Clinical treatment decisions are often made in the setting of incomplete or conflicting evidence, which leads to uncertainty. Physicians' decisions are affected by a variety of factors when making choices with uncertainty. We sought to determine if patient age affects physician decision making under different levels of risk (when outcomes have known probabilities) and ambiguity (when outcomes have unknown probabilities).
We designed a pilot study to test the feasibility of physician participants using a computer-based behavioral experiment with hypothetical patient vignettes. We recruited physicians from a single academic medical center using convenience sampling with snowballing. Inclusion criteria included completion of residency training in internal, family, or emergency medicine. We stratified physician assignment to one of two groups based on gender and field of residency. The experiment exposed participants to a vignette in which they chose a treatment for a patient with traumatic brain injury due to motor vehicle crash. The age of the patient varied by group assignment: 74-years-old for Group 1, 22-years-old for Group 2. Each participant made 256 treatment decisions for the patient, each decision between two treatment options. The first option was a treatment with a single, deterministic outcome (100% probability of occurrence, e.g. mild headache). The second option was an “experimental” therapy, which had varied risk and ambiguity in each decision. For example, we varied the second option to include some superior and some inferior outcomes to the single outcome associated with the first treatment option (e.g. 25% probability of severe headache and 75% probability of recovery). Some options also had ambiguous outcomes probabilities (e.g. 25% probability of severe headache, 25% probability of recovery, and 50% unknown probability that could be attributed to either outcome). Half of the decisions focused on treatment benefits (varying cognitive improvement), and half focused on adverse effects (varying severity of headaches). Participants also completed instruments that characterized inhibition, numeracy, risk-taking, impulsiveness, and demographic information. We calculated risk seeking using the proportion of choices in which participants chose the experimental over the certain therapy. We calculated ambiguity aversion using the difference in percentage of experimental choice between a specific level of ambiguity (24%, 50%, or 74% of outcome probabilities unknown) and baseline (0% ambiguity, or all outcome probabilities known). We calculated group means for each level of risk and ambiguity, and then compared means using a Student's t-Test.
Twenty-two physicians were stratified into two groups, which had similar mean age and proportion of gender and field of residency. When considering potential benefits of treatment, Group 1 was more likely to choose the experimental therapy when it had a low probability of cognitive recovery than was Group 2 [73% versus 57% (p=0.04) with 38% probability of recovery, 68% versus 43% (p=0.006) with 25% probability, 54% versus 30% (p=0.07) with 13% probability]. Group 2 was more ambiguity averse than Group 1 (p=0.05 at 24% ambiguity level, p=0.06 at 50%, p=0.03 at 74%). The two groups showed similar risk seeking and ambiguity aversion profiles when considering potential adverse effects of treatment.
Physicians were more risk seeking and ambiguity tolerant when considering potential cognitive benefits for older compared to young patients.
Poster Number: EI 58
This research was funded by: Funding for this project provided by the Yale Hartford Center of Excellence in Aging.
© 2014 Published by Elsevier Inc.