The coronavirus disease 2019 (COVID-19) pandemic, caused by severe acute respiratory syndrome coronavirus (SARS-CoV-2), has led to high rates of morbidity and mortality worldwide [1]. At the start of the pandemic in Wuhan, China in 2020, hospitals were overwhelmed by the excessive number of critical patients requiring intensive care, which was greater than the available capacity of these hospitals. Therefore, a significant number of patients stayed in emergency areas or were transferred to general wards until intensive care unit (ICU) beds became available [2]. Some of these areas were serviced by registered nurses who were not trained to treat ICU patients. Isolation of patients required nega-tive pressure rooms in wards that were specifically created to deal with the pandemic. Many hospitals established negative pressure rooms and ICU beds outside of designated ICU areas. As a result, many critical patients were transferred between wards when a negative pressure room or ICU bed became available. Whenever these mechanically ventilated patients are transferred, they require a change from ventilator to mobile ventilator, which may result in a shifting in the endotracheal tube position or loss of positive end expiratory pressure (PEEP). These transfers may require a temporary hold on dynamic monitoring equipment (clamping arterial line) or infusion pumps of medications (sedatives). Schwe-bel et al. [3] documented that patients with intrahospital transport are 1.9 times more likely to develop a complication (atelectasis, ventilator-associated pneumonia, hypoglycaemia, or hyperglycaemia). Braman et al. [4] also documented increased complications from intrahospital transport in critically ill patients and concluded that these complications can be prevented. The impact of these transfers is largely unknown for COVID-19 patients with acute respiratory distress syndrome (ARDS); as such, the number of patient transfers was unprecedented. We aimed to record the occurrence of transfers and evaluate the impact of these transfers on the clinical outcome of mortality and length of stay in the ICU.
We evaluated the effect of the care of critically ill COVID-19 patients in undesignated ICU wards, with a particular focus on observing the occurrence of transfer of patients between wards before reaching designated ICU areas, on the clinical outcomes of mortality and length of stay in the ICU (LOSICU).
METHODS
We retrospectively collected the data from electronic medical records of all critically ill patients with COVID-19 pneumonia meeting ICU admission criteria, who were admitted to Dubai Hospital between 1 January 2020, and 30 June 2020. All patients who transferred to wards other than the medical intensive care unit (MICU) or surgical intensive care unit (SICU) constituted the cases, because they were cared for in undesignated ICU areas. All other patients who were transferred to the MICU or SICU directly from the emergency department (without going to any other ward) constituted controls. The decision about where the patient was admitted was based upon the availability of an ICU bed at the time of admission, because most patients had a waiting time in the emergency department before admission to the ICU bed. Two investigators recorded data on the number of transfers from transfer notes written by intensivists or anaesthesia registrars, because the hospital protocol required each ventilated patient to be escorted by teams supervised only by these assigned doctors. The demographics recorded were as follows: age; sex; body mass index (BMI); nationality; clinical parameters recorded; positive swab sample polymerase chain reaction (PCR) test result for SARS-CoV-2; number of swabs; number of days to negative PCR test result; number of days of symptoms; and presence of symptoms such as cough, fever, dyspnoea, gastric complaints on admission to ICU, and first set of vitals. Data on comorbidities included diabetes, hypertension, coronary artery disease, renal failure, and outpatient dialysis. Inpatient clinical data on admission to the ICU, including fever (temperature > 38.0°C), tachycardia (pulse > 100 per minute), hypotension (systolic BP less than 90 mmHg), hypoxia (SpO2 < 92%), use of oxygen (L min–1), mechanical ventilation, use of pressors, and inpatient dialysis. Laboratory parameters on the day of admission to the ICU included disease activity markers or inflammatory markers (C-reactive protein [CRP], ferritin, and procalcitonin levels [PCT]), haematological indices (white blood and platelet counts), chemistries (electrolyte levels), and culture results of sputum, blood, pleural or peritoneal fluid, or pneumonia panels because secondary bacterial infection impacts clinical outcomes. Data on therapeutic agents, including chloroquine and antivirals, were also recorded because they can be significant confounding factors. We calculated APACHE-2 scores within 24 hours of admission to the ICU, to assess the severity of illness. Because many patients later in the course worsened and required transfer to the designated ICU for extracorporeal membrane oxygenation (ECMO) provision, the initial APACHE-2 score may not be truly reflective of severity, which means that the designated ICU probably had a high severity of illness. Medical management was provided by the same team of physicians for both groups; therefore, the same treatment was provided in terms of medication (steroids, tocilizumab, anticoagulation) or timing for intubation or continuous renal replacement therapy (CRRT). Some undesignated ICU rooms lack the materials necessary for CRRT, so they had to be transferred to dialysis sections as needed, which resulted in a greater number of transfers. Similarly, an undesignated ICU room was not suitable for ECMO cannulation, so patients were transferred to the designated ICU if ECMO was prescribed. Therefore, more sick patients were probably transferred to the designated ICU at some point if a bed was available at that time. Similarly, bedside tracheostomies were performed in the operating theatre for patients from undesignated ICUs, while designated ICU patients may have bedside percutaneous tracheostomy in the ICU. As the number of licensed ICU nurses (LICUN) was insufficient to service the presented load of patients, all ICU ward nurses were supervised by LICUN. Relatively more LICUNs were assigned to undesignated ICUs to offset the other factors (non-critical care nurses were less familiar with critical care procedures such as ventilator management, haemodialysis, and central line placements). The undesignated ICUs had some rooms that were smaller in size, and some rooms were cubicles that had 2 patients, while all designated ICU rooms were single beds. The designated ICU rooms had fixed ventilators while the undesignated ICU rooms had similar but moveable ventilators.
Because the rooms in undesignated ICUs did not have the capability for in-room dialysis or interventional procedures (i.e. haemodialysis, bronchoscopy, and bedside tracheostomy), patients in undesignated ICUs had more transfers.
LOSICU includes the total number of days in ICU regardless of designated or undesignated ICU or a combination of the 2. For example, if a patient stayed 5 days in an undesignated ICU and was then transferred to the designated ICU and stayed there for 10 days, the total LOSICU for this patient was 15 days.
The study was approved by the Dubai Scientific Research Ethics Committee (DSREC), Dubai Health authority on 10 June 2021 (approval number DSREC-05/2021_18). Written informed consent was waived because of the retrospective nature of the analysis.
Statistical analysis
Sample characteristics were compared between the group of patients who went to non-ICU wards before reaching the MICU or SICU for administrative reasons (lack of available beds in the MICU or SICU) and the group that was admitted directly to the designated ICU (MICU or SICU). Chi-square tests were performed for categorical variables, and the Mann-Whitney U test was performed for continuous variables because the data were found to be non-normally distributed.
Initially, univariate logistic regression analysis determined that the variables were significant predictors of mortality. Multiple logistic regression analysis was performed by including only those variables found to be significant predictors in univariate analysis. Similarly, univariate linear regression was performed to determine significant variables as predictors of LOSICU admission. Multiple linear regression was used by including only those variables found to be significant in univariate regression analysis to assess predictors of LOSICU. A P-value of 0.05 was considered significant. All analyses were performed with SPSS version 27 (IBM Corp., Armonk, N.Y., USA).
RESULTS
The characteristics of the total sample (N = 239) and the 2 groups (undesignated ICU ward/extra transfers [cases], n = 107; and direct transfer to designated ICU bed [controls], n = 132) are shown in Table 1 for categorical variables and in Table 2 for continuous variables. The sample included 208 (87%) males and 31 (12.2%) females. The mean age of the sample was 49 years. Mechanical ventilation was required in 203 (85.2%) patients (107 in designated vs. 98 in undesignated, P = 0.08). Vasopressors were required in 188 (79%) subjects (100 in designated vs. 88 in undesignated ICUs, P = 0.266). One-third of patients (30.3%) developed renal injury requiring CRRT (38 in designated vs. 34 in undesignated, P = 0.644). The prevalence of documented secondary bacterial infection was 106 (44.3%), which was similar between the 2 groups (59 in designated vs. 47 in undesignated, P = 0.905). Steroids were prescribed in 188 (79.3%) patients (99 in designated and 89 in undesignated ICU patients, P = 0.18). All patients who received ECMO were in designated ICU beds because it cannot be provided at an undesignated ICU bed.
TABLE 1
Sample characteristics (categorical variables)
TABLE 2
Sample characteristics – continuous variables
Details of how many extra transfers were made are reported in Table 3. The crude mortality was better for patients in the undesignated ICUs: 45 (37.8%) vs. designated ICUs 74 (62.2%), P = 0.02. Patients with care at an undesignated ICU and with extra transfers spent more days on mechanical ventilation [median (IQR) 18 (19) vs. 11 (14), P = 0.001], LOSICU [21.5 (19) vs. 15 (14), P = 0.001], more days alive in the first 30 days after admission [24.5 (13) vs. 11 (17), P = 0.001], LOSH [32 (28) vs. 21 (26), P = 0.001], and had a lower prevalence of secondary bacterial infection [47 (43.9%) vs. 59 (44.6%), P = 0.905]. They also had a higher occurrence of being sedated [102 (95.2%) vs. 109 (83.2%), P = 0.003] and being paralyzed [98 (91.6%0 vs. 104 (79.4%), P = 0.009] than patients in the designated ICU. Univariate logistic regression showed the interaction of each clinically significant variable with mortality (Table 4A). After inclusion of only significant variables in univariate analysis, a model for multiple logistic regression was developed, which showed that patients with care at an undesignated ICU bed had better survival, with an odds of death of 0.347 (CI of 0.178–0.676), P = 0.002. The use of steroids also predicted better survival, with an odds of death of 0.166 (CI of 0.058–0.474), P = 0.001. Patients who had tracheostomy also had better survival – odds of death 0.08 (CI of 0.026–0.247), P ≤ 0.001. Those receiving CRRT had worse survival, with an odds of death of 4.5 (CI of 2.047–10.236), P ≤ 0.001 (Table 4B).
TABLE 3
Numbers of administrative transfers
TABLE 4
A. Univariate logistic regression analysis of clinically relevant predictors of mortality
TABLE 4
B. Multiple logistic regression analysis (predictors of mortality)
For the outcome of length of stay, we conducted 2 analyses: LOSICU and days the patient was alive in the first 30 days after admission to the ICU. First, univariate simple regression was performed to assess the relationship of variables to LOSICU (Table 5A). Only those variables that were found to be significant predictors of LOSICU were included in the final model of multiple linear regression, which showed that care at an undesignated ICU was associated with a longer LOSICU of 4.206 days (CI of 1.290–7.122), P = 0.005 (Table 5B). Other factors predictive of longer stay were occurrence of secondary bacterial infection, tracheostomy, and provision of ECMO. Although these factors were more prevalent in designated ICUs, they did not have any significant collinearity among them (VIF < 4) (Table 5B). A similar analysis for days alive in the first 30 days after admission was performed, with similar results (Table 5C).
TABLE 5
A. Univariate linear regression for factors determining length of stay in ICU
TABLE 5
B. Multiple linear regression of variables determining length of stay in ICU
TABLE 5
C. Multiple linear regressions of variables determining number of days alive in hospital
DISCUSSION
The COVID-19 pandemic exhausted health care systems worldwide. Hence, an obvious concern was whether care provided under these unprecedented conditions affected the quality of care and clinical outcomes.
We addressed this issue and found that care in an undesignated ICU was associated with better survival and longer stays in the ICU. This was not expected. We believe this may have resulted from any combination of the following factors: sicker patients who required complex therapies such as ECMO and CRRT were preferably treated at designated ICUs because the provision of these therapies was only possible in designated ICUs. Data on ARDS from COVID-19 treated with ECMO therapy showed higher mortality compared to non-COVID-19 ARDS treated with ECMO therapy [5]. Similarly, patients with COVID-19 ARDS who developed acute kidney injury had higher mortality [6]. The primary indication for choosing CRRT over intermittent haemodialysis (HD) is haemodynamic instability; therefore, patients treated in undesignated ICU areas are more likely to be haemodynamically stable than those with haemodynamic instability (who are more likely to be treated with CRRT in designated ICU areas) [7]. In contrast to our findings, Yung et al. [8] found a beneficial effect of CRRT on mortality in COVID-19 patients with renal failure. Their sample size was very small – only 36 patients were included. We believe that their trial was underpowered and that they may have overestimated the survival. Ng et al. [9] retrospectively examined patients with end-stage kidney disease and concluded that they had a higher rate of in-hospital death than those without end-stage kidney disease (31.7% vs. 25.4%; OR 1.38; 95% CI: 1.12–1.70). To date, there has been no large prospective trial addressing the effect of CRRT on mortality among critically ill COVID-19 patients. Secondary bacterial or fungal infections are more likely to be from a resistant organism (carbapenem resistance or methicillin resistance) in designated ICU areas in comparison to undesignated ICU areas, owing to the difference in composition of microbial inhabitance in designated and undesignated ICU areas. For example, Stenotrophomonas are rarely isolated from respiratory secretions in patients occupying undesignated ICU areas or general medical wards. Stenotrophomonas are more common in designated ICU areas, and they can be acquired through the shared use of ICU equipment between ICU patients. A study documented the use of bronchoscopy and calorimetry in the transmission of this organism among ICU patients [10].
Braman et al. [4] documented complications from intrahospital transport in 1987. At that time, technology to assist intrahospital transport was not as advanced as it is today. Szem et al. [11] observed that mortality was elevated in high-risk patients involving transfer, but intrahospital transfers were not the direct reason for the high mortality. The positive outcomes in our study may be the result of better organization of undesignated ICU beds, use of newer and more advanced portable monitors and ventilators, and improved preparation by escorting staff (because the COVID-19 risk of exposure led to extraordinary precautions taken by staff, this may have positively affected the process).
Regarding the length of stay in the ICU, we also found that patients cared for at undesignated ICU beds had longer stays in the ICU. Other factors predictive of longer stay were tracheostomy, steroids, and ECMO, for obvious reasons. Secondary bacterial infection was associated with a shorter length of stay because they had higher mortality and may have died early in the course of illness. Similar results have been documented by others [12]. Schwebel et al. [3] documented that intrahospital transfers increase the LOSICU. However, their study was of non-COVID-19 patients. Ng et al. [9] documented that patients with end stage kidney disease (ESKD) have increased odds of having more than 7 days of LOSICU compared to those without ESKD.
We identified the following limitations of our study: it was a single-centre, retrospective study with a small sample size, and the population was predominantly young males. Therefore, the findings of our study may not be applicable to other communities. We did not record the composition of complications (loss of PEEP, displacement of endotracheal or chest tube, etc.) that might have helped us to understand this process. This is the question we are presently studying in an ongoing study. We also did not record details of isolated pathogens that could affect the outcome. Similarly, details of the strategy of mechanical ventilation were not recorded. Finally, we did not record the nurse-to-patient ratio for either clinical setting. Nonetheless, this was the first step to document and reassure that extra transfers and care outside designated ICU areas do not adversely affect the clinical outcome.
