key: cord-0689093-qhg78gin
authors: Falandry, Claire; Bitker, Laurent; Abraham, Paul; Subtil, Fabien; Collange, Vincent; Balança, Baptiste; Haïne, Max; Guichon, Céline; Leroy, Christophe; Simon, Marie; Malapert, Amélie; Pialat, Jean-Baptiste; Jallades, Laurent; Lepape, Alain; Friggeri, Arnaud; Thiolliere, Fabrice
title: Senior-COVID-Rea Cohort Study: A Geriatric Prediction Model of 30-day Mortality in Patients Aged over 60 Years in ICU for Severe COVID-19
date: 2022-04-01
journal: Aging Dis
DOI: 10.14336/ad.2021.1004
sha: 7f2b210f811cb9ea15144fa8ae16f465fdf89c75
doc_id: 689093
cord_uid: qhg78gin

The SARS-COV2 pandemic induces tensions on health systems and ethical dilemmas. Practitioners need help tools to define patients not candidate for ICU admission. A multicentre observational study was performed to evaluate the impact of age and geriatric parameters on 30-day mortality in patients aged ≥60 years of age. Patients or next of kin were asked to answer a phone questionnaire assessing geriatric covariates 1 month before ICU admission. Among 290 screened patients, 231 were included between March 7 and May 7, 2020. In univariate, factors associated with lower 30-day survival were: age (per 10 years increase; OR 3.43, [95%CI: 2.13-5.53]), ≥3 CIRS-G grade ≥2 comorbidities (OR 2.49 [95%CI: 1.36-4.56]), impaired ADL, (OR 4.86 [95%CI: 2.44-9.72]), impaired IADL8 (OR 6.33 [95%CI: 3.31-12.10], p<0.001), frailty according to the Fried score (OR 4.33 [95%CI: 2.03-9.24]) or the CFS ≥5 (OR 3.79 [95%CI: 1.76-8.15]), 6-month fall history (OR 3.46 [95%CI: 1.58-7.63]). The final multivariate model included age (per 10 years increase; 2.94 [95%CI:1.78-5.04], p<0.001) and impaired IADL8 (OR 5.69 [95%CI: 2.90-11.47], p<0.001)). Considered as continuous variables, the model led to an AUC of 0.78 [95% CI: 0.72, 0.85]. Age and IADL8 provide independent prognostic factors for 30-day mortality in the considered population. Considering a risk of death exceeding 80% (82.6% [95%CI: 61.2% - 95.0%]), patients aged over 80 years with at least 1 IADL impairment appear as poor candidates for ICU admission.

At the end of December 2019, the severe acute respiratory syndrome coronavirus 2 (SARSCoV-2) and the disease it causes, coronavirus disease 2019 (COVID- 19) , became rapidly an emerging pandemic and a public health challenge [1] . The older population was rapidly identified as having the highest risk for severe complications including acute respiratory distress syndrome (67-71%), acute kidney injury (20-29%), acute cardiac injury (23-33%) or liver dysfunction (15-29%) , and a 5-fold increased risk for premature death [1] [2] [3] ; this leads to a frequent theoretical indication for intensive care unit (ICU) transfer in this population. However, the first published data suggested a low benefit of ICU admission for patients aged over 70 years, and that this was almost absent in patients aged over 80 years [2, 4, 5] , but this does not take into account the heterogeneity of this population [6] and therefore the risk/benefit balance of an ICU transfer may be questioned at the individual level, independently of care resources.

Cardiovascular comorbidities [7] , laboratory covariates [8, 9] , the time since the first symptoms [5] , and ICU-specific scores [5] have been found to be associated with ICU mortality in COVID-19 patients. The identification of such variables led to the development of prognostic models; these were compiled in a systematic review by Wynants et al. and had AUC values ranging from 0.68 to 0.99, however the authors highlighted the low quality of the published studies [10] and appealed for a broader use of Transparent Reporting of a multivariable prediction model for Individual Prognosis or Diagnosis (TRIPOD) reporting guidelines [11] .

Among the available geriatric parameters, frailty is frequently used to explain the heterogeneity of the older population; it is defined as a state of increased vulnerability to poor resolution of homoeostasis after a stressor event, which increases the risk of adverse outcomes [12] . Some consensus has been reached on its definition, but there are two conceptual views as to its operational criteria: a multidomain view of frailty [13] , and a phenotypical view of frailty, linked to malnutrition and sarcopenia [14] . In an even more pragmatic view of frailty, the Clinical Frailty Score (CFS) was developed; this stratifies older patients in to distinct levels of fitness according a rapid "at a glance" assessment based on the diagnosis of a specifically trained geriatrician [15] . Interestingly, according to recent evidence from 2 studies, disease outcomes of COVID-19 patients admitted to hospital would be better predicted by frailtyusing different multidomain models of frailty assessmentsthan either age or comorbidity [16, 17] . In parallel, on March 20, 2020, the United Kingdom's National Institute for Health and Care Excellence (NICE) published a COVID-19 rapid guideline; this indicated that only patients with a CFS <5 should be considered for critical care [18] , as the threshold of 5 was previously shown to predict a higher mortality in (non-COVID) older patients admitted to ICU [19] . This guideline applies to those with the indication for ICU admission, and therefore should be considered as a non-admission decision help tool as opposed as an admission decision help tool. The geriatric community promptly reacted to these guidelines, pointing out the risk of drift in assessing the CFS, designed to be performed by trained geriatricians [20, 21] , and the ethical dilemma of transforming the frailty spectrum into a binary covariate, considering that the inter-rater variability may be high between CFS scores 4 and 5 [21] . Moreover, two studies evaluating the impact of CFS score in COVID-19 versus non-COVID-19 populations found that CFS was not a good discriminator of prognosis in COVID-19 populations [22, 23] . There is therefore still a lack of nonadmission decision help tools that may be applied for such patients. In line with that proposed by Christian et al. to deal with mass casualty events such as flu epidemics, exclusion criteria should be defined to identify patients who are not candidates for ICU admission including those: with a poor prognosis despite care in an ICU, requiring resources that cannot be provided, whose underlying illness has a poor prognosis with a high likelihood of death, and who are "too well" [24] .

To better define the individual risk/benefit ratio of ICU admission, we conducted a multicentre observational study to determine the covariates predictive of mortality in the population of patients aged ≥60 years admitted to ICU, with a specific attention paid to their geriatric parameters 1 month before ICU admission.

The study protocol was extensively described elsewhere [25] .

The primary objective was to evaluate the impact of age on 30-day mortality after ICU admission. A secondary objective was to construct a prognostic model for 30-day mortality based on co-morbidities, functional status of the patient 1 month before COVID-19 infection, laboratory data [8, 9, 26] , radiological data [27], ICU parameters, and time since the first symptoms [25] .

This analysis focused on the prognostic and discriminatory performance of the following geriatric covariates: number of grade ≥2 comorbidities according to the cumulative illness rating scale-geriatric (CIRS-G), activities of daily living (ADL) score [28] , instrumental ADL, 8 variables (IADL8) score [29] , Fried score, CFS, 6-month fall history. The CFS score ranges from 1 to 9; 1: very fit, 2: well, 3: managing well, 4: vulnerable, 5: mildly frail, 6: moderately frail, 7: severely frail, 8: very severely frail, 9: terminally ill.

The Senior-COVID-Rea study was a multicentre observational cohort study. The study protocol (V1.0 of April 7, 2020) was approved by the ethics committee of the Hospices Civils de Lyon and declared on the ClinicalTrials platform (NCT04422340). According to the patient's clinical condition, non-opposition was collected from the patient or next of kin if this was not possible, in accordance with the International Council for Harmonisation (ICH) Harmonised Guideline For Good Clinical Practice. The study follows the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement for the reporting of cohort studies [30] (Supplementary Table 1 ) and the TRIPOD guidelines for the reporting of prediction models [11] (Supplementary Table 2 ).

All patients aged ≥60 admitted to the participating ICUs with a diagnosis of COVID-19 were screened centrally using computerized medical records. Diagnostic criteria were laboratory-confirmed SARS-CoV-2-positive swabs and/or a radiological diagnosis made by lung computed tomography (CT)-scan. Patients were included when non-opposition was collected. Patients who were transferred from one ICU to another were identified and their stay was considered as a single ICU admission. Screening logs of eligible participants were retained at each site.

After inclusion, a telephone-administered questionnaire explored the functional status of the patient 1 month before ICU admission. CFS assessment was either routinely fulfilled by the physicians in charge during the course of patient care or by a geriatrician based on the analysis of the medical and functional charts.

Data were collected across 7 ICUs in the Auvergne-Rhone-Alpes Region, France, 4 in referral university hospitals and 3 in primary care hospitals. Due to territorial collaborations, patients could be transferred from one ICU to the other during their stay and duplicate were deleted. A standardized case report form was used to collect data, including: comorbidities 1 month prior to infection (CIRS-G grade ≥2 [31] ) and more specifically cardiac and vascular comorbidities (CIRS-G grade ≥2); the functional status 1 month before infection, assessed by the caregiver using the CFS and the ADL and IADL8 and 4 scores; nutritional data (weight at hospital and ICU admission, weight loss in the month and 6 months before infection, presence of mild or severe anorexia); laboratory data at ICU admission (LDH: lactate dehydrogenase, CRP: C reactive protein, and creatinine levels, as well as lymphocyte and neutrophil counts, and Sysmex haematological analyzer data [immature granulocyte count: IG;: high fluorescent lymphocyte count: HFLC; Sysmex, Kobe, Japan] (6-8)); chest imaging data (COVID-19 lung extension rated as minimal, moderate, extensive, severe, or critical according to the French Radiology Society guidelines [27]); and resuscitation parameters at ICU admission (arterial oxygen pressure/fraction of inspired oxygen [PaO2/FiO2] ratio, indice de gravité simplifié IIsimplified acute physiology score II [IGS II-SASP II] [32] ) and/or sepsis-related organ failure assessment (SOFA) score (a posteriori estimate based on IGS II-SASP II [33] ), and interval between the first signs of infection and admission to ICU.

The primary outcome was the 30-day mortality after ICU admission. No secondary outcome was considered herein.

The first hypothesis of Senior-COVID-Rea was based on the first Chinese retrospective results [2] : considering a single analysis variable (age), with expected mortality of 30% in patients <70 years of age, and 70% in patients ≥70 years of age (with 40% of patients ≥70 years of age), a total of 130 patients was expected to show a significant difference with a power of 90% (bilateral alpha risk of 5%). Since the model of survival proposed in Senior-COVID-Rea analysis considered the integration of a maximum of 15 factors, aiming at a R 2 of 0.5, and to achieve an optimism (the fact that the model accuracy is overestimated when it is assessed on the same sample as the one used to build the model) of ≤10%, 185 patients were to be included (criterion 1 of Riley and Snell [34] ). After the publication of data on mortality in ICU in the Lombardy region of Italy [4] , considering a risk of insufficient statistical power and selection bias, the study scientific committee decided that all the patients admitted to ICU before the May 7, 2020 should be screened and invited to participate in the study. This sample size calculation was modified on Clinicaltrials.gov site accordingly (July 28, 2020).

Continuous variables were described by the mean, standard deviation (SD), and range. Categorical variables were described by the frequency and percentage for each level. Commonly used thresholds were applied: CFS ≥5 [35] , ADL <6 [28] , IADL8 <8 [29] , Fried score >2 [14] .

The effect of factors on day-30 mortality risk was quantified by odds ratios (OR; with their associated 95% confidence Interval, 95% CI), the objective of the study being to take decisions for patients at a fixed-delay horizon. Factors with a p-value <0.20 in univariate were included in the multivariate analyses (logistic regression). A backward approach was used to simplify the model. The overlap between the different categorized factors was analysed using a Venn diagram. During the multivariate analyses, the collinearity between factors was analysed using variance inflation factors (VIF); with a threshold >1.5. The ability of the last model of 30-day mortality prediction was quantified by the area under the receiver operating characteristic (ROC) curve (AUC). A 5-fold cross-validation of the AUC of the model was performed to assess the optimism of this model.

No imputation of missing variables was performed. P-values <0.05 were considered significant. Analyses were performed using R software, version 3.3.2 (R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL http://www.R-project.org/). 

Between March 7 and May 7, 290 patients were screened among whom 3 were admitted to >1 ICU and therefore 287 individual patients were identified; 40 were not included for lack of availability of the teams, and 16 were excluded because of patient or next of kin refusal. A total of 231 patients were included (Supplementary Fig. 1 ). There was no significant difference in the age, sex, and 30-day mortality between those included to those not included for reason of team availability. (Table 1) . At admission, the median IGSII/SASP2 score was 39 (range 18-93), median PAO2/FiO2 ratio was 120 (range 49-532) and the median interval between the first COVID-19 symptoms and ICU admission was 9 days (range 0-41).

At day 30, 60 (26.0%) patients had died; the proportion ranged from 0 to 42.9% among the 7 ICUs (Table 1) .

In univariate analysis, the OR of death for each additional decade was 3.43 (95% CI [2.13; 5.53], p<0.001). The day-30 mortality rate in patients aged <70 years was 15%, and that in those aged ≥70 years was 31.8%. Comorbidities (grade ≥2 CIRS-G) were significantly associated with death at 30-days (OR: 2.48, 95% CI [1.36; 4 .56]), as were all geriatric factors tested; the point estimates of which (OR) ranged from 3.46 (6 months fall history) to 6.33 (IADL8 score <8) ( Table 2 ). There was a high overlap between IADL8 <8, ADL <6, Fried score >2, and CFS ≥5 on Venn diagram (Supplementary Fig. 2) . The different multivariate models built are presented in Table 3 . In the initial model including all variables significant in univariate analysis, the VIF associated with ADL, IADL, Fried scores, and CFS suggested collinearity between these factors. Table 3 ). No significant interaction was found between age and IADL8 (p=0.827). The results of these analyses were independent from the thresholds chosen for the different studied factors; in fact, considering continuous factors (instead of categorized factors) led to the same final model (Supplementary Table 3 )."

The logistic model including the age and IADL8 score as continuous variables led to an AUC of 0.78 [95% CI: 0.72, 0.85] (Fig. 1) 

In the present study age and IADL8 provided the best multivariate prediction model of 30-day mortality, with a good discriminative ability. These are highly objective parameters and poorly sensitive to inter-rater variation and provide an easy-to-use tool. Accordingly, patients aged <80 years with no IADL impairment may benefit from ICU stay, and whether it is fair to admit COVID-19 patients aged ≥80 years with ≥1 IADL impairment is questionable. Conversely, it is less straight forward for the remaining patients. It is of note that the impact of functional impairment appears from 60 years of age; this may be considered as surprising since patients aged between 60 and 70 years are generally considered fit for ICU admission independently of any functional assessment, but are in line with previous reports that indicate that the first increase in mortality rate Falandry C., et al.

Aging and Disease • Volume 13, Number 2, April 2022 is found in patients aged ≥60 years [36] . For those aged ≥80 years without any IADL8 impairment, ICU admission is associated with the same benefit as younger patients with functional impairment and should therefore not be excluded based only on their age. Taken together, the results indicate that, in addition to age, IADL8 provides additional information to decide on ICU admission but there remains relative uncertainty for the patients of intermediary risk.

In such contexts and again according Christian et al, there is a need to provide a process and structure to mitigate chaos and improve the effectiveness of actions taken and triage criteria should be objective, ethical, transparent, applied equitably and publicly disclosed Interestingly, IADL8 was significantly associated with 30-day in multivariate analysis while frailty scales were not. This absence of significant association is in relative contradiction with the literature highlighting the relationship between COVID-19 mortality and frailty [16, 17] , and may be explained by several points. The most important is that the previously published study on the impact of CFS on ICU mortality included more than 6fold more patients identified in a vast multicenter database [16] , leading to a greater statistical power. In addition, as the present study included only those admitted to ICU, frail patients were poorly represented impairing de facto discrimination properties of frailty scores, and some discrepancies may lay between investigators when assessing the CFS [20, 21] leading to potential evaluation bias. However, these results are in line with previous data reported by Miles et al. , who questioned the real impact of frailty facing COVID-19-infected patients, compared to non-COVID-19 patients [22] . Similarly, in their description of CFS in 1071 hospitalized older adults, Owen et al. demonstrated that COVID-19 infection was the main risk factor for death in the population, CFS appearing to make little incremental contribution to the hazard of dying in older people hospitalized with COVID-19 [23] . Another point to highlight is the absence of clearcut discriminative threshold for CFS [16] , as confirmed by the data presented herein, leading to question the arbitrary threshold of 5 proposed by the NICE [18] since the clinical distinction between CFS 4 and 5 may be considered as subjective [21] . The IADL8 score provided, in the present study, a good discriminative score, which is concordant with previous studies investigating ICU outcomes in older patients [37, 38] . For instance, according to Chelluri et al., the prehospitalization functional status (evaluated using IADL score) was significantly associated with short-term mortality, whereas age and comorbidities were associated with longterm mortality [37] . More recently, Giannotti et al. reported that, in the context of elective gastrointestinal oncogeriatric surgery, pre-morbid functional status (IADL) and cancer stage were the most significant predictors of one-year mortality, after having considered Frailty Index (FI) in the construction of the multivariate model [38] . Moreover, IADL integrates cognitive susceptibility, in addition to pure physical performance as demonstrated by Shimada et al. [39] , and cognitive impairment is reported to be associated with short-term mortality (in-hospital death) after ICU admission [40] . A strength of the present cohort study lays in the low rate of missing data in the a posteriori assessment of geriatric covariates of the patients 1 month before their admission to ICU. One of the limitations of this study may lie in the heterogeneity in the management of older patients during their stay in ICU; for instance it is widely documented that geriatric vulnerability may induce medical limitations during ICU stay, impacting mortality [41] [42] [43] . Another limitation is the change in the management of COVID-19 patients in ICU that have occurred since the first wave of the pandemic, and which may change the crude mortality risks in each risk category. In addition, differences in ICU admission criteria and care during ICU stay between the hospitals have also to be acknowledged.

To conclude, in addition to age, IADL8 provides information for the orientation of COVID-19 patients aged ≥60 years when discussing ICU admission. This tool could be of value as a non-admission decision help tool.

All the authors declare grants from Hospices Civils de Lyon, during the conduct of the study, no other competing interests with the considered topic.

CF reports outside of the present work consulting/advisory roles for GSK, Leo Pharma, Pfizer, MSD Oncology, Teva, AstraZeneca, Baxter, Eisai, Janssen, and Novartis; research funding from Chugai Pharma, Pfizer, Pierre Fabre, and Astellas Pharma; and travel/accommodation/expenses from Janssen Oncology, Pierre Fabre, and Leo Pharma. The Hospices Civils de Lyon gave funding for Clinical Research Assistants, statistical analysis and manuscript editing.

The funder of the study had no role in study design, data collection, data analysis, data interpretation. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication. The corresponding author drafted the manuscript. All authors critically reviewed the manuscript and approved the final version for submission.

Estimating clinical severity of COVID-19 from the transmission dynamics in Wuhan, China

Clinical course and outcomes of critically ill patients with SARS-CoV-2 pneumonia in Wuhan, China: a single-centered, retrospective, observational study

Characteristics and Outcomes of 21 Critically Ill Patients With COVID-19 in Washington State

Baseline Characteristics and Outcomes of 1591 Patients Infected With SARS-CoV-2 Admitted to ICUs of the Lombardy Region

Clinical characteristics and day-90 outcomes of 4244 critically ill adults with COVID-19: a prospective cohort study

Geriatric Syndromes and SARS-COV-2: More than Just Being Old

Risk Factors for Intensive Care Unit Admission and In-hospital Mortality Among Hospitalized Adults Identified through the US Coronavirus Disease

Lactate dehydrogenase, a Risk Factor of Severe COVID-19 Patients

D-dimer as an indicator of prognosis in SARS-CoV-2 infection: a systematic review

Prediction models for diagnosis and prognosis of covid-19 infection: systematic review and critical appraisal

Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD)

Frailty in elderly people

A comparison of two approaches to measuring frailty in elderly people

Frailty in older adults: evidence for a phenotype

A global clinical measure of fitness and frailty in elderly people

The effect of frailty on survival in patients with COVID-19 (COPE): a multicentre, European, observational cohort study

Frailty index predicts poor outcome in COVID-19 patients

Admission to critical care | COVID-19 rapid guideline: critical care in adults | Guidance | NICE

The contribution of frailty, cognition, activity of daily life and comorbidities on outcome in acutely admitted patients over 80 years in European ICUs: the VIP2 study

COVID-19 and the Challenges of Frailty Screening in Older Adults

Decision-Making in COVID-19 and Frailty

Outcomes from COVID-19 across the range of frailty: excess mortality in fitter older people

Comparing associations between frailty and mortality in hospitalised older adults with or without COVID-19 infection: a retrospective observational study using electronic health records

Prognostic Criteria Associated With Survival in Patients Over 60 Admitted to ICU for Severe COVID Infection: the Senior-COVID-Rea Multicentric Survey protocol

Metabolic disturbances and inflammatory dysfunction predict severity of coronavirus disease 2019 (COVID-19): a retrospective study

Studies of Illness in the Aged. The Index of Adl: A Standardized Measure of Biological and Psychosocial Function

Assessment of older people: self-maintaining and instrumental activities of daily living

Observational Studies: Getting Clear about Transparency

CUMULATIVE ILLNESS RATING SCALE

A new Simplified Acute Physiology Score (SAPS II) based on a European/North American multicenter study

The SOFA (Sepsis-related Organ Failure Assessment) score to describe organ dysfunction/failure. On behalf of the Working Group on Sepsis-Related Problems of the European Society of Intensive Care Medicine

Minimum sample size for developing a multivariable prediction model: PART II -binary and time-to-event outcomes

The impact of frailty on intensive care unit outcomes: a systematic review and meta-analysis

Development and validation of a 30-day mortality index based on preexisting medical administrative data from 13,323 COVID-19 patients: The Veterans Health Administration COVID-19 (VACO) Index

Long-term mortality and quality of life after prolonged mechanical ventilation

Frailty assessment in elective gastrointestinal oncogeriatric surgery: Predictors of one-year mortality and functional status

Impact of Cognitive Frailty on Daily Activities in Older Persons

Variability of intensive care admission decisions for the very elderly

Frailty in Critical Care Medicine: A Review

Retrospective frailty determination in critical illness from a review of the intensive care unit clinical record

Withholding or withdrawing of life-sustaining therapy in older adults (≥ 80 years) admitted to the intensive care unit

The authors thank the patients and their families, the centres that participated in this trial, the Hospices Civils de Lyon Direction for Research and Innovation, the clinical research assistants in all investigations centres (Lyon Sud Hospital, Croix Rousse Hospital, Edouard Herriot Hospital and Lyon Est Hospital from the Hospices Civils de Lyon, Lyon-Villeurbanne Mé dipôle Nord-Ouest Villefranche-sur-Saône Hospital and Emile Roux Hospital of Le Puy en Velay) who contributed to the design, building and accrual of the study, the Hospices Civils de Lyon for granting and sponsoring the trial (Grant N°2020-05-12), Philip Robinson (DRCI, Hospices Civils de Lyon) for manuscript editing and Thomas Gilbert for a friendly reviewing of the manuscript. 

Aging and Disease • Volume 13, Number 2, April 2022

The Supplementary data can be found online at: www.aginganddisease.org/EN/10.14336/AD.2021.1004.