Older adults have been disproportionately impacted by SARS-CoV-2, frequently requiring more intensive medical care and experiencing worse outcomes.1,2 Atypical syndromes such as functional decline and weakness, and absence of classic symptoms of fever or cough may result in delayed diagnoses.1 The higher prevalence of comorbidities, combined with the physiologic effects of aging and frailty, may contribute to different clinical courses, treatment effects, and outcomes.1,2

Outcomes of older patients with COVID-19 critical illness have been reported from the USA and other countries, but the Canadian experience has not been well characterized.3 Prior studies have shown that older age, need for invasive mechanical ventilation (IMV), acute kidney injury (AKI), and acute respiratory distress syndrome (ARDS) are independently associated with fewer days alive.3,4 A meta-analysis of all COVID-19 patients also revealed significant regional differences in patient outcomes, highlighting the need to formulate region-specific management strategies reflecting variable resources, population health status, and/or viral strains.4

Across Canada, between January 2020 and December 2021, more than 25% of COVID-19-related hospital stays included admission to the intensive care unit (ICU), and 60% of these patients required mechanical ventilation.2 Patients who did not survive their hospital stay were older than those who were discharged alive (median age, 73 vs 57 yr).2 In a Canadian retrospective study of 1,671 adults older than 65 yr of age who were hospitalized for COVID-19, 24.1% required ICU admission and 23.5% died; advanced age, need for ICU admission, frailty and delirium were independently associated with hospital mortality.1

Given the large proportion of older adults requiring ICU care and the implications for health systems, we aimed to describe the management and outcomes of COVID-19 patients ≥ 65 yr of age across five Toronto ICUs and to identify predictors of mortality. Such data on outcomes and risk factors can guide management and inform patient and clinician decision-making regarding ICU admission and organ support.

Methods

Study design

This was an ICU substudy of a multicentre retrospective cohort study, which investigated the atypical presentation of COVID-19 among older adults.1 The protocol was prospectively published in Open Science Framework (https://osf.io/k4g7a/) and the study was reported according to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines (Electronic Supplementary Material [ESM] eTable 1).5 Research Ethics approval was obtained through Clinical Trials Ontario (3186-OPIA-Apr/2020-38044) and individual patient consent was not required.

Setting and participants

The study was conducted in five academic medical centres in Toronto, ON, Canada (Mount Sinai Hospital, University Health Network – Toronto General Hospital and Toronto Western Hospital, St. Michael’s Hospital, and Sunnybrook Health Sciences Centre). We included consecutive adults ≥ 65 yr who were admitted between 11 March 2020 and 30 June 2021 to one of these ICUs with a primary ICU diagnosis of polymerase chain reaction-confirmed acute SARS-CoV-2 infection. We excluded patients who were 1) readmitted more than 24 hr after an index ICU admission for COVID-19 and 2) required ICU management for non-COVID-19-related conditions, as defined by the site’s ICU and infection control teams (e.g., admitted for an elective procedure).

Data collection

Patients meeting the eligibility criteria were identified by the data analytics service at each site, using the same case detection protocol for public health reporting. Deidentified data were extracted from electronic patient records and entered into a REDCap database by a trained chart assessor (K. H.) using a standardized data abstraction form.6 The chart assessor (K. H.) was trained by physician investigators at each hospital site (S. M., E. W., R. N.). The first five charts were extracted in duplicate with the physician investigator, and additional charts were reviewed as needed.

From the electronic patient records, we recorded the following: 1) patient characteristics including age, sex, place of residence, code status, clinical frailty score,7 and comorbidities; 2) daily respiratory parameters (e.g., lowest ratio of arterial partial pressure of oxygen and fraction of inspiratory oxygen [PaO2/FIO2], median tidal volume, lowest arterial pH); 3) ICU complications as defined by the ICU team (e.g., ARDS, AKI, bacterial pneumonia, myocardial infarction [MI] delirium, prolonged fever); 4) analgosedation (sedatives, opioids, antipsychotics) and neuromuscular blockers; and 5) laboratory variables such as hemoglobin, creatinine, and bilirubin at ICU admission. Prolonged fever was defined as a body temperature > 38.3 °C for more than seven days. We calculated illness severity scores (Sequential Organ Failure Assessment [SOFA] and Acute Physiological and Chronic Health Evaluation [APACHE] II) using data from the first 24 hr of ICU admission if arterial blood gas results were available.8,9,10 Hospital and ICU outcomes included delirium prevalence as defined by validated tools used clinically at each site11,12 and a validated chart review method,13 duration of mechanical ventilation, length of stay, and mortality. Delirium was assessed during the entire hospitalization (ICU and wards), with at least one positive delirium score indicating delirium.

Statistical analysis

Patient demographics and clinical characteristics are described using proportions (frequencies) and means (standard deviation) or medians [interquartile range (IQR)] as appropriate. We used the Chi square test to determine statistical significance for proportions. We identified statistical significance using ANOVA and the Kruskal–Wallis test for normally distributed and skewed continuous data, respectively.14

To identify the association between patient risk factors and in-hospital mortality, we used a multivariable logistic regression model. The following variables were chosen a priori for clinical relevance guided by previously reported associations with death in ICU patients: age,15 sex,16 number of comorbidities,17 APACHE II score,10 clinical frailty score,18 and the need for organ support (i.e., one or more of mechanical ventilation, continuous renal replacement therapy [CRRT], and vasopressors).19 We used multiple imputation to address missing data in the models (“mice” R package; R Foundation for Statistical Computing, Vienna, Austria), with the assumption that variables were missing at random.20 The data are presented using unadjusted and adjusted odds ratios (ORs) with corresponding 95% confidence intervals (CIs) and P values. All data were analyzed using R version 4.2.1.21

Results

Baseline characteristics

During the study period, 273 patients ≥ 65 yr who were admitted to the ICU for COVID-19 were included. The median [IQR] age was 74 [69–80] yr and 104 (38.1%) patients were female. Across the cohort, 199 (81.6%) were full-code at admission, 29 (11.9%) had a do-not-resuscitate (DNR) designation, and 16 (6.6%) were for intubation only (Table 1). The median [IQR] number of comorbidities was 2 [1–3], with hypertension (194, 71.1%) and diabetes (129, 47.4%) being among the most prevalent. A large proportion of patients developed complications (228, 83.5%) including ARDS, AKI, prolonged fever, and bacterial pneumonia (Table 2). Patients required prolonged median [IQR] hospital (19 [11–35] days) and ICU (9 [5–22] days) stays, with 142 (52.0%) surviving their hospital stay (Table 3).

Table 1 Patient characteristics, stratified by hospital outcome
Full size table
Table 2 Intensive care unit interventions and organ dysfunction, stratified by hospital outcome
Full size table
Table 3 Intensive care unit and hospital outcomes
Full size table

Compared with hospital survivors, patients who died in hospital were older (median [IQR], 74 [70–82] vs 73 [68–78] yr; P = 0.03), and a smaller proportion of nonsurvivors were female (39/131, 29.8% vs 65/142, 45.8%; P < 0.01). There were no significant differences in the prevalence of comorbidities, except for chronic kidney disease, which was more frequent among nonsurvivors than survivors (44/131, 33.6% vs 16/142, 11.3%; P < 0.001). On ICU admission, nonsurvivors had significantly lower hemoglobin, neutrophils, and platelets, and higher creatinine compared with survivors (Table 1).

At ICU admission, nonsurvivors had lower mean (SD) Glasgow Coma Scores (8.7 [5.3] vs 10.1 [5.3]; P = 0.03), and higher SOFA (10.3 [5.0] vs 8.8 [4.9]; P = 0.03) and APACHE II (25.0 [8.0] vs 22.0 [7.8]; P < 0.01) scores (Table 1). In addition, more nonsurvivors had a DNR order (20/131, 15.3% vs 9/142, 6.3%; P = 0.02).

Intensive care unit interventions and organ dysfunction

There were no differences in the proportion of patients receiving different COVID-19 treatments across the two cohorts (Table 2). There was no difference between survivors and nonsurvivors in the proportion of patients requiring IMV (79/142, 55.6% vs 85/131, 64.9%; P = 0.15). Patients who died in-hospital had significantly higher rates of ARDS (119/131, 90.8% vs 105/142, 73.9%; P < 0.01), AKI (74/131, 56.5% vs 36/142, 25.4%; P < 0.001), and MI (13/131, 9.9% vs 3/142, 2.1%; P = 0.01), and a larger proportion required CRRT (35/131, 26.7% vs 9/142, 6.3%; P < 0.001), compared with those who survived. There was no difference in the prevalence of delirium between survivors and nonsurvivors (93/142, 65.5% vs 77/131, 58.8%; P = 0.24).

There were no between-group differences in the proportion of patients requiring high-flow nasal oxygen (HFNO) or prone positioning (ESM eTable 2). Nevertheless, across patients receiving HFNO, survivors had a longer duration of use than those who died (median [IQR], 5 [3–7] vs 3 [2–4] days; P < 0.001). During the first seven ICU days, nonsurvivors had lower arterial pH and mean arterial blood pressure (ESM eTable 2; ESM eFigs 1–6). Nonsurvivors also had higher plateau pressures; however, this was no longer significant after Bonferroni correction. There were no significant differences in the PaO2/FIO2 ratio, tidal volume, or positive end-expiratory pressure between the two groups (ESM eTable 2; ESM eFigs 1–6).

A significantly greater proportion of nonsurvivors received midazolam (61/131, 46.6% vs 39/142, 27.5%; P < 0.01) and hydromorphone (59/131, 45.0% vs 38/142, 26.8%; P < 0.01) infusions. Among those receiving midazolam and/or hydromorphone, survivors received longer durations of infusion than nonsurvivors did (midazolam mean [SD] 12.1 [9.3] days vs 7.6 [6.3] days; P < 0.01, and hydromorphone 15.0 [11.4] vs 10.7 [8.0] days; P = 0.04). There was also a greater need for vasopressor support in the nonsurvivor cohort than the survivors cohort (95/131, 72.5% vs 80/142, 56.3%; P < 0.01). While restraint use was similar, survivors had a longer duration of restraint use (9.1 [8.6] days vs 4.6 [4.0] days; P < 0.01). Survivors were more likely to have a tracheostomy performed (29 [20.4%] vs 9 [6.9%]; P < 0.01) than nonsurvivors were (Table 2). Across the cohort, 170 (62.3%) patients developed delirium during their hospital stay; with comparable rates in survivors and nonsurvivors (93/142, 65.5% vs 77/131, 58.8%; P = 0.24).

Patient outcomes

There were no differences in the durations of IMV or ICU stay between the two groups (Table 3). Hospital stay was significantly longer among survivors than nonsurvivors (median [IQR], 27 [15–55] vs 14 [9–24] days; P < 0.001; Table 3). Of the 131 nonsurvivors, 103 (78.6%) died in the ICU and 28 (21.4%) died in hospital following ICU discharge.

Using a multivariable logistic regression model, older age (adjusted OR [aOR], 1.42; 95% CI, 1.16 to 1.72; P < 0.001), higher APACHE II score (aOR, 1.05; 95% CI, 1.00 to 1.09; P = 0.04), and the need for organ support (aOR, 2.10; 95% CI, 1.04 to 4.27; P = 0.04) were independently associated with increased odds of in-hospital mortality after adjusting for total number of comorbidities, sex, and clinical frailty scale (Table 4). Need for organ support was defined as IMV, CRRT, and/or vasopressors. Female sex (aOR, 0.42; 95% CI, 0.24 to 0.73; P < 0.01) was independently associated with decreased odds of in-hospital mortality.

Table 4 Unadjusted and adjusted analyses for factors associated with in-hospital mortality in older adults admitted to the intensive care unit with COVID-19
Full size table

Discussion

In this multicentre Toronto cohort of 273 patients ≥ 65 yr admitted to the ICU with COVID-19, 60% required IMV for an average of over two weeks, 38% died in the ICU, and 48% died in the hospital. Nonsurvivors were older, had higher severity of illness, and were more likely to have ARDS, AKI, and MI than survivors were. A higher APACHE II score was independently associated with higher hospital mortality while female sex was associated with lower mortality. Furthermore, increasing age and the need for organ support were also associated with greater odds of mortality.

Comparison with existing evidence

A recent narrative review identified 19 studies reporting data on critically ill older adults (≥ 60 yr) with COVID-19; however, only one study exclusively focused on this population.22 Consistent with our findings, across seven studies, the proportion of older patients requiring IMV ranged from 52% to 91%.22 The high mortality rate in our cohort is comparable to that in other studies. A multicentre retrospective study of 239 patients reported an overall 60-day mortality rate of 61.5%, and noted higher mortality in patients older than 65 yr compared with those younger than 65 yr (73% vs 51%).23 The high prevalence of comorbidities is also consistent with a large cohort study which found that all patients over 80 yr and 76% of those over 60 yr had at least one comorbidity.24 The greater prevalence of ARDS among nonsurvivors in our cohort is also consistent with some prior studies,25,26 while other studies have not reported this finding.27,28 Our modelling revealed that older age was independently associated with a higher odds of hospital mortality, similar to a retrospective study of 1,591 critically ill adults with COVID-19, 436 of whom were older than 60 yr.24 Another study also found significantly higher 90-day mortality rates among those over 80 yr compared with those aged 75–79 yr and 70–74 yr (67% vs 47% vs 39%; P < 0.001, respectively).29 In contrast to our findings, they reported higher 90-day mortality among those requiring intubation, and patients with higher clinical frailty scores. We also found that female sex was associated with decreased in-hospital mortality. This is consistent with a meta-analysis, which reported that male sex was an important risk factor for COVID-19-related adverse events.30 Another study noted that this difference may be attributable to multiple factors, including differential hormonal regulation of immune responses including a greater antiviral response in females, greater dysregulated inflammation in males, and the higher prevalence of underlying cardiovascular comorbidities in males.31

Across the cohort, more than 60% of patients experienced delirium during their hospital stay; however, we did not find a significant difference in the prevalence of delirium between survivors and nonsurvivors. A cohort study of 213 critically ill patients ≥ 18 yr with COVID-19 reported that the mortality rate among those who experienced delirium was not significantly different from those without delirium (25% vs 34%; P = 0.17).32 In contrast, a retrospective study of 235 noncritically ill adults ≥ 65 yr hospitalized with COVID-19 reported that delirium was associated with increased mortality (hazard ratio, 2.1; 95% CI, 1.2 to 3.7; P = 0.01).33 Similarly, in a large cohort of noncritically ill older adults admitted to the same five Toronto hospitals, delirium increased the odds of mortality overall.1 This raises the concern that delirium assessment in critically ill patients may not be accurate because of confounding medications and sepsis.34 Nevertheless, this finding must be interpreted with caution given the limited sample size, and the potential for the presence of hypoactive delirium and detection bias.

We found a significantly higher use of midazolam infusions among hospital nonsurvivors than survivors. This may be attributable to the higher illness severity, higher ARDS incidence, or palliation, which necessitated the use of additional sedative agents. The high use of midazolam in the entire cohort is inconsistent with the Choosing Wisely Canada guidelines, which recommend avoidance of benzodiazepines in older adults, particularly given the suggested association with increased mortality.35 A multicentred, propensity matched study of mechanically ventilated patients ≥ 18 yr reported lower hospital mortality in patients sedated with propofol vs midazolam (RR, 0.76; 95% CI, 0.69 to 0.82).36 A prospective, multicentre study also found that a higher cumulative dose of benzodiazepines was associated with increased odds of 90-day mortality.37

More nonsurvivors had a DNR designation at ICU admission than survivors did. This could be attributable to care teams and families recognizing the risk of poor outcomes among these patients. Nevertheless, nonsurvivors had long stays in the ICU before they died, and approximately half of them died in the ICU following withdrawal of life support. Our data reinforces the importance of early goals-of-care conversations in providing patient-centred care, as well as time-limited ICU trials.

Strengths and limitations

This study is among the first to characterize the management and outcomes of critically ill, older adults with COVID-19. The multicentre retrospective design allowed us to include a larger cohort of consecutive patients across different Toronto ICUs, improving the precision and representativeness of the findings. A study investigator supervised the chart review at each site to ensure consistent data abstraction. Our study has limitations. The data were collected retrospectively, and some data were missing. While techniques such as multiple imputation were used to address missing data, the accuracy and availability of our data were also limited to electronic medical records. We were unable to evaluate the impact of different COVID-19 variants on patient management and outcomes. Furthermore, because only a small proportion of patients received a COVID-19 vaccine, we were unable to investigate the role of vaccination status on outcomes. The impact of other sociodemographic variables such as race and ethnicity were not assessed. We also lack data on morbidity and mortality beyond the hospital stay, and on quality-of-life following recovery from serious COVID-19 illness.

Conclusions

In this retrospective cohort of older adults admitted to the ICU with COVID-19, approximately half died in hospital. Across all patients, over half required IMV and vasopressors. A higher APACHE II score, increasing age, and the need for organ support were associated with increased odds of hospital mortality while female sex was associated with decreased odds. Further prospective research is required to better understand the risk factors for adverse events and provide more long-term data on physical and cognitive outcomes.