aversion as assessed by the AA-Med scale has been shown to be associated with diminished interest in a hypothetical ambiguous cancer screening test (4). For the current study, wording of the original scale was adapted to make the items specific to cancer screening. AA-Med scores were calculated by averaging scores of individual scale items (range 1–4); participants with ≥2 missing items (n = 16) were dropped from analysis, leaving 1,057 participants in the analytic sample.

Table 1.

AA-Med scalePlease imagine that you are considering having a medical test that checks for cancer. Experts have conflicting opinions about this medical test. Tell me if you would strongly agree, agree, disagree, or strongly disagree with the following statements

1. I would still be willing to have the test. 
2. The conflicting opinions would make me upset. 
3. The conflicting opinions would lower my trust in the experts. 
4. I would be afraid of trying the test. 
5. I would avoid making a decision about the test. 
6. I would not have confidence in the test. 
1. I would still be willing to have the test. 
2. The conflicting opinions would make me upset. 
3. The conflicting opinions would lower my trust in the experts. 
4. I would be afraid of trying the test. 
5. I would avoid making a decision about the test. 
6. I would not have confidence in the test. 

NOTE: All items rated on 4-point Likert scale ranging from strongly disagree-strongly agree; item 1 reverse-coded.

Cancer screening cognitions assessed in the current study pertained to three different screening tests that vary in their associated ambiguity: colonoscopy for colon cancer screening, mammography for breast cancer screening, and PSA testing for prostate cancer screening. The comparative level of ambiguity—judged with respect to the strength of existing evidence and the degree of consensus among clinical practice guidelines—is greatest for PSA testing, least for colonoscopy, and intermediate for mammography (given the lack of definitive evidence about its mortality benefit in women ages 40–49 years; ref. 14). Perceived benefit of screening was assessed using two items. The first assessed the perceived mortality reduction associated with screening: “Having a (colonoscopy/mammogram/PSA test) will reduce my chances of dying from (colon cancer/breast cancer/prostate cancer).” The second assessed the perceived reassurance from undergoing screening: “Having a normal [colonoscopy/mammogram/PSA test] would reassure me that I do not have (colon cancer/breast cancer/prostate cancer).” Perceived harm of screening was also assessed using two items. The first assessed perceived discomfort associated with screening: “Having a (colonoscopy/mammogram/PSA test) is unpleasant.” The second assessed fear of screening: “Thinking about having a (colonoscopy/mammogram/PSA test) would make me nervous about what they might find.” Ambivalence about screening was assessed with a single item: “I have mixed feelings about having a (colonoscopy/mammogram/PSA test).” All items used a 4-point response scale ranging from “strongly disagree” to “strongly agree.” Intentions for future screening were assessed with a single item with a “yes/no” response option: “Do you plan on having a (colonoscopy/mammogram/PSA test)?” Both women and men completed questions about colonoscopy, whereas only women completed questions about mammography and only men completed questions about PSA testing.

Sociodemographic characteristics included age, sex, race (Black/African American, Asian/Pacific Islander, White, Hispanic, Other), and education (high school or less, some college, bachelor's degree, graduate degree).

Data analyses

Simple descriptive statistics were computed for all study variables. ANOVA was used to test univariate associations between both health cognitions and sociodemographic factors (independent variables) and AA-Med scores (dependent variable), and ANCOVA using general linear models was used to test multivariate associations adjusting for sociodemographic factors. We also conducted sensitivity analyses limited to the subpopulation of participants ≥50 years old, given that for adults ages 40 to 49 years, existing clinical practice guidelines do not recommend colonoscopy or PSA screening, and vary in their recommendations for mammography.

Among the total sample (N = 1067), 453 (42.5%) had a colonoscopy within the previous 5 years (earlier colonoscopy utilization data were unavailable), while among all females (N = 630), 485 (77.0%) had a mammogram within the previous one year, and among all males (N = 437), 211 (48.3%) had a PSA test within the previous year. Other characteristics of the study population are listed in Table 2. The sample had substantial sociodemographic diversity; 51.6% were non-White and 26.5% had an educational level of high school or less. The AA-Med scale demonstrated acceptable internal consistency reliability (α = 0.77).

Table 2.

Sociodemographic characteristics and their association with individual-level ambiguity aversion as assessed by the AA-Med scale

Ambiguity aversion (AA-Med score)a
N%Mean (SD)P
Age    0.12 
 40–49 191 18.1 2.31 (0.44)  
 50–59 394 37.3 2.24 (0.39)  
 60–69 448 42.4 2.23 (0.38)  
 70–74 24 2.3 2.18 (0.41)  
Sex    0.31 
 Female 624 59.0 2.26 (0.39)  
 Male 433 41.0 2.23 (0.40)  
Education    0.03 
 High school or less 279 26.5 2.30 (0.40)  
 Some college 289 27.4 2.23 (0.39)  
 Bachelor's degree 258 24.5 2.24 (0.42)  
 Graduate degree 229 21.7 2.21 (0.37)  
Race    0.13 
 Black/African American 145 13.7 2.25 (0.42)  
 Asian/Pacific Islander 141 13.3 2.27 (0.40)  
 White 512 48.4 2.25 (0.40)  
 Hispanic 196 18.5 2.27 (0.37)  
 Other/unknown 63 6.0 2.12 (0.44)  
Ambiguity aversion (AA-Med score)a
N%Mean (SD)P
Age    0.12 
 40–49 191 18.1 2.31 (0.44)  
 50–59 394 37.3 2.24 (0.39)  
 60–69 448 42.4 2.23 (0.38)  
 70–74 24 2.3 2.18 (0.41)  
Sex    0.31 
 Female 624 59.0 2.26 (0.39)  
 Male 433 41.0 2.23 (0.40)  
Education    0.03 
 High school or less 279 26.5 2.30 (0.40)  
 Some college 289 27.4 2.23 (0.39)  
 Bachelor's degree 258 24.5 2.24 (0.42)  
 Graduate degree 229 21.7 2.21 (0.37)  
Race    0.13 
 Black/African American 145 13.7 2.25 (0.42)  
 Asian/Pacific Islander 141 13.3 2.27 (0.40)  
 White 512 48.4 2.25 (0.40)  
 Hispanic 196 18.5 2.27 (0.37)  
 Other/unknown 63 6.0 2.12 (0.44)  

aMean scores on AA-Med measure (range, 1–4); higher scores indicate greater ambiguity aversion.

Table 2 also shows the associations between sociodemographic characteristics and individual-level AA, as assessed by the AA-Med scale. Higher AA was associated with lower education but not with other sociodemographic characteristics. Table 3 shows the adjusted associations between individual-level AA and different cancer screening cognitions. The pattern of associations was broad and nonspecific; higher individual-level AA was significantly associated (P < 0.05) with nearly all cognitions examined, including lower perceived benefits (mortality reduction, reassurance), greater perceived harms (unpleasantness, fear of screening), and greater ambivalence about screening. Notably, the associations between individual-level AA and these multiple screening-related cognitions were also broadly consistent across all three cancer screening tests (colonoscopy, mammography, PSA testing). The exceptions were perceived reassurance from PSA screening and perceived unpleasantness of mammography, which were not associated with individual-level AA, and future intentions for cancer screening. Higher individual-level AA was associated with lower intentions for only colonoscopy (the least ambiguous of the three tests), and not for mammography or PSA testing (the most ambiguous of the tests). Sensitivity analyses (not shown) restricting the analytic sample to participants ≥50 years old showed no significant differences in either the direction or the magnitude of the observed associations.

Table 3.

Associations between cancer screening cognitions and individual-level ambiguity aversion as assessed by the AA-Med scale

Ambiguity aversion (AA-Med score)a
N%Mean (95% CI)P
Perceived benefit: mortality reduction 
Colonoscopy    0.007 
 Strongly agree 238 23.3 2.14 (2.08–2.21)  
 Agree 614 60.1 2.22(2.17–2.28)  
 Disagree 138 13.5 2.24 (2.16–2.31)  
 Strongly disagree 31 3.0 2.36 (2.22–2.51)  
Mammography    0.002 
 Strongly agree 125 20.1 2.08 (1.98–2.17)  
 Agree 380 61.2 2.22 (2.14–2.29)  
 Disagree 96 15.5 2.25 (2.14–2.35)  
 Strongly disagree 20 3.2 2.29 (2.11–2.48)  
PSA    <0.001 
 Strongly agree 39 15.2 1.99 (1.86–2.13)  
 Agree 164 63.8 2.24 (2.15–2.33)  
 Disagreeb 54 21.0 2.24 (2.12–2.36)  
Perceived benefit: reassurance 
Colonoscopy    <0.001 
 Strongly agree 214 21.0 2.13 (2.06–2.19)  
 Agree 749 73.4 2.24 (2.18–2.29)  
 Disagreeb 58 5.7 2.29 (2.18–2.40)  
Mammography    <0.001 
 Strongly agree 100 16.1 2.05 (1.95–2.15)  
 Agree 447 71.8 2.23 (2.15–2.30  
 Disagreeb 76 12.2 2.22 (2.11–2.34)  
PSA    0.07 
 Strongly agree 25 9.7 2.05 (1.89–2.21)  
 Agree 188 73.2 2.23 (2.14–2.33)  
 Disagreeb 44 17.1 2.20 (2.07–2.33)  
Perceived harm: unpleasant 
Colonoscopy    <0.001 
 Strongly agree 190 18.7 2.21 (2.14–2.28)  
 Agree 513 50.6 2.25 (2.19, 2.30)  
 Disagree 276 27.2 2.20 (2.14–2.27)  
 Strongly disagree 35 3.5 1.90 (1.77–2.03)  
Mammography    0.91 
 Strongly agree 123 19.7 2.18 (2.08–2.28)  
 Agree 341 54.7 2.21 (2.13–2.29)  
 Disagree 149 23.9 2.19 (2.10–2.28)  
 Strongly disagree 10 1.6 2.19 (1.93–2.44)  
PSA    <0.001 
 Agreec 33 13.8 2.38 (2.24–2.52)  
 Disagree 156 65.3 2.20 (2.12–2.29)  
 Strongly disagree 50 20.9 1.99 (1.86–2.11)  
Perceived harm: fear 
Colonoscopy    <0.001 
 Strongly agree 39 3.8 2.30 (2.17–2.42)  
 Agree 380 37.2 2.32 (2.26–2.37)  
 Disagree 503 49.3 2.18 (2.13–2.23)  
 Strongly disagree 99 9.7 1.88 (1.79–1.96)  
Mammography    <0.001 
 Strongly agree 12 1.9 2.26 (2.03–2.49)  
 Agree 176 28.3 2.28 (2.20–2.37)  
 Disagree 352 56.5 2.22 (2.14–2.29)  
 Strongly disagree 83 13.3 1.98 (1.88–2.09)  
PSA    <0.001 
 Agreec 62 24.1 2.37 (2.26–2.48)  
 Disagree 152 59.1 2.20 (2.11–2.29)  
 Strongly disagree 43 16.7 1.93 (1.80–2.05)  
Future screening intentions 
Colonoscopy    0.004 
 Yes 891 87.5 2.19 (2.14–2.24)  
 No 127 12.5 2.30 (2.22–2.37)  
Mammography    0.15 
 Yes 609 98.4 2.19 (2.12–2.27)  
 No 10 1.6 2.37 (2.12–2.62)  
PSA    0.87 
 Yes 211 87.9 2.19 (2.10–2.28)  
 No 29 12.1 2.18 (2.02–2.34)  
Ambivalence about screening 
Colonoscopy    <0.001 
 Strongly agree 40 3.9 2.37 (2.24–2.49)  
 Agree 268 26.3 2.30 (2.24–2.36)  
 Disagree 573 56.1 2.21 (2.15–2.26)  
 Strongly disagree 140 13.7 1.94 (1.87–2.02)  
Mammography    <0.001 
 Strongly agree 1.4 2.71 (2.46–2.96)  
 Agree 79 12.7 2.40 (2.30–2.50)  
 Disagree 421 67.6 2.22 (2.15–2.29)  
 Strongly disagree 114 18.3 1.95 (1.85–2.04)  
PSA    <0.001 
 Agreec 40 15.6 2.37 (2.24–2.50)  
 Disagree 167 65.0 2.24 (2.15–2.33)  
 Strongly disagree 50 19.4 1.96 (1.84–2.08)  
Ambiguity aversion (AA-Med score)a
N%Mean (95% CI)P
Perceived benefit: mortality reduction 
Colonoscopy    0.007 
 Strongly agree 238 23.3 2.14 (2.08–2.21)  
 Agree 614 60.1 2.22(2.17–2.28)  
 Disagree 138 13.5 2.24 (2.16–2.31)  
 Strongly disagree 31 3.0 2.36 (2.22–2.51)  
Mammography    0.002 
 Strongly agree 125 20.1 2.08 (1.98–2.17)  
 Agree 380 61.2 2.22 (2.14–2.29)  
 Disagree 96 15.5 2.25 (2.14–2.35)  
 Strongly disagree 20 3.2 2.29 (2.11–2.48)  
PSA    <0.001 
 Strongly agree 39 15.2 1.99 (1.86–2.13)  
 Agree 164 63.8 2.24 (2.15–2.33)  
 Disagreeb 54 21.0 2.24 (2.12–2.36)  
Perceived benefit: reassurance 
Colonoscopy    <0.001 
 Strongly agree 214 21.0 2.13 (2.06–2.19)  
 Agree 749 73.4 2.24 (2.18–2.29)  
 Disagreeb 58 5.7 2.29 (2.18–2.40)  
Mammography    <0.001 
 Strongly agree 100 16.1 2.05 (1.95–2.15)  
 Agree 447 71.8 2.23 (2.15–2.30  
 Disagreeb 76 12.2 2.22 (2.11–2.34)  
PSA    0.07 
 Strongly agree 25 9.7 2.05 (1.89–2.21)  
 Agree 188 73.2 2.23 (2.14–2.33)  
 Disagreeb 44 17.1 2.20 (2.07–2.33)  
Perceived harm: unpleasant 
Colonoscopy    <0.001 
 Strongly agree 190 18.7 2.21 (2.14–2.28)  
 Agree 513 50.6 2.25 (2.19, 2.30)  
 Disagree 276 27.2 2.20 (2.14–2.27)  
 Strongly disagree 35 3.5 1.90 (1.77–2.03)  
Mammography    0.91 
 Strongly agree 123 19.7 2.18 (2.08–2.28)  
 Agree 341 54.7 2.21 (2.13–2.29)  
 Disagree 149 23.9 2.19 (2.10–2.28)  
 Strongly disagree 10 1.6 2.19 (1.93–2.44)  
PSA    <0.001 
 Agreec 33 13.8 2.38 (2.24–2.52)  
 Disagree 156 65.3 2.20 (2.12–2.29)  
 Strongly disagree 50 20.9 1.99 (1.86–2.11)  
Perceived harm: fear 
Colonoscopy    <0.001 
 Strongly agree 39 3.8 2.30 (2.17–2.42)  
 Agree 380 37.2 2.32 (2.26–2.37)  
 Disagree 503 49.3 2.18 (2.13–2.23)  
 Strongly disagree 99 9.7 1.88 (1.79–1.96)  
Mammography    <0.001 
 Strongly agree 12 1.9 2.26 (2.03–2.49)  
 Agree 176 28.3 2.28 (2.20–2.37)  
 Disagree 352 56.5 2.22 (2.14–2.29)  
 Strongly disagree 83 13.3 1.98 (1.88–2.09)  
PSA    <0.001 
 Agreec 62 24.1 2.37 (2.26–2.48)  
 Disagree 152 59.1 2.20 (2.11–2.29)  
 Strongly disagree 43 16.7 1.93 (1.80–2.05)  
Future screening intentions 
Colonoscopy    0.004 
 Yes 891 87.5 2.19 (2.14–2.24)  
 No 127 12.5 2.30 (2.22–2.37)  
Mammography    0.15 
 Yes 609 98.4 2.19 (2.12–2.27)  
 No 10 1.6 2.37 (2.12–2.62)  
PSA    0.87 
 Yes 211 87.9 2.19 (2.10–2.28)  
 No 29 12.1 2.18 (2.02–2.34)  
Ambivalence about screening 
Colonoscopy    <0.001 
 Strongly agree 40 3.9 2.37 (2.24–2.49)  
 Agree 268 26.3 2.30 (2.24–2.36)  
 Disagree 573 56.1 2.21 (2.15–2.26)  
 Strongly disagree 140 13.7 1.94 (1.87–2.02)  
Mammography    <0.001 
 Strongly agree 1.4 2.71 (2.46–2.96)  
 Agree 79 12.7 2.40 (2.30–2.50)  
 Disagree 421 67.6 2.22 (2.15–2.29)  
 Strongly disagree 114 18.3 1.95 (1.85–2.04)  
PSA    <0.001 
 Agreec 40 15.6 2.37 (2.24–2.50)  
 Disagree 167 65.0 2.24 (2.15–2.33)  
 Strongly disagree 50 19.4 1.96 (1.84–2.08)  

aMean scores on AA-Med measure (range 1–4) from multivariate ANCOVA models adjusted for age, sex, race, and education; higher scores indicate greater ambiguity aversion; mammography and PSA analyses limited to females and males, respectively, and not adjusted for sex.

bWhen N ≤ 5 for “strongly disagree,” collapsed and included in “disagree.”

cWhen N ≤ 5 for “strongly agree,” collapsed and included in “agree.”

This study sheds new light on the nature and mechanisms of individual differences in aversion to ambiguity about medical tests and treatments. Individual-level AA was broadly and simultaneously associated with multiple pessimistic appraisals—including perceived benefits and harms and greater ambivalence—pertaining to multiple cancer screening tests (colonoscopy, mammography, PSA testing). These associations have several implications about the mechanisms by which individual-level AA influences behavior.

First, the overall directions of the observed associations are consistent with the phenomenon of AA, and thus provide indirect support for the hypothesis that individual-level ambiguity aversion promotes more pessimistic appraisals of ambiguous choice options. This conclusion is preliminary because definitive evidence for a moderating role of individual-level AA can only be obtained by directly assessing how people with different levels of this attribute respond to ambiguous information. Our study, however, provides initial supportive evidence that individual-level AA moderates people's responses to ambiguous medical interventions, and raises the need for further research to confirm these findings.

The more significant implications of our findings relate to the breadth of the potential effects of individual-level AA. Overall, our findings suggest that these effects are insensitive to the degree of ambiguity associated with medical interventions. Patients with higher AA-Med scores perceived lower benefits and greater harms and reported greater ambivalence about all three cancer screening tests, including not only PSA testing (a clearly ambiguous intervention), but also mammography and colonoscopy—both of which are well supported by existing scientific evidence and clinical practice guidelines, and hence less ambiguous.

In addition, our findings suggest that the potential effects of individual-level AA are nonspecific with respect to the ultimate cause of ambiguity surrounding cancer screening tests. From a normative perspective, any existing ambiguity surrounding the three cancer screening tests (colonoscopy, mammography, PSA testing) pertains primarily to the mortality benefit of the test. Yet patients with higher individual-level AA demonstrated more pessimistic appraisals not only of the tests' mortality benefit, but also their reassurance value and associated harms (perceived discomfort, aversive findings). Both the nonspecific and ambiguity-insensitive pattern of these associations suggests that the negative cognitive appraisals associated with individual-level AA result from heuristic rather than systematic reasoning processes. Individual-level AA as measured by the AA-Med scale may ultimately constitute a nondeliberative cognitive bias against medical interventions in general.

At the same time, our findings also suggest that the effects of this anti-interventionist bias on decision making are not straightforward. Among all of the health cognitions measured in our study, future cancer screening intentions were the most direct proxy for actual behavior and indicator of AA. Yet contrary to predictions, higher AA-Med scores were negatively associated with lower cancer screening intentions only for colonoscopy (the least ambiguous of the three tests examined) and not for PSA screening (the most ambiguous) or mammography. These findings suggest that the effects of individual differences in AA on screening intentions (and ultimately behaviors) are moderated and/or mediated by other, unmeasured factors. In the case of PSA screening, for example, prevailing standards of care and the attitudes of physicians and patients may increase screening intentions despite the ambiguity surrounding the net benefits of this intervention. Notably, data collection for this study occurred before release of the 2012 revised U.S. Preventive Services Task Force recommendation against PSA screening, and the highly-publicized controversy about this recommendation. Nevertheless, the ambiguity-insensitive nature of the associations between individual-level AA and future cancer screening intentions is counterintuitive, and raises the need for more research to understand the factors that modify the effects of individual-level AA.

More research is also needed to address other study limitations and alternative explanations for our findings. The current study was part of a larger study in which participants were exposed to other spoken health messages related to cancer prevention and screening; this exposure may thus have influenced participants' perceptions of the three screening tests. We do not believe that any such influence was significant, however, given that the AA measure was administered before participants were exposed to cancer prevention and screening messages. One might also argue that all of the cancer screening tests examined in our study are ambiguous to at least some degree, and that firm inferences thus cannot be drawn about how individual-level AA affects patients' responses to unambiguous interventions. Mammography for breast cancer screening in women ages 40 to 49 has been a matter of controversy, for example, as has the question of the optimal screening test for colon cancer. We believe there are still meaningful differences in the degree of ambiguity associated with these tests; however, more research—examining a greater number of medical tests and treatments of varying ambiguity—is required to establish how individual-level AA affects health cognitions and behaviors. Furthermore, the degree of ambiguity that patients themselves perceive—regardless of whatever ambiguity exists from the perspective of experts—remains an open question and the ultimate determinant of patients' judgments and decisions. We did not assess these or related perceptions, such as patients' familiarity with the three screening tests or any associated controversies. Assessing these perceptions is an important focus for future research.

One might also argue that the observed associations between individual-level AA and at least some of the cognitions examined in this study—for example, ambivalence and lower intentions about breast, colon, and prostate cancer screening—are simply a measurement artifact attributable to the fact that the AA-Med scale ascertains closely related cognitions. We believe that this explanation is unlikely, however, given that the AA-Med scale assesses attitudes toward an unspecified “medical test that checks for cancer”—not colonoscopy, mammography, PSA testing, or any other specific cancer screening test (Table 1). Furthermore, the AA-Med scale focuses not on any test whatsoever but specifically on ambiguous tests for which there are conflicting expert opinions; it neither mentions breast, colon, or prostate cancer screening tests, nor attributes conflicting expert opinions to these tests. Nevertheless, more research—measuring individual-level AA in different ways and with differing levels of domain specificity—is needed to understand the extent to which this attribute affects health cognitions and behaviors.

Other study limitations include the sample population, which was sociodemographically diverse but limited to members of large, mature, integrated health care delivery systems that provide a high level of access to health information and services. Although the consistency of our findings with past theory and empirical evidence on AA supports their validity, more research is needed to determine the generalizability of our findings to other patient populations. Corroborating this need, the current study diverged from the original AA-Med scale development study in demonstrating significant associations between AA-Med scores and education level only, and not age, sex, or race (4). At least some of these negative findings are arguably unexpected; for example, one might have sex-based differences in AA given that females are often more engaged in decision making than males (14, 15). Notably, the former study notably utilized a national sample with a substantial proportion of patients aged less than 40 years (39%) and greater than 70 years (10%), as well as smaller proportions of minorities (31%) and college graduates (31%) than the current study. The influence of sociodemographic factors on ambiguity aversion is thus an important focus for future research, particularly since the observed influence of factors such as education level raises the interesting question of whether individual-level AA might be alterable through educational interventions. Finally, our study also assessed behavioral intentions rather than actual cancer screening behaviors. Behavioral intentions predict behaviors and are a central element in health behavior theories such as the Theory of Planned Behavior; however, multiple factors influence their effects (16). Further research examining actual screening behaviors is needed to definitively establish the predictive validity of the AA-Med scale and the relationship between individual differences and actual manifestations of AA.

Despite these limitations, the current study provides important preliminary evidence that individual-level aversion to ambiguity about medical tests and treatments constitutes a broad cognitive bias against medical intervention, characterized by pessimistic cognitive appraisals that are both insensitive to the degree and nonspecific with respect to the causes of ambiguity. The study thus sheds new light on an old problem. Nearly four decades ago, Sackett and Holland famously characterized people who oppose screening interventions in the absence of definitive scientific evidence as “snails,” in contrast with “evangelists” who advocate interventions despite uncertainty (17). The individual differences separating snails and evangelists have since remained poorly understood, but the current work furnishes seminal evidence that AA constitutes a measurable, wide-ranging cognitive bias that contributes to these differences. It remains for further research to elucidate the causes and mechanisms of this bias, and the ways in which clinicians might help patients make better decisions under conditions of ambiguity.

No potential conflicts of interest were disclosed.

Conception and design: P.K.J. Han, A.E. Williams, K.M. Mazor

Development of methodology: P.K.J. Han, A. Haskins, K.M. Mazor

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A.E. Williams, K.M. Mazor

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): P.K.J. Han, A. Haskins, F. Lee Lucas, W.M.P. Klein

Writing, review, and/or revision of the manuscript: P.K.J. Han, A.E. Williams, A. Haskins, C. Gutheil, F. Lee Lucas, W.M.P. Klein, K.M. Mazor

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): P.K.J. Han, C. Gutheil, K.M. Mazor

Study supervision: P.K.J. Han, K.M. Mazor

This work was supported by the National Cancer Institute (grant U19 CA079689; to K.M. Mazor, A.E. Williams, and P.K.J. Han).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Han
PK
,
Klein
WM
,
Arora
NK
. 
Varieties of uncertainty in health care: a conceptual taxonomy
.
Med Decis Making
2011
;
31
:
828
38
.
2.
Ellsberg
D
. 
Risk, ambiguity, and the Savage axioms
.
Quart J Econ
1961
;
75
:
643
69
.
3.
Camerer
C
,
Weber
M
. 
Recent developments in modeling preferences: uncertainty and ambiguity
.
J Risk Uncertainty
1992
;
5
:
325
70
.
4.
Han
PK
, et al
Aversion to ambiguity regarding medical tests and treatments: measurement, prevalence, and relationship to sociodemographic factors
.
J Health Commun
2009
;
14
:
556
72
.
5.
Chang
EC
,
Chang
R
,
Sanna
LJ
. 
Optimism, pessimism, and motivation: relations to adjustment
.
Social Pers Psych Compass
2009
;
3
:
494
506
.
6.
Hall
MA
,
Dugan
E
,
Zheng
B
,
Mishra
AK
. 
Trust in physicians and medical institutions: what is it, can it be measured, and does it matter
?
Milbank Q
2001
;
79
:
613
39
.
7.
Ganther
JM
,
Wiederholt
JB
,
Kreling
DH
. 
Measuring patients' medical care preferences: care seeking versus self-treating
.
Med Decis Making
2001
;
21
:
133
40
.
8.
Einhorn
HJ
,
Hogarth
RM
. 
Ambiguity and uncertainty in probabilistic inference
.
Psychol Rev
1985
;
92
:
433
61
.
9.
Viscusi
WK
. 
Alarmist decisions with divergent risk information
.
Econ J
1997
;
107
:
1657
70
.
10.
Han
PK
,
Moser
RP
,
Klein
WM
. 
Perceived ambiguity about cancer prevention recommendations: relationship to perceptions of cancer preventability, risk, and worry
.
J Health Commun
2006
;
11
:
51
69
.
11.
Mazor
KM
,
Haley
HL
,
Sullivan
K
,
Quirk
ME
. 
The video-based test of communication skills: description, development, and preliminary findings
.
Teach Learn Med
2007
;
19
:
162
7
.
12.
Mazor
KM
,
Roblin
DW
,
Williams
AE
,
Greene
SM
,
Gaglio
B
,
Field
TS
, et al
Health literacy and cancer prevention: two new instruments to assess comprehension
.
Patient Educ Couns
2012
;
88
:
54
60
.
13.
Mazor
KM
,
Rogers
HJ
,
Williams
AE
,
Roblin
DW
,
Gaglio
B
,
Field
TS
, et al
The cancer message literacy tests: psychometric analyses and validity studies
.
Patient Educ Couns
2012
;
89
:
69
75
.
14.
Nelson
HD
,
Tyne
K
,
Naik
A
,
Bougatsos
C
,
Chan
BK
,
Humphrey
L
, et al
Screening for breast cancer: an update for the U.S. Preventive Services Task Force
.
Ann Intern Med
2009
;
151
:
727
37
,
W237
42
.
15.
Arora
NK
,
McHorney
CA
. 
Patient preferences for medical decision making: who really wants to participate
?
Med Care
2000
;
38
:
335
41
.
16.
Connor
M
,
Norman
P
, eds. 
Predicting Health Behavior: Research and Practice With Social Cognition Models
.
Open University Press
:
Buckingham
; 
2001
.
230
.
17.
Sackett
DL
,
Holland
WW
. 
Controversy in the detection of disease
.
Lancet
1975
;
2
:
357
9
.