Effect on patient safety of a resident physician schedule without 24-hour shifts

Over the last 4 decades, a body of literature has studied the effect of extended shifts (working ≥24 hours) for resident physicians on their alertness and performance, and by extension, patient safety. Since then, the results of further studies have been inconsistent. One recent study, by the authors of this current study, found that 36% more serious errors occurred among those working extended shifts than those who worked a cycle of day and night shifts of 16 hours or less. Other studies, including the Flexibility in Duty Requirements for Surgical Trainees and iCOMPARE (Individualized Comparative Effectiveness of Models Optimizing Patient Safety and Resident Education), have shown no changes in patient mortality rates between those who worked extended shifts and those who did not. The aim of this study was to examine the effect on patient safety of eliminating extended work shifts for resident physicians with shorter, cycling work shifts. The ROSTER (Randomized Order Safety Trial Evaluating Resident-Physician Schedules) is a cluster-randomized, crossover trial conducted at 6 US pediatric intensive care units (PICUs) from July 1, 2013, to March 5, 2017. Each of the PICUs had schedules for residents that included extended work shifts (the control schedule). Each site began a 2-year trial at different intervals within the study period. Sites were first paired based on the start of their trials, with one site randomly assigned to start with an extended-shift schedule and the other site to begin a schedule that cycled residents through day and night shifts lasting no more than 16 hours (intervention schedule). After a 4-month wash-in period and 8 months of data collection, each site then crossed over to the other schedule. Another 4-month wash-in period took place, and 8 months of data were collected, thereby allowing each site to serve as its own control. Data were collected on all adverse events, near misses, and errors associated with little or no harm, based on chart reviews and direct observation. Data on resident-physician workloads were also collected. Residents who worked intervention schedules make more serious medical errors than those who worked extended shifts (unadjusted rates 97.1 vs 79.0 per 1000 patient days; adjusted relative risk, 1.53 [95% confidence interval (CI), 1.37-1.72]). Wide discrepancies on the effect of intervention across sites were observed, with an increase in serious errors by residents during the intervention schedule at 3 sites, no differences at 2 sites, and fewer errors at 1 site. The number of unit-wide serious errors whether by residents or not were higher during intervention schedules than control schedules (unadjusted rates 181.3 vs 131.5 per 1000 patient days; adjusted relative risk, 1.56 [95% CI, 1.43-1.71]). A secondary analysis, adjusting for the number of patients per resident, showed a 0.54 relative risk (95% CI, 0.35-0.85) of serious error during the intervention schedule versus the control schedule. Introducing an intervention schedule at the PICUs included in this study resulted in a significant increase in medical errors by resident physicians, although the effect of the intervention schedule varied widely by site. A secondary analysis showed that intervention schedules were not associated with increased medical errors when adjusted for the number of patients per resident; however, further research is needed on this topic.