Paper Session 17: Decision Making

Background
Adherence to medical recommendations is critical for health outcomes, yet patients frequently resist advice—even when it comes from trusted experts. This research identifies cognitive flexibility (CF), the ability to generate alternatives and adapt to new information, as a key driver of health advice uptake.
Objective
We examined whether individuals higher in CF are more likely to adopt health therapies and adhere to recommendations, and whether this effect depends on the source of the advice.
Methods
Across four studies (total N > 2,300), we measured CF using the Cognitive Flexibility Inventory and assessed uptake of health therapies and recommendations. Study 1 surveyed 1,570 patients with type-2 diabetes regarding medication, diet, and exercise adherence. Study 2 experimentally tested CF’s impact on uptake intentions for a novel stress-relief remedy (N=192). Study 3 randomized participants (N=392) to receive treatment advice from either a doctor or a friend. Study 4 (N=388) used a serial mediation design to examine whether perceived knowledge differentials explain the link between CF and advice uptake.
Results
Among patients with type-2 diabetes, CF positively predicted adherence to medication (β=0.14), diet (β=0.34), and exercise (β=0.30), all ps<.001. In Study 2, CF increased the number of benefit-related thoughts, which mediated uptake intentions (indirect effect ab=.077, 95% CI [.010, .203]). Study 3 revealed that higher CF increased uptake intentions when advice was provided by a doctor (β=0.25, p=.020), but not when provided by a non-expert friend, ns. Study 4 showed that CF increased perceived knowledge differential between the self and the physician (β=0.21, p<.001), which enhanced perceptions of physician expertise (β=0.09, p<.001), leading to greater advice adherence intentions (β=0.33, p<.001).
Conclusions
Individuals higher in cognitive flexibility are more likely to follow expert medical advice, driven by their recognition of expertise and their ability to generate potential benefits of therapies. This suggests that interventions prompting patients to articulate benefits or emphasizing provider expertise may improve patient outcomes.

Co-Author - Tony Stoval, Ph.D., Indiana University, Kelley School of Business
Co-Author - Ryan Cruz, Ph.D., Thomas Jefferson University
Co-Author - Lauren Friedly, M.D., Mills-Peninsula Medical Center

Personalized health risk estimates may help people make informed health decisions. However, interpreting numeric information can depend on one’s numeracy. We investigated how people responded to personalized melanoma risk estimates and the influence of subjective numeracy (belief about one’s numeracy skills and liking of math) on these responses. Participants were recruited via Prolific. After subjective numeracy was assessed, objective risk factors were used to calculate a personalized melanoma risk estimate. Then, participants who opted to learn their risk (n=838, 74.7% of the total sample; Mage=40.4, SDage=11.7, 48% women) were shown their personalized melanoma risk (ranging from 0.7% to 14.8%) in one of four numeric formats (we collapsed across condition in the present analyses). Then, they were asked about their agreement with and perceived personalization of the estimate, as well as their perceived risk (cognitive and affective perceived likelihood). Finally, participants’ prior sun protection behavior, as well as priority of and intentions to protect themselves from sun exposure were assessed. Linear regressions tested whether risk estimate shown, numeracy, and the risk estimate by numeracy interaction predicted agreement, personalization, perceived risk, prioritization of risk, and sun protection intentions (the latter two accounting for past behavior). Higher risk estimates shown were associated with greater agreement (β=.08, p=.026), higher perceived risk (β=.51, p<.001) and higher prioritization (β=.19, p<.001). Higher numeracy was associated with lower perceived risk (β=-.09, p=.008) and prioritization (β=-.08, p=.038). The interaction between subjective numeracy and risk estimate shown was only significant for perceived personalization (β=.08, p=.036)—when presented with high risk, perceived personalization was similar across subjective numeracy levels. When presented with low risk, more numerate participants perceived lower personalization than less numerate participants. Results suggest that magnitude of risk estimates shown and subjective numeracy can impact responses to personalized risk.

Author - Ivy Cheng, BSS, MA, Kent State University
Co-Author - Jeremy Foust, PhD, Converse University
Co-Author - Charles Fitzsimmons, PhD, University of North Florida
Co-Author - Pooja Sidney, PhD, University of Kentucky
Co-Author - Clarissa Thompson, PhD, Kent State University
Co-Author - Jennifer Taber, PhD, Kent State University

Background: Over 40% of U.S. adults have obesity, often linked to greater delay discounting--preferring immediate over larger future rewards. Episodic Future Thinking (EFT), which involves imagining vivid, personally meaningful future events, can reduce delay discounting and unhealthy behaviors (eg., fast-food demand) based on previous studies. A key EFT component is generating detailed future event cues, usually guided by a clinician. While effective, this is costly; self-administered surveys are more scalable but often less engaging and vivid. We developed a GPT-4-based AI-EFT tool (EFTeacher) that interactively helps users create richer cues. This study tested whether EFTeacher more effectively reduced delay discounting and fast-food demand compared to the self-administered EFT and a control.

Methods: We recruited 120 participants with obesity from Prolific (an online crowdsourcing platform) who completed two sessions. Session 1 included baseline assessments of fast-food demand (a measure of reinforcing value) and delay discounting (a measure of impulsivity). In Session 2, participants were randomly assigned to EFTeacher, self-administered EFT, or the control group to generate their cues and repeated the delay discounting and food-demand tasks while engaging in their cues. Mixed factorial ANOVAs tested effects of group and session on delay discounting and fast-food demand.

Results: For delay discounting, there was a significant session-by-group interaction (p < .001). Post-hoc tests showed that EFTeacher and self-administered EFT groups significantly reduced delay discounting from Session 1 to Session 2 (EFTeacher: p < .001, d= 0.71; Self-administered EFT: p = .003, d=0.33), while the control group showed no change (p=.53, d= 0.07). In contrast, fast-food demand showed no significant effects.

Conclusions: EFTeacher successfully reduced impulsivity on the delay discounting task, matching or exceeding effects of self-administered EFT. However, no changes were observed in fast-food demand, which may reflect limitations of our demand task or suggest that EFTeacher's impact on food behavior requires further investigation. These preliminary results support the feasibility and promise of EFTeacher as a scalable tool to target impulsivity linked to obesity. More work is needed to understand the effects of EFT on other outcomes such as food demand.

Co-Author - Nicki Rohani, BS, Virginia Tech Carilion School of Medicine
Co-Author - Allison Tegge, PhD, Virginia Tech
Co-Author - Sareh Ahmadi, PhD, Virginia Tech
Co-Author - Haylee Downey, BS, Virginia Tech
Co-Author - Edward Fox, PhD, Virginia Tech
Co-Author - Jeffrey Stein, PhD, Virginia Tech

Intro: Effective communication during post-test genetic counseling is crucial for ensuring patients begin to psychologically adapt to their genetic test results and feel equipped to adhere to medical recommendations. Patients’ decisions regarding what information is shared with and what is withheld from their clinician are integral in achieving effective communication and adaptation to illness in other clinical contexts, though the consequences of disclosure decisions have not been explored in post-test genetic counseling. This study applies the disclosure decision-making model to test how sharing and withholding communication behaviors occur in post-test genetic counseling and how these relate to psychological health outcomes.

Methods: We recruited 9,479 US-based potential participants to take a screening survey developed for this study. After our quality control and screening pipeline, 230 clinical genetic testing patients remained. This represents ~2% of the initial recruitment and matches the expected prevalence of genetic testing in the US adult population. Participants completed online anonymous questionnaires about information they did and did not disclose during their post-test counseling, their efficacy for disclosing this information, anxiety levels, and psychological outcomes of testing.

Results: Latent class analysis revealed two distinct patient populations. Latent class 1 (n=177) was characterized by more sharing (p=.01), less withholding (p<.01), higher efficacy for disclosure (p<.01), and lower anxiety (p<.01) as compared with latent class 2 (n=53). Logistic regression revealed that patients were significantly more likely to be in latent class 2 if they had more uncertainty about their genetic test results (OR=1.88, p=.02) and reported more avoidant thoughts related to testing (OR=2.69, p=.02).

Conclusions: Approximately 1 in 4 patients were classified in latent class 2, which was characterized by more withholding and higher anxiety. These patients were more likely to have worse psychological outcomes after genetic counseling (more uncertainty, more avoidant thoughts). This aligns with research in other clinical contexts and supports the notion that patients benefit from having post-test genetic counseling where they feel comfortable communicating with their clinician. Clinicians providing genetic testing results should seek to actively elicit and address topics of potential worry or concern, even for those who receive negative results.

Author - Liesl Broadbridge, ScM, CGC, Rutgers University
Co-Author - Jada Hamilton, PhD, MPH, FSBM, Memorial Sloan Kettering Cancer Center
Co-Author - Kathryn Greene, PhD, Rutgers University