Understanding Stress, Burnout, and Behavioral Patterns in Medical Residents Using Large-scale Longitudinal Wearable Recordings

TL;DR

Medical residency imposes intense cognitive, physical, and emotional demands that contribute to stress and burnout. The authors combine longitudinal wearable sensing (Fitbit Charge 3 and proximity badges) with self-report ecological momentary assessments to track 43 residents across a 3-week ICU rotation, capturing movement, rest-activity, computer use, and mentor interactions. They show distinct behavioral patterns by training year, with junior/senior residents walking more and engaging more with attending doctors, and find that mentor interactions correlate with stress among senior residents while computer use correlates with stress among interns. A Random Forest classifier using multimodal behavioral features can predict midday stress, end-of-day stress, and job satisfaction, with multimodal signals especially informative for job satisfaction and uni-modal cues driving midday stress. Together, these findings support using wearable-enabled monitoring to inform wellness interventions and program design in medical residency.

Abstract

Medical residency training is often associated with physically intense and emotionally demanding tasks, requiring them to engage in extended working hours providing complex clinical care. Residents are hence susceptible to negative psychological effects, including stress and anxiety, that can lead to decreased well-being, affecting them achieving desired training outcomes. Understanding the daily behavioral patterns of residents can guide the researchers to identify the source of stress in residency training, offering unique opportunities to improve residency programs. In this study, we investigate the workplace behavioral patterns of 43 medical residents across different stages of their training, using longitudinal wearable recordings collected over a 3-week rotation. Specifically, we explore their ambulatory patterns, the computer access, and the interactions with mentors of residents. Our analysis reveals that residents showed distinct working behaviors in walking movement patterns and computer usage compared to different years in the program. Moreover, we identify that interaction patterns with mentoring doctors indicate stress, burnout, and job satisfaction.

Figures (11)

• Figure 1: Sensor setup example in the hospital unit. Red and blue rectangular indicate the BT hubs and BT beacons, respectively. Each resident was instructed to wear a social badge and a Fitbit tracker, while the doctors were asked to carry BT beacons.
• Figure 2: Example of the walking movement sessions.
• Figure 3: Gender and year in the residents' program in this work.
• Figure 4: Comparisons of pres-study assessments and EMAs.
• Figure 5: Correlations of computer access and mentoring doctor interaction with pre-study survey responses.