Background: In a previous randomized trial we found that shorter ward rotations reduced attending physician burnout at rotation-end. This improvement was offset, however, by negative education outcomes among trainees. Therefore solutions to burnout other than simply shortening rotations are needed to preserve the educational mission. We examined additional data collected (but not yet presented) to determine if temporal patterns of burnout, when assessed repeatedly and in real-time, predicted summative recall assessments at rotation-end.

Methods: We conducted a crossover trial with attending physicians as the unit of crossover randomization to 2-week (‘short’) and 4-week (‘long’) rotations during the 2009 academic year. Participants staffed at least 6 weeks of the general medicine inpatient service. On the same day of every 4-day call cycle of each rotation, attending physicians rated their real-time intensity of burnout. Then, at the end of each rotation, they rated their summative recall of burnout across the entire rotation. We determined how many days occurred between peak intensity of burnout and rotation-end. We used multilevel logistic regression to model how the proximity of peak burnout intensity to rotation-end was associated with summative recall.

Results: 62 attending physicians were assigned to randomized sequences of rotation durations and completed a median of 1 long (range 1 to 3) and a median of 2 short (1 to 6) rotations. At rotation-end, attending physicians reported burnout in 36% of 76 long and 17% of 130 short rotations. The median duration from peak intensity to rotation-end was 0 days (interquartile range 0 to 8 days) and did not differ by rotation length (p-value 0.14). The relationship between the proximity of peak burnout intensity to rotation-end and summative recall of burnout did not differ by rotation duration (p-value for interaction 0.4). After adjustment for rotation length, each additional day from peak intensity to rotation-end was associated with a reduction in the odds of summative recall of burnout (OR 0.77, 95% CI 0.62 to 0.94). For example, at the end of a long rotation the probability of burnout when the peak intensity occurred at rotation-end was 44% (95% CI 28 to 60%) whereas the probability of burnout when peak intensity occurred 1 week before rotation-end was 28% (95% CI 19 to 36%).

Conclusions: More time between peak burnout intensity and rotation-end protected against summative recall of burnout regardless of rotation length. This suggests that system-level interventions that reduce burnout intensity toward the end of a rotation may make longer rotations more tolerable, thereby improving the educational experiences of trainees. Such system-level interventions may include 1) offloading attending physicians’ non-ward-related duties that pile up until rotation-end and 2) overlapping attending physicians’ schedules to ease the transitions off-service.