A 2×2 between-subject randomized message experiment was conducted. A total of 320 students in medical and health-related disciplines were included in the analysis. The questionnaire system randomly presented each participant with one of four sleep health messages combining gain versus loss framing with autonomy-supportive versus controlling expression. All main measures were assessed immediately after message exposure. Intention to adopt sleep-management recommendations was the primary outcome, and the main effect of expression style was the primary effect; a 2×2 analysis of variance was performed, with marginal mean differences reported. The six items assessing intention to use sleep-support resources were secondary outcomes and were analyzed separately using 2×2 linear models with Holm adjustment; generalized estimating equations(GEE) were used as an aggregate supplementary analysis. The statistical mediation relationship involving psychological reactance was examined exploratorily using bootstrap analysis.

For the primary outcome, the marginal mean difference between autonomy-supportive and controlling expression was 0.661 points(95% CI: 0.530-0.791; F=99.18, P < 0.001, partial η²=0.239), whereas the marginal mean difference between gain and loss framing was 0.185 points(95% CI: 0.055-0.316; F=7.80, P=0.006, partial η²=0.024).The difference between the two estimated main effects was 0.475 points(95% CI: 0.292-0.659, P < 0.001), and the interaction was not statistically significant(F=1.46, P=0.227, partial η²=0.005). Across the six items assessing intention to use sleep-support resources, all coefficients for autonomy-supportive expression were positive(β=0.444-0.711) and remained statistically significant after Holm adjustment(all adjusted P < 0.001); neither the outcome-framing main effects nor the expression-style×outcome-framing interactions remained statistically significant after Holm adjustment. The aggregate GEE analysis showed a consistent direction of association(β=0.574, 95% CI: 0.450-0.698, P < 0.001). In the bootstrap analysis adjusted for covariates, psychological reactance statistically mediated the relationship between expression style and intention to adopt sleep-management recommendations(ab=0.512, 95% CI: 0.353-0.677).

The estimated difference associated with expression style is larger than that associated with outcome framing. Autonomy-supportive expression is associated with greater intention to adopt sleep-management recommendations and greater intention to use each sleep-support resource. The statistical mediation relationship involving psychological reactance provides a clue for understanding this difference.