Assessing the influence of after-hours surgery: concerns with the confounders and conclusion

Abstract

Oh et al. [1] used a large sample size to analyse short-term complications and long-term outcomes of after-hours surgery, providing valuable insights into operating theatre management. The use of propensity scores to adjust for the higher likelihood of high-urgency patients and those with severe comorbidities being managed after-hours is commendable. However, critical potential confounders and analytical errors introduce complexities that hinder clear interpretation of the results.

First, there are concerns regarding clinicopathological characteristics and their adjustment. As expected, emergency surgeries were performed more frequently after-hours. However, Table S5 reveals that, even after adjustment, there was a four-fold higher rate of emergency procedures after-hours (4.1% vs. 16.0%) [1]. The authors report that the absolute standardised mean difference was >0.1 after adjustment, but our calculations using the following formula show that the absolute standardised mean difference for emergency surgery was 0.40 (${\widehat{p}}_{\mathit{\text{after}}-\mathit{\text{hours}}}$ and ${\widehat{p}}_{\mathit{\text{in}}-\mathit{\text{hours}}}$, denoting the prevalence of the dichotomous variables in the after-hours and in-hours groups, respectively) [2].

$$d=\frac{\left({\widehat{p}}_{\mathit{\text{after}}-\mathit{\text{hours}}}-{\widehat{p}}_{\mathit{\text{in}}-\mathit{\text{hours}}}\right)}{\sqrt{\frac{{\widehat{p}}_{\mathit{\text{after}}-\mathit{\text{hours}}}\left(1-{\widehat{p}}_{\mathit{\text{after}}-\mathit{\text{hours}}}\right)+{\widehat{p}}_{\mathit{\text{in}}-\mathit{\text{hours}}}\left(1-{\widehat{p}}_{\mathit{\text{in}}-\mathit{\text{hours}}}\right)}{2}}}$$

Given the well-established impact of emergency surgeries on patient outcomes [3, 4], it is likely that the increased frequency of emergency surgeries contributed to the poorer after-hours outcomes observed. Additional discrepancies between the exposure rates and absolute standardised mean difference results for other factors, such as solid organ transplantation or surgeries for severe trauma, highlight the need for further clarification. Table S5 includes the essential background factors for discussing the generalisability of the results [1]. We hope that these errors will be addressed.

Second, the dataset lacks detailed information on the surgical procedures. Certain surgeries, particularly life-saving ones, are more likely to occur after-hours [5], which could lead to an overestimation of the association between after-hours surgery and increased 90-day and 1-year mortalities. The authors should consider identifying the most common surgeries performed in-hours and after-hours. If this is not feasible owing to database limitations, it should be explicitly acknowledged as a significant constraint.

Finally, the conclusion, “The results of our study serve as evidence to inform policy decisions on which surgeries should be performed after-hours, considering patient safety”, may not apply to this patient group, which includes emergency surgeries. These, by nature, require immediate intervention regardless of the time of the day, making it challenging to derive policy recommendations based on this dataset. A more appropriate approach would be to analyse patient groups restricted to elective surgeries. As they can be scheduled flexibly, obtaining similar results in these cases would provide a stronger foundation for policy decisions. To address the concern regarding the influence of emergency surgeries on outcomes, an additional analysis could further strengthen the validity of the conclusions.

We concur with the authors that their discussion on the impact of anaesthetist and surgeon fatigue, limited night-time resources and circadian rhythm effects is crucial for optimising surgical scheduling to improve patient safety. Fixing the errors and adding more analyses would strengthen this argument greatly.