Article Text
PDF
Abstract
Objectives Direct comparisons between COVID-19 and influenza A in the critical care setting are limited. The objective of this study was to compare their outcomes and identify risk factors for hospital mortality.
Design and setting This was a territory-wide, retrospective study on all adult (≥18 years old) patients admitted to public hospital intensive care units in Hong Kong. We compared COVID-19 patients admitted between 27 January 2020 and 26 January 2021 with a propensity-matched historical cohort of influenza A patients admitted between 27 January 2015 and 26 January 2020. We reported outcomes of hospital mortality and time to death or discharge. Multivariate analysis using Poisson regression and relative risk (RR) was used to identify risk factors for hospital mortality.
Results After propensity matching, 373 COVID-19 and 373 influenza A patients were evenly matched for baseline characteristics. COVID-19 patients had higher unadjusted hospital mortality than influenza A patients (17.5% vs 7.5%, p<0.001). The Acute Physiology and Chronic Health Evaluation IV (APACHE IV) adjusted standardised mortality ratio was also higher for COVID-19 than influenza A patients ((0.79 (95% CI 0.61 to 1.00) vs 0.42 (95% CI 0.28 to 0.60)), p<0.001). Adjusting for age, PaO2/FiO2, Charlson Comorbidity Index and APACHE IV, COVID-19 (adjusted RR 2.26 (95% CI 1.52 to 3.36)) and early bacterial-viral coinfection (adjusted RR 1.66 (95% CI 1.17 to 2.37)) were directly associated with hospital mortality.
Conclusions Critically ill patients with COVID-19 had substantially higher hospital mortality when compared with propensity-matched patients with influenza A.
- COVID-19
- Adult intensive & critical care
- Respiratory infections
Data availability statement
Data are available upon reasonable request. The datasets used and/or analysed in this study are available from the corresponding author on reasonable request.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
STRENGTHS AND LIMITATIONS OF THIS STUDY
Retrospective analysis with historical control.
None of COVID-19 patients were fully vaccinated.
Territory-wide study of all publicly funded adult general intensive care units.
Used propensity scored matching.
Objectives
The SARS-CoV-2 virus is the cause of COVID-19.1 The pandemic has led to high demand for intensive care unit (ICU) services. Between 5% and 32% of hospitalised COVID-19 patients needed ICU care.1–7 In addition, 41.9%–88% of ICU cases required invasive mechanical ventilation (MV).8–20 In a systematic review, the estimated duration on MV and ICU length of stay (LOS) was 8.4 (95% CI 1.6 to 13.7) and 9.0 (95% CI 6.5 to 11.2) days, respectively.21 The described hospital mortality ranges from 11.5% to 61.5% with risk factors including: advance age, male sex, body mass index, diabetes mellitus, hypertension, coronary artery disease, malignancy, immunocompromised state, organ failure and raised D-dimer.8–18 20–23 Differences in healthcare systems, capacity, admission triage and COVID-19 burden may account for the marked variation in reported ICU admission and mortality rates.
Influenza A was responsible for annual epidemics and occasional pandemics in the pre-COVID-19 era. The reported mortality of ICU patients infected with influenza A ranges from 20.7% to 37.1%.24–28 Comparison between patients hospitalised for influenza and COVID-19 patients showed that the latter are more likely to require ICU admission and MV, and have higher hospital mortality.29–34 During the COVID-19 pandemic, healthcare systems were stressed and overwhelmed. As a result, the observed case fatality of COVID-19 patients would be higher than expected if healthcare systems managed well within capacity. Furthermore, there have been few direct comparisons between SARS-CoV-2 and influenza A in the critical care setting.20 23 35–40
During the first year of the COVID-19 pandemic, the incidence of COVID-19 was relatively low in Hong Kong.41 In the context of relatively low burden of COVID-19, we performed a territory-wide, retrospective, propensity-matched, study to compare the outcomes and resource usage of critically ill patients suffering from SARS-CoV-2 infection with those suffering from influenza A. The objectives were to compare the hospital mortality and time to death or discharge between ICU patients with COVID-19 and influenza A; as well as to determine the direct associations between type of respiratory infection and early bacterial-viral coinfection with hospital mortality.
Methods
Study design and cohort selection
This was a territory-wide, retrospective study on all adult (≥18 years old) patients admitted to public hospital ICUs in Hong Kong with either SARS-CoV-2 or influenza A infection. Patients were identified using the Hospital Authority’s electronic health database called the Clinical Data Analysis and Reporting System (CDARS).42 Patients were included if they tested positive for SARS-CoV-2 or influenza A on PCR during ICU stay. The inclusion period for COVID-19 patients was between 27 January 2020 and 26 January 2021, corresponding to the first critical case of COVID-19 in Hong Kong. The inclusion period for influenza A patients was between 27 January 2015 and 26 January 2020. The inclusion periods did not overlap because there were no influenza A patients admitted to our ICUs after the start of the COVID-19 pandemic. Only the first ICU admission of the hospital episode (defined as admission to hospital until discharge home or death) was included for analysis.
Hong Kong public general ICUs
There are 15 publicly funded adult general ICUs in Hong Kong with over 14 000 admissions and overall Acute Physiology and Chronic Health Evaluation IV (APACHE IV) adjusted standardised mortality ratio (SMR) of 0.65 in 2019.42 In Hong Kong public hospitals, all ICU admission data is prospectively collected by trained nurses for the Intensive Care Unit Outcomes Monitoring and Improvement Programme. Decisions on ICU admissions during the COVID-19 pandemic followed the Hong Kong COVID-19 adult ICU triage tool.43
Data collection and parameter definitions
Demographics data such as age, gender, Charlson Comorbidity Index (CCI) were collected from CDARS. CCI was calculated by diagnostic codes from medical history.44 APACHE IV score, PaO2/FiO2 on admission to ICU, use of MV, vasopressors and renal replacement therapy (RRT) during ICU stay were also retrieved from CDARS. The APACHE IV score is a severity of illness scoring system for predicting hospital mortality of ICU patients. It uses physiological parameters such as temperature, heart rate, mean arterial blood pressure, urea, albumin and other physiological parameters, chronic health conditions and admission information to assess severity of illness.45 It has been previously validated for prediction of hospital mortality in a large cohort of ICU patients in Hong Kong.42
Any positive respiratory tract (sputum, tracheal aspirate, bronchial alveolar lavage), pleural or blood bacterial culture results obtained within 48 hours of ICU admission were used to define early bacterial-viral coinfection with SARS-CoV-2 or influenza A. The threshold was set at 48 hours from ICU admission as this has been used to define early bacterial-viral coinfection in critically ill patients with COVID-19 and influenza A.46 Furthermore, time from ICU admission rather than hospital admission was used to define window of early bacterial-viral coinfection for three reasons. First, the median time to development of critical illness from onset of viral infection is different between influenza A (5 days) and COVID-19 (10 days).47 48 Second, all patients in Hong Kong who had confirmed COVID-19 were hospitalised for airborne isolation regardless of severity during the first year of the pandemic. Taken together, duration of hospitalisation in COVID-19 may be longer than influenza A based on natural history of the two different viral infections and public health policy. Third, the purpose of this study was to evaluate the association between early bacterial-viral coinfection in critically ill patients with COVID-19 and influenza A at time of ICU admission rather than the whole hospitalisation.
Clinical endpoints of time of death and discharge, ICU and hospital mortality were collected for outcome analysis. ICU LOS, hospital LOS and organ support during ICU stay (MV, RRT, vasopressors and extra-corporeal membrane oxygenation (ECMO) requirement) were used to evaluate resource usage.
Patient and public involvement
Patients or the public were not involved in the design, or conduct, or reporting, or dissemination plans of our research.
Statistics
Continuous variables were expressed as median IQR, and categorical variables were described as count and percentage for summary statistics. Propensity matching was performed to match the COVID-19 patients to influenza A patients using the following baseline variables at time of ICU admission: age, gender, types of comorbidities based on Charlson Comorbidty classification, CCI score, APACHE IV score, PaO2/FiO2 and early bacterial-viral coinfection.44 These baseline variables were chosen based on known associations with mortality to reduce potential bias between the COVID-19 and influenza A groups.8–18 22 49–51 The propensity score was calculated using multivariable logistic regression, and matching was performed by the matchit package in R, using different matching ratios by the nearest-neighbour approach within a calliper distance of 0.2. After matching, standardised mean differences (SMD) of the baseline variables between the COVID-19 and influenza A groups were calculated, and an SMD<0.1 was considered as satisfactory balance.52 The best matching ratio that resulted in the largest sample size with the most matched variables with SMD<0.1 was used.
Age, male gender, PaO2/FiO2, CCI, APACHE IV, early bacterial-viral coinfection are known factors associated with mortality in viral pneumonia.8–18 22 49–51 All of these variables were included in direct acyclic graphs (DAG) to identify the minimum set of adjustment factors required to assess for the direct associations between type of respiratory viral infection and early bacterial-viral coinfection on hospital mortality (online supplemental figures 1 and 2). DAGs showed that adjustment for age, PaO2/FiO2, CCI, APACHE IV with virus type or bacterial-viral coinfection (without gender) were sufficient to assess the direct effect of virus type or early bacterial-viral coinfection on hospital mortality. Univariate analysis was performed to assess the association between individual factors with hospital mortality. Multivariate analysis of all variables using Poisson log-linear regression and relative risk (RR) was used to assess for direct associations between respiratory virus infection (COVID-19 or influenza) and bacterial-viral coinfection on hospital mortality. Subsequently, direct association of early bacterial-viral coinfection with hospital mortality was also separately assessed for COVID-19 and influenza A (using all 1371 unmatched patients) groups. The time-to-event profiles of hospital mortality and discharge were analysed and visualised by using Kaplan-Meier estimator with two terminal outcomes.53 SMR was calculated by dividing the actual number of deaths by the APACHE IV predicted number of deaths in each cohort. Differences in SMR, ICU and hospital mortality between COVID-19 and influenza A patients were tested by the Fisher’s exact test. Difference in proportion of patients who required MV, vasopressor, RRT and ECMO were also assessed using Fisher’s exact test. Difference in LOS was assessed using Wilcoxon ranked-sum test. Annual APACHE IV adjusted SMRs of influenza A patients (unmatched) were calculated for each calendar year between 2015 and 2020, and linear regression was used to assess its trend over time. Similarly, change in monthly APACHE IV adjusted SMR for COVID-19 group was assessed by linear regression. Statistical analyses were performed in SPSS (V.21.0, IBM) and R (V.4.2.2, R Foundation for Statistical Computing).
Supplemental material
Supplemental material
Results
Characteristics of study cohort
Between 27 January 2020 and 26 January 2021, 373 COVID-19 patients were admitted to publicly funded ICUs in Hong Kong and 1371 critically ill patients were admitted for influenza A between 27 January 2015 and 26 January 2020 (online supplemental figure 3 and online supplemental table 1). Using 1:1 matching ratio, the baseline characteristics of 373 COVID-19 and 373 influenza A patients were evenly matched by known confounding variables at ICU admission (table 1 and online supplemental table 1). Matching ratio of 1:2 resulted in less satisfactory matching as target SMD<0.1 was not achieved for all confounding variables (online supplemental table 2). The SMR of all influenza A patients (n=1371) progressively reduced from 1.42 (95% CI 1.19 to 1.67) to 0.72 (95% CI 0.40 to 1.20) by −0.11 /year (95% CI −0.17 to −0.06, p=0.004) (online supplemental table 3). Up to 61/373 (16.3%) COVID-19 and 123/1371 (9.0%) influenza A patients did not have bacterial culture tests performed within 48 hours of ICU admission. The median (IQR) hospitalisation time prior to ICU admission was 2.8 (0.7–6.0) days in COVID-19 group and 0.3 (0.0–1.2) days in the matched influenza A cohort. The most common bacterial pathogens isolated in the COVID-19 group included Coagulase negative Staphylococci species (9/70), Klebsiella pneumonia (7/70) and Staphylococcus aureus (4/70). In the matched influenza A cohort, the most common respiratory bacterial pathogens isolated were Pseudomonas aeruginosa (12/81), S. aureus (7/81) and K. pneumonia (6/81).
Supplemental material
Supplemental material
Supplemental material
Supplemental material
Baseline characteristics of COVID-19 and propensity matched influenza A cohort
Clinical outcomes
The need for MV, vasopressors, RRT and ECMO during ICU stay are shown in table 2. More patients in the COVID-19 group (130/373, 34.9%) required at least two organ support during their ICU stay compared with the influenza group (103/373, 27.6%) (p=0.033). Unadjusted ICU mortality was higher and LOS longer in patients with COVID-19 compared with matched influenza A patients (table 2). COVID-19 patients also had higher unadjusted hospital mortality than matched patients with influenza A (17.5% vs 7.5%, p<0.001). The APACHE IV adjusted SMR for COVID-19 and influenza patients were 0.79 (95% CI 0.61 to 1.00) versus 0.42 (95% CI 0.28 to 0.60) (p<0.001). The monthly APACHE IV adjusted SMR for COVID-19 patients did not change overtime during the study period (+0.01 /month (95% CI −0.08 to 0.11), p=0.756).
Organ support and clinical outcomes
Kaplan-Meier hospital survival and discharge curves of COVID-19 and influenza A patients are illustrated in figure 1. Median time to death was shorter for influenza A patients at 9.5 (IQR 5.8–18.8) days than COVID-19 patients at 25.5 (IQR 12–37) days (p=0.002). Similarly, median time to discharge in hospital survivors was shorter for influenza A patients at 10 (IQR 6–17) days compared with patients with COVID-19 at 25 (IQR 18–39) days (p<0.001). Longer time to discharge of patients with COVID-19 compared with influenza A persisted when stratified by age (online supplemental table 4). In the subgroup of patients aged >65 years old, median time to death was longer in COVID-19 than influenza A patients ((27.0 (IQR 12.8–39.0) vs 8.0 (IQR 5.2–16.5)) days, p=0.003) (figure 2A and online supplemental table 4). In contrast, median time to death was similar for patients aged ≤65 years old ((12.0 (IQR 6.8–19.2) vs 15.0 (IQR 5.8–21.2)) days, p=0.949) (figure 2B and online supplemental table 4).
Supplemental material
Kaplan-Meier plot of hospital death and discharge in COVID-19 and influenza A intensive care unit patients. Time to hospital death or discharge from hospital admission for critically ill patients with COVID-19 (n=373) and influenza A (n=373).
Kaplan-Meier plot of hospital death and discharge based on 65 years old age threshold. (A) Time to hospital death or discharge from hospital admission for critically ill patients aged ≤65 years old with COVID-19 (n=169) and influenza A (n=225). (B) Time to hospital death or discharge from hospital admission for critically ill patients aged >65 years old with COVID-19 (n=204) and influenza A (n=148).
In patients with SARS-CoV-2 infection, unadjusted hospital mortality rate of patients with early bacterial-viral coinfection was 21/71 (29.6%) compared with 43/302 (14.2%) in those with SARS-CoV-2 infection alone (p=0.002). In unmatched influenza patients (n=1371), unadjusted hospital rate for early bacterial-viral coinfection was 405/921 (44.0%) compared with 107/343 (23.7%) in those with influenza A infection alone (p<0.001). In matched patients with influenza A, unadjusted hospital mortality rate was 10/80 (12.5%) in those with early bacterial-viral coinfection compared with 18/293 (6.1%) in those with influenza A infection alone (p=0.056).
Direct effect of type of virus infection and early bacterial-viral coinfection on hospital mortality
Age, PaO2/FiO2 ratio, CCI, APACHE IV, COVID-19 and early bacterial-viral coinfection were associated with hospital mortality on univariate analysis (table 3). Adjusting for significant factors such as age, PaO2/FiO2, APACHE IV based on DAG, COVID-19 and early bacterial-viral coinfection both had direct associations with hospital mortality (table 3). In subgroup analysis, early bacterial-viral coinfection was independently associated with hospital mortality in COVID-19 (adjusted RR 2.10 (95% CI 1.39 to 3.18), p<0.001) and unmatched influenza A (adjusted RR 1.44 (95% CI 1.22 to 1.68), p<0.001) patients (online supplemental tables 5 and 6).
Supplemental material
Supplemental material
Univariate and multivariate analysis of factors associated with hospital mortality
Discussion
In this territory-wide, retrospective study, critically ill COVID-19 patients had a higher hospital mortality compared with a propensity-matched historical cohort of ICU patients with influenza A infection. The median time to death and hospital discharge were both longer for patients with COVID-19. After adjustment of age, PaO2/FiO2, CCI and APACHE IV, COVID-19 and early bacterial-viral coinfection were both directly associated with hospital mortality.
The 17.5% unadjusted hospital mortality rate of COVID-19 patients is comparable to previous reports.9–17 19–21 23 54 Compared with other Asia-Pacific regions with similar healthcare settings and COVID-19 burden, critically ill patients in South Korea had a higher in-hospital mortality of 42.7% while those in Japan and Australia had a lower mortality of 15.5% and 11.5%, respectively.13 19 23 However, comparison of unadjusted mortality rates across different healthcare systems and cohorts is challenging due to differences in baseline characteristics, many of which are associated with COVID-19 mortality. For example, compared with Hong Kong, the South Korean cohort had older patients with more underlying diabetes and need for organ support.
The APACHE IV adjusted SMR of 0.79 for COVID-19 patients suggests that survival was worse than that of the average ICU patient in Hong Kong (SMR of 0.65 in 2019). A few factors should be considered when interpreting our results. First, the alpha and delta SARS-CoV-2 variants were the dominant strains during the study period. Second, Hong Kong started its COVID-19 vaccination programme on 26 February 2021, and thus none of the COVID-19 patients in this study were fully vaccinated. Third, there were a total of 10 223 confirmed cases in Hong Kong with only 375 (3.6%) requiring adult ICU care.41 This suggests our healthcare system was challenged but not overwhelmed during the first year of the COVID-19 pandemic. This is important because strain on ICU capacity is associated with increased mortality of ICU patients with COVID-19.55 Furthermore, we have shown that, even among survivors and non-survivors, both ICU and hospital LOS in COVID-19 patients are more than double than those with influenza A. This is despite a higher proportion of influenza A patients who required MV. Overall requirement for two or more organ support (MV, vasopressor, RRT and ECMO) were also higher in the COVID-19 group. This affects future critical care and hospital resource planning if SARS-CoV-2 burden changes to a seasonal pattern.
There is conflicting evidence on difference in mortality between critically ill patients with COVID-19 and influenza A.20 23 31 35–38 40 Two studies found a higher hospital mortality among COVID-19 than influenza patients.20 31 In contrast, three studies found similar unadjusted mortality for the two viruses, although one of them found a higher severity-adjusted mortality in H1N1 ICU patients.23 35 40 Furthermore, most studies comparing outcomes of COVID-19 or influenza patients who required ECMO for acute respiratory distress syndrome found no difference in mortality between the two viruses.36–38 However, a common limitation of published evidence is the presence of unequal baseline characteristics among patients with the two different viruses, making direct comparison difficult. For example, if outcomes were compared in this study between COVID-19 and unmatched influenza A patients in 2020, the conclusion would be that they had similar APACHE IV adjusted SMR of 0.79 and 0.72, respectively. In contrast, we found a substantially higher unadjusted ICU and hospital mortality among patients with COVID-19 when compared with propensity matched influenza A patients. The APACHE IV adjusted SMR of COVID-19 patients (0.79) was also substantially higher than that of those infected with influenza A (0.42). Finally, after adjustment for age, severity of illness, PaO2/FiO2 and early bacterial-viral coinfection, multivariate analysis confirmed that COVID-19 patients had an adjusted RR of 2.3 for hospital mortality compared with those with influenza A. Interestingly, the absolute gap between unadjusted ICU and hospital mortality was similar in both virus groups, suggesting that efforts to improve ICU mortality should translate to lower hospital mortality in COVID-19. Our data provides further impetus to promote COVID-19 vaccination for unvaccinated individuals as survival from critical COVID-19 is much lower than that of seasonal influenza.
Use of antibiotics is not recommended in patients with mild to moderate COVID-19.56 However, we showed that early bacterial-viral coinfection in critically ill patients was directly associated with hospital mortality in both COVID-19 (adjusted RR 2.07) and influenza A (adjusted RR 1.43) even after adjustment for severity of illness and comorbidities. This is in line with previous studies which demonstrated increased mortality in influenza A patients with bacterial-viral coinfection.50 57 In our cohort, 18.8% of COVID-19 cases had early bacterial-viral coinfection, which is consistent with studies which reported rates of 5%–28% bacterial COVID-19 coinfection within 48 hours of ICU admission.46 58–60 Similar to other reports, we found that S. aureus and K. pneumonia were the most common bacterial copathogens in COVID-19.61 This highlights the importance of routine bacterial cultures to identify bacterial-viral coinfection in any patient with viral pneumonia. Empirical antibiotics in critically ill patients with COVID-19 at ICU admission is generally recommended with a plan to review antibiotic discontinuation after a few days based on culture results and the interpretation of procalcitonin concentrations when available.62
This study had several limitations. First, this was a retrospective analysis of prospectively collected data from a cohort of COVID-19 patients that was compared with a historical cohort of influenza A patients in the ICU. To minimise selection bias, all critically ill patients with COVID-19 in the territory during the study period were included. In addition, patients from all 15 general ICUs in the territory were represented in both viral groups, reducing organisational and healthcare factors that may affect outcomes. Second, propensity matching is only possible for known confounding factors, and even when known, matching may not fully eliminate the potential confounding effects of known factors between the matched cohorts. It should be specifically noted that we were unable to capture data on body mass index which may have led to such an imbalance. The mortality of patients with influenza in our cohort may have been lower than that of COVID-19 because of the protective effect of seasonal influenza vaccinations.63 Unfortunately, we were unable to verify the influenza vaccination status in this study. However, since the 2020 uptake rate of seasonal influenza vaccination in this study’s median age group was only 11%, it is less likely to be a significant confounder.64 Third, although availability of evidence-based treatment may have affected case mortality rate over time, we found that APACHE IV adjusted SMR of COVID-19 patients remained unchanged over time during the first year of the pandemic. Fourth, general hospitals beds were used as isolation facilities during this COVID-19 pandemic surge in Hong Kong. This may have prolonged the hospital LOS among COVID-19 survivors compared with influenza A due to need for isolation rather than clinical severity. However, the effect of this isolation policy is likely low since ICU and hospital LOS for survivors and non-survivors were consistently longer in COVID-19 patients. Fifth, up to 16.3% and 9.0% of COVID-19 and influenza A patients included in the cohort did not have bacterial culture test 2 days before or after ICU admission, which may have underestimated the proportion of early bacterial-viral coinfections with either virus or affected propensity matching. Matching for time interval from infection onset to bacterial-viral coinfection was not performed as initial time of infection for each patient was not retrievable. In addition, hospitalisation time prior to ICU admission was not matched between the two virus groups as public health policy required hospitalisation of COVID-19 patients even if they were asymptomatic. Thus, it is possible that some patients had superinfections instead of early bacterial-viral coinfection. However, this is less likely since the median hospitalisation time prior to ICU admission was less than 3 days in both virus groups, and all positive cultures 2 days prior to 2 days after ICU admission were counted as bacterial-viral coinfection. Sixth, there was a significant portion of COVID-19 patients with Coagulase negative Staphylococcus species which has been reported in other studies.65 66 Although we could not systematically classify whether each case was a contaminant by manual case review, their inclusion would have underestimated the direct association between bacterial-viral coinfection on mortality in COVID-19. Seventh, we were unable to include lymphocyte count in the model as a known prognostic factor as this data was not consistently available. Eighth, as 1:1 matching ratio was used instead of higher ratios to optimise matching, this reduced the overall sample size which may have limited the power to detect smaller differences in outcomes between SARS-CoV-2 and influenza. Ninth, COVID-19 vaccination and predominance of the omicron variant has significantly reduced the global case fatality rate of COVID-19.67 Therefore, the conclusions drawn between influenza and COVID-19 are only valid for the first year of the COVID-19 pandemic.
Conclusion
Critically ill patients with COVID-19 had substantially higher hospital mortality and longer time to death or discharge when compared with propensity-matched patients with influenza A. After adjustment of age, PaO2/FiO2, CCI and APACHE IV, COVID-19 and early bacterial-viral coinfection were both directly associated with hospital mortality.
Data availability statement
Data are available upon reasonable request. The datasets used and/or analysed in this study are available from the corresponding author on reasonable request.
Ethics statements
Patient consent for publication
Ethics approval
This study involves human participants and was approved by The Joint Chinese University of Hong Kong-New Territories East Cluster Clinical Research Ethics Committee (2021.026). The Hong Kong East Cluster Research Ethics Committee (HKECREC-2021-038); The Institutional Review Board of The University of Hong Kong/Hospital Authority Hong Kong West Cluster (UW21-321); The Research Ethics Committee of Kowloon Central/Kowloon East (KC/KE-21-0112/ER-1); The Kowloon West Cluster Research Ethics Committee (KW/EX-21-104(161-04)); New Territories West Cluster Research Ethics Committee (NTWC/REC/21031). No consent was required because this was a retrospective study.
Acknowledgments
We thank Dr Chun Ming Ho for his leadership in ICUOMP and collection of data for this study.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
Twitter @lingling7
RBHC and SZ contributed equally.
Contributors LL designed the study. RBHC, JZZ and LL recruited the different ICUs and collected the data. SZ, JZZ, MHW and LL performed the data analysis. LL and RBHC wrote the first draft of the manuscript together. KCKC, PYN, CC, KMF, SYA, AWTY, JKHC, HHT, KIP, FLC, KNL, KMC, MD, WTW and GMJ reviewed the data and helped revise the manuscript. The final manuscript was read and approved by all authors. LL is responsible for the overall content as the guarantor.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests None declared.
Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.