Physical Activity And Model-Based Decision-Making In Trauma-Exposed Adults

Many decision-making contexts require navigating multi-step choices to obtain a desired outcome (e.g., reward). This kind of sequential decision making is thought to arise from a combination of 1) simple trial-and-error learning (i.e., model-free [MF] learning), and 2) more sophisticated learning where an abstract representation of the environment is formed, thereby facilitating prospective predictions about likely outcomes based on different choices (i.e., model-based [MB] learning). MB learning is thought to be advantageous in many contexts as it provides a more informed cognitive map. However, emerging evidence suggests that MB learning may be impaired among trauma exposed adults. Although greater physical activity (PA) is associated with enhanced cognitive function, no prior studies have examined the specific domain of MB decision-making. PURPOSE: To examine whether greater PA is associated with greater MB decision-making in pursuit of reward. METHODS: Eighty-four trauma-exposed adults (62% women, 55% white, M ± SD age = 28 ± 9 y) completed the IPAQ-SF and a two-stage Markov task capable of quantifying MF vs MB decision-making. Logistic regression mixed models were used to determine if PA (quartiles of total MET-min/wk) promotes greater engagement in MB behavioral strategies during the task (i.e., magnitude of beta-coefficient for reward x transition interaction) while controlling for variables that may influence MB behavior (e.g., working memory, age, sex), with random effects by subject. RESULTS: Participants reported a M ± SD of 2157 ± 1365 total MET-min/wk (quartile 1 = 706 ± 258; quartile 2 = 1455 ± 251; quartile 3 = 2380 ± 242; quartile 4 = 4088 ± 922). Participants from quartile 2 (β = 0.17, 95%CI = 0.11-0.23), quartile 3 (β = 0.27, 95%CI = 0.21-0.33), and quartile 4 (β = 0.23, 95%CI = 0.17-0.30) exhibited greater MB decision-making compared to participants from quartile 1 (β = 0.08, 95%CI = 0.02-0.14), with participants from quartile 3 exhibiting greater MB decision-making compared to quartile 2. CONCLUSION: PA is positively associated with a greater propensity to utilize MB behavioral strategies during decision-making in pursuit of reward. Future research should examine whether changes in PA over time predict subsequent changes in MB behavior (i.e., examine temporal relationship). Supported by: K01MH132545