Mitigating the impact of coronavirus disease 2019 on emergency stroke care: an original study and meta-analysis

Ethical approval

This study was conducted in accordance with the Declaration of Helsinki and was approved by the ethical review board in West China Hospital of Sichuan University (2019/319).

Study design

This retrospective, multi-center study was conducted at three designated hospitals for COVID-19 including West China Hospital of Sichuan University, Chengdu Fifth People’s Hospital, and West China-Guang’An Hospital of Sichuan University. The West China Hospital of Sichuan University is the largest hospital with around 4300 beds and 6700 healthcare workers in the southwest of China. The hospital annual census has exceeded five million patients. The Chengdu Fifth People’s Hospital served the majority of patients in the satellite towns of Chengdu, with over 2400 staffs. The West China-Guang’An Hospital of Sichuan University is located in Guang’An, a big city which is nearest to Hubei Province, the epicenter in China. Under the city lockdown, people in Guang’An could be in great fear of affection and paid more attention on keep social distance than others.

During the pandemic, medical practice strictly followed hospitals contingency plans, and the recently developed expert consensus on the stroke treatment announced by China National Health Commission and Chinese Medical Association (COVID-19 Neurologists Consensus Collaborators 2020; National Health Commission Stroke Expert Committee of China 2020a, 2020b, 2020c). Experience from other cities in China were also used (Zhao et al. 2020a, 2020b). Figures 1 and 2 summarized the changes in the process of emergency stroke treatment before and during the crisis period. These included:

1. All staff were equipped with standardized personal protection including goggles, protective gowns, surgical masks, and gloves when practicing in emergency department;

2. All patients had temperature screening at the entrance to the hospital and checked again before treatment was started. Epidemiological investigation (questionnaires about whether they had been to epicenter of the outbreak, or in contact with infected patients) was conducted at entrance;

3. Patients were all required to take computed tomography (CT) test for chest to screen viral pneumonia and identify suspected case;

4. Before conducting endovascular thrombectomy, the attending doctor checked the temperature again. A negative pressure operation room was established for patients suspected or positive for COVID-19;

5. Suspected patients were kept in an isolation room for medical observation. After the confirmation by quantitative real-time polymerase chain reaction (qRT-PCR) and/or CT test of chest, negative patients would be admitted to the normal neurology ward. The temperature was taken again before entering the ward;

6. For the hospital located nearest to the epicenter, the qRT-PCR test was required for all patients.

7. One-caregiver demand. Each patient could only have one caregiver for company. Everyone in the ward had to wear a mask.

Figure 1:

The protocol of stroke emergency treatment during the pre-epidemic stage. AIS, acute ischemic stroke; ED, emergency department; CT, computer tomography.

Figure 2:

The protocol of stroke emergency treatment during the peak-epidemic stage. T, temperature; AIS, acute ischemic stroke; ED, emergency department; CT, computer tomography; qRT-PCR, quantitative real-time polymerase chain reaction; IVT, intravenous thrombolysis; MT, mechanical thrombectomy.

Participants

The most suspenseful COVID-19 period in Sichuan was defined from Dec 1, 2019 to Mar 11, 2020, 102 days in total. According to the governmental announcement, the first COVID-19 case in Sichuan was reported on Jan 21, 2020, so we further divided the COVID-19 period into early epidemic stage (Dec 1, 2019 to Jan 20, 2020) and peak epidemic stage (Jan 21 to Mar 11, 2020). Moreover, we also explored the late epidemic (Mar 12 to May 1, 2020) and pre-epidemic stages (Jan 21 to Mar 11, 2019). All stages last 51 days. We consecutively registered patients visiting emergency department of stroke center diagnosed with acute ischemic stroke during the four stages according to the criteria of World Health Organization (Recommendations on stroke prevention, diagnosis, and therapy 1989). Patients who presented with hemorrhagic stroke or other neurological diseases were excluded from the study.

Study variables and outcomes

Information about the stroke emergency visits, demographics and processes of care including patient volume, reperfusion therapy rates, delays in visits and outcome after recanalization were obtained from the database of these three stroke centers. The National Institutes of Health Stroke Scale (NIHSS) score was used to assess severity of neurologic deficit. Both NIHSS at admission and discharge were collected to compare the neurological improvements after treatments including reperfusion therapies. We calculated admission rates for acute ischemic stroke, rates of reperfusion therapies [including intravenous thrombolysis (IVT) and mechanical thrombectomy (MT), onset to admission time, door to needle time (DNT), door to puncture time (DPT), and neurological improvements during the baseline and COVID-19 periods as primary outcomes. Changes in the stroke care in two representative hospitals during peak-epidemic stage were set as secondary outcomes.

Statistical analyses

Comparisons were performed using SPSS, Version 20.0 (IBM, Armonk, New York). The increase and decrease in percentages were presented to show the changes during the whole crisis period. Continuous variables including age and patient volume were reported as mean ± standard deviation and compared by Kruskal–Wallis H-test or ANOVA after Normal distribution test and homogeneity test of variance. Categorical variables including sex distribution, time from onset to admission, disease severity, rates of reperfusion therapies, and neurological improvements were presented as frequency (percentage), using Chi-Square test or Fisher exact test. Bonferroni correction was performed to adjust for multiple tests (ɑ = 0.05/6 = 0.0083). Missing data within 5% of all cases were allowed, and the missing parts would be deleted when analyzing. All P-values are two-sided, with P < 0.05 considered statistically significant.

Systematic literature search and meta-analysis

We strictly followed the MOOSE (Meta-Analysis of Observational Studies in Epidemiology) guidelines throughout the design, implementation, analysis and reporting processes in this study (Stroup et al. 2000).

Search strategy and data extraction

A systematic literature search was conducted in PubMed, Ovid Embase and Cochrane online scientific publication databases for English language publications from Dec 2019 to July 2020 to identify potentially relevant observational studies. The following medical subject heading (MeSH) terms and/or text words were used “stroke or ischemic stroke or cerebrovascular incident or cerebrovascular diseases” and “COVID-19 or the 2019 novel coronavirus disease or 2019-nCoV or SARS-Cov-2 or severe acute respiratory syndrome coronavirus 2”.

Study selection

Two qualifies investigators independently screened the titles and abstracts from the initial search to assess potential relevance (JW and YH). Discrepancies were resolved by discussion. Furthermore, full-text articles were obtained to reevaluate the eligibility and determine the final inclusion. Included studies should: 1) report data from an original, peer-reviewed observational human study of all designs, 2) present the change in stroke care quantitatively before and during the COVID-19 pandemic. We excluded studies that with duplicate citations or unavailable data. When studies had been published more than one time on the same topic and theme, only the most recent study was selected to extract data, others were listed and analyzed as similar studies. References of the included studies were also reviewed.

Study quality assessment

Two investigators (JW and YH) independently assessed the quality of each study using the Agency for Healthcare Research and Quality (AHRQ) including following items: source of information, inclusion and exclusion criteria, study period indication, consecutive subjects or not, evaluators of subjective components of study were masked to other aspects of the status of the participants, assessments undertaken for quality assurance purposes, patients exclusion from analysis, confounding assessment, missing data, patients response rates and completeness of data collection, follow-up related (Rostom et al. 2004). An item would be scored 0 if it was answered no or unclear; inversely the item would be scored 1 with a yes answer. Total score was 11. High quality was defined as 8–11. Discrepancies were resolved by consensus with input from a third investigator (MM).

Data extraction and outcome variables

The following information was extracted from the included studies: authors’ name, publication year, study countries (regions), participants’ ages and sex ratios, number of participants, study periods and quality scores. Outcome parameters included the change in the rates of reperfusion therapies during pandemic.

Statistical analysis

The pooled outcomes were meta-analyzed if sources were available from a minimum of three studies using a random or fixed effect model according to the heterogeneity. The 95% CI of the odds ratio (OR) and their 95% CIs were calculated and presented by forest plot. Heterogeneity was quantified with the I² test (values of 0–40%: not important, 30–60%: moderate heterogeneity, 50–90%: substantial heterogeneity). Once substantial heterogeneity was assessed, analyses predefined by stratifying original estimate (countries/regions) would be conducted to further investigate the sources of heterogeneity between studies. Qualitative analyses using funnel plots were performed to assess publication bias (10 studies at least). We extracted data from specific studies using Chi-Square test or Fisher exact test to compare percentage change. Bonferroni correction was used to adjust for multiple tests (ɑ = 0.05/3 = 0.00167). The analyses were performed using the metaphor package for Review Manager software version 5.3 and SPSS, Version 20.0 (IBM, Armonk, New York). P < 0.05 was considered statistically significant.

Participant characteristics

The mean age (and standard deviation) of the patients seen in the stroke center for emergencies during the pre-, early, peak and late epidemic stages were comparable with 69.2 (13.2), 70.2 (12.6), 69.6 (12.4) and 71.2 (11.5) years respectively (Bonferroni adjusted P > 0.05). Male patients accounted for 55.7% in pre-epidemic stages, 55.8% in early epidemic stage, 56.9% male patients in peak epidemic stage and 53.9% in late epidemic stage (Bonferroni adjusted P > 0.05). None of stroke patients were diagnosed with COVID-19.

Comparison between four epidemic stages

Details of the comparison between four epidemic stages appear in Table 1. There were no significant changes in patient volume, stroke severity, rate of intravenous thrombolysis rate (IVT) and neurological improvements in the whole COVID-19 period. However, rates of mechanical thrombectomy (MT) in early and peak epidemic stages were significantly higher than pre-epidemic period and more patients came to hospital within 6 h after stroke onset during peak pandemic (all Bonferroni adjusted P < 0.05). Descriptive data especially on early and peak epidemic stages are depicted in Figure 3.

Figure 3:

Numbers of patient visits in stroke emergencies during the whole COVID-19 period. The x-axis represents time weekly started from Dec 7 the first week to Mar 11 the last week in COVID-19 period. The y-axis represents number of patients visit multiple stroke centers for emergency care.

Single-center analysis based on different characteristics

Literature search

Based on the pre-specified selection criteria, 583 studies were excluded after the initial screen of title. 276 studies were excluded after reading the abstracts. 43 studies were excluded after a review of the full text review for several major reasons including: studies were presented as case reports, reviews or editorial articles; no pre-pandemic data or no separate analysis on ischemic stroke care. In total, 11 published articles, together with our current study were included for meta-analysis (Ahmad et al. 2020; Diegoli et al. 2020; Frisullo et al. 2020; Kristoffersen et al. 2020; Pop et al. 2020; Rudilosso et al. 2020; Siegler et al. 2020; Meza et al. 2020a,b; Silva et al. 2020). All data were successfully extracted from the published studies. The selection process is provided in Figure 4.

Figure 4:

Study screening flow diagram.

Study characteristics of eligible studies

The meta-analysis was performed on 12 studies (i.e., 11 previously published studies identified by literature search as described in the methods section as well as the current original study) with a total of 4,789 patients (3,149 patients in pre-pandemic and 1,640 patients during pandemic). The number of participants ranged from 57 (Silva et al. 2020) to 2737 (Meza et al. 2020a). Six studies were performed in European countries (Frisullo et al. 2020; Kristoffersen et al. 2020; Pop et al. 2020; Rudilosso et al. 2020; Meza et al. 2020a,b), two were conducted in the United States (Ahmad et al. 2020; Siegler et al. 2020), two were in Brazil (Diegoli et al. 2020; Silva et al. 2020), and another 2 in the China mainland (current original study) and Hong Kong respectively (Teo et al. 2020). Data collection was all based on cross-sectional design. A significant reduction in thrombolysis occurred in one study (Frisullo et al. 2020), total reperfusion therapies in another study (Pop et al. 2020). Others reported no significant association (Ahmad et al. 2020; Diegoli et al. 2020; Kristoffersen et al. 2020; Rudilosso et al. 2020; Siegler et al. 2020; Meza et al. 2020a,b; Teo et al. 2020; Silva et al. 2020) and current original study. A total of three studies had a low risk of bias (Diegoli et al. 2020; Siegler et al. 2020; and current original study), and nine had a moderate of bias (Ahmad et al. 2020; Frisullo et al. 2020; Kristoffersen et al. 2020; Pop et al. 2020; Rudilosso et al. 2020; Meza et al. 2020a,b; Teo et al. 2020; Silva et al. 2020). The characteristics of included studies are summarized in Table 4.

Change in the rates of IVT and MT treatments

The summarized characteristics of reperfusion therapies and stroke severity were reported in Table 5. There were significantly more severe stroke cases (OR = 1.49, 95%CI 1.01, 2.19; P = 0.044) and less minor stroke admission (OR = 0.75, 95%CI 0.58, 0.96; P = 0.025) after the outbreak as reported. While we found an increase in the number of MT treatment, the reductions in IVT and total reperfusion therapies were presented. However, none of them showed statistical difference. Moreover, as conducted in different countries/regions with completely different epidemiologic situation and health services, included studies with IVT/MT changes were further divided into three major parts (Europe, Asia and the United States) to compare data before and during the epidemic. In Table 6, the rate of IVT was significantly higher in Europe (32.6%) than in Asia (22.7%) and the United States (12.0%) before the epidemic. After COVID-19 outbreak, Asia started to show the highest rate of IVT (26.6%) among these three parts (13.2% for Europe and 5.7% for the United States) of world (Bonferroni adjusted P < 0.05). The United States presented lowest IVT rates in both periods. However, things changed when it came to the rate of MT. The highest rate of MT was significantly showed in Asia (20.3%) before the epidemic while the United States (34.1%) during the crisis (Bonferroni adjusted P < 0.05).

Investigators in totally 11 studies respectively reported on the change in the rates of IVT or MT treatments during the epidemic (Figure 5). Results showed the forest plot of the pooled OR for the effect of the crisis on stroke emergency care. With an OR of 0.87 (95% CI 0.74, 1.02), it showed a trend toward great impact on thrombolysis treatment, but not significantly (P = 0.09). Similarly, analysis was conducted by fixed-effect model among studies focused on MT treatment during the crisis, with an OR of 1.11(95% CI 0.96, 1.29; P = 0.17), presenting no statistical significance either. Heterogeneities were mild (I² = 21% for IVT and 36% for MT) between studies.

Figure 5:

Pooled estimate of change in the rates of IVT and MT treatments.

Change in the rates of total reperfusion therapies

Ten studies that explored change in the rates of total reperfusion therapies caused by the COVID-19 pandemic were included in the meta-analysis, as shown in Figure 6. The pooled data under random-effect model showed an OR of 0.89 (95% CI 0.68, 1.15; P = 0.36), with substantial heterogeneity (I² = 51%, P = 0.03). Thus, a stratified analyses restricted to European studies were conducted to investigate potential source of heterogeneity (Figure 7). With the strict inclusion of only European studies, the heterogeneity between studies was reduced to I² = 34% (P = 0.20), with an OR = 0.92 (P = 0.27).

Figure 6:

Pooled estimate of change in the rates of total reperfusion therapies.

Figure 7:

Pooled estimate of analyses stratified by nation.

Publication bias

Publication bias assessments were carried out on studies focused on IVT and MT treatments separately. Funnel plots suggested possible publication bias as asymmetric graph was observed (Figure 8). We aimed for a comprehensive search, thus published and unpublished studies were all included. Moreover, we noticed that six of our included studies had small sample sizes (<200), which could somewhat reduce the statistical power of analyses and increase the publication bias. However, the limited number of studies did not permit many of our planned subgroup analyses to furthermore detect potential biases.

Figure 8:

Funnel plot for publication bias.