Introduction

The severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is a risk factor for acute respiratory distress syndrome (ARDS) that is currently a major healthcare challenge worldwide. The prognosis of this disease widely varies between countries, the age of the patients, the characteristics of the population studied, and the severity of the ARDS [1]. Then, the case fatality rates observed in ARDS-related SARS-CoV-2 is close to 30–40% [2,3,4], but can reach 70% in the older patients [5,6,7]. Given the heavy burden of ARDS-related SARS-CoV-2 infection in terms of health care resources and the worrisome prognosis of this disease, the pandemic has raised several ethical questions. One of them is the decision to admit the oldest patients in the ICU [8], which should be guided by robust data on the outcomes of that population. Therefore, there is an urgent need to provide consistent data on the management and prognosis of the elderly patients in the intensive care unit (ICU) [9]. These data may serve policymakers to properly and fairly allocate health care resources to that population and also to provide transparent information to the patient and caregivers. To date, few studies specifically reported the management and prognosis of the elderly patients in the context of SARS-CoV-2 lower respiratory tract infection [10, 11], but none were focused on a population admitted in ICU. In a large German study enrolling 10,021 patients, 923 (9%) patients over 70 years old received ventilatory support which was associated with 63% in-hospital mortality in those 70–79 years [4]. This result concurred with the dismal prognosis reported in previous studies focused on elderly patients with ARDS not related to SARS-CoV-2 infection [12, 13]. As the debate is still active whether the management of COVID-19 should differ from ARDS related to other causes [14], the specific ICU management and outcomes of the old patients with SARS-CoV-2 related ARDS has not been fully described so far. We sought to assess the characteristics, management, and prognosis of the patients over 70 years enrolled in the international COVID-ICU cohort [15].

Methods

Study design, patients

We performed an ancillary analysis of the COVID-ICU study. COVID-ICU was a multi-center, observational, and prospective cohort study conducted in 149 ICUs from 138 centers, across three countries (France, Switzerland, and Belgium) and has been described elsewhere [15]. It received approval from the ethical committee of the French Intensive Care Society (CE-SRLF 20-23) and Swiss and Belgium ethical committees following local regulations. All patients or close relatives were informed that their medical data were anonymously included in the COVID-ICU cohort. Patients and relatives had the possibility not to participate in the study. In case of refusal, the data were not collected accordingly. This manuscript follows the STROBE statement for reporting cohort studies.

For this analysis, we restricted the study population to patients who were 70 and above 70 years of age at the time of the admission to the participating ICU between February 25, 2020, and May 4, 2020, with laboratory-confirmed SARS-CoV-2 infection, and available Day-90 vital status. Laboratory confirmation for SARS-Cov-2 was defined as a positive result of real-time reverse transcriptase-polymerase chain reaction (RT-PCR) assay from either nasal or pharyngeal swabs, or lower respiratory tract aspirates [16].

Data collection

Full description of data collection is provided in the Additional file 1. Baseline information collected at ICU admission were: age, sex, body mass index (BMI), active smoking, Simplified Acute Physiology Score (SAPS) II score [17], worse Sequential Organ Failure Assessment (SOFA) [18] during the first 24 h, comorbidities, immunodeficiency (if present), Clinical Frailty Scale [19], the date of the first symptom, and dates of the hospital and ICU admissions. The Clinical Frailty Scale was collected upon ICU admission by the physician in charge of the patient during the medical examination. If the patient was not able to communicate, the physician obtained the information from the relatives. The Clinical Frailty Scale is an ordinal hierarchical scale of 9 ranks, with a score of 1 being very fit, 2 well, 3 managing well, 4 vulnerable, 5 mildly frail, 6 moderately frail, 7 severely frail, 8 very severely frail, and 9 terminally ill. We also collected modes of ventilation and oxygenation and complications over the ICU stay. Patient outcomes included duration of mechanical ventilation, vital status at ICU and hospital discharge, and 28, 60, and 90 days after ICU admission. Lastly, life-sustaining treatment decisions were also collected.

Statistical analyses

Characteristics of patients were described as frequencies and percentages for categorical variables, whereas continuous variables were reported as mean and standard deviation or median and interquartile range. Categorical variables were compared by Chi-square or Fisher's exact test, and continuous variables were compared by Student's t-test or Wilcoxon's rank-sum test. Kaplan–Meier overall survival curves until Day-90 were computed, and were compared using log-rank tests. Detailed statistical analysis is provided is the Additional file 1.

Baseline risk factors of death at Day-90 were assessed using univariate and multivariate Cox regression model stratified on the center variable. Proportional hazard assumption was assessed by inspecting the scaled Schoenfeld residuals and Harrell’s test [20]. To assess invasive mechanical ventilation effect on Day-90 mortality, we used a Cox proportional hazard model weighted on inverse probability of treatment weighting (IPTW) using propensity score (PS) defined as the predictive probability of invasive mechanical ventilation conditional on measured baseline covariates [21]. A multivariate logistic regression model was performed to estimate the PS for each patient in that population. To assess the balance of measured covariates between treatment groups, we used the standardized mean differences before and after PS weighting [22]. Then, a Cox proportional hazard model weighted on IPTW was performed to estimate the average treatment effect in the entire eligible population [21]. Hazard ratio and its 95% confidence interval were then estimated for the Day-90 mortality associated with invasive mechanical ventilation at Day-1. This analysis was performed on the complete cases data set, and a sensitivity analysis was performed using multiple imputations due to missing data.

All analyses were performed at a two-sided α level of 5% and conducted with R version 3.5.1 (R Foundation for Statistical Computing, Vienna, Austria).

Results

Characteristics of patients at ICU admission

From the 4244 patients enrolled in the COVID-ICU dataset, 1199 (28%) (1115, 41, 43 patients in France, Switzerland, and Belgium, respectively) met the inclusion criteria of the present study (i.e., age over 70 years old) (see the Additional file 1: Figure S1). The main descriptors of the patient’s characteristics are presented in Table 1. The median (IQR) age was 74 (72–78) years. Fifty-three percent of the patients were 70–74 years old, 31% were 75–79 years old and 16% were over 80 years old. The majority of the patients were male (73%). The most frequent comorbidities were chronic hypertension (62%), diabetes (30%), and chronic respiratory disease (25%). Noticeably, the median (IQR) Clinical Frailty Scale was 3 (2–3), with only 160/1085 (15%) vulnerable patients (i.e., Clinical Frailty Scale 4), and 99/1085 (9%) frail patients (i.e., Clinical Frailty Scale 5–9). The time between first symptoms and ICU admission was 8 (6–12) days. SAPS II and SOFA scores at ICU admission were 43 (35–54) and 5 (3–8), respectively.

Table 1 Demographic characteristics and management during the first 14 days of ICU according to their Day-90 survival status
Full size table

Mortality was 41%, 45%, and 46% at Day-28, Day-60, and Day-90, respectively (Additional file 1: Table S1). Mortality at Day-90 increased with the age and the Clinical Frailty Scale (Fig. 1). Indeed, Day-90 mortality increased from 39% in the patients between 70 and 74 years to 47% and 67% in the groups of patients between 75 and 79 years and those over 80 years old, respectively (p < 0.001) (Fig. 2a). Similarly, mortality at Day-90 was 40%, 61%, and 71% in the patients' groups with Clinical Frailty Scale from 1–3; 4; and ≥ 5, respectively (p < 0.001) (Fig. 2b). The mortality was also higher in patients intubated during their ICU stay ranging from 44 to 74% (Additional file 1: Figure S2). Of note, during the period of the first 14 days following the ICU admission, 253/1,199 (21%) of the patients had a life-sustaining treatment limitation decision, whom 223 (88%) died at day 90 (207 (82%) while in the ICU).

Fig. 1
figure 1

Day-90 mortality according to age and Clinical Frailty Scale

Full size image
Fig. 2
figure 2

Kaplan–Meier survival estimates during the 90 days following ICU admission, according to a age (70–74 years, 75–79 years and > 80 years), b Clinical Frailty Scale (1–3; 4; >  = 5) and c PaO2/FiO2 ratio at Day-1 of ICU admission

Full size image

Predictive factors of mortality at Day-90

Results of the multivariable analysis are reported in Table 2. Because of multicollinearity observed between age and Clinical Frailty Scale, invasive mechanical ventilation at Day-1 and PaO2/FiO2 ratio, renal replacement therapy and the renal component of the SOFA, only Clinical Frailty Scale, PaO2/FiO2 ratio, and the renal component of the SOFA were retained in the model. Day-1 patients’ characteristics significantly associated with a higher 90-Day mortality rate identified by the Cox regression model after center stratification were older age, diabetes, higher cardiovascular component of the SOFA score, lower PaO2/FiO2, and a shorter time between first symptoms and ICU admission (Table 2). The same analysis re-run of missing after multiple imputations data (Additional file 1: Table S2) yielded similar conclusions. Interestingly, being admitted to the ICU after March 29 was also associated with a better outcome (Additional file 1: Figure S3). Kaplan–Meier survival estimates according to age categories, Clinical Frailty Scale, and PaO2/FiO2 ratio at Day-1 of ICU admission are provided in Fig. 2.

Table 2 Predictive patient factors associated with Day-90 mortality in critically ill patients older than 70 years old with COVID-19 stratified on the center variable
Full size table

Propensity score analysis

Six hundred and forty-four patients had a cardiovascular component of the SOFA < 2, comprising 425 patients intubated on Day-1 and 219 initially treated without invasive mechanical ventilation. These two groups differed in several respects (Additional file 1: Table S3). Patients intubated on Day-1 had a higher SOFA cardiovascular component and were more likely admitted to the ICU before March 28. Interestingly, their Clinical Frailty Scale, their BMI, the time between first symptoms and ICU admission, and the PaO2/FiO2 ratio were not different. After weighting on the Inverse Probability Weighting Treatment using propensity score estimated in 269 patients with no missing values, 123 non-intubated patients were compared to 146 patients intubated at Day-1 with a similar medical history and initial severity Additional file 1: Table S3). We found a significantly different Day-90 mortality (28% in the non-intubated group vs. 42% in the intubated group; HR 1.68; 95% CI 1.24–2.27; p < 0.001) (Fig. 3). A similar analysis performed after multiple imputations of missing data (i.e., 644 patients) yielded similar conclusions (HR 1.33; 95% CI 1.11–1.59; p = 0.002).

Fig. 3
figure 3

Kaplan–Meier survival estimates during the 90 days following ICU admission in propensity score-matched patients

Full size image

Discussion

Herein, we report the characteristics, management, and outcomes of a large prospective cohort of old critically ill patients during the first wave of the COVID-19 outbreak. Patients over 70 years represented 28% of the COVID-19 population admitted during that period of 8 weeks in the participating ICUs. Their overall Day-90 mortality was 46%, which increased with the age and the Clinical Frailty Scale and reached 67% for the patients over 80 years. Older age, diabetes, a longer time between first symptoms and ICU admission, a SOFA cardiovascular component ≥ 3, a lower PaO2/FiO2 ratio, and being admitted to the ICU during the first month of the pandemic were independent risk factors of Day-90 mortality. Noticeably, our propensity score analysis suggests that an early invasive mechanical ventilation strategy seemed associated with a worse prognosis in that population.

The mortality of elderly patients admitted in the ICU for SARS-Cov-2-related ARDS varied from 77 to 84% [1]. These mortality rates appear very high compared to those reported in ARDS outside COVID-19 [12, 23]. For instance, the Large Observational Study to Understand the Global Impact of Severe Acute Respiratory Failure (LUNG SAFE) reported Day-90 mortality rates of 47%, 51%, and 50% for the 70–74 years, 75–79, and > 80 years old patients, respectively (unpublished data, personal communication from the authors) [24]. Our Day-90 mortality (46%) contrasts with early reports (1–3) and the large German cohort of 10,021 patients (923 patients over 70 years) [4] despite a large proportion of patients intubated during their ICU stay in our study (78%). It was, however much higher than the 25% Day-90 mortality observed in the rest of the population of the COVID-ICU cohort (i.e., patients < 70 years old) [15]. Besides, the mortality of our patients over 80 years old seems higher when compared with same-age patients with non-COVID-19-related ARDS, planned [25], or unplanned ICU admission [26]. Several factors such as triage policy before ICU admission, ICU resources at the time of the pandemic, ICU case volume [27] and patients’ comorbidities may explain these discrepancies.

Before the context of COVID-19, frailty as measured with the Clinical Frailty Scale in elderly critically ill patients was strongly associated with Day-30 mortality [26]. This tool was even a better predictor of mortality than SOFA score [25] or classical geriatric scales [26]. Recently, in a large observational study performed in the United Kingdom that enrolled 1564 COVID-19 patients with a median age of 74 years, and more than 50% of the population with a Clinical Frailty Scale > 4, the crude hazard ratio (95% confidence interval) for mortality were 3.12 (2.05–4.76) and 4.41 (2.90–6.71) for those with a Clinical Frailty Scale of 5–6 and 7 to 9, respectively [11]. However, the overall low Clinical Frailty Scale reported in our study and our low proportion of vulnerable or frail patients suggest that a significant triage was performed before ICU admission [28]. No national ICU admission criteria policy was provided at the time of the study, and the ICU admission decision was left to the discretion of the physicians in charge of the patient. Whether this triage resulted from intensivist’s evaluation, non-intensivists practitioner’s judgment, ICU beds occupancy, or the patients themselves should be further investigated.

Old patients admitted to the ICU with COVID-19 are at increased risk of death [3, 29] and the decision of ICU admission can be challenging [8]. The use of the Clinical Frailty Scale has proven to be helpful in this context [9]. Besides, the respect of the patient’s wishes and values, expressed directly by the patient via advance directives or reported by the healthcare surrogate should have to be taken into consideration [30]. In old patients with an uncertain prognosis, it can be particularly difficult to decide whether or not to admit to the ICU and provide invasive treatments such as mechanical ventilation. In such circumstances, an “ICU-trial of limited-time” has been proposed [31]. However, in the context of COVID-19, this strategy could be challenging as a long invasive mechanical duration is often required to see any clinical improvement. In other words, an ICU trial with a too-short limited-time could lead to misinterpretation and ethical misconduct. This important point is reinforced by the extremely long durations of invasive mechanical ventilation, and ICU length of stay observed in our surviving patients.

Beyond the admission of elderly patients in the ICU, the decision of the timing of intubation remains crucial. The majority of our patients (62%) were intubated on ICU Day-1. Interestingly, apart from obvious reasons such as hemodynamic instability, relevant clinical differences were scarce between patients who were intubated upon admission and those who were not. For instance, their Clinical Frailty Scale, time between first symptoms and ICU admission, and PaO2/FiO2 ratio were not significantly different, suggesting that the decision of intubation on admission was mainly driven by the experience of the physicians and the limited knowledge of this new disease at that time. As reported by others [32], the proportion of patients being intubated upon ICU admission during the first period of the study decreased from 67 to 56% during the last month (after March 29th, 2020), with being admitted in that latter period independently associated with a lower Day-90 mortality. An early intubation strategy was even associated with a poorer outcome in our matching analysis while further studies are warranted to confirm this finding. Less reluctance of the caregivers to provide non-invasive oxygen strategies along the first COVID-19 wave has been reported [15], but the benefit in terms of survival is still uncertain [33]. These strategies seem promising in that at-risk population where patients receiving invasive mechanical ventilation are more likely to experience long-term physical, neuropsychiatric, and quality of life impairments [34, 35].

Our study is a large international cohort of old critically ill patients with detailed characteristics and Day-90 outcome. However, despite a large number of participating ICUs, our population sample may be prone to selection biases that may limit generalizability. Since the study was mainly conducted in France (1115, 41 and 43 patients in France, Switzerland, and Belgium, respectively) during a period with high pressure on the health system and before the publication of several core randomized trials [36, 37], our findings may differ during subsequent COVID-19 outbreaks, and in countries with different public health care organizations, ICU admission policy, or ICU resources [4]. Comparison with further studies from other countries will help to better allocate health care resources and determine the indications and contra-indications of non-invasive ventilatory strategies in this specific population. Besides, we only provided data on patients who were admitted to the ICU, and no information was available on treatments before ICU admission nor on patients for whom an ICU admission was denied in the participating ICUs. Besides, important detailed information is also lacking regarding therapy limitations. This information would have allowed a thorough investigation of ICU-admission criteria used during this surge of ICU resources.

Conclusions

During the first COVID-19 pandemic wave, patients over 70 years old represented more than a quarter of the COVID-19 population in the participating ICUs of that study. Their overall Day-90 mortality was 46% with a dismal prognosis in patients older than 80 years old. Given the very long duration of mechanical ventilation as well as a prolonged ICU and hospital stay in the survivors, further studies are urgently warranted to evaluate the long-term psychological, neurocognitive, and functional outcomes of this high-risk and vulnerable population.