Patient Safety under Flexible and Standard Duty-Hour Rules

To the Editor:

In the iCOMPARE (Individualized Comparative Effectiveness of Models Optimizing Patient Safety and Resident Education) trial, Silber et al. (March 7 issue)¹ prespecified an unreasonable noninferiority threshold based on an increase in 30-day mortality from 12.5% to 13.5%. Even the highest estimates would project that fewer than a quarter of deaths (and probably far fewer) in hospitals are due to medical errors.^2,3 An absolute increase of 1 percentage point in error-related mortality (i.e., from approximately 3% to approximately 4%) would therefore represent an increase of more than 30% in error-related mortality. It was unreasonable to set such a high threshold for noninferiority, since this is not an acceptably small degree of harm. First-year resident physicians make up only a small fraction of the providers of direct patient care in teaching hospitals, and although fatigue is an important root cause of error, it is but one of many. The authors’ data revealed a 15% higher rate of death related to medical errors among patients who received care in hospitals that were allowed to schedule first-year resident physicians to work extended-duration (>24-hour) shifts than among those who received care in hospitals where first-year resident physicians were not allowed to work more than 16 consecutive hours. This was a degree of difference for which the authors had insufficient power to assess significance. The authors’ conclusion that allowing physicians to work extended-duration shifts was noninferior was thus based on unreasonable criteria. Moreover, this conclusion is contrary to extensive data showing that extended-duration shifts endanger resident physicians and their patients.^4,5 Large as it was, the iCOMPARE trial was underpowered by an order of magnitude to determine the effects of work-hour limits on in-hospital mortality.