Short-Term Effects of Canceled Elective Procedures Due to COVID-19

Evidence From the Veterans Affairs Healthcare System

Linda Diem Tran, PhD, MPP; Liam Rose, PhD; Tracy Urech, MPH; Aaron Dalton, MA, MSW; Siqi Wu, MPH; Anita A. Vashi, MD, MPH, MHS


Annals of Surgery. 2021;274(1):45-49. 

In This Article


Data and Population

We used data from the VA Corporate Data Warehouse (CDW),[10] a database of all VA electronic health records. We assessed Current Procedural Terminology (CPT) codes, procedure dates (ie, scheduled, and completed or canceled), and reasons for cancellations from the CDW surgery domain for calendar years 2018–2020. Cancellation reasons included changes in medical condition and other nonhealth reasons, but inconsistencies in data entry during the early period of the pandemic prevented data from being sufficiently precise to differentiate from COVID-19 and other reasons for cancellation.[2]

To examine the impact of cancellations due to the pandemic rather than changes in patient condition, we selected cases canceled on March 13–19, 2020. Cancellations spiked dramatically during this seven-day period, which supports our claim that these cases were canceled due to the nationwide order (Figure 1). The Supplementary Figure 1, illustrates that most canceled cases were scheduled for that week, but some were scheduled for more than a month later. We limited cancellations to cases canceled for the following reasons: environmental issue (2585, 71.2%), clinic-related issue (279, 7.7%), and patient-related issue (517, 14.2%). We excluded surgical cases moved to earlier dates and cases labeled urgent or emergent — which should not have been affected by the national order — (29, 0.8%), and cancellations due to patient-health related issues (201, 5.5%). For comparison, nearly 30 percent of VA surgical cancellations were attributed to changes in medical condition before the pandemic.[11] Twenty (0.6%) canceled procedures with missing reasons or missing cancellation reasons were also excluded. In supplementary analyses, we limited the sample to only those cases canceled on March 16th or 17th.

Figure 1.

Time profile of the number of elective procedures canceled between February 1 and May 31, 2020. Gray area indicates March 13–19, 2020.

The study cohort included patients with therapeutic and diagnostic elective procedures scheduled by all VA surgical service lines. We limited the sample to surgical procedures defined as CPT codes 10004–69979. We converted principal CPT codes to clinically meaningful procedures using the Agency for Healthcare Research and Quality's Clinical Classifications Software and used the 179 classifications as surgical case categories.[12] To classify the operative complexity of surgical procedures, we used invasive procedure complexity designations assigned by the VA Surgical Quality Improvement Program (VASQIP), a nationwide quality improvement effort responsible for measuring and improving the quality of surgical outcomes within the VA and the progenitor to the National Surgical Quality Improvement Program.[13] VASQIP categorizes procedures using the following complexity designations: standard, intermediate, complex, or not in complexity matrix.

Finally, all-cause mortality data were extracted from the CDW, which is updated nightly. Date of death information come from numerous sources, including data in Social Security Administration Death Master File, from the Department of Defense, the National Cemetery Administration, VA medical facilities, and spousal or family notification. Lags between date of notification and date of death vary by source. The latest surgical case in the COVID-19 group was scheduled for June 2, 2020, and the latest data extraction on September 5, 2020 included death records as recent as September 3, 2020.


We examined 30- and 90-day ED use and 30- and 90-day mortality from patients' scheduled surgery dates.

Covariates included patient age, sex, race, ethnicity, marital status, and VA priority group assignment (high disability: priority groups 1 and 4, low/moderate disability: 2, 3, 6, nondisabled, co-pay required: 7 and 8, and low-income: 5). Clinical measures included Elixhauser comorbidity index score[14] and quarterly Nosos score. The Nosos risk score indicates the patient's healthcare expense level compared to the average risk score in the VA population (eg, Nosos score of 1.5 indicates the patient is 50% more expensive compared to the average risk score).[15] Finally, we included a covariate for the VA medical center associated with the scheduled or completed procedure.

Statistical Analysis

To assess whether procedural cancellations due to COVID-19 had immediate adverse consequences on patient health, we compared the outcomes of patients who had elective procedures canceled due to COVID-19 (COVID-19 group) to the outcomes of similar patients who had the same procedure completed in March–June of 2018 and 2019 (comparison group).

We summarized patient sociodemographic and clinical characteristics for the COVID-19 group, the comparison group before matching, and the matched comparison group. Standardized differences were calculated to assess imbalance between the COVID-19 group and matched comparison groups. We used nearest neighbor matching with exact matching on surgical case category to estimate potential outcome means for patients who had procedures canceled due to COVID-19.[16] Agency for Healthcare Research and Quality's Clinical Classifications Software surgical case categories were matched exactly, and nearest neighbors were selected on patient covariates, VA facility, and procedure year based on Mahalanobis distance. COVID-19 canceled procedures that had fewer than two exact matching cases needed for robust standard error estimation (52, 2.2%) or that had missing values (3, 0.1%) were dropped. We applied bias-correction for matching on two or more continuous covariates and estimated robust standard errors.[17,18] Finally, we stratified the analyses by the following: whether the surgical case was assessed by the VASQIP, cases with intermediate operative complexity as defined by VASQIP, and cases with standard operative complexity as defined by VASQIP. All analyses were performed using Stata version 15.1 (StataCorp LLC, College Station, TX). This study was approved by the Stanford University institutional review board.