Eunpyeong St. Mary's Hospital was established in April 2019. In February 2020, after a transport agent of the hospital was diagnosed with COVID-19, the hospital had to be closed under the direction of the Korean Disease Control and Prevention Agency. The hospital was reopened from March 2020 after the COVID-19 protection protocol was established. We retrospectively analyzed the mechanical thrombectomy outcomes before and after the establishment of the protection protocol. Between April 2019 and December 2020, 109 patients with large vessel occlusion were treated with mechanical thrombectomy for an acute stroke. Of these, 55 were diagnosed between April 2019 and February 2020 and 54 between March 2020 to December 2020. Of the total patients, five were excluded because a stenotic lesion without occlusion was present and only angioplasty was performed. Seven patients were excluded because they had only undergone diagnostic angiography and their occluded vessels were judged to have spontaneously recanalized, or they had chronic occlusions that did not need further intervention. We further excluded nine patients who underwent mechanical thrombectomy during hospitalization. Finally, 88 patients were included in this study who underwent mechanical thrombectomy for large vessel occlusions on an emergency basis. Of these 88 patients, 45 were managed using the conventional protocol and 43 were treated as per the COVID-19 protocol (Fig. 1).

In this study, we investigated the last-normal-to-door time, first-abnormal-to-door time, door-to-imaging time, door-to-puncture time, puncture-to-recanalization time, total procedure time, Thrombolysis in Cerebral Infarction (TICI) grade, and modified Rankin score (mRS) at discharge.

Multiphase computed tomography (CT) angiography (CTA) and perfusion CT (CTP) were performed to identify the location of an occlusion site (SOMATOM Force; Siemens, Erlangen, Germany) to evaluate the extracranial and intracranial vessel lesions. If enhanced imaging was not possible, a non-enhanced brain CT scan and diffusion magnetic resolution image (MRI) with perfusion were recorded with time-of-flight MR angiography to detect vessel lesions (MAGNETOM Vida; Siemens). A biplane angiography equipment, Siemens’ Artis Q biplane, was used for mechanical thrombectomy.

The conventional protocol was applied before the COVID-19 pandemic. When a patient visited the emergency room (ER) of our hospital, an emergency medical doctor first triaged the patient, and a “severe emergency call” was activated if the patient had developed neurological symptoms within the last 24 hours. Simultaneously, tissue plasminogen activator (tPA) was prepared according to the patient’s weight, and an intravenous line was secured for CTA. Once the stroke team arrived at the ER, brain images were evaluated, and if the patient had arrived at the hospital within 4.5 hours from the last-normal time, intravenous tPA was administered, except for patients for whom it was contraindicated [13].

Mechanical thrombectomy was considered in the cases where anterior circulation (internal carotid artery or middle carotid artery) occlusions or posterior circulation (basilar artery or posterior cerebral artery) occlusions, confirmed as small infarctions with diffusion/perfusion mismatch, occurred within 24 hours or 6 hours from the last-normal time [12,13].

Diffusion MRI was required to confirm the infarction volume in selective cases. Mechanical thrombectomy was performed in the angiography room or the hybrid operation room depending on the availability of staff. If mechanical thrombectomy was conducted in the hybrid operation room, the interventionist determined whether to apply general anesthesia.

Under the COVID-19 protocol, each patient who arrived at the ER was initially assessed by an emergency medical doctor in the outdoor triage room. Then, the patient was moved to an isolated negative pressure room and photographed using a portable chest radiography machine. Any patient with consolidation or ground-glass opacity on chest radiograph or upper respiratory symptoms or fever was suspected to be COVID-19 positive (+). Otherwise, the patient was assumed to be COVID-19 negative (-). Even if the patient was assumed to be COVID-19 (-), a confirmative test was conducted. For all patients, an isolated route was used for all subsequent movements, tests, and procedures before the COVID-19 test result was confirmed. A negative pressure tent was used outside the negative pressure room for suspected COVID-19 (+) suspected patients. For suspected COVID-19 (-) patients, a Kf94 (Dr. Puri^® KF [Korea filter] 94-fine dust, yellow dust mask; KM Healthcare Corp., Guri, Korea) or N95 (3 M™ Health Care Particulate Respirator 1860; 3 M Health Care™, St. Paul, MN, USA) filtering facepiece respirator and a bouffant cap were worn by the treating staff [6]. When patients were taken to the CT room for brain imaging, non-enhanced chest CT was also conducted to check for lung lesions as it is more precise than radiography. If a lung lesion that was not seen on the radiograph was found on chest CT, patients were treated as suspected COVID-19 (+) cases thereafter [8]. Following the new protocol, only CTP was used to evaluate the hypoperfusion area in suspected COVID-19 (+) patients since there was a high chance of time delay for other patients because of the need for complete blockage and cleaning of the MRI room, which usually took more than 30 minutes. To protect the medical staff, our hospital established a PPE system according to the guidelines of the Korean Disease Control and Prevention Agency [7]. When treating suspected COVID-19 (+) cases, level D protection gown with double gloving, protection goggles, and an N95 mask were worn by the medical staff, and a positive air purifying respirator was added during procedures. When treating suspected COVID-19 (-) patients, basic PPE, including a disposable plastic apron, a Kf94 or N95 mask, a surgical cap, nitrile gloves, and a face shield (5-set) were worn by the medical staff [7,18]. In our facility, the angiography room was located in a separate space with a separate air-conditioning system. In contrast, the hybrid operation room was not isolated from other operating rooms. Therefore, when performing mechanical thrombectomy in patients who were suspected to be COVID-19 (+), only the angiography room was used. If a patient was suspected to be COVID-19 (-), the hybrid operation room was used, but the air-conditioning was blocked to prevent any cross-contamination. Once mechanical thrombectomy was performed, patients were transferred to a neuro-intensive care unit (NCU) via an isolated route under a negative pressure tent if suspected to be COVID-19 (+) or with a Kf94/N95 mask and bouffant cap if suspected to be COVID-19 (-). Patients were kept in a negative pressure isolation room inside the NCU until COVID-19 test results were confirmed. Once a patient was confirmed to be COVID-19 (-), the patient was transferred via the routine route and routine care was provided (Fig. 2).

Patients were classified into two groups according to the protocol used. The baseline characteristics of these groups are listed in Table 1. A chi-square test or Fisher’s exact test was used for statistical analysis according to the variable factors of the two groups. A Wilcoxon-Mann-Whitney test was used to compare continuous time variables. Before the comparisons, we evaluated the statistical power of the Wilcoxon-Mann-Whitney test by the method proposed by Shieh et al. [16]. Assuming that the time variables of the two groups followed an exponential distribution, the effect size (the probability that the value in the first group is less than the value in the second group) was calculated to be 0.7. A two-sided comparison was performed with a significance level of 0.05, and the calculated statistical power was over 0.8 when the number of patients in each group exceeded 40. When comparing the time variables among the three groups, analysis was conducted using a Kruskal-Wallis test followed by a post hoc Dunn’s test with Bonferroni p-value adjustment [2]. We also conducted a multivariate analysis to identify the variables that have a significant influence on the procedure time. In the multivariate analysis, linear regression models with log-normal errors and identity link were used [10] because we observed that the time variables showed evidence of heteroscedasticity and, among the generalized linear models with various distributions, the log-normal regression models fitted the data properly with the lowest Akaike’s information criterion. We set up the multivariate models using a stepwise algorithm [9], and the p-value for each variable was calculated through a likelihood ratio test between the models including and excluding the variable. A variable was included in the model if the p-value was less than 0.10. All statistical analyses were performed using the R software (version 4.0.3; R Foundation for Statistical Computing, Vienna, Austria). All p-values were two-sided.

The age distribution of patients between the two groups was similar (p=0.7101), while the proportion of females was higher in the conventional group than in the COVID-19 group (48.9% vs. 27.9%, p=0.0716). Compared to that before the pandemic, significantly lower proportion of patients received tPA outside the center after the COVID-19 outbreak (No tPA, tPA at our center, tPA at other centers : 62.2%, 20%, and 17.8%, respectively, in the conventional group and 65.1%, 32.6%, and 2.3%, respectively, in the COVID-19 group; p=0.0405). No statistically significant differences were found in the last-normal-to-door time and the first-abnormal-to-door time between the groups (p=0.6612 and 0.1406, respectively) (Table 1).

In the conventional group, mechanical thrombectomy was more frequently performed in the hybrid operating room than in the angiography room (62.2% vs. 37.8%). In contrast, for the COVID-19 group, thrombectomy was more frequently performed in the angiography room than in the hybrid operation room (14.0% vs. 86.0%), and this difference was significant (p<0.0001). Furthermore, mechanical thrombectomy was performed more frequently under local anesthesia in the COVID-19 group than in the conventional group (83.7% vs. 31.1%; p<0.0001). Among the time indices, there was a significant increase in the door-to-imaging time in the COVID-19 group compared to that in the conventional group (median time in the conventional group and COVID-19 group was 14 minutes [interquartile range (IQR), 6–18] and 15 minutes [IQR, 12.0–25.2], respectively; p=0.0257). Among patients who received tPA, the difference in the door-to-tPA time was borderline between the two groups (median 51 minutes [IQR, 49–57] in the conventional group and 48 minutes [IQR, 45.0–51.8] in the COVID-19 group; p=0.0941). It is noteworthy that the door-to-imaging time in the COVID-19 group tended to be longer than that in the conventional group, but no significant difference was observed in the door-to-puncture time between the two groups. The median door-to-puncture time in the conventional and COVID-19 groups was 145 minutes (IQR, 136–162) and 140 minutes (IQR, 107.5–168.0; p=0.5042), respectively. The puncture-to-recanalization time was found to be shorter in the COVID-19 group than in the conventional group (conventional group : median 60 minutes [IQR, 47.0–94.5]; COVID-19 group : median 43 minutes [IQR, 26.5–74.0]; p=0.0093). Moreover, the difference in the total procedure time of the recanalized patients between the groups was borderline significant (median 209 minutes [IQR, 188.0–270.5] in the conventional group and 189 minutes [IQR, 158.0–241.5] in the COVID-19 group; p=0.0580). The postoperative TICI was superior in the COVID-19 group than in the conventional group (0, 2a, 2b, 3 : 0%, 4.7%, 25.6%, 69.8% in the conventional group and 4.4%, 4.4%, 48.9%, 42.2% in the COVID-19 group, respectively; p=0.0238). The patient mRS at discharge were not significantly different between the two groups (p=0.1043) (Table 2).

The patients treated under the COVID-19 protocol were classified into two groups : suspected COVID-19 (-) and COVID-19 (+) patients. We compared the door-to-imaging time and door-to-puncture time according to the subgroups (conventional, suspected COVID-19 [-], and suspected COVID-19 [+] groups). In this analysis, we found significant differences in the door-to-imaging time between the three groups (p=0.0048, Fig. 3), with a significant increase in the suspected COVID-19 (+) group (post hoc p=0.0043 and 0.0511, between the conventional group and suspected COVID-19 [-] group). The median door-to-imaging time in the suspected COVID-19 (+) and COVID-19 (-) groups was 28.5 minutes (IQR, 22–32) and 15 minutes (IQR, 11.2–19.0), respectively. There was no significant difference in the door-to-puncture time between the three groups (p=0.7436).

Multivariate analysis was conducted to identify factors that have a significant impact on the door-to-puncture time and total procedure time (Table 3). In our study, the door-to-puncture time was significantly influenced by the occlusion site (p=0.0356), the last-normal-to-door time (p=0.0068), Glasgow coma scale (GCS) score at arrival (p=0.0045), and the door-to-imaging time (p=0.0097). Type of anesthesia showed borderline significance (p=0.0694). In the model for total procedure time, the same variables were shown to be significant, except GCS score (p-value for occlusion site, last-normal-to-door time, GCS score, door-to-imaging time, and anesthesia type : 0.0001, 0.0231, 0.0574, 0.0103, and 0.0207, respectively). Among the hospital-based variables, the door-to-imaging time was the factor that contributed to a significant increase in the door-to-puncture time and total procedure time. The expected increment of the door-to-puncture time per 10-minute increase in the door-to-imaging time was 11.28 minutes (95% confidence interval [CI], 0.54–17.29), and the expected increase in the total procedure time per 10-minute increase in the door-to-imaging time was 15.4 minutes (95% CI, 3.65–27.76). A significant delay in the total procedure time was observed among patients who received general anesthesia. In our multivariate analysis, expected delay in the total procedure time was 28.39 minutes (95% CI, 4.45–52.76) in patients who underwent mechanical thrombectomy under general anesthesia when compared to those who received local anesthesia.