**4. Discussion**

The COVID-19 outbreak upended traditional health system practices amid an environment that demanded an accelerated pace of innovation. Health systems were faced with difficult decisions as to how to safely resume margin-producing elective orthopedic surgery in the midst of the pandemic. Many have suggested transitioning more joint arthroplasty procedures to the outpatient setting [4,9,10], but whether this actually has taken place is unclear. This study showed that shoulder arthroplasty following the resumption of elective surgery during the COVID-19 pandemic was associated with a shorter LOS and higher rate of same-day discharge.

The finding that baseline preoperative patient characteristics remained unchanged compared to before the outbreak suggests that the observed changes in discharge patterns may indeed be a direct consequence of COVID. There is recent evidence that sociodemographic and psychological factors may have more influence than patient infirmity and technical issues in the variation in LOS and discharge disposition after shoulder arthroplasty [11]. Although this requires formal investigation, it is possible that patients may be more motivated to go home after surgery during the pandemic to minimize risk of contagion [8]. Indeed, it has been our experience during the pandemic that patients are more invested in making arrangements for going home the same day of surgery. Health systems may also be pushing for early elective surgery discharges to limit exposure and reallocate resources to sicker patients [9].

The observation that 90-day readmissions, reoperations, and ED visits did not increase following the resumption of elective surgery during the pandemic is reassuring. This is consistent with the growing realization that shorter postoperative stays after shoulder arthroplasty are safe. Shorter LOS and/or same-day discharge following shoulder arthroplasty do not seem to increase the risk of postoperative mortality and morbidity [12,13]. The important addition of the current study to this literature is the fact that both cohorts represented the majority of the shoulder arthroplasty population in the surgeon's practice. The near 90% utilization of same day discharge in the post-COVID cohort indicates that there was limited potential for patient selection bias. In other words, outpatient shoulder arthroplasty is safe in not only selected patients, but in the majority of cases based on the findings of the current study.

Interestingly, we found that the rate of ED visits decreased significantly from 13.3% (pre-pandemic) to 7.4%. It may be that patients are now more likely to use and rely on technology (e.g., emails with image exchange, telehealth) to address postoperative concerns that would traditionally warrant a visit to the ED. The observed reduction in ED visits may indicate that some of them are preventable with the use of technology and improved postoperative care coordination. This subject deserves further study. Although another explanation could be that patients were more fearful of postoperative ED visits, this is not supported by the lack of change in the 90-day complication or re-operation rate.

The principal strengths of our study include its relatively large sample size and the fact that all procedures were performed by the same experienced surgeon, thus reducing surgeon variation in perioperative protocols. Nonetheless, our analysis was subject to several shortcomings that might be addressed in future research. First, the retrospective nature of this study does not allow causal inference. Therefore, we can only determine associations between COVID and the outcomes of interest. Second, because this study was performed at a private practice with a high pre-COVID rate of same day discharge shoulder arthroplasty, the results may lack generalizability. However, one might expect an even greater increase in the rate of outpatient shoulder arthroplasty among practices with traditionally higher rates of inpatient procedures. Future studies should evaluate and compare shoulder arthroplasty discharge patterns pre- and post-COVID across different practices and regions. Third, while we collected data on multiple markers of postoperative resource use (e.g., LOS, discharge disposition, ED visits, readmissions, reoperations), we did not assess patient experience and functional outcomes to better define the value equation. Fourth, there was a trend towards a potentially clinically relevant (+5.5% difference) higher rate of diabetes in the pre-COVID cohort compared to the post-COVID cohort which, while not ye<sup>t</sup> significant, may affect results in larger samples. Finally, our study was limited in follow-up duration (90 days) due to the recency of the pandemic.
