|BEC||Behavioral Economics||ESP||Applied Health Economics, Services, and Policy Research|
|DEC||Decision Psychology and Shared Decision Making||MET||Quantitative Methods and Theoretical Developments|
* Candidate for the Lee B. Lusted Student Prize Competition
Purpose: To examine the effect of narrative content and emotional valence on decisions about treatments for early stage breast cancer.
Method: 263 women were asked to imagine they had been diagnosed with early stage breast cancer, needed to choose between two surgical treatments (lumpectomy with radiation versus mastectomy), and were provided with one of five computer-administered sets of information about these two surgeries. In the control condition, participants viewed a table containing descriptions of the surgeries, the length of recovery time, need for radiation, and other decision relevant facts. In the four remaining conditions, participants viewed the same table plus four videotaped narratives, which varied in structure by a 2 (narrative content: process or experience) x 2 (emotional valence: positive or mixed) factorial design. Process narratives discussed the factors a woman considered when making her surgical decision, whereas experience narratives described what it was like to go through the surgery itself. Two narrative conditions used only positive narratives while the other two contained equal numbers of positive and negative narratives. After reviewing all materials, participants were asked to make a hypothetical treatment decision and complete several measures of reactions to the narratives and confidence in the decision process. Participants also completed the Subjective Numeracy Scale, the Need for Cognition scale, the Decision Quality Index and the Decision Conflict Scale.
Result: Providing narratives to participants had no effect on treatment decisions; approximately two-thirds of participants in all groups preferred lumpectomy and radiation. However, participants in the narrative conditions reported somewhat less uncertainty than participants in the control condition, F (1, 261) = 3.66, p = .057. Experience narratives were better than process narratives at increasing decisional confidence, feelings of preparedness, and the ability to imagine what it would be like to have a lumpectomy Fs (1, 205) >4.65, ps < .05. The mix of positive and negative narratives was perceived to be more emotional, t (206) = -2.78, p = .006, and produced a greater connection to the breast cancer survivors, t (206) = -1.96, p = .05, than positive narratives alone.
Conclusion: While providing narratives did not change participants’ treatment intentions, narratives appeared to lower decisional uncertainty, and a mixture of positive and negative experience narratives may be most helpful to decision makers.
Purpose: To test whether using 1) an avatar (a figure representing an individual) and 2) animations of randomness in a pictograph help people better understand a personal health risk by explicitly showing 1) how population-based statistics apply to individuals and 2) the random element of risk estimates.
Methods: 3676 adults in a demographically diverse US-based online sample (mean age 53, 55% female, 78% white, 54% no college degree) with no history of cardiovascular disease entered their personal health information in a validated model that calculates 10-year risk of general cardiovascular disease (CVD risk). The median 10-year risk of CVD within this population was 8% (interquartile range 11%). Risk levels were classified as low if <5% (24% of participants), moderate if 5-9% (32%) and high if 10% or higher (45%). Participants were randomized to different versions of an animated pictograph showing their CVD risk. Pictographs either included an avatar or not, and were either standard versions that grouped all event rectangles together or versions that first displayed event rectangles randomly distributed in the pictograph before transitioning to a standard version. Participants answered a brief set of questions about their risk perceptions (how large or small the risk feels and how likely do they think they are to have CVD in the next ten years) and their behavioral intentions in the next 30 days. At the conclusion of the survey, participants were asked to recall their risk estimate.
Results: Using an avatar in the graphic increased perceptions of CVD likelihood for those at moderate and high risk (F(1,2792)=8.45, p=.004), but not for those at low risk. Using animated randomness made lower risks feel smaller and less likely, and higher risks feel larger (F(2,3623)=3.40, p=.03) and more likely (F(2,3669)=4.28, p=.01). Both avatars (F(2,3648)=6.03, p=.002) and animated randomness (F(2,3648)=3.95, p=.02) resulted in people at lower risk reporting lower intentions and those at higher risk reporting higher intentions to see a doctor in the next 30 days. Neither avatars nor animated randomness affected recall.
Conclusions: Using avatars and animated randomness can help convey difficult concepts in personal health risk. These types of design features are straightforward to implement in an online environment, require minimal viewing time, and suggest potential to improve the effectiveness of health risk communication methods.
Purpose: Experts question whether certain decision-making biases are caused by low literacy. In this study, we explore whether decision-making biases are caused by low literacy per se, or if these biases can instead be explained by larger cultural factors, which are related to both literacy and patients’ medical beliefs.
Method: 574 men were recruited for a study about prostate cancer decisions. All of the men were undergoing prostate biopsies following a high PSA test. As a part of a larger questionnaire, each patient was asked to respond to a hypothetical cancer scenario. They were asked to choose between having surgery and accepting a 10% chance of dying from cancer, versus not having surgery (watchful waiting) and accepting a lower, 5% chance of dying from cancer. Past research has shown that a surprising number of people (~60%) choose the dominated surgery in this scenario. Just prior to this scenario, participants’ literacy (REALM) and numeracy (Subjective Numeracy Scale) were assessed. Patients were also asked questions about their beliefs about cancer treatment. These questions specifically assessed the patients’ bias toward active treatment options (e.g. “How important is it to treat cancer, whether or not it makes a difference in survival?”).
Result: 65% of the patients chose the dominated surgery option. As expected, participants who choose the dominated option were significantly lower in literacy than participants who chose the more rational treatment option (p < .01; numeracy did not predict choice, p > .10). However, the relationship between literacy and choice was mediated by participants’ desire for more active treatment. That is, literacy was not related to choice when controlling for participants’ desire for active treatment.
Conclusion: In the present scenario, the proximate cause of irrational decision making was patients’ desire for active treatment, rather than low literacy per se. Literacy predicted patients’ tendency to choose the dominated option, but only because literacy was related to general attitudes about active treatment. These data suggest that attempts to improve patient comprehension will not be successful at debiasing those patients. The fact that low literacy is related to preferences for active treatment suggests that there may be larger cultural factors at work that cause the present decision bias.
Purpose: All else equal, rare outcomes should be given relatively little weight in decision making. But, when strong emotions like disgust are present, objectively unlikely outcomes may feel more likely than they really are. We examined this possibility, which could account for preference-inconsistent decisions.
Method: In two pilot studies, we asked 3428 participants to rate 24 descriptions of health states on several dimensions to determine states that were rated the same on quality-of-life but differently on how disgusting they were. The pair that best fulfilled these criteria – chronic diarrhea and severe fatigue – was used in subsequent studies. In Study 1, we asked a different group of 3094 participants, "If you had to choose, would you prefer a [x]% chance of [condition], or a [x]% chance of death?" where x = 4% or 100% for both outcomes, and the condition was chronic diarrhea or severe fatigue. If people overweigh small probabilities of disgusting events like chronic diarrhea, we would expect them to choose death over diarrhea more often at 4% probability than at 100%. Substituting a less disgusting outcome, severe fatigue, should then lead to less inconsistency across probabilities. In Study 2, we presented another group of 300 participants with two hypothetical medical treatments, one with a 4% chance of death, and the other with a 4% chance of a complication, either chronic diarrhea or severe fatigue. Participants rated how likely each possible outcome felt, how vulnerable they felt to each outcome, and estimated their own particular chance of experiencing each outcome. We examined the relationship between these ratings and participants' trait disgust sensitivity.
Result: In Study 1 (see figure), people chose death over diarrhea significantly more often at 4% than at 100% (40% vs. 28%, chi-square=15.89, p<0.01), while preferences for fatigue versus death did not change across probabilities (30% vs. 34%, chi-square=1.92, p=0.18). This difference between conditions was more pronounced among lower-numeracy participants. In Study 2, trait disgust sensitivity significantly predicted both likelihood (r = 0.19, p < 0.01) and vulnerability (r = 0.12, p = 0.02) ratings across both conditions.
Conclusion: These results support the idea that disgust, a medically-relevant emotion, exaggerates people's tendency to overweight small probabilities. This occurs independent of quality-of-life concerns, and appears more prevalent among lower-numeracy individuals.
Purpose: To determine whether the number of decimal places in a personal health risk estimate influences the extent to which people believe and remember the estimate.
Methods: 3422 adults in a demographically diverse US-based online sample (mean age 50, 52% female, 74% white, 56% no college degree) were asked to imagine they were visiting an online risk calculator hosted by a prominent university’s medical school. We designed a mock calculator similar to existing calculators available online. The calculator asked a series of health questions relevant to kidney cancer and returned a hypothetical estimate of lifetime risk of kidney cancer. In this between-subjects experiment, participants were assigned one of seven risk estimates close to the average lifetime risk of kidney cancer in the US. Participants who were randomized to the no decimals condition received an estimate of 2%. Those in the one, two or three decimals conditions received an estimate of 2.1% or 1.9% (one decimal), 2.13% or 1.87% (two decimals), or 2.133% or 1.867% (three decimals). Participants were asked to indicate how believable they found the estimate to be on a six-point scale anchored by labels, “not at all,” and, “extremely.” Then, after completing a second, unrelated survey (median time for this task was 8 minutes), they were asked to recall to the best of their ability the kidney cancer lifetime risk estimate they had been given earlier.
Results: Risk estimates expressed as integers were judged as the most believable (F(3, 3384)=2.94, p=.03). Compared to estimates with decimal places, integer estimates were judged as highly believable (defined as the top two points of the six-point scale) by 7 to 10% more participants (Chi-squared(3)=17.82, p<.001). Recall was highest for integer estimates. Odds ratios for correct approximate recall (defined generously as being within 50% of the original estimate) were, for one decimal place, OR=0.65 (95% CI 0.49, 0.86), for two decimal places, OR=0.70 (95% CI 0.53, 0.94), and for three decimal places, 0.61 (95% CI 0.45, 0.81). Exact recall showed a similar pattern, with larger effects.
Conclusions: Using decimals in risk calculators offers no benefit and some cost. Rounding to the nearest integer is likely preferable for communicating risk estimates so that they might be remembered correctly and judged as believable.
Purpose: We developed a web-based Geriatric Diabetes Decision Aid (GDDA) which combines a decision analytic model of DM complications with a geriatric life expectancy prediction tool. To date, little is known about the best ways to display the risk and benefits of varying levels of glycemic control to older patients with DM and their providers. We present the patients and provider acceptability testing of the GDDA.
Methods: 9 patients and 12 providers from local federally qualified health centers were interviewed utilizing qualitative methods regarding computer usage patterns, patient risk comprehension, as well as their opinions on methods of visually displaying the lifetime risk of amputation at different glycemic targets (A1c of 7, 8, and 9%). Options included a bar graph, tables, and pictograms. Patients and providers were also asked questions about the website’s overall usability and design. Interviews were audio recorded and transcribed for accuracy and theme saturation. Patients and providers used the website throughout the interview.
Results: Mean patient age was 68 and 56% were female. Nine providers were male. All the providers were either in family or internal medicine. Four patients owned and used a computer regularly, three regularly used but did not own and two did not own or use computers. When tested on their knowledge of risk of amputation, only two patients failed to understand. Risk display results were different between patients and providers. Six patients preferred tables which showed the incidence of events per thousand patients. Seven providers thought patients would prefer pictograms for the different A1c targets. Patients and providers agreed that the use of color, pictures, large print, simple wording and easy to operate navigation and scroll buttons were a necessary part of the website design. All patients agreed that the GDDA is a tool that could assist in learning about A1c and discussing treatment goals with their doctor. All providers thought the GDDA could be a useful tool to stimulate conversation regarding A1c targets with their patients.
Conclusions: The GDDA is an instrument that may be able to assist patients and providers in determining individualized glycemic control targets. Pictures, simple wording, and easy navigation buttons can increase usability. Provider opinions should not be used as a proxy for patient opinions in determining the acceptability of website design.