In this cross-sectional study of interns in cognitive specialties at an academic health center, working >80 h a week did not have a statistically significant association with interns’ perceived stress levels, quality of sleep, or fatigue. In fact, lower perceived teamwork functioning was associated with increased stress, and a higher level of stress was associated with decreased quality of sleep (Fig. ). Higher stress and low quality of sleep, in turn, were associated with increased fatigue.
Our cohort reported fairly high frequencies of fatigue-related behaviors, e.g., feeling drowsy or the need to rest, and poor sleep quality behaviors, e.g., getting insufficient sleep. Respondents also reported moderate overall stress levels. Despite this, teamwork-enhancing attributes such as feeling part of a team and feeling a sense of camaraderie with coworkers was reported frequently. The developmental sample (n
3,053) for the MOS sleep scale had a mean score of 28.3, SD of 18.2. While our mean score is much higher (59.2), indicating more sleep problems, our standard deviation (15.3) is similar to that of the much larger developmental sample. Normative data for the Chalder Fatigue Scale and Cohen Perceived Stress Scale are not available.21
Our findings suggest that solely reducing the raw number of hours worked by residents does not achieve the core goal of the duty hour reduction mandates – to reduce resident fatigue and its negative consequences. To reduce fatigue, programs should consider strategies that enhance the quality of trainees’ sleep and sense of teamwork, and decrease perceived stress.
Future interventions should be evidence-based and might include scheduling maintenance naps for on-call interns25
and improving quality of sleep by restructuring schedules to minimize sleep-wake cycle disturbances.7,26
Examining the effects of a change to a shift-work system is warranted. Increasing autonomy and support27
and decreasing workload might also decrease stress levels.20,28,29
Our finding linking poor quality teamwork to stress and, in turn, to fatigue may provide further evidence of poor teamwork’s contribution to errors.30
Curricula teaching formal teamwork structure and process,31
high fidelity simulation-based team training,32
and training incorporating didactic instruction with interactive participation33
have been shown to improve communication and enhance safety. The efficacy of these training interventions are likely generalizable to residents and might reduce levels of stress and fatigue.
We recognize several limitations of the study. Our sample size is moderate, with a 56% response rate, but our sample represents residents across cognitive specialties, permitting wider generalizability of our findings. We attempted to diminish the extent of recall bias by anchoring our questions with short time frames. Our study was performed at a single, large academic center. Although this may limit generalizability, many of our duty hour reduction strategies are similar to those reported elsewhere.34–36
Because data were collected over 4 months, responses could reflect a seasonal effect. However, we believe that collecting data from February through May optimized the chance of detecting our outcome measure. We also acknowledge that fatigue itself could have affected the accuracy of survey responses.
Seventy percent of respondents were female (n
46), compared with 56% of nonrespondents (n
23), likely indicating a response bias. Non-responding residents and preliminary interns may have shared certain characteristics, such as increased fatigue, and their exclusion could lead to bias. Our moderate sample size and the relatively small proportion of interns who reported working >80 h may have precluded us from finding a relationship between duty hours and fatigue. Additionally, there may be a continuous relationship between work hours and stress, diminished quality of sleep, and fatigue. However, because our measurement of work hours was categorical, and because responses were limited almost exclusively to two of the categories, we were not able to conduct an analysis to address that potential relationship. Last, we recognize the problem of causal inference and emphasize that our results simply identify associations.