Critically Ill Patients With Severe Acute Respiratory Syndrome

Context Severe acute respiratory syndrome (SARS) is a newly recognized infectious disease capable of causing severe respiratory failure.

Objective To determine the epidemiological features, course, and outcomes of patients with SARS-related critical illness.

Design, Setting, and Patients Retrospective case series of 38 adult patients with SARS-related critical illness admitted to 13 intensive care units (ICUs) in the Toronto area between the onset of the outbreak and April 15, 2003. Data were collected daily during the first 7 days in the ICUs, and patients were followed up for 28 days.

Main Outcome Measures The primary outcome was mortality at 28 days after ICU admission. Secondary outcomes included rate of SARS-related critical illness, number of tertiary care ICUs and staff placed under quarantine, and number of health care workers (HCWs) contracting SARS secondary to ICU-acquired transmission.

Results Of 196 patients with SARS, 38 (19%) became critically ill, 7 (18%) of whom were HCWs. The median (interquartile range [IQR]) age of the 38 patients was 57.4 (39.0-69.6) years. The median (IQR) duration between initial symptoms and admission to the ICU was 8 (5-10) days. Twenty-nine (76%) required mechanical ventilation and 10 of these (34%) experienced barotrauma. Mortality at 28 days was 13 (34%) of 38 patients and for those requiring mechanical ventilation, mortality was 13 (45%) of 29. Six patients (16%) remained mechanically ventilated at 28 days. Two of these patients had died by 8 weeks' follow-up. Patients who died were more often older, had preexisting diabetes mellitus, and on admission to hospital were more likely to have bilateral radiographic infiltrates. Transmission of SARS in 6 study ICUs led to closure of 73 medical-surgical ICU beds. In 2 university ICUs, 164 HCWs were quarantined and 16 (10%) developed SARS.

Conclusions Critical illness was common among patients with SARS. Affected patients had primarily single-organ respiratory failure, and half of mechanically ventilated patients died. The SARS outbreak greatly strained regional critical care resources.

Severe acute respiratory syndrome (SARS) is a newly recognized illness that has rapidly spread throughout Asia, North America, and Europe. As of June 9, 2003, 8241 people in 30 countries have developed SARS leading to 784 deaths.¹ The morbidity and mortality associated with SARS has led to international concern.

The epidemiological findings and clinical presentation of SARS for the initial cases in Canada and Hong Kong have been described.^2-5 SARS produces an acute respiratory illness with 23% to 32% of patients becoming critically ill.^4,6 The burden of illness, clinical features, and outcome may be different from acute lung injury due to other etiologies. In addition, these outbreaks have caused a significant strain on the health care system by the influx of patients and the human resources issues related to quarantine and SARS infection in health care workers (HCWs).

The objectives of this study were to characterize the epidemiology, clinical characteristics, and 28-day outcomes of critically ill patients with SARS, and to evaluate the impact of SARS transmission from critically ill patients to HCWs. A better understanding of SARS-related critical illness will allow for improved resource planning and better protection of HCWs and may suggest effective interventions for the patients most seriously affected by SARS.

We retrospectively studied consecutive critically ill adult patients with suspected and probable SARS in the Toronto area who were admitted to intensive care units (ICUs) between the onset of the Toronto outbreak and April 15, 2003 (Figure 1). We included 13 hospitals (5 university, 8 community) known to care for SARS patients. (A list of the participating hospitals appears at the end of this article.) Identification of all critically ill SARS patients in these institutions was achieved by collaboration with the Ontario Hospital Association and the City of Toronto Department of Public Health (which were responsible for the mandatory reporting of SARS), by communication among an ad hoc Toronto SARS Critical Care group and by cross-reference with a database from a previous study that reported the general characteristics of patients with SARS in 10 of our study hospitals.⁵

Suspected and probable SARS was defined according to the definitions issued by the World Health Organization as of April 20, 2003.⁷ Suspected SARS was defined by the presence of fever greater than 38°C, respiratory symptoms, and a history of travel to a geographic location associated with SARS transmission or close contact with a known SARS patient. Probable SARS required the addition of lung infiltrates on chest radiograph. We defined critically ill patients as those admitted to the ICU requiring mechanical ventilation, inspired oxygen concentration on face mask greater than or equal to 60%, or inotropic medication. To evaluate the proportion of patients with suspect or probable SARS who became critically ill, we compared critically ill patients with the total number of patients diagnosed with probable or suspected SARS treated at any of the participating hospitals by April 15, 2003.

Data collection forms were created with input from a multidisciplinary group of HCWs. Following approval from each hospital's research ethics boards, experienced research assistants abstracted data retrospectively from the medical records. Data were checked for errors by a second investigator through manual and electronic inspection using prespecified range limits. The authors of a recent report of SARS patients in Toronto provided a database of their general characteristics during a similar period.⁵ This database was used only to compare noncritically ill with critically ill SARS patients. Twenty-nine (76%) of the patients discussed in this article were part of this earlier study,⁵ which did not address SARS-related critical illness. When occupational transmission of illness was reported in any of the 5 university ICUs, we identified the number of HCWs who were quarantined and who developed suspected or probable SARS. The number of ICU bed closures resulting from SARS transmission or quarantine was tracked in all ICUs.

The following information was collected for each patient: age, sex, occupation (HCW or non-HCW), time course of fever or respiratory symptoms, contact or travel to a SARS-affected area, medical comorbidities, date of hospital and ICU admission and discharge, date of initiation of and liberation from mechanical ventilation, and the Acute Physiology and Chronic Health Evaluation (APACHE) II and sepsis-related organ failure assessment (SOFA) scores.^8,9

For each of the first 7 days the patient was admitted to the ICU, physiological markers of organ dysfunction, ventilatory, radiographic, and treatment-related variables were recorded. These variables included mode of ventilation, fraction of inspired oxygen (FIO₂), tidal volume supplied, positive end-expiratory pressure, peak airway pressure, plateau airway pressure, mean airway pressure, respiratory rate, and adjuncts to ventilation such as neuromuscular blockade, prone positioning, inhaled nitric oxide, or surfactant administration. Daily arterial blood gas values included pH, PaCO₂, PaO₂, bicarbonate concentration, and oxygen saturation. Radiographic findings recorded included the number of involved quadrants, the presence of unilateral or bilateral disease, and barotrauma (presence of interstitial emphysema, pneumothorax, subcutaneous emphysema, pneumomediastinum, or pneumopericardium). Organ dysfunction was defined using an organ SOFA score of more than 2 (ie, cardiovascular: requiring dopamine >5 ug/kg per minute or any dose of norepinephrine or epinephrine; renal: urine output <500 mL/d, or creatinine level >3.4 mg/dL[>299 µmol/L]; hematologic: platelet count <100 × 10³/µL; and liver: bilirubin level >5.9 mg/dL [>101 µmol/L]). Microbial culture results were recorded as were specific treatments, including administration of ribavirin, corticosteroids, antibiotics, activated protein C, intravenous immunoglobulin, and plasmapheresis.

Acute respiratory distress syndrome (ARDS) was defined according to established criteria.¹⁰ To evaluate whether lung protective ventilatory support was provided, we identified patients who received a threshold tidal volume greater than 8 mL/kg (actual body weight) or had a peak airway pressure greater than 35 cm H₂O on any 2 consecutive days during the first 7 days in the ICU.¹¹ We used the threshold tidal volume of 8 mL/kg because we did not have measured height and predicted body weight. When multiple daily measurements were performed, those closest to 08:00 hours were recorded because this time corresponded with the majority of daily measurements.

The primary outcome was mortality at 28 days after ICU admission. Secondary outcomes included the proportion of SARS-related critical illness; patient location and ventilation requirements at day 28; the number of tertiary care medical-surgical ICUs placed under quarantine, and in those institutions, the number of HCWs contracting SARS secondary to ICU SARS transmission. Intensive care unit HCWs who became ill were followed up to determine their need for critical care support.

To determine association between variables and mortality, the Fisher exact test was used for categorical variables and univariable logistic regression was used for continuous variables. Variables considered included age; sex; occupation; medical comorbidity (diabetes mellitus, ischemic cardiac disease, chronic obstructive pulmonary disease); clinical features at admission (temperature, heart rate, respiratory rate, presence of nonproductive cough, dyspnea, oxygen saturation); admission laboratory values (lymphocyte count, lactate dehydrogenase [LDH], calcium, creatine kinase); admission chest radiograph findings characterized by unilateral, bilateral, or no disease; treatment with ribavirin, corticosteroids, or antibiotics; tidal volume (per kilogram of actual body weight); and peak pressure while receiving mechanical ventilation. When the clinical laboratory data were sparse, exact logistic regression was used. The Kaplan-Meier method was used to determine the probability of survival over the duration of follow-up and to generate survival curves. All statistical tests were 2-tailed. Factors were considered statistically significant at α less than .05. The SAS System for Windows version 8.2 (SAS Institute Inc, Cary, NC) was used for all analyses.

Of 196 patients with probable or suspected SARS in the 13 Toronto area hospitals during the study period, 38 (19%) met inclusion criteria. Two additional patients were admitted to ICUs for negative pressure isolation but did not meet our definition for critical illness; hence, they were excluded.

Demographic characteristics of critically ill SARS patients are presented in Table 1. Health care workers comprised 18% of the critically ill patients. The median (interquartile range [IQR]) age of all critically ill patients was 57.4 years (39.0-69.6 years) compared with 45 years (34-57 years) reported for all patients with SARS.⁵ There was a predominance of older, non-HCWs (82%) in the critically ill group with a median (IQR) age of 61 years (44-75 years) compared with a preponderance of HCWs (58%) in the noncritically ill patients, who had a lower median (IQR) age of 42 years (31-50 years).⁵ There was a high rate of prior comorbidity, particularly diabetes, among patients with SARS-related critical illness. The durations between initial symptoms of SARS, hospital admission, ICU admission, and death are presented in Table 2.

Of the 38 patients admitted to the ICU, 31 (82%) met diagnostic criteria for ARDS. Mechanical ventilation was required for 29 patients (76%), representing 15% of all SARS patients at study hospitals. The median (IQR) time from hospital admission to institution of mechanical ventilation was 4.0 days (1.0-5.0). Barotrauma occurred in 10 of these ventilated patients (34%). Other organ dysfunction (defined as organ SOFA score >2) was less common—14 patients (37%) developed cardiovascular dysfunction; 8 (21%), hepatic dysfunction; and 4 (11%), renal dysfunction in the first 7 days in the ICU. Neuromuscular blockade was used in 15 patients (representing 52% of ventilated patients), high-frequency oscillatory ventilation in 3 patients (10%), nitric oxide in 8 patients (28%), and prone positioning in 1 patient (3%). Inotropic or vasoactive medications were required by 14 patients (37% of ICU patients) and 2 patients (5%) required hemodialysis.

Mortality, mechanical ventilation dependency, and patient location at day 28 are presented in Table 3. The median (IQR) ICU length of stay was 10.5 days (5-28 days). The 28-day mortality rate for SARS patients admitted to the ICU was 34%, and was 45% (13 of 29 patients) for those requiring mechanical ventilation. However, 6 patients (16%) remained on ventilatory support at 28 days. Late follow-up at 8 weeks revealed that 15 patients had died for a mortality rate of 39% (or 52% for those requiring mechanical ventilation), with 3 patients remaining on ventilatory support.

Table 4 compares characteristics of surviving and nonsurviving critically ill patients with SARS with characteristics of patients with SARS who did not require ICU admission.⁵ We found that older age, a history of diabetes mellitus, admission tachycardia, and elevated creatine kinase were associated with poor outcome. The presence of bilateral radiographic lung infiltrates at admission was more common among patients subsequently needing ICU care. Figure 2 presents the probability of survival over time for all critically ill patients with SARS and for patients older than 65 years vs patients 65 years or younger. Only 3 of 10 patients (30%) older than 65 years were alive at day 28 while 22 of 28 patients (79%) aged 65 years or younger were alive.

Median peak airway pressure among ventilated critically ill patients was similar in survivors (30 cm H₂O; IQR, 21.5-33.0 cm H₂O) and nonsurvivors (30 cm H₂O; IQR, 26-35 cm H₂O). Median tidal volume (IQR) among survivors was 6.0 mL/kg (5.5-7.0 mL/kg) and 7.8 mL/kg (5.9-8.0 mL/kg) among nonsurvivors. Tidal volume exceeded 8 mL/kg for 2 consecutive days in 6 (21%) of the patients who received mechanical ventilation, of whom 5 died (P = .06).

On 2 separate occasions, ICU patients transmitted SARS to HCWs in 2 of the 5 university medical-surgical ICUs in Toronto. The first episode occurred when a patient with unsuspected SARS was treated for 30 hours in the absence of respiratory precautions. When SARS was recognized, 69 HCWs were quarantined. Seven HCWs subsequently developed SARS. All were hospitalized, but none became critically ill. The second episode involved exposure during endotracheal intubation of a hospitalized SARS patient. Although infection-control precautions with N-95 respirator mask, gloves, and gowns were used, 9 HCWs developed SARS likely related to a prolonged and difficult endotracheal intubation of a combative patient. Eight of these HCWs were hospitalized, but none became critically ill. This episode required 95 HCWs to be quarantined.

These 2 events led to 10-day closures of 35 critical care beds, representing 38% of the tertiary care university medical-surgical ICU beds in Toronto. The loss of critical care capacity resulted in the cancellation of surgery that would have required perioperative critical care monitoring, including cardiovascular surgery and transplantation. In addition, there were 38 concurrent bed closures due to ICU SARS transmission and quarantine of HCWs in 4 of 8 study community hospitals. This represented 33% of the Toronto community medical-surgical ICU-bed capacity. Other events also affected critical care bed availability. In 2 institutions, HCWs developed symptoms that were investigated for SARS and subsequently found not to have met the World Health Organization definitions. However, during the investigation period 1 ICU was closed for several days and HCWs were quarantined.

In this study, we identified that a high proportion of patients with probable and suspected SARS became critically ill. We found that the median time from symptom onset to death was 19 days, with many deaths occurring beyond the follow-up time of previously reported SARS epidemiological studies.^2-5 Although recent media reports have suggested an apparently increasing SARS mortality, we hypothesize that these results are due to longer follow-up studies like ours, rather than a changing epidemiology of SARS. Our data confirm previous observations that mortality is associated with older age,⁴ and, that HCWs, who are often younger than other SARS patients, are less likely to die.

We found that SARS-related critical illness predominantly involved a single-organ system, respiratory failure. A much smaller proportion of patients exhibited cardiovascular instability, and very few developed other organ failure during the first 7 days of critical illness. Mortality at 8 weeks was 52% among patients with SARS requiring mechanical ventilation. This mortality rate is similar to the mortality rate of a large unselected series of patients with ARDS requiring mechanical ventilation.¹² High tidal volumes administered to some patients with SARS may have contributed to this mortality rate.¹³ We examined ventilation for 7 days to reflect management during the initial phase of acute lung injury due to SARS. The proportion of patients with SARS who developed barotrauma in our study (34% of ventilated patients) is higher than reported for other forms of acute lung injury or ARDS.^13,14

We observed that diabetes mellitus was a common comorbidity among those with SARS-related critical illness as was shown in the previous Toronto cohort, which included many of our patients.⁵ The association between tachycardia on admission and mortality likely reflects increased severity of disease and is a common component of severity of illness scales.⁸ A high-serum LDH level on admission to the hospital appeared to be associated with increased mortality among the critically ill patients. However, the LDH level of noncritically ill and critically ill SARS patients was not different. Lee et al⁴ have demonstrated an association between peak LDH levels and mortality in SARS. Increased serum LDH has previously been associated with several pulmonary infections.^15,16 Although this finding is usually nonspecific, the observation that higher levels of LDH may be associated with increased mortality in SARS is similar to the experience with patients who have Pneumocystis carinii pneumonia.^17,18

Critically ill patients with SARS are at high risk of infecting HCWs, likely related to high-risk procedures predisposing to droplet spread and to larger viral loads in these patients. Viral shedding appears to peak relatively late in the course of the disease when patients become critically ill.⁶ For clinicians on the front lines caring for patients with SARS, it is of concern that almost one fifth of critically ill patients discussed in our study were HCWs. Although none have died, one continued to require ventilatory support at 8 weeks. In a previous series of 138 patients with a shorter follow-up period, a similarly good outcome in HCWs was noted.⁴ Whether the apparently better outcome in HCWs is an age-related phenomenon, due to a lower viral inoculum or other as yet unexplained factors, remains to be determined.

Early identification of patients likely to require critical care services and the relatively slow progression to intubation that we observed should allow for optimal management of these patients. Uncontrolled exposure to infecting agents during invasive procedures such as intubation can be averted by the early use of universal infection-control precautions and appropriate HCW training and education. Early transfer to the ICU, avoidance of noninvasive ventilation, and controlled endotracheal intubation with enhanced infection control precautions¹⁹ will hopefully minimize the occupational hazard. During this outbreak, infection control precautions changed on an almost daily basis as new evidence emerged. A number of HCWs were infected during intubations while using precautions that were thought to be adequate at the time (N-95 respirator masks, gowns, gloves, goggles). This has led us to use more stringent infection control precautions including powered air-purification respirators for high-risk procedures.^19,20 Increased awareness of the risk of contact spread has led to the use of double gowns and gloves. The risk of infection and concern about transmitting disease to family resulted in significant stress among HCWs.²¹

Our study has a number of important limitations. The calculation of the number of patients becoming critically ill may have been affected by transfers to our study centers. However, since no other hospitals in the Toronto area had significant numbers of SARS patients and interhospital transfer of SARS patients was strongly discouraged during the study period, we believe that our report is accurate. Our definition of critical illness required admission to the ICU, and it is possible that a higher proportion of patients with SARS were critically ill but cared for in hospital wards. Although this is a comprehensive series of critically ill patients with SARS admitted to study ICUs, the sample size is limited and we were unable to precisely identify all variables that may influence outcomes such as specific therapies.

How is the critical illness and respiratory failure associated with SARS different from severe respiratory failure due to other viral illnesses such as influenza? The mortality rate for patients with SARS who require mechanical ventilation may be similar to that observed during severe influenza outbreaks.²² However, a prominent difference is that SARS is much more likely to progress from a mild to a severe disease in young, otherwise healthy individuals—an uncommon feature of influenza infection.²³ Furthermore, we identified the need for prolonged mechanical ventilation and supportive critical care in critically ill patients with SARS. This is important information for clinicians, health administrators, and governments planning for ongoing and future outbreaks of SARS. In Toronto, critical care resources were significantly strained during the SARS outbreak as a result of the influx of SARS patients, the closing of several institutions for quarantine, and illness or quarantine of HCWs. Affected health districts in the future will need to increase their capacity to treat critically ill patients in respiratory isolation. We highly recommend that all bedside clinicians have the necessary equipment, protective devices, and most important, training and experience to use such devices prior to the onset of a SARS outbreak.