Estimated Risk of Adverse Surgical Outcomes Among Patients With Recent COVID-19 Infection Using Target Trial Emulation Methods

This cohort study emulates a trial within a large national veteran population to assess the risk of adverse postoperative outcomes among patients with recent COVID-19 infection.


eMethods. Study Design and Data Sources
This was a retrospective observational database study using the target trial emulation framework. The hypothetical trial is one where preoperative patients randomly acquire Covid-19 infection in the 30 or 60 days prior to surgery, and their postoperative outcomes are compared with a control group that was without Covid-19 infection within the same timeframe. While a randomized control trial (RCT) of this nature is obviously impossible, the emulation approach can reduce certain types of bias in observational studies. The Veterans Affairs Surgical Quality Improvement Program (VASQIP) was the data source for index surgeries, demographic and clinical characteristics, and adverse postoperative outcomes. The VA Covid-19 Shared Data Resource was used to assess vaccination and prior infection status for all participants. Approval was obtained from VA Boston Institutional Review Board, and STROBE guidelines were followed in reporting this study.

Eligibility and Baseline Characteristics
Eligibility criteria was major surgery performed during January 1, 2021 -September 30, 2021 reviewed by the Veterans Affairs Surgical Quality Improvement Program (VASQIP). The start date was chosen to coincide with widespread vaccine availability, as there was likely an overall change in perceptions of safety in operating on vaccinated patients, making comparisons with the previous year difficult. The end date reflects the most recent VASQIP data available. The specialties of anesthesia, endoscopy, ophthalmology, oral, proctology, and transplantation were excluded due to low volume. We obtained data on patient factors that were known on day 60 ("time zero" for pseudo-randomization) prior to the index surgery and were thought to be predictive of allocation to the exposure groups of interest.
Demographics included age, race, and sex. Clinical characteristics were American Society of Anesthesiologists Classification, body mass index (BMI), serum albumin, smoking, and diabetes. A 5-level geographical region was included to control for possible differences in SARS-CoV-2 strain prevalence in different parts of the US. Vaccination status was defined as a 3-level variable: "fully" if the final dose had been administered at least 14 days before time zero, "partial" if only the 1st in a 2-dose series was administered, otherwise "none." All VA enrollees were encouraged and eligible to receive vaccines at no cost in order to reduce risk of transmission, severe illness, hospitalization, and death.

Exposures
We defined 3 mutually exclusive exposure groups based on the most recent infection leading up to surgery: infection days 1-30, in days 31-60, or none. Infections prior to day 60 were not assessed. Covid-19 infection was defined as any positive/detected polymerase chain reaction (PCR) or rapid antigen (e.g. Binax) test. National VA policy during the study was to perform universal preoperative screening for Covid-19 infection. The dominant SARS-CoV-2 strains affecting our cohort were Alpha in early 2021, followed by Delta later that summer, both of which were notable for contagiousness and severe disease compared with the original strain. At the time, there were no oral outpatient therapies available, and outpatient use of Remdesivir had not been authorized. Therefore, we did not control for treatments as post-exposure confounders.

Outcomes
Selected 30-day postoperative adverse outcomes were then identified. Infectious outcomes consisted of pneumonia, sepsis, and superficial, deep wound, or organ space surgical site infections (SSI). Respiratory outcomes were reintubation or failure to wean from ventilator. The "any" adverse outcome category consisted of infectious or respiratory outcomes, as well as 30-day mortality, cardiac outcomes (myocardial infarction or cardiac arrest), nervous system outcomes (peripheral nerve injury or coma lasting > 24 hours), and vascular outcomes (pulmonary embolism, graft failure, cerebrovascular accident/stroke, deep vein thrombosis, or bleeding requiring > 4 units packed red blood cells).