The impact of surgical time of day and surgical site infection

Circadian rhythms—biological and physiological changes that follow a roughly 24-hour cycle—exist in all organisms. These rhythms are driven by the body’s internal circadian clock and are influenced by environmental cues such as light and temperature. The most apparent circadian rhythm in humans is the sleep–wake cycle. However, we also know that the clock has a profound effect on innate and acquired immunity,^1,2 and that disruption to the clock has impacts on health and wellbeing, including impaired immune function³ and wound healing.⁴ It is plausible that the timing of surgical procedures could impact the body’s response to trauma and susceptibility to infections.

A retrospective cohort study of 56,920 general and vascular operations found a time of day effect on morbidity, which included infection.⁵ However, a similar study of 21,985 cases found there was no correlation between time of surgery and post-operative infection.⁶ Our preliminary study aimed to investigate the possibility of surgical time affecting post-operative infection using New Zealand data.

Methods

Using the Health Quality & Safety Commission (HQSC) data on surgical site infections (SSI)⁷ we sought to investigate any possible time of day effects on the incidents of recorded infections. With ethics approval (AH22794), all recorded data from June 2017 to July 2021 for cardiac (including coronary artery bypass grafts) and orthopaedic (hip and knee arthroplasty) operations in New Zealand were extracted from the HQSC.⁸ Admissions classified as “arranged” or “acute” were considered acute admissions; those classified as “waiting list” were considered elective.⁹ The incidence of SSI was calculated for each 4-hour bin of time across the day, using the knife-to-skin time. Cases where this was recorded as being exactly at midnight were removed as likely artefacts. Any case where the age of the patient was under 18 years was also removed, as were cases with unknown or unrecorded American Society of Anesthesiologists scores. A logistic regression model was used to determine any time of day effects (with reference to the data between 12 pm and 4 pm), while controlling for age, acute admission and ASA scores.

Results

There was a total of 91,165 cases in the dataset. Of these, 3,480 were removed for being recorded at exactly midnight, a further 1,939 were removed for being under 18 and 2,192 more cases were excluded for having missing ASA data. This left 87,034 cases, of which 1,327 (1.5%) had an infection. Univariately, there was a significant time of day effect, with operations after 8 pm being 3.91 times more likely to have an infection (p<0.001), and operations between 4 pm and 8 pm being 0.71 times more likely (p=0.03) to have an infection compared to operations conducted between 12 pm and 4 pm. However, when age, acuity and ASA scores were included in the model, there were no differences in the rate of infection by time of day. There were no significant differences in the rate of infection by age, but there were increasing rates of infection as ASA score increased, with patients with ASA 2 being 1.81 (p=0.003), ASA 3 being 3.99 (p<0.001), ASA 4 being 7.28 (p<0.001) and ASA 5 being 9.26 (p<0.001) times more likely to have an infection compared to patients with ASA 1; acute admissions were 1.60 (p<0.001) times more likely to have an infection compared to non-acute cases.

The acute cases were more likely to have higher ASA scores (Kruskal–Wallis p<0.001), and there were more acute cases operated on outside of business hours: between the hours of 8 am and 8 pm, the percentage of acute cases was approximately 35%. This rose to between 60% and 65% outside of those hours, reflecting the more urgent nature of the operations. Similarly, during 8 am–8 pm and in the non-acute cases, there was some evidence that patients with a higher ASA score were scheduled earlier: the percentage of patients with ASA <3 rose from 62.4% to 65.4% and the percentage of patients with ASA >3 dropped from 7.7% to 5.2% between 8 am–12 pm and 12pm–4 pm (Chi-squared p<0.001).

Discussion

Simple, univariate analysis of time of day effects may mask more complex mechanisms. This preliminary work suggests that surgical time of day does not significantly impact incidence of surgical infection when other variables are controlled for.

This may partly reflect the decisions of list schedules, in which patients at more risk of infection are prioritised for earlier surgery, and patients with a large number of comorbidities and more acute need are seen out of normal hours.

We also did not access data about the experience of surgeons or other team members for each operation despite possible correlations with post-operative infection. This work has also not considered the administration of antibiotics, circadian disruption pre-surgery or other factors that may also be important.

We included all ASA levels—with appropriate preliminary analysis—as they were recorded, and we have not done any work to verify them.

Future work is required to link the SSI data with the National Minimum Dataset to investigate these other factors and conduct a more comprehensive analysis.