Association of COVID-19 With Achieving Time-to-Surgery Benchmarks in Patients With Musculoskeletal Trauma

Key Points Question Were resource constraints due to the COVID-19 pandemic associated with a delay in urgent fracture surgery beyond national time-to-surgery benchmarks? Findings In this cohort pre-post study that included 3589 patients, there was no association between time to surgery and COVID-19 in either open fracture or closed femur/hip fracture cohorts. Meaning Despite concerns that the unprecedented challenges associated with the COVID-19 pandemic would delay acute management of urgent surgery, many hospital systems within the US were able to implement strategies in keeping with time-to-surgery standards for orthopedic trauma.


Introduction
Since December 2019, the novel SARS-CoV-2 virus has led to more than 123 million worldwide infections and has claimed the lives of more than 2 million people. 1,2 In response to the COVID-19 pandemic, hospital systems were forced to reduce operating room capacity and reallocate resources.
The outcomes of these policies on a hospital's ability to care for patients with COVID-19 and maintain emergency services have not been adequately reported. The outcomes of these unprecedented policy decisions on the care of acutely injured patients are equally unclear.
The purpose of this study was to evaluate whether the COVID-19 pandemic and subsequent COVID-19-related policies were associated with a delay in urgent fracture surgery beyond national time-to-surgery benchmarks. A 24-hour time from injury to fracture surgery benchmark is recommended for the treatment of open fractures, hip fractures, and femoral shaft fractures to prevent severe infections and mortality. We hypothesized that COVID-19 policies would be associated with a lower proportion of acutely injured patients receiving care within this national guideline and that this association would vary depending on the rate of regional COVID-19 cases.

Study Design and Procedures
This cohort study was a secondary analysis of data collected in the Program of Randomized Trials to presenting with an open fracture of an extremity requiring surgery or a closed lower extremity or pelvis fracture requiring surgery. This study has a blanket approval to use the data collected for secondary analysis from all participating institutional review boards, including the Methods Centre REB. In addition, when data sharing agreements are prepared, the Methods Centre REB is required to sign off. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cohort studies.
The PREP-IT trials compare iodophor vs chlorhexidine-based antiseptic skin preparation solutions. Clinical sites were randomized to determine which study solution to begin using, and each site crossed over to the alternative study solution every 2 months. The primary benefit of the cluster randomized crossover trial design is that all patients treated at the recruiting hospitals will receive the predetermined study intervention prior to patient enrollment. This efficiency maximizes recruitment feasibility because study consent does not need to occur prior to the patient's urgent surgery and it minimizes selection bias to improve internal and external study validity. Because of this, study participation has no effect on time to surgery. Study participant race and ethnicity were based on self-report and verified in the electronic health record.
For the current retrospective cohort study, all patients included in the PREP-IT program at sites who had begun enrolling pre-COVID-19 who sustained an open fracture, femur fracture (Orthopaedic Trauma Association classification 32A, 32B, 32C, 33A, 33B, 33C), or hip fracture (Orthopaedic Trauma Association classification 31A, 31B, 31C) were eligible for inclusion. The outcome investigated was time to surgery bifurcated into surgery within 24 hours of hospital admittance or greater than 24 hours from hospital admittance. No patients were excluded from the analysis. Patient cohorts examined were included from multiple institutions as part of the PREP-IT program. Institutions were eligible for this secondary analysis if they reported treating eligible patients both prior to and during the threshold when the COVID-19 policy period started because this is when a federally mandated declaration of national emergency was announced . Patients admitted to 1 of the PREPARE   participating sites between the dates of March 13, 2019, and March 12, 2020, were dummy coded as   the control group representing the pre-COVID-19 period, and patients admitted between the dates   of March 13, 2020, and January 31, 2021, were dummy coded as the intervention group representing   the COVID-19 pandemic period subject to emergency policy changes. Additional analyses were conducted to determine whether regional rates of COVID-19 infection were associated with the same binary time-to-surgery outcome. To determine regional rates of infection, US county population estimates 4 and Canadian census data collected from specified hospital site regions as well as number of daily positive COVID-19 viral test results for each county or region in which a participating institution sits were populated from March 13, 2020, through January 2021 using the continuously collected data set compiled by the New York Times 5 and for each Canadian region. The ratio of positive viral test results to estimated population size for each county was determined for each day. Change in day-over-day rates were used as the focal independent variable for this analysis in the closed fracture cohorts.

Data Transformations
All analyses were conducted in the R environment, version 4.0.3 (R Foundation for Statistical Computing). 6 Univariate tests were conducted across relevant patient characteristics stratified by period (ie, pre-COVID-19 vs COVID-19) for both open and closed fracture cohorts. Continuous variables were reported as means and SDs, and comparisons were performed with use of independent-sample t tests. Categorical variables were reported as frequencies and percentages.
Missingness was assessed for all statistical models. Significant missingness was discovered for the injury severity score variable for both open and closed fracture cohorts. Multiple imputation by chained equations was used to remediate missingness in this injury severity score, accounting for the hierarchical structure of the data.

Statistical Modeling
Binary logistic mixed effects modeling with random intercepts was used to investigate all research questions. Considering that patient data were derived from multiple institutions, hospital site was entered into each model as a random effect. Due to minimal sizes in certain categories of the American Society of Anesthesiologists (ASA) class variable, categories were collapsed for the closed cohort model. Age was centered on hospital site mean. For the open fracture cohort, injury severity score underwent a log transformation. For the closed fracture cohort, fracture location and an interaction term between fracture location (femoral shaft vs proximal femur) and COVID-19 period were included as independent variables. The interaction term was further probed to understand the association of fracture location and COVID-19 period with the time-to-surgery outcome.
For the analysis investigating regional COVID-19 infection rates for the closed fracture cohort, age, ASA class, injury severity score, fracture location, and COVID-19 regional rate were included as control variables (same as the prior regression model). Because of some sites having so few cases that did not meet 24-hour time-to-surgery benchmarks, we had insufficient power and were unable to investigate regional COVID-19 infection rate for the open fracture cohort.
Considering the use of multiple imputation procedures, Rubin's rules for pooled estimates are reported for all analyses. 7

Results
A total of 3598 patients from 20 medical centers across the US and Canada were included. Patient demographics and injury characteristics are summarized in Table 1 for the open fracture cohort and       fracture location was also not significantly associated with time to surgery (OR, 0.88; 95% CI, 0.54-1.45; P = .62). Among femur and hip fractures, when evaluating the association between meeting less-than-24-hour time-to-operating room benchmarks and regional COVID-19 case rate, there was still no association between regional COVID-19 rate and meeting time-to-operating room benchmarks (OR, 1.07; 95% CI, 0.70-1.64; P = .76) ( Table 5).

JAMA Health Forum | Original Investigation
When the association between race and delay beyond benchmarks were assessed, there was no association between race and meeting 24-hour time-to-surgery benchmarks in the open fracture cohort. However, non-White race was associated with increased odds of not meeting 24-hour timeto-surgery benchmarks in the hip/femur fracture cohort (OR, 2.93; 95% CI, 2.04-4.27; P < .001) (eTables 2 and 3 in Supplement 1).
There was a reduction in the number of patients screened who sustained study-eligible traumatic injuries and the number of patients enrolled in March and April, which rebounded in May and June, reflecting both the association of COVID-19 with research infrastructure and patient enrollment as well as the volume of trauma at these centers (eFigure 4 in Supplement 1).

Discussion
In this cohort study, there was no association between meeting time-to-surgery benchmarks in either open fracture or closed femur/hip fracture during the COVID-19 pandemic compared with before the pandemic. The COVID-19 pandemic has created unprecedented challenges to our global and national health care systems, stressing our resources and shifting protocols around clinical management strategies. As a result of increasingly limited resources, many hospital systems were forced to reduce operating room capacity and reallocate resources, which raised concern around delays in acute management of patients with traumatic injury. However, contrary to the hypothesis, in this cohort study of patients enrolled at 1 of 20 sites into the PREP-IT program, the COVID-19 pandemic was not associated with the ability of hospital systems to meet 24-hour time-to-surgery benchmarks for open fractures, hip fractures, and femur fractures during the COVID-19 pandemic. It is particularly striking that there was a rate of only 3% for delay to the operating room beyond 24 hours in the open fracture group. In the femur/hip fracture cohort, there was a 22% rate of delay beyond 24 hours.
There are several potential reasons for these results, which likely vary across institutions.
Overall, it is reassuring that hospitals included in this study were able to effectively manage operating room resources to ensure timely care of acutely injured patients. Owing to the large scale of this  study and its retrospective nature, it was not possible to report which specific components of trauma workflow may have contributed to the preservation of standard time-to-operating room protocols.

JAMA Health Forum | Original Investigation
It is possible that a reduction in trauma volumes contributed to this, as many fields saw an acute reduction in the number of patients presenting for acute issues, such as stroke, 8,9 myocardial infarction, 10,11 and emergency surgery. 12 In the UK and Scotland, trauma admissions were reported to be down during the lockdown compared with prior years. 13,14 This was thought to be associated with collateral effects of social distancing and avoidance of health care institutions due to fear of contracting the virus. 15 There was a reduction in the number of patients who were screened and who were enrolled into the PREP-IT cohort after COVID-19. However, this may be, in part, due to The 3% rate of delay beyond 24 hours in the open fracture group was reassuring and was actually so uncommon that it limited our analytic power. This study did identify a relatively high rate of greater-than-24-hour time to operating room of 22% in the hip/femur fracture group, which was similar in both the hip and femur fracture subgroups. Delayed time to operating room in the open fracture cohort had an independent association with less severe fracture contamination and higher Injury Severity Score. These associations are not surprising, as surgeons likely feel increasingly motivated to ensure early time to operating room in more highly contaminated injuries and necessary surgery is most frequently delayed by increasing injury severity in remote regions, resulting in critical systemic condition. In the hip/femur fracture cohort, delayed time to operating room was independently associated with higher ASA class, which reflects delays due to medical complexity.

Limitations
There are several limitations associated with this study. Because failure to meet time-to-surgery benchmarks was rare in the open fracture group (only 3%), there was a concern about the potential for underpower. However, a post hoc power analysis was performed and demonstrated that even in significant), which substantiates the conclusion that COVID-19-related resource reallocation does not appear to have had a detrimental association with meeting time-to-surgery benchmarks.
Furthermore, the low rate of delayed time to surgery throughout the entire study time is reassuring, broadly speaking. While this study did not capture data on the presence of COVID-19 infection in acutely injured patients or at specific institutions enrolling patients, the rate of infection is likely related to the regional rates, which was factored into the second multivariate model in closed fractures. Because of the large scale of this study and its retrospective nature, it was not possible to report hospital-specific policies. Additional research is needed to evaluate the outcome of specific policies on ability to meet benchmarks. Although there are no published or accepted values for a clinically important difference, the less than 3% differences between the groups in both cohorts are unlikely to represent a clinically relevant difference. Because the rate of not meeting the less-than-24-hour time-to-surgery benchmarks was low in open fracture patients, there was not enough statistical power to evaluate the association between regional COVID-19 case rate and time to operating room. Although the reasons underlying the less-than-24-hour time-to-operating room standards in closed femur and closed hip fractures differ, these were evaluated in a single closed fracture cohort. A single model approach was used for both hip and femur fractures because the same set of independent variables was relevant for inclusion in the regression model. This has the benefit of a larger sample size and thus more statistical power, which helps to alleviate issues associated with imbalance in outcome (the majority of patients received surgery within 24 hours). An interaction term was included, which evaluates the independent association of fracture type (femur vs hip) with COVID-19 variables, and there was no independent association of this interaction term with the time-to-surgery metric. There was a high degree of missingness in Injury Severity Scores, which is an important factor in considering time to surgery. However, the risk of bias associated with this was mitigated using multiple imputation procedures. Because the PREPARE trial was not designed to assess the outcomes of the COVID-19 pandemic, the groups are not evenly distributed and there are differences between the groups. However, multivariate regression was used to control for these differences, as is indicated for retrospective or observational research.
There were several strengths associated with this study. The high-quality prospectively collected data represent a substantial strength. Furthermore, 2 multivariate models were created in an effort to answer this question, assessing any change in time-to-operating room benchmarks from pre-COVID-19 to after the federally mandated declaration of a national emergency was announced and also based on regional COVID-19 rates derived from the Centers for Disease Control and Prevention. This analysis included a large sample size adding to the power and inclusion of 20 sites, which expands generalizability. The multicenter nature of this study allowed us to evaluate the broad associations of the COVID-19 pandemic on a national scale, as opposed to a single center.

Conclusions
Although the COVID-19 pandemic has created unprecedented challenge to health care systems far and wide, this cohort study suggests that hospital systems throughout the US and Canada were able to effectively mitigate these challenges such that 24-hour time-to-surgery benchmarks were unchanged from before the pandemic to after the pandemic.