Cardiac surgery outcome during the COVID-19 pandemic: a retrospective review of the early experience in nine UK centres

All cardiac surgery centres in the UK were invited to participate in this study. The WHO defined Covid-19 as a pandemic on 11th March 2020. However, clinical services in the UK had already started planning and responding to the crisis, had already treated a number of patients with Covid-19 and almost all elective cardiac surgery had been suspended. Thus, to capture the start of the clinical cardiac surgery response to the pandemic, all patients undergoing any form of adult (> 18 years old) cardiac surgery between 1st March 2020 and 30th April 2020 were included.

Data from the National Adult Cardiac Surgery Audit (NACSA) database was linked to Intensive Care National Audit and Research Centre (ICNARC) data and local existing databases (for Covid-19 related data) at each participating centre. The linked dataset was anonymised locally and submitted to St Bartholomew’s Hospital.

A standardised national Covid-19 swabbing protocol was not in place, and within centres, diagnosis protocols were not uniform. Most relied on one negative swab result anytime between one to 7 days prior to surgery. None undertook CT scanning at that time. Thus, for the purposes of this study a diagnosis of Covid-19 was defined as having a positive swab test or, where no test was performed, suspected clinically at any time pre- or post-surgery during the index hospital admission. This was reflective of current practice at this time.

The anonymised dataset from each centre was transferred via an AWS (Amazon Web Service) S3 bucket. Submission to S3 via the S3 console is secure, using HTTPS/SSL protocols. Each centre was allocated a user login, restricted only to the uploading of data files to the bucket. Access to download and archive of the submitted files into a designated archive sub-folder was restricted to the St Bartholomew’s research team only.

Results are presented as mean (SD) or median [IQR] for continuous variables and groups are compared using Mann-Whitney U test. Categorical data are presented as N(%) and compared using chi-squared or Fisher’s exact tests. Propensity score matching was used to assess the impact of Covid-19 on mortality and length of stay after accounting for potential covariates. We used optimal matching with a caliper width of 0.2 standard deviations. Three controls were matched to each Covid-19 case using age, centre, sex, ethnicity, diabetes, BMI, procedure type and urgency, resulting in improved balance between the groups as measured by the standardised mean differences (Fig. 1). Missing values were imputed prior to matching using a random forest algorithm [11]. Matched analysis was undertaken using conditional logistic regression and hierarchical quantile regression. Sensitivity analysis conducted using Firth logistic regression and quantile regression in a complete case analysis confirmed the results.

Table 1 details the baseline characteristics of the study cohort. Overall, patients were predominantly male (76.4%), of white background (87.1%) with a median age of 66 [57–72] years. Most presented for urgent (require surgery on this admission) (52.3%) or emergency (7.0%) first-time cardiac surgery (95.3%), staying in ICU for a median of 2 days and in hospital after the surgery for 6 days.

Comparing those with (n = 53) and without (n = 702) Covid-19 found those with Covid-19 were more likely to be male (96.0% v 75.2%, p = 0.004), of white background (90.9% v 86.8%, p = 0.046), diabetic (39.6% v 25.3%, p = 0.023), have higher pre-operative serum creatinine levels (96umol/L [82.5umol/L v 115umol/L], v 86umol/L [73umol/L − 101.5umol/L], p < 0.001) and greater calculated EuroSCORE II (3.91 [1.7–6.5] v 2.26 [1.4–4.9], p < 0.015) (Table 1). Those with Covid-19 were also more likely to undergo urgent (74.5% v 50.7%, p < 0.001) and more complex surgery (for example, combined CABG and valve surgery 26.8% v 8.8%; major aortic surgery 22.0% v 14.5%, p = 0.004) with longer cardiopulmonary bypass (116 [87–153] v 102 [75–127], p = 0.02) and aortic cross clamp times (84 [62–119] v 70 [51–100], p = 0.024) as a result. While there were no differences in ICU length of stay (LOS) or in-hospital post-operative morbidity, those diagnosed with Covid-19 remained in hospital a median of 5 days longer after surgery than those without Covid-19 (11 [6–15] v 6 [5–9], p = 0.001), and had an overall median hospital stay of 6 days more than those without it (15 [12–21 v 9 [7–15], p < 0.0001). Furthermore, the in-hospital mortality rate was seven times higher in those with (24.5%) than without (3.5%) Covid-19 (p < 0.0001) and for those who died the median time to death was significantly longer for those with (14 days [8–21 days] Covid-19 than without it (4 days [1–15 days], p = 0.02).

Of the patients with a post-operative diagnosis of Covid-19 (n = 35), 29 (82.9%) tested negative for Covid-19 at the time of surgery, where the Covid-19 status was not known or not tested in the remaining patients (likely due to being urgent/emergency cases). Similarly, of those with a pre-operative Covid-19 diagnosis (n = 17), all underwent urgent or emergency surgery.

Statistically significant differences are shown Table 2, highlighting that patients with a post-operative diagnosis of Covid-19 were older, had greater BMI, more likely to experience post-operative atrial fibrillation/flutter and had longer aortic cross clamp time than those diagnosed pre-operatively. They also were more likely to have experienced longer mechanical ventilation, and remain in hospital for 5 days longer for a total of 12 post-operative days than those diagnosed pre-operatively. Interestingly, there was no difference observed in ICU LOS or post-operative morbidity outcome. In-hospital mortality was 0% in those with pre-operatively diagnosed Covid-19, but 37.1% in those with a post-operative diagnosis (p < 0.0005). All other variables from Table 1 were explored and were statistically non-significant.

This study has four main limitations. Firstly, not all UK centres participated in this study (reasons include lack of capacity to prepare and submit data) so our results may not be representative of the whole of the UK. Nevertheless we do have wide geographical representation covering England, Wales and Northern Ireland, including three large centres (Liverpool Heart and Chest Hospital, James Cook University Hospital and St Bartholomew’s Hospital). One of these (St Bartholomew’s Hospital) became the command centre for all cardiac surgery across London and parts of the South East of England [20], providing a similar ‘hub and spoke’ system referred to previously. Secondly, as previously eluded to, there were variable swabbing protocols across centres and Covid-19 diagnosis was initially reliant on a single negative swab pre-surgery. Due to the high false-negative rate this could account for the high incidence of post-operative Covid-19 diagnosis, particularly if surgery does indeed accelerate Covid-19 progression as suggested by Lei and colleagues [7]. Equally, to reflect clinical practice at the time we defined patients as covid-19 positive if no test were performed but was suspected clinically. Although only very few patients were included this way (one patient pre-operative diagnosis and five with a post-operative diagnosis), there is a potential that these could reflect other infectious complications. Nevertheless, as they have been recorded as covid-19 positive, they would have been clinically managed as such, and so we have also analysed them this way. Thirdly, our study reflects the initial snapshot of the outcomes of cardiac surgery during the Covid-19 pandemic, at a time when services, and the country, were adapting to the impact of Covid-19 in the UK and when personal protective equipment and staff and public Covid-19 tests were limited. Finally, as with any retrospective database analysis we are limited by the data collected within those existing datasets. Data pertaining to cause of death and also the timing of post-operative Covid-19 diagnosis, for example, would have been beneficial, but were not available at the time of analysis. To mitigate against some of these limitations, further work being undertaken focuses on the subsequent cardiac surgery response post April 2020, the longer-term implications and outcomes of having cardiac surgery during this time and whether the impact of a post-operative diagnosis remains following the revision of diagnosis processes and improvements in swab detection.