ORAL ABSTRACTS: DECISION BIASES, AFFECT AND INFORMATION PROCESSING
Victoria A. Shaffer, PhD
University of Missouri
School of Health Professions; Department of Psychological Sciences
Evidence is accumulating that risk feelings are more strongly related to health decision making than risk cognitions, but little research has examined possible reasons for these effects. One possibility is that, because risk feelings reflect a more accessible way of thinking about probabilities compared to cognitive probability judgments, people may be less uncertain about their risk when thinking in affective terms. The present study investigates this in the context of don’t know (DK) responding to risk perception survey questions. We expected the frequency of DK responding to be higher for cognitive risk perceptions compared to affective risk perceptions.
Through secondary analysis of data collected for a risk communication experiment (N=835), the present study investigated the frequency of DK responding to cognitive and affective perceived risk measures related to colon cancer and other physical activity-related diseases (i.e., diabetes, stroke, heart disease). Absolute risk perceptions and comparative risk perceptions were assessed for each of the measures, resulting in a total of eight risk perception items. All items included an explicit DK response option. The chi-squared, one variable test was used to test the difference in proportions.
DK responding was 24.3% vs. 15.0% for cognitive and affective absolute risk towards colon cancer respectively. DK responding was 19.0% vs. 17.0% for cognitive and affective comparative risk towards colon cancer respectively. For other diseases, DK responding was 13.2% vs. 10.5% for cognitive and affective absolute risk respectively and 12.6% vs. 11.3% for cognitive and affective comparative risk respectively. The differences in proportions were significant for the absolute risk questions, (i.e., p<0.001 and p=0.02 for colon cancer and other diseases respectively) and not significant for the comparative risk questions (i.e., p=0.13 and p=0.25 for colon cancer and other diseases respectively).
Evidence that affective risk beliefs might be more accessible resulting in less DK responding compared to cognitive risk beliefs was only found for how people feel about their absolute risk of disease and not for their comparative risk of disease. Additional research is needed to better understand these contextual differences to advance theory and the development of more effective communication practices that better serve the informational needs of the general public.
Method: US nationally representative samples of adult (n=736) and teen (n=469) smokers were randomly assigned to view either text-only warning labels, warnings pairing text with images that elicited little emotional response, or warnings pairing text with images that elicited a significant negative emotional response. Participants viewed nine warnings from their experimental condition four times over two weeks. Participants reported their emotional reactions to the warnings on their first and fourth exposure. Either immediately or six weeks after the fourth exposure, participants reported their smoking risk perceptions and quit intentions.
Result: Consistent with predictions, participants in the high-emotion condition reported more arousal than participants in the text-only condition (bAdult = .21; bTeen = .27, p’s<.004). Participants in the low-emotion condition reported less arousal than participants in the text-only condition (bAdult = -.18; bTeen = -.22, p’s<.018). Greater arousal led to increased risk perceptions in both samples (bAdult =.66; bTeen = .85, p’s<.001) and quit intentions among adults (bAdult = 1.00, p<.001). Compared to text-only warnings, low-emotion warnings led to reduced risk perceptions and quit intentions whereas high-emotion warnings led to increased risk perceptions and quit intentions.
Conclusion: Similarly sized graphic cigarette warning labels are not all created equal. Although larger than text alone, graphic warnings which did not elicit a stronger emotional response from smokers led to reduced risk perceptions and quit intentions vs. text-only warnings. Conversely, graphic warnings which elicited a significant negative emotional response led to increased risk perceptions and quit intentions vs. text-only warnings. This suggests that the ability to elicit negative emotions may drive the effectiveness of health warnings, and warnings with images that do not elicit an emotional response are unlikely to be more beneficial than text-only warnings. Results have implications for communication of health-risk information in other medical domains as well.
Patients have difficulty understanding medication risks and communication approaches to risk often lack theoretical motivation. Applying fuzzy-trace theory and empirical research, we examined how theoretically and empirically motivated enhancements of risk messages influence three outcome variables: ordinal risk perception, worry about experiencing side effects, and willingness (intentions) to start a new medication.
We contacted 1453 patients with a chronic inflammatory rheumatic disease. 465 patients (mean age 58.99 years, 79.7% female) completed and returned a mail survey. Patients considered the acceptability of a medication with a realistic type and level of risk (a 2% or 0.2% chance of developing a serious infection). We also varied factorially the presence/absence of an icon array (IA) and a balance image (BI). Thus, patients were randomly assigned to one of eight conditions. Benefit, cost and route of administration were held constant. Data were analyzed using structural equation modeling.
Neither quantitative level of risk nor the presence of IA or BI affected willingness to start medication by themselves. Instead, perceived gist of the acceptability of the medication's risks, perceived benefits relative to risks, and gist principles (values/beliefs applied to perceptions of options) such as “Better safe than sorry” influenced outcome variables: Judging the risks as unacceptable and as outweighing the benefits of taking the medication increased ordinal risk perception and feelings of worry, and decreased intentions to start treatment. Gist principles supporting medication decreased feelings of worry and increased willingness to start. For outcome variables, higher ordinal risk perceptions heightened worry, and worry in turn lowered starting intentions. The resulting SEM model was robust to controlling for demographic variables and illustrated that willingness to start a new medication was mediated by the gist of risk acceptability, perceived benefits relative to risks, and gist principles (Figure 1). Alternative models were tested and ruled out.
Across different ways of communicating risk, the subjectively extracted overall gist of the options – and not the actual magnitude of risk – influenced ordinal risk perception, worry, and willingness to start a medication. Consistent with theory, models built on representations of gist (related to acceptability of risks, benefits relative to risks, and gist principles) performed better than models which include only affect (worry) and risk perception to predict medication intentions.
Method: 2701 Spanish participants completed an Internet survey after reading a hypothetical newspaper article describing a serious influenza epidemic in Spain, and advocating vaccination. Participants were randomized to 1 of 3 conditions varying in expressed ambiguity regarding the effectiveness of vaccination: No Ambiguity, Ambiguity, and Normalized Ambiguity (accompanied by language reinforcing the routine, expected nature of ambiguity). After reading their assigned scenarios, participants completed measures of vaccination intentions, influenza risk perceptions and worry, and perceived vaccination effectiveness, as well as individual differences in subjective literacy, optimism, and tolerance of uncertainty, ambiguity, and risk. General linear models (MANOVA and ANOVA) adjusting for age, sex, and education were used to assess the effects of communicating ambiguity on both vaccination intentions and related health cognitions, and to explore potential moderators and mediators.
Result: Vaccination intentions were significantly lower (p<0.0001) for both the Ambiguity (M=4.72, 95% CI: 4.57-4.87) and Normalized Ambiguity (M=4.66, 95% CI: 4.51-4.81) groups, compared with the No Ambiguity group (M=5.08, 95% CI: 4.93-5.24), consistent with the phenomenon of “ambiguity aversion.” Subjective health literacy—but no other individual differences—moderated this effect; higher health literacy was associated with higher ambiguity aversion (more reduced vaccination intentions) for individuals in both the Ambiguity and Normalized Ambiguity groups, compared to the No Ambiguity group (p=0.0015). Overall, the Ambiguity and Normalized Ambiguity groups also demonstrated higher perceived severity of influenza, higher perceived influenza risk, higher influenza worry, and lower perceptions of the effectiveness of vaccination than the No Ambiguity group (p≤.01). Perceptions of influenza severity and vaccination effectiveness mediated the negative effect of ambiguity on vaccination intentions.
Conclusion: Communicating ambiguity about the effectiveness of influenza vaccination reduces vaccination intentions. Ambiguity-normalizing language does not decrease this effect, and health literacy increases it; more literate individuals demonstrate greater ambiguity aversion. The negative effect of ambiguity communication on vaccination intentions is mediated by its effects on perceptions of both the severity of the vaccine-preventable disease and the effectiveness of vaccination. The mechanisms of ambiguity aversion in public health communication are complex, and greater understanding of these mechanisms may facilitate more effective management of this problem.
Method: We administered a nationally-representative 10-minute survey to parents of 6-18 year old children (n = 410) though AmeriSpeak®, a division of the National Opinion Research Center. We asked parents to estimate the chances that (a) their child and (b) “a typical child in their community” would be overweight or obese at age 30, complete a 4-year college degree by age 30, get married by age 30, and the child’s likely future income at age 30 on a sliding visual analog scale. We collected data on family demographic and health characteristics, and parent numeracy. Fixed effects models were employed to assess within-parent variance about parents’ future predictions for their child versus predictions for a typical child in their community.
Result: Approximately 69% of children were normal weight and 31% of children were overweight or obese. Compared to the typical child in the community, parents were less likely to predict that their child would be overweight in adulthood (-24.8 percentage points, 95% CI: -27.3, 22.3), more likely to predict that their child would complete a four year degree by age 30 (28.3 percentage points, 95% CI: 25.4, 31.2), and more likely to predict that their child would have a higher income in adulthood ($14,529, 95% CI: 12,758, 16,300). There were no significant differences in parent predictions for their child and the typical child in terms of the probability of being married by age 30 (1.8 percentage points, 95% CI: -0.7, 4.4).
Conclusion: Parents exhibit equal levels of optimism bias around their child’s future health, education, and economic outcomes, but not social outcomes such as marriage. Accurate parental perception of a child’s future outcomes may motivate parents to engage in behavior change to ensure these positive expectations come to fruition. Child development practitioners should develop risk communication approaches that can compensate for optimism bias and improve parent assessment of health and socioeconomic outcomes.
Previous research on choice architecture suggests that consumers will have a difficult time choosing an optimal healthcare plan with the current structure of healthcare.gov (e.g. Barnes et al., 2012). This research is designed to examine how choice complexity in this naturalistic context influences information search and subsequent decision quality.
Across three studies, participants read a series of hypothetical vignettes describing individuals or families with specific healthcare needs. Their task was to select the optimal healthcare plan (lowest cost, while accounting for health history) for each individual/family from a set of three plans. For each vignette, participants viewed the health plan choices in an option (i.e., plan) by attribute (e.g., deductible) grid similar to the ‘compare plans' function on healthcare.gov.
In Study 1 and 2, participants (Internet sample recruited via Amazon Mechanical Turk, N=201, and college students, N=200), completed this task for four vignettes. In Study 3, college students (N=159) completed the task with only two vignettes but this time viewed the plans using MouselabWEB, an interactive table that displays option by attribute information and tracks search behavior. We hypothesized that: 1) accuracy in health plan selection would decrease with choice complexity; and 2) increased search of relevant information would increase task accuracy.
For Study 1 and 2, participants' ability to choose the optimal healthcare plan decreased as the complexity of the health history increased, X2(3) = 127.88, p<.001. For example, in Study 1, the majority of participants (70%) chose the optimal plan for an individual with few healthcare needs, while only 25% of participants were able to choose the optimal plan for the family with the greatest healthcare needs.
In Study 3, search behavior predicted task accuracy. Participants who correctly identified the optimal health plans adopted a search strategy focused on the most relevant information for each vignette, whereas those who could not identify the optimal plan used a more diffuse search strategy (Figure 1).
A large portion of the population may be ill equipped to choose optimal health plans, and the amount of information within a health plan may be a distraction to consumers. Choosing health plans that do not fit individual needs may lead to overpayment for some consumers and avoidance of care due to unexpected cost for others.