Intensive Care Med (2020) 46:57–69 https://doi.org/10.1007/s00134-019-05853-1 ORIGINAL 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 Bertrand Guidet1* , Dylan W. de Lange2, Ariane Boumendil3, Susannah Leaver4, Ximena Watson5, Carol Boulanger6, Wojciech Szczeklik7, Antonio Artigas8, Alessandro Morandi9, Finn Andersen10, Tilemachos Zafeiridis11, Christian Jung12, Rui Moreno13, Sten Walther14, Sandra Oeyen15, Joerg C. Schefold16, Maurizio Cecconi17,18, Brian Marsh19, Michael Joannidis20, Yuriy Nalapko21, Muhammed Elhadi22, Jesper Fjølner23, Hans Flaatten24,25 for the VIP2 study group © 2019 Springer-Verlag GmbH Germany, part of Springer Nature Abstract Purpose: Premorbid conditions affect prognosis of acutely-ill aged patients. Several lines of evidence suggest geriat- ric syndromes need to be assessed but little is known on their relative effect on the 30-day survival after ICU admis- sion. The primary aim of this study was to describe the prevalence of frailty, cognition decline and activity of daily life in addition to the presence of comorbidity and polypharmacy and to assess their influence on 30-day survival. Methods: Prospective cohort study with 242 ICUs from 22 countries. Patients 80 years or above acutely admitted over a six months period to an ICU between May 2018 and May 2019 were included. In addition to common patients’ characteristics and disease severity, we collected information on specific geriatric syndromes as potential predic- tive factors for 30-day survival, frailty (Clinical Frailty scale) with a CFS > 4 defining frail patients, cognitive impairment (informant questionnaire on cognitive decline in the elderly (IQCODE) with IQCODE ≥ 3.5 defining cognitive decline, and disability (measured the activity of daily life with the Katz index) with ADL ≤ 4 defining disability. A Principal Com- ponent Analysis to identify co-linearity between geriatric syndromes was performed and from this a multivariable model was built with all geriatric information or only one: CFS, IQCODE or ADL. Akaike’s information criterion across imputations was used to evaluate the goodness of fit of our models. Results: We included 3920 patients with a median age of 84 years (IQR: 81–87), 53.3% males). 80% received at least one organ support. The median ICU length of stay was 3.88 days (IQR: 1.83–8). The ICU and 30-day survival were 72.5% and 61.2% respectively. The geriatric conditions were median (IQR): CFS: 4 (3–6); IQCODE: 3.19 (3–3.69); ADL: 6 (4–6); Comorbidity and Polypharmacy score (CPS): 10 (7–14). CFS, ADL and IQCODE were closely correlated. The multivari- able analysis identified predictors of 1-month mortality (HR; 95% CI): Age (per 1 year increase): 1.02 (1.–1.03, p = 0.01), ICU admission diagnosis, sequential organ failure assessment score (SOFA) (per point): 1.15 (1.14–1.17, p < 0.0001) and *Correspondence: bertrand.guidet@aphp.fr 1 Sorbonne Université, INSERM, Institut Pierre Louis d’Epidémiologie et de Santé Publique, Saint Antoine Hospital, AP-HP, Hôpital Saint-Antoine, service de réanimation, F75012 Paris, France Full author information is available at the end of the article 58 CFS (per point): 1.1 (1.05–1.15, p < 0.001). CFS remained an independent factor after inclusion of life-sustaining treat- ment limitation in the model. Conclusion: We confirm that frailty assessment using the CFS is able to predict short-term mortality in elderly patients admitted to ICU. Other geriatric syndromes do not add improvement to the prediction model. Since CFS is easy to measure, it should be routinely collected for all elderly ICU patients in particular in connection to advance care plans, and should be used in decision making. Keywords: Critical care, Outcome, Prediction, Elderly, Frailty, Cognitive functioning, Activities of daily living, Comorbidity Introduction Take‑home message In this prospective multinational study of 3920 very old intensive The number of very old patients admitted to the intensive care patients (≥ 80 years), clinical frailty scale alone described geriat- care unit (ICU) is rising, which parallels the increase in ric syndromes. Cognition and activity of daily life did not add to the the proportion of elderly people in most developed coun- prediction model for 1-month survival after ICU admission. tries [1]. In the past decade this has been identified as a potential health challenge because these patients aged 80 and above consume a disproportionally large part of Methods the health care budget, while the mortality rates are con- Design and setting sistently reported to be higher than in younger patients This was a prospective observational study in 242 ICUs [2]. However, it is important to separate unplanned from from 20 European countries, plus Turkey and Libya. The planned ICU admissions, since the latter group has a participating ICU characteristics were: mean number of much better outcome [3]. In unplanned admissions the beds: 15; mean number of ICU admission in year 2017: overall 30-day mortality of ICU patients is approximately 955; percentage of patients 80 years of age or older admit- 40% [4]. Long-term mortality (6 months or 12 months) is ted in 2017: 17%. Each participating ICU is listed in elec- most often 50% or above [5]. This high mortality trans- tronic supplement material (ESM1) and recruitment per lates into suffering in patients and relatives and moral country is presented in ESM2. The inclusion period was distress in care-givers. It is, therefore, of utmost impor- from May 2018 to May 2019. Most of the patients were tance to better identify elderly patients with an expected included during the 2018–2019 winter season (ESM3). good outcome including a post discharge acceptable The 30-day follow-up ended on July 15 2019. quality of life. The study was coordinated through the health services Previous research has shown that age and severity resource and outcome (HSRO) section of the European of illness at admission only partly explain the chance society of intensive care medicine (ESICM). Each coun- of survival of elderly patients and traditional prognos- try had a national coordinator (NC) responsible for ICU tic scoring systems are often inaccurate [6]. In the very recruitment and application for national or regional ethi- old patients the ability to cope with severe stressors like cal and regulatory study approval. Institutional research critical illness seems to be more related with the geriatric ethic board approval was obtained from each study site. syndromes, like frailty [4], cognitive decline [7, 8] and a Five countries were allowed to recruit patients without reduced performance on the activity of daily life (ADL) prior written informed consent while others had to col- scale in addition to comorbidity. However, the impact lect informed consent from patient or their legal rep- and relationship of this premorbid functioning on the resentative [9]. Individual ICUs were asked to include outcome of acutely admitted elderly ICU patients have consecutive patients for a 6-month period in the 1-year not been established [7, 8]. study period and were allowed to stop after 20 included In this study, using the European Very elderly Intensive patients. All patients were followed until 30 days after Patient (VIP) network, our primary aim was to document ICU admission. A dedicated web site (www.VIP2s​tudy. the prevalence of three common geriatric syndromes: com) was set up to facilitate information about the frailty, cognitive impairment and disability in addition to study and study progress and to allow for data entry the presence of comorbidity and polypharmacy, and to using an electronic case record form CRF (supplemen- assess their influence on 30 days’ survival. tal 1). The study is registered on ClinicalTrials.gov (ID: NCT03370692). There was no specific funding except an ESICM award received by HF. 59 Participants day of death after ICU admission. Information source for Only emergency (acute) ICU admissions in vital status at day 30 was documented. patients ≥ 80 years of age were eligible. All reasons for The Case Record Form and database were hosted on a acute ICU admissions were accepted. secure server located on the campus of Aarhus Univer- sity, Denmark. Data collection Data collection at admission Bias For each eligible patient, demographic data were col- The ICUs were asked to include all acutely admitted, con- lected: age, sex, place of living before admission to the secutive patients irrespective of the estimated duration of hospital, and reason for admission according to a prede- ICU stay. fined list (ESM4). Second, the study collected mandatory data on the Study size patient’s geriatric conditions prior to this hospital admis- We had no formal calculation in this purely observational sion, including the clinical frailty scale (CFS) [10]. For the study. We estimated a 30-day mortality of 40% in a simi- assessment of frailty, we defined the frailty level present lar population [17]. before hospital admission and not affected by the acute illness. The information necessary to perform this assess- Statistical analysis ment were given by patients or proxy or patient records, Baseline characteristics of patients were analyzed as fre- and the assessor’s profession was reported. The Clinical quencies and percentages for categorical variables and as Frailty Scale visual and with simple description were used medians and interquartile ranges (IQRs) for continuous with permission (ESM5) [10]. The scale is composed of variables. 9 classes from very fit to terminally ill. Prefrail patients Principal Component Analysis (PCA) was used to have a CFS of 4, while frail patients have a CFS of 5 or represent the variation present in the dataset and show above. We also recorded Katz activities of daily living association between geriatric and severity variables (age, (Katz ADL) (ESM6) [11] with ADL score ≤ 4 defining SOFA score, CPS, IQ code, Katz’s scale and clinical frailty disability. Short form of Informant Questionnaire on scale). For visual analysis, two-dimensional projection of Cognitive Decline in the Elderly (IQCODE) [12] (ESM7) the sample was constructed having the axes (principal was used. The information to calculate the IQCODE components, PC) as the factors. Each PC is a linear com- was given by caregivers with intimate knowledge of bination of the original variables (that retain some corre- the patient for the past 10 years. We defined cognitive lation among each other) and PCs are orthogonal to each decline as an IQCODE ≥ 3.5 [13]. Comorbidity and poly- other. pharmacy score (CPS) were calculated [14]. A CPS above The crude overall survival at 30 days was estimated by 15 was considered as high. the Kaplan–Meier method and compared using a log- rank test. We used multiple imputations for participants Data collection of variables during ICU admission with missing data, using predictive mean matching for Within the first 24 h of ICU admission the sequential continuous variables, logistic regression for binary data, organ failure assessment (SOFA) score was calculated and polynomial regression for (unordered) categorical [15]. The total SOFA score on admission was calculated data. The cumulative baseline hazard was approximated using an online calculator in the eCRF. Length of ICU by the Nelson-Aalen estimator and included in the impu- stay (LOS) was recorded as the number of hours between tation model. Five imputations were drawn. admission and discharge and latter converted to days in Multivariable analyses were performed using the the analysis. Any period on non-invasive, invasive (with Cox proportional hazards model for overall survival at endotracheal intubation or tracheotomy) mechanical 30 days. All patients’ characteristics listed in the CRF ventilation, use of vasoactive drugs and renal replace- were included in our models including LST limitation. ment therapy was registered with start (ICU day) and We used mean Akaike’s information criterion (AIC) duration of the treatment (in hours). across imputations to evaluate the goodness of fit of Limitation of life sustaining therapies (LST) such as our models. AIC = − 2 Log likelihood + 2p, where p is withholding or withdrawing life-supporting treatments the number of parameters [18]. Models with lower AIC was documented as performed and on which ICU day, or are interpreted as models that best fit the data. Last, we if not performed [16]. included limitation decisions as a time-dependent vari- Outcome was measured as survival in the ICU and able to test whether geriatric parameters kept their prog- 30 days after ICU admission with information on the nostic value. Hazard ratio (HR) given by the Cox model 60 describe the relative risk of death based on comparison individuals. The weighted Kappa was 0.85 (95%CI 0.84– of event rates in different groups. HR gives the change in 0.87) which is excellent (ESM12). risk of death per each unit increase for continuous varia- Frailty (CFS ≥ 5) was present in 40.2% (1568/3903). bles and for one specific category versus a reference cate- IQCODE was 3.19 (3–3.69) with 940/3920 (24%) miss- gory for categorical variables. HR > 1 suggests an increase ing values. Patients with missing values of IQCODE in the risk of death, HR < 1 suggest a decrease in the risk were different from patients with documented IQCODE of death. A Cox model was used to analyze treatment (ESM123). IQCODE above 3.5 indicating a cognitive limitation as a time to event variable. decline was observed in 30.2%. Katz ADL was 6 (4–6) Adjusted survival curves were produced using an with 447/3920 (11.4%) missing values. An ADL of 4 or inverse probability-weighted Kaplan–Meier estima- less, as a marker of functional decline was observed in tion [19]. Significance was tested using a Cox regression 27.7% (962/3473). Comorbidity and polypharmacy score model weighted by the same weights (inverse probabil- (CPS) was 10 (7–14) with 8/3920 (0.2%) missing values. ity-weighted Cox). Inter-rater variability for assessing A CPS above 15 was present in 17.7% (693/3912) of the Clinical Frailty Scale was analyzed with Med-Calc for patients. Windows 15.0 (MedCalc Software, Ostend, Belgium) and presented as Weighted kappa [20]. All analyses were per- formed by AB with R software, version 3.2.2 (R Founda- Outcome data tion for Statistical Computing) (packages FactorMineR LST limitations were documented in 34.4% (1332/3872) and missMDA for PCA, MICE for multiple imputations, including withholding in 29.4% (1140/3872) of patients survival for survival analysis, ipw for inverse-probability with a decision made at day 1 (IQR 1–4) and treatment weighting, rms for adjusted survival curves). withdrawal (with or without prior withholding) in 14% (545/3872) with a median decision time of 4 days (IQR Results 2–7) after ICU admission. Patients with LST limitation Participants were frailer and more severely ill compared to patients We included 3920 patients (median age 84 years) from with no LST limitation (Table 4). The multivariable anal- 22 countries and 242 ICUs. Most patients lived at home ysis identified age, habitat, SOFA score, CPS and CFS without help (72.8%,) or with care-givers (11.1%) before as independent factors predicting decision to limit LST being admitted to the hospital. Admissions related to res- (Table 5). piratory and/or circulatory failures and sepsis (as defined The ICU and 30-day vital status were, respectively, by sepsis 3 criteria) accounted for 63.2% of the total available for 3900/3920 (99.9%) and 3841/3920 (99.4%) while emergency surgery represented 13.8% of the total patients. The vital status was assessed by checking the admissions. The patients were severely ill as evidenced hospital records (81.4%, 2780/3415), direct contact with by a median SOFA score of 6 (IQR 4–9) (Table 1). Most the patient (12.3%, 420/3415) or checking a national reg- patients had at least one organ support (Table 2). The istry (3.3%, 111/3415). median ICU LOS was 3.9 days (IQR: 1.83–8). The overall proportion of patients surviving ICU was No difference was found in patients’ characteristics 72.5% (2828/3900). Survival after ICU admission was in ICUs with a recruitment period above vs. below the 96.3% at day one (95.7–96.9), 74.2% (72.8–75.6) at day 10, median of 60 days (ESM8) ruling out possible bias due 65.9% (64.5–67.5) at day 20 and 61.2% (59.7–62.7) at day to different recruitment times. Recruitment per ICU was 30. excellent (ESM9) but patients included in countries with no need for informed consent were sicker with a SOFA Predictive factors for 30‑day survival score almost two points higher (ESM10). In our PCA analysis the first two PC explain as much The geriatric syndromes (Tables 1, 3) were as fol- as 56.79% of the total variation in the sample; the anal- lows: Median (IQR): CFS: 4 (3–6) with only 17 miss- ysis was restricted to these first two dimensions. PC1 ing values. Information to assess CFS was given by the retained 39% of data variation and differentiates the indi- patient in 29.7% (1159/3901) and by the family in 51.7% viduals according to Katz ADL, Clinical Frailty Scale and (2016/3901). The professional assessor was an ICU physi- IQCODE. Similarly, PC2 explained another 18% of vari- cian (59.8%, 2335/3901), an ICU nurse (13.7%, 535/3901) ability in the original data and separates the individuals or research staff (23.7%, 925/3901). CPS values were based on severity (SOFA) and age. Figure 1 represents similar when measured by different assessors (ESM11). the projection of variables on the factor-plane and shows The CFS was measured by two different raters in 1924 correlation between the three geriatric syndromes IQ code, Katz ADL and Clinical Frailty Scale. Except for 61 Table 1 Patients characteristics Variables Categories All Survivors Non survivors p-value N 3920 2327 1514 Age Median (range) (IQR) 84 (range: 80–104) (81–87) 84 (range 80–104) (81–87) 84 (range 80–98) (82–87) 0.04 Gender Male 2089 (53.3%) 1210 (52%) 836 (55.2%) 0.06 Female 1831 (46.7%) 1117 (48%) 678 (44.8%) Place of living Own home (including if with 2852 (72.8%) 1774 (76.3%) 1031 (68.1%) < 0.0001 spouse) Other home with family or 434 (11.1%) 228 (9.8%) 191 (12.6%) caregivers Nursing home 210 (5.4%) 112 (4.8%) 96 (6.3%) Hospital ward 361 (9.2%) 179 (7.7%) 174 (11.5%) Other 35 (0.9%) 22 (0.9%) 11 (0.7%) Unknown 27 (0.7%) 11 (0.5%) 11 (0.7%) Reason for admission Respiratory failure 944 (24.1%) 583 (25.1%) 343 (22.7%) < 0.0001 Circulatory failure 541 (13.8%) 303 (13%) 228 (15.1%) Combined respiratory/circulatory 449 (11.5%) 186 (8%) 253 (16.7%) failure Sepsis (according to Sepsis3) 539 (13.8%) 287 (12.3%) 239 (15.8%) Multitrauma without head injury 87 (2.2%) 58 (2.5%) 27 (1.8%) Multitrauma with head injury 75 (1.9%) 35 (1.5%) 37 (2.4%) Isolated head injury 71 (1.8%) 33 (1.4%) 38 (2.5%) Intoxication 23 (0.6%) 17 (0.7%) 6 (0.4%) Non-traumatic cerebral pathology 190 (4.8%) 103 (4.4%) 84 (5.5%) Emergency surgery 541 (13.8%) 369 (15.9%) 163 (10.8%) Other causes 459 (11.7%) 352 (15.1%) 96 (6.3%) SOFA Median (range) (IQR) 6 (range: 0–20) (4–9) 5 (range 0–18) (3–8) 8 (range 0–20) (5–11) < 0.0001 CFS Median (range) (IQR) 4 (range: 1–9) (3–6) 4 (range 1–9) (3–5) 4 (range 1–9) (3-6) < 0.0001 IQ Code Median (range) (IQR) 3.19 (range: 1–5) (3–3.69) 3.19 (range 1–5) (3–3.56) 3.31 (range 1–5) (3–4) < 0.0001 ADL (Katz) Median (range) (IQR) 6 (range: 0–6) (4–6) 6 (range 0–6) (5–6) 6 (range 0–6) (3–6) < 0.0001 Comorbidities Median (range) (IQR) 4 (range: 0–20) (3–6) 4 (range 0–20) (3–5) 4 (range 0–20) (3–6) 0.013 Drugs taken daily Median (range) (IQR) 6 (range: 0–26) (4–9) 6 (range 0–21) (4–9) 6 (range 0–26) (4–9) 0.17 CPS Median (range) (IQR) 10 (range: 0–36) (7–14) 10 (range 0–36) (7–14) 11 (range 0–35) (7–14) 0.05 Vital status was assessed 1 month after ICU admission. 79 patients had no vital status reported at 1 month SOFA sequential organ failure assessment, CFS clinical frailty scale, IQCODE informant questionnaire on cognitive decline in the elderly, ADL activity of daily living, CPS Co-morbidity and Polypharmacy score CPS, each geriatric syndrome had impact on 1-month in the risk of death per each unit increase of CFS of 1.08 survival (Fig. 2). (1.05–1.12, p < 0.001) (Table 7). The multivariable analysis identified the usual predic- tors of 30-day outcome (HR; 95% CI): Age (increase in Discussion risk of death per 1 year increase): 1.02 (1–1.03, p = 0.01), In the ICU community there is a growing need for evi- ICU admission diagnosis, SOFA (increase in risk of death dence and guidelines to ease the decision-making process per one-point increase): 1.15 (1.14–1.17, p < 0.0001) and for an informed ICU admission in the very old including CFS (increase in risk of death per one point increase): 1.1 ICU triage, and this should be supported by best available (1.05–1.15, p < 0.001). The model including all geriatric evidence [21]. In that perspective, this large prospective parameters did not perform better than the model with study of non-scheduled, acute ICU admissions provides CFS only (Table 6). The CPS was not independently asso- important information. ciated with 30-day outcome. When including limitation We confirmed that 30-day survival is very low (61.2%) decision as a time-dependent variable in the model, CFS in unselected patients 80 years or older who are acutely still has a prognostic impact on survival with an increase admitted to the ICU. Age-related symptoms as frailty, 62 Table 2 ICU stay characteristics Variables Categories All Survivors Non survivors p-value N 3920 2327 1514 Intubation Yes 1953 (49.9%) 939 (40.4%) 974 (64.5%) < 0.0001 Tracheostomy Yes 264 (6.8%) 168 (7.2%) 90 (6%) 0.14 Non invasive ventilation Yes 903 (23.1%) 547 (23.6%) 342 (22.6%) 0.51 Vaso active drugs Yes 2329 (59.5%) 1169 (50.3%) 1114 (73.6%) < 0.0001 Renal replacement therapy Yes 429 (11%) 163 (7%) 259 (17.1%) < 0.0001 Withholding of treatment Yes 1140 (29.4%) 390 (16.9%) 738 (49.2%) < 0.0001 Withdrawing of treatment Yes 545 (14%) 27 (1.2%) 517 (34.4%) < 0.0001 ICU LOS in days Median (range) (IQR) 3.88 (0.04-30) (1.83-8) 3.67 (0.04-30) (1.92–7.71) 4 (0.04-30) (1.54–8.46) 0.75 ICU survival Missing 13 10 3 Alive 2828 (72.5%) 2285 (98.6%) 471 (31.2%) < 0.0001 30 days’ survival Missing 79 Alive 2346 (61.1%) Day of death Median (range) (IQR) 7 (0–30)(3–14) Non-survivors are defined by 30-day mortality Table 3 Scores assessed as categories Variables Categories All Survivors Non survivors p-value CFS Fit (CFS 1-3) 1544 (39.6%) 1011 (43.5%) 509 (33.9%) < 0.0001 Vulnerable (CFS 4) 791 (20.3%) 483 (20.8%) 287 (19.1%) Frail (CFS 5-8) 1568 (40.2%) 830 (35.7%) 704 (46.9%) IQCODE categories <3 178 (6%) 125 (6.3%) 48 (5.1%) < 0.0001 3 709 (23.8%) 483 (24.5%) 212 (22.3%) > 3–3.5 1194 (40.1%) 859 (43.6%) 316 (33.3%) > 3.5 899 (30.2%) 505 (25.6%) 373 (39.3%) CPS categories 0–9 1697 (43.4%) 1024 (44%) 644 (42.6%) 0.12 10–15 1522 (38.9%) 913 (39.3%) 575 (38.1%) > 15 693 (17.7%) 389 (16.7%) 292 (19.3%) KATZ categories 0–4 962 (27.7%) 500 (22.6%) 441 (36.8%) < 0.0001 5 440 (12.7%) 295 (13.3%) 135 (11.3%) 6 2071 (59.6%) 1417 (64.1%) 623 (52%) SOFA categories SOFA 0–5 1741 (44.5%) 1304 (56.2%) 410 (27.1%) < 0.0001 6–9 1286 (32.9%) 699 (30.1%) 555 (36.7%) > 10 882 (22.6%) 317 (13.7%) 547 (36.2%) cognitive decline and disability were strongly associated ICU patients [4] and was closely related to both ICU and with 30-day mortality and were more important than age 30-day mortality. In the present study, frailty remained alone. Comorbidity and polypharmacy, however, had no independently associated with both ICU and 30 days predictive value when measured as the CPS. Interest- mortality, and this strengthens its importance in pre- ingly, the addition of cognitive impairment and disabil- ICU assessment. Frailty affects several important body ity to the prognostic analysis did not increase prognostic functions like the endocrine system, brain, muscular and value compared with the model with frailty alone. The immune function, all very important in the ICU context, CFS was reported in 99.5% of the cases and the inter- and was recently discussed in a clinical review [22]. The rater reproducibility was very good confirming that this link between frailty and health-related quality of life was tool is robust and easy to use. documented in a prospective Canadian study [23]. Frail In a previous prospective study in Europe, frailty was patients were found to report reduced quality of life both found to be present in 46% of acute admitted elderly in the physical and mental subset of EuroQol Health 63 Table 4 Patient characteristics according to decision to limit life sustaining treatment: withholding and/or withdrawing Variables Categories All No limitation Withdraw and/or withhold p-value N 3920 2540 1332 Age Median (range) (IQR) 84 (range: 80–104) (IQR: 83 (range 80–100) (IQR 81–86) 84 (range 80–99) (IQR 82-87) < 0.0001 81–87) SOFA Missing 11 3 6 Median (range) (IQR) 6 (range: 0–20) (IQR: 4–9) 6 (range 0–20) (IQR 3–9) 7 (range 0–19) (IQR 5–10) < 0.0001 CFS Missing 17 7 9 Median (range) (IQR) 4 (range: 1–9) (IQR: 3–6) 4 (range 1–9) (IQR 3–5) 4 (range 1–9) (IQR 3–6) < 0.0001 IQ Code Missing 940 504 416 Median (range) (IQR) 3.19 (range: 1–5) (IQR: 3–3.69) 3.19 (range 1–5) (IQR 3–3.57) 3.31 (range 1–5) (IQR < 0.0001 3.06–3.88) Katz Missing 447 219 215 Median (range) (IQR) 6 (range: 0–6) (IQR: 4–6) 6 (range 0–6) (IQR 5–6) 6 (range 0–6) (IQR 4–6) < 0.0001 Comorbidities Missing 7 1 2 Median (range) (IQR) 4 (range: 0–20) (IQR: 3–6) 4 (range 0–20) (IQR 3–5) 4 (range 0–20) (IQR 3–6) 0.005 Drugs daily Missing 8 1 2 Median (range) (IQR) 6 (range: 0–26) (IQR: 4–9) 6 (range 0–21) (IQR 4–8) 7 (range 0–26) (IQR 4–9) < 0.0001 CPS Missing 8 1 2 Median (range) (IQR) 10 (range: 0–36) (IQR: 7–14) 10 (range 0–36) (IQR 7–14) 11 (range 0–35) (IQR 8–5) < 0.0001 ICU LOS in days (all Missing 11 3 1 patients alive/ Median (range) (IQR) 3.88 (range: 0.04–120) (IQR: 3.96 (range 0.04–120) (IQR 3.67 (range 0.04–85.5) (IQR 0.03 dead) 1.83–8) 1.96–8) 1.54–7.85) N = 48 patients with missing information on LST limitation (WH/WD) Table 5 Multivariate analysis of treatment limitation (Cox model analyzing treatment limitation as a time to event vari‑ able) Variables Categories HR (95%CI) p-value Age One-point increase 1.04 (95% CI 1.03–1.06) < 0.00001 Habitat (ref = not home) Own home 1.27 (95% CI 1.12–1.45) 0.0004 Gender (ref = male) Female 1.08 (95% CI 0.97–1.21) 0.17 Reason for admission (ref = circulatory Combined respiratory/circulatory failure 0.9 (95% CI 0.7–1.11) 0.32 failure) Emergency surgery 0.67 (95% CI 0.5–0.83) 0.0003 Multitrauma with head injury 0.92 (95% CI 0.–1.41) 0.70 Multitrauma without head injury 0.77 (95% CI 0.–1.18) 0.23 Other 0.8 (95% CI 0.6–0.97) 0.02 Respiratory failure 1 (95% CI 0.8–1.19) 0.96 Sepsis (according to Sepsis3) 0.92 (95% CI 0.75–1.12) 0.40 SOFA One-point increase 1.07 (95% CI 1.05–1.09) < 0.00001 CPS One-point increase 1.01 (95% CI 1–1.02) 0.009 CFS One-point increase 1.11 (95% CI 1.08–1.15) < 0.00001 Questionnaire and in the SF-12. The use of the CFS has required to investigate the possible link between these probably expanded lately, and different acute care settings two factors on ICU survival [27]. find it useful to evaluate elderly patients in such a quick Cognitive decline has been found to affect mortality manner, even without patient contribution [24–26], mak- in many clinical settings [28, 29] and several methods ing it of particular interest in the unresponsive or uncon- are used to diagnose and assess the severity of cognitive scious patient. There is also an emerging line of research decline and dementia. Most of these tests require active trying to link frailty with delirium. Future studies are patient involvement and are, for obvious reasons, not 64 full geriatric assessment in addition to disease severity scores [32]. They found parameters related to quality of life and functional status to be independently correlated with mortality, but not cognitive decline. However, most of their patients had IQCODE values within the normal range. Although IQCODE did not increase the predictive value when added to frailty, it was shown to have impact on 30-day outcome. It is possible that other more specific measures on cognitive decline would have performed better, but such may be difficult to implement in an acute setting. Even the IQCODE proved the most difficult measure to obtain in our cohort with 24% missing data. The CPS score is a simple score, which measures the impact of comorbidity by adding the number of comor- bidities to the number of daily drugs [33]. Different thresholds have been used, but often a CPS score of < 8 has been considered to be categorized as minor, 8–14 as moderate and ≥ 15 as major comorbidity. In our cohort the median CPS value was 10, and 22% had a Fig. 1 Principal component analysis (PCA). Two-dimensional CPS ≥ 15. We have not found data from its previous use projection of the sample was constructed having the axes (principal components, PC) as the factors. Each PC is a linear combination of the in ICU patients, but using the CPS in geriatric trauma original variables and PCs are orthogonal to each other. The angles (age ≥ 65 years), the investigators found CPS to pre- between the vectors tell us how variables correlate with one another: dict mortality better than the more traditional Charl- when two vectors are close, forming a small angle, the two variables son Comorbidity Index (CCI) [34]. Other studies have they represent are positively correlated. If they meet each other at found the CPS to predict hospital mortality and read- 90 °, they are not likely to be correlated and when they diverge and form a large angle (close to 180 °), they are negatively correlated. The missions [35, 36] and to predict in hospital complica- length of the vector shows how much weight a specific variable has tions in older burn patients [37]. However, in our study on each principal component this score had no impact on outcome prediction. Our main aim was to document the impact from the clinical situation present before ICU admission on out- applicable to patients who are acutely admitted to the come in very old ICU patients. Such knowledge could ICU. Hence, we must rely on information from someone support the important clinical decision of pre-ICU tri- close to the patient in order to approach cognitive func- age: Will this patient benefit from being admitted to the tion if there is no clear evidence from patient records. The ICU? [21]. Despite associations between the geriatric IQCODE was developed as a response to such demand syndromes and outcome, we will probably not be able [30] and has been used frequently as a simple screening to answer this important question just by using evalua- tool based on information from caregivers. The IQCODE tion tools and measurement. Often the patient is admit- has been analyzed in a Cochrane review with regard to its ted to the ICU in order to observe response to initial accuracy in finding patients with dementia [13]. A thresh- treatment. Important information related to survival old of 3.3 had 0.83 sensitivity and 0.80 specificity while a can also be found during the first 48–72 h in the ICU. threshold of 3.5 had 0.82 sensitivity and 0.84 specificity. Hence the degree of organ dysfunction and vital organ Studies where IQCODE has been applied prior to ICU support could be highly relevant due to what frequently admission are scarce. An early study of IQCODE com- is called the “in ICU triage”. Such an evaluation, formal pared with Modified Blessed Dementia Rating Scale or informal, often leads to treatment being withheld (MBDRS) was performed in 130 ICU patients ≥ 65 years and/or withdrawn. As a matter of fact, LST limitations [31]. They found that an IQCODE > 3.3 was present in occurred in 34.4% of the patients. Severity (SOFA) and 31% of their patients. Using the same threshold, we found also frailty were independent factors predicting deci- 44% having an IQCODE > 3.3, and the higher percentage sion to limit LST. In the Ethicus-2 study, LST limita- probably mirrors the older age group in our study. With tions were documented in 13.1%, but the patients were a threshold of 3.5; the prevalence of cognitive impair- younger (median 70 years) and age was found to be one ment was 30.2%. In a more recent study from Spain, ICU of the determinants for EOL decision [38]. patients were prospectively studied with data on cogni- In our study, some form of organ support was given tion (IQCODE), activity of daily life, comorbidity and a in 80% of all ICU admissions, most often as ventilator 65 Fig. 2 Adjusted survival curves according to geriatric symptoms: Adjusted survival curves for geriatric variables were produced using an inverse probability-weighted Kaplan–Meier estimation [19]. Variables used to adjust the curves were age, gender, place of living, reason for ICU admission and SOFA score. Significance was tested using a Cox regression model weighted by the same weights (inverse probability-weighted Cox). a Survival curves according to CFS. b Survival curves according to Katz ADL. c Survival curves according to IQCODE. d Survival curves according to CPS and cardiovascular support. This indicates that in most influence the external validity. An unforeseen limita- of the cases, there is no therapeutic limitation, and the tion was a huge delay in ethical and regulatory clearance very old ICU patients in our study were given active in many countries caused by the General Data Protec- organ support when admitted to an ICU. tion Regulation (GDPR) introduced in all EU countries from May 2018. This made the actual recruitment period Strengths and limitations shorter for many ICUs. However, when comparing ICU The main strength of this study is the number of par- with short versus long recruitment period, the patients’ ticipants derived from several European countries, with characteristics were very similar (see ESM 8). The esti- different cultural and legal settings, and the prospective mated percentage of included patients to the potential multicenter observation study design using the same def- candidates was 71.7% (ESM 9), but patients’ characteris- initions of the geriatric syndromes. Limitation of LST is tics were different in countries with a need for informed common practice in most ICUs in Europe and contrib- consent compared to patients enrolled in countries with utes to the observed mortality. Patients with such deci- no-informed consent (ESM10). This study did not focus sion differ from other patients (Table 4), but excluding on pre-ICU triage, and hence we cannot exclude some those patients would introduce a bias to this “real life” selection bias, in particular given the high number of observational study. When LST limitation was included patients living in their own home prior to admission. ICU in the multivariable model, CFS still has a prognos- triage has been the focus of other studies [39]. There is tic impact on survival. The study was conducted within only scarce evidence of the inter-rater variability of the 1 year. Hence, we assume that trends in time do not CFS between the caregiver and the patient when the 66 Table 6 Multivariable analysis (Cox regression analysis) Variables Categories ALL covariates CFS Katz IQ code HR (95%CI) p-value HR (95%CI) p-value HR (95%CI) p-value HR (95%CI) p-value Age One-point increase 1.02 (1–1.03) 0.01 1.02 (1.01– 0.006 1.02 (1.01– 0.002 1.02 (1.01– 0.003 1.03) 1.04) 1.04) Habitat Own home 1.01 (0.9–1.14) 0.83 1 (0.89–1.12) 0.96 0.98 (0.87–1.1) 0.73 0.95 (0.84– 0.43 (ref = not 1.08) home) Gender Female 0.98 (0.88– 0.68 0.98 (0.88– 0.67 1 (0.9–1.11) 0.96 1.01 (0.91– 0.79 (ref = male) 1.09) 1.09) 1.13) Reason for Combined respira- 1.07 (0.89– 0.46 1.07 (0.89– 0.46 1.09 (0.91– 0.34 1.09 (0.91– 0.37 admission tory/circulatory 1.29) 1.29) 1.31) 1.31) failure (Ref = circula- Emergency surgery 0.64 (0.52– < 0.0001 0.64 (0.52– < 0.0001 0.64 (0.52– < 0.0001 0.63 (0.51– < 0.001 tory failure) 0.78) 0.78) 0.78) 0.77) Multitrauma w/wo 1.16 (0.81– 0.41 1.15 (0.81– 0.42 1.12 (0.79-1.59) 0.51 1.09 (0.77– 0.63 head injury 1.64) 1.64) 1.54) Multitrauma with- 0.86 (0.58– 0.46 0.87 (0.58– 0.48 0.83 (0.56– 0.35 0.78 (0.53– 0.23 out head injury 1.29) 1.29) 1.23) 1.16) Other 0.81 (0.67– 0.025 0.81 (0.67– 0.03 0.8 (0.66–0.97) 0.02 0.79 (0.66– 0.016 0.97) 0.98) 0.96) Respiratory failure 0.81 (0.69– 0.017 0.82 (0.69– 0.02 0.82 (0.69– 0.02 0.82 (0.69– 0.03 0.96) 0.97) 0.98) 0.98) Sepsis (according 0.71 (0.59– 0.0003 0.71 (0.59-0.86) 0.0003 0.72 (0.6-0.87) 0.0005 0.71 (0.59-0.86) 0.0004 to Sepsis3) 0.86) SOFA One-point increase 1.15 (1.14– < 0.00001 1.15 (1.14– < 0.00001 1.15 (1.14– < 0.001 1.15 (1.14-1.17) < 0.00001 1.17) 1.17) 1.17) CPS One-point increase 1 (0.99–1.01) 0.75 1 (0.99–1.01) 0.74 1 (0.99–1.01) 0.65 1.01 (1-1.02) 0.28 CFS One point increase 1.1 (1.05–1.15) 0.0001 1.12 (1.09– < 0.00001 1.15) Katz One point increase 0.98 (0.94– 0.40 0.93 (0.9–0.95) < 0.00001 1.02) IQ code One-point increase 1.03 (0.91– 0.63 1.18 (1.06– 0.0029 1.18) 1.31) Mean AIC across 23,993.92 23,991.31 24,016.93 24,025.87 imputations HR gives the change in risk of death per each unit increase for continuous variables and for one specific category versus a reference category for categorical variables. HR > 1 suggests an increase in the risk of death, HR < 1 suggests a decrease in the risk of death SOFA sequential organ failure assessment, CFS clinical frailty scale, IQCODE informant questionnaire on cognitive decline in the elderly, ADL activity of daily living, CPS Co-morbidity and Polypharmacy score AIC Akaike’s information criterion. AIC was used across imputations to evaluate the goodness of fit of our models. AIC = − 2 Log likelihood + 2p, where p is the number of parameters CFS is performed at the same time. However, one study Another limitation is the lack of other explanatory vari- concluded that CFS scores can be reliably assigned retro- ables like individual socio-economic status, education spectively [40] and another study, comparing CFS using and the absence of delirium evaluation on admission and different assessors, concluded that “CFS scores can be during the ICU stay. We focused on mortality but other generated using medical chart review and can be reli- outcomes such as long-term ventilation, chronic critical ably completed by ICU clinicians and research staff“ [41]. illness and cognitive dysfunction which are patient-rele- Inter-rater was excellent in our study with a weighted vant outcomes should also be assessed in future studies. kappa above 0.8. The IQCODE was missing in 24% and was based on informant data collected in a life-threaten- ing situation suggesting that it is not a good tool for criti- Conclusions cally ill patients. We found that Clinical frailty scale alone was an inde- pendent prognostic factor for 1-month survival after ICU admission. Adding other geriatric syndromes and 67 Table 7 Multivariable analysis including life-sustaining treatment limitation Variables Categories HR (95%CI) p-value Age One-point increase 1 (0.99–1.02) 0.79 Habitat (ref = not home) Own home 0.96 (0.85–1.08) 0.47 Gender (ref = male) Female 0.94 (0.84–1.04) 0.24 Reason for admission (ref = circulatory Combined respiratory/circulatory failure 1.13 (0.94–1.36) 0.19 failure) Emergency surgery 0.68 (0.56–0.84) 0.0003 Multitrauma w/wo head injury 1.21 (0.85–1.73) 0.28 Multitrauma without head injury 0.91 (0.61–1.36) 0.64 Other 0.88 (0.73–1.06) 0.18 Respiratory failure 0.82 (0.69–0.97) 0.021 Sepsis (according to Sepsis3) 0.67 (0.55–0.81) 0.0003 SOFA One-point increase 1.16 (1.14–1.17) < 0.0001 CPS One-point increase 0.99 (0.98–1) 0.05 CFS One-point increase 1.08 (1.05–1.12) < 0.0001 Withholding or withdrawal Yes vs no 4.25 (3.8–4.74) < 0.0001 scores did not improve the model. This is an important University Innsbruck, Innsbruck, Austria. 21 European Wellness International, ICU, Luhansk, Ukraine. 22 Alkhums Hospital, ICU, Tripoli, Libya. 23 Department finding suggesting that collection of only one score is of Intensive Care, Aarhus University Hospital, Aarhus, Denmark. 24 Dep. of Clini- feasible in circumstances of rapid processing of the cal Medicine, University of Bergen, Bergen, Norway. 25 Department of Anaes- decision making to admit or refuse an old patient in thesia and Intensive Care, Haukeland University Hospital, Bergen, Norway. ICU. Predicting the future, look at the past! [42]. Acknowledgement This study was endorsed by the ESICM. Free support for running the electronic Electronic supplementary material database and was granted from the dep. of Epidemiology, University of The online version of this article (https​://doi.org/10.1007/s0013​4-019-05853​-1) Aarhus, Denmark. Financial support for creation of the e-CRF and mainte- contains supplementary material, which is available to authorized users. nance of the database was possible from a grant (open project support) by Western Health region in Norway) 2018 who also funded the participating Norwegian ICUs. DRC Ile de France and URC Est helped conducting VIP2 in Author details 1 France. The VIP2-study collaborators are listed in the ESM 1. Sorbonne Université, INSERM, Institut Pierre Louis d’Epidémiologie et de Santé Publique, Saint Antoine Hospital, AP-HP, Hôpital Saint-Antoine, service Author contributions de réanimation, F75012 Paris, France. 2 Department of Intensive Care Medi- BG, DDL, HF designed the study; AB performed the statistical analysis; JF cine, University Medical Center, University Utrecht, Utrecht, The Netherlands. 3 run the database and the eCRF; BG, DDL, HF, AB drafted the manuscript; AM AP-HP, Hôpital Saint-Antoine, service de réanimation, F75012 Paris, France. 4 provided geriatric expertise in designing the protocol. All other authors were Research Lead Critical Care Directorate St George’s Hospital, London, UK. 5 country coordinators and validated the manuscript. St George’s University Hospital, London, UK. 6 Chair NAHP Section ESICM, Intensive Care Unit, Royal Devon & Exeter NHS Foundation Trust, Exeter, UK. 7 Compliance with ethical standards Intensive Care and Perioperative Medicine Division, Jagiellonian University Medical College, Kraków, Poland. 8 Department of Intensive Care Medicine, Conflicts of interest CIBER Enfermedades Respiratorias, Sabadell and Critical Care Department, Joerg C. Schefold declares that the Dept. of Intensive Care Medicine Bern has/ Corporacion Sanitaria Universitaria Parc Tauli, Autonomous University of Barce- had research and/or development/consulting contracts with (full disclosure): lona, Sagrado Corazon-General de Cataluña University Hospitals, Quiron Salud, Orion Corporation, Abbott Nutrition International, B. Braun Medical AG, CSEM Barcelona, Spain. 9 Department of Rehabilitation Hospital Ancelle di Cremona SA, Edwards Lifesciences Services GmbH/SA, Kenta Biotech Ltd, Maquet Criti- Italy, Geriatric Research Group, Brescia, Italy. 10 Department of Anaesthesia cal Care AB, Omnicare Clinical Research AG, and Nestlé. Educational grants and Intensive Care, NTNU, Dep of Circulation and Medical Imaging, Ålesund were received from Fresenius Kabi; GSK; MSD; Lilly; Baxter; Astellas; AstraZen- Hospital, Trondheim, Ålesund, Norway. 11 Intensive Care Unit General Hospital eca; B. Braun Medical AG, CSL Behring, Maquet, Novartis, Covidien, Nycomed, of Larissa Tsakalof Larissa, Larissa, Greece. 12 Division of Cardiology, Pulmonol- Pierre Fabre Pharma (Roba Pharma); Pfizer, Orion Pharma. The money went ogy and Vascular Medicine, University Hospital Düsseldorf, Heinrich-Heine- into departmental funds. No personal financial gain applies. All other authors University, Düsseldorf, Germany. 13 Faculdade de Ciências Médicas de Lisboa do not have any conflict of interest to declare related to this manuscript. (Nova Médical School), Unidade de Cuidados Intensivos Neurocríticos e Trauma. Hospital de São José, Centro Hospitalar Universitário de Lisboa Central, Lisbon, Portugal. 14 Linkoping University Hospital, Linkoping, Sweden. 15 Publisher’s Note Department of Intensive Care 1K12IC Ghent University Hospital, Ghent, Bel- Springer Nature remains neutral with regard to jurisdictional claims in pub- gium. 16 Department of Intensive Care Medicine, Inselspital, Universitätsspital, lished maps and institutional affiliations. University of Bern, Bern, Switzerland. 17 Department of Anesthesia and Inten- sive Care Medicine, Humanitas Clinical and Research Center – IRCCS, Via Received: 12 August 2019 Accepted: 1 November 2019 Alessandro Manzoni 56, 20089 Rozzano, MI, Italy. 18 Department of Biomedical Published online: 29 November 2019 Sciences, Humanitas University, Pieve Emanuele, Rozzano, MI, Italy. 19 Mater Misericordiae University Hospital, Dublin, Ireland. 20 Division of Intensive Care and Emergency Medicine, Department of Internal Medicine, Medical 68 References 20. Kottner J, Audigé L, Brorson S, Donner A, Gajewski BJ, Hróbjartsson A, 1. Margaras V (2019) Demographic trends in EU regions. European Parlia- Roberts C, Shoukri M, Streiner DL (2011) Guidelines for reporting reli- mentary Research Service, January 2019. https​://ec.europ​a.eu/futur​ium/ ability and agreement studies (GRRAS) were proposed. J Clin Epidemiol en/syste​m/files​/ged/eprs-brief​i ng-63316​0-demog​raphi​c-trend​s-eu-regio​ 64:96–106. https​://doi.org/10.1016/j.jclin​epi.2010.03.002 ns-final​.pdf 21. Guidet B, de Lange DW, Flaatten H (2018) Should this elderly patient 2. Leblanc G, Boumendil A, Guidet B (2017) Ten things to know about be admitted to the ICU? Intensive Care Med 44:1926–1928. https​://doi. critically ill elderly patients. Intensive Care Med 43:217–219. https​://doi. org/10.1007/s0013​4-018-5054-7 org/10.1007/s0013​4-016-4477-2 22. Flaatten H, Jung C, Vallet H et al (2019) How does frailty affects outcome? 3. Jung C, Wernly B, Muessig JM et al (2019) A comparison of very old Curr Anesthesiol Rep 9:144 patients admitted to intensive care unit after acute versus elective 23. Bagshaw SM, Stelfox HT, Johnson JA, McDermid RC, Rolfson DB, Tsuyuki surgery or intervention. J Crit Care 52:141–148. https​://doi.org/10.1016/j. RT et al (2015) Long-term association between frailty and health-related jcrc.2019.04.020 quality of life among survivors of critical illness: a prospective multi- 4. Flaatten H, de Lange DW, Morandi A, Andersen F, Artigas A, Bertolini G center cohort study. Crit Care Med 43:973–982. https​://doi.org/10.1097/ et al (2017) The impact of frailty on ICU and 30-day mortality and the CCM.00000​00000​00086​0 level of care in very elderly patients (≥ 80 years). Intensive Care Med 24. Hewitt J, Carter B, McCarthy K, Pearce L, Law J, Wilson FV et al (2019) 43:1820–1828. https​://doi.org/10.1007/s0013​4-017-4940-8 Frailty predicts mortality in all emergency surgical admissions regard- 5. Flaatten H, Garrouste-Orgeas M (2015) The very old ICU patient: a never- less of age. An observational study. Age Ageing 48:388–394. https​://doi. ending story. Intensive Care Med 41:1996–1998. https​://doi.org/10.1007/ org/10.1093/agein​g/afy21​7 s0013​4-015-4052-2 25. Hewitt J, Long S, Carter B, Bach S, McCarthy K, Clegg A (2018) The 6. Guidet B, Vallet H, Boddaert J, de Lange DW, Morandi A, Leblanc G, prevalence of frailty and its association with clinical outcomes in general Artigas A, Flaatten H (2018) Caring for the critically ill patients over 80: a surgery: a systematic review and meta-analysis. Age Ageing 47:793–800. narrative review. Ann Intensive Care 8:114. https​://doi.org/10.1186/s1361​ https​://doi.org/10.1093/agein​g/afy11​0 3-018-0458-7 26. Theou O, Campbell S, Malone ML, Rockwood K (2018) Older adults in the 7. Yu WC, Chou MY, Pen LN, Lin YT, Liang CK, Chen LK (2017) Synergistic emergency department with frailty. Clini Geriatr Med 34:369–386. https​:// effects of cognitive impairment on physical disability in all-cause mortal- doi.org/10.1016/j.cger.2018.04.003 ity among men aged 80 years and over: results from longitudinal older 27. Persico I, Cesari M, Morandi A, Haas J, Mazzola P, Zambon A, Annoni G, veterans study. PLoS One 12:e0181741. https​://doi.org/10.1371/journ​ Bellelli G (2018) Frailty and delirium in older adults: a systematic review al.pone.01817​41 and meta-analysis of the literature. J Am Geriatr Soc 66:2022–2030. https​ 8. Level C, Tellier E, Dezou P, Chaoui K, Kherchache A, Sejourné P, Rullion-Pac ://doi.org/10.1111/jgs.15503​ Soo AM (2018) Outcome of older persons admitted to intensive care unit, 28. Hayat SA, Luben R, Dalzell N, Moore S, Hogervorst E, Matthews FE et al mortality, prognosis factors, dependency scores and ability trajectory (2018) Understanding the relationship between cognition and death: within 1 year: a prospective cohort study. Aging Clin Exp Res 30:1041– a within cohort examination of cognitive measures and mortality. Eur J 1051. https​://doi.org/10.1007/s4052​0-017-0871-z Epidemiol 33:1049–1062. https​://doi.org/10.1007/s1065​4-018-0439-z 9. de Lange DW, Guidet B, Andersen FH, Artigas A, Bertolini G, Moreno 29. Crosby G, Culley DJ, Hyman BT (2011) Preoperative cognitive assess- R et al (2019) Huge variation in obtaining ethical permission for a ment of the elderly surgical patient: a call for action. Anesthesiology non-interventional observational study in Europe. BMC Medical Ethics 114:1265–1268. https​://doi.org/10.1097/ALN.0b013​e3182​1b1bc​8 20(1):39. https​://doi.org/10.1186/s1291​0-019-0373-y 30. Jorm AF (1994) A short form of the Informant Questionnaire on Cogni- 10. Rockwood K, Song X, MacKnight C, Bergman H, Hogan DB, McDowell tive Decline in the Elderly (IQCODE): development and cross validation. I et al (2005) A global clinical measure of fitness and frailty in elderly Psychol Med 24:145–153 people. CMAJ 173:489–495. https​://doi.org/10.1503/cmaj.05005​1 31. McNicoll L, Redlich CA (2003) Screening for preexisting cognitive impair- 11. Katz S (1983) Assessing self-maintenance: activities of daily living, mobil- ment in older intensive care unit patients: use of proxy assessment. J Am ity, and instrumental activities of daily living. J Am Geriatr Soc 31:721–727. Geriatr Soc 51:689–693. https​://doi.org/10.1034/j.1600-0579.2003.00215​.x https​://doi.org/10.1111/j.1532-5415.1983.tb033​91.x 32. Sacanella E, Pérez-Castejón JM, Nicolás JM, Masanés F, Navarro M, Castro 12. Jorm A (2004) The Informant Questionnaire on Cognitive Decline in the P, López-Soto A (2009) Mortality in healthy elderly patients after ICU Elderly (IQCODE): a review. Int Psychogeriatr 16:1–19 admission. Intensive Care Med 35:550–555. https​://doi.org/10.1007/s0013​ 13. Quinn TJ, Fearon P, Noel-Storr AH, Young C, McShane R, Stott DJ (2014) 4-008-1345-8 Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE) 33. Kalra S et al (2015) Comorbidity polypharmacy score and its clinical for the diagnosis of dementia within community dwelling populations. utility: a pragmatic practitioner’s perspective. J Emerg Trauma Shock. Cochrane Database of Syst Rev. https​://doi.org/10.1002/14651​858.cd010​ 8:224–231. https​://doi.org/10.4103/0974-2700.16165​8 079.pub2 34. Nossaman VE, Larsen BE, DiGiacomo JC et al (2018) Mortality is predicted 14. Evans DC, Cook CH, Christy JM, Murphy CV, Gerlach AT, Eiferman D et al by comorbidity polypharmacy score but not charlson comorbidity (2012) Comorbidity-polypharmacy scoring facilitates outcome prediction index in geriatric trauma patients. Amer J Surgery 216:42–45. https​://doi. in older trauma patients. J Am Geriatr Soc 60:1465–1470. https​://doi.org/ org/10.1016/j.amjsu​rg.2017.09.011 10.1111/j.1532-5415.2012.04075​.x 35. Housley BC, Stawicki SP, Evans DC, Jones C (2015) Comorbidity-polyp- 15. Vincent JL, Moreno R, Takala J, Willatts S, De Mendonça A, Bruining H, harmacy score predicts readmission in older trauma patients. J Surg Res Reinhart CK, Suter PM, Thijs LG (1996) The SOFA (Sepsis-related Organ 199:237–243. https​://doi.org/10.1016/j.jss.2015.05.014 Failure Assessment) score to describe organ dysfunction/failure. On 36. Sammy I, Lecky F, Sutton A, Leaviss J, O’Cathain A (2016) Factors affecting behalf of the working group on sepsis-related problems of the ESICM. mortality in older trauma patients-a systematic review and meta-analysis. Intensive Care Med 22:707–710 Injury 47:1170–1183. https​://doi.org/10.1016/j.injur​y.2016.02.027 16. Guidet B, Flaatten H, Boumendil A et al (2018) Withholding or withdraw- 37. Justiniano CF, Evans DC et al (2015) Comorbidity-polypharmacy score ing of life-sustaining therapy in older adults (≥ 80 years) admitted to predicts in-hospital complications and the need for discharge to the intensive care unit. Intensive Care Med 44:1027–1038. https​://doi. extended care facility in older burn patients. J Burn Care Res. 36:193–196. org/10.1007/s0013​4-018-5196-7 https​://doi.org/10.1097/BCR.00000​00000​00009​4 17. Flaatten H, de Lange DW, Artigas A et al (2017) The status of ICM research 38. Sprung CL, Ricou B, Hartog CS et al (2019) Changes in end-of-life prac- and a future agenda for very elderly patients in the ICU. Intensive Care tices in european intensive care units from 1999 to 2016. JAMA 2:1–12. Med 43:1319–1328. https​://doi.org/10.1007/s0013​4-017-4718-z https​://doi.org/10.1001/jama.2019.14608​ 18. Pan W (2001) Akaike’s information criterion in generalized estimating 39. Guidet B, Leblanc G, Simon T et al (2017) Effect of systematic intensive equations. Biometrics 57:120–125. https​://doi.org/10.1111/j.0006- care unit triage on long-term mortality among critically ill elderly patients 341X.2001.00120​.x in France: a randomized clinical trial. JAMA 318:1450–1459. https​://doi. 19. Cole SR, Hernán MA (2004) Adjusted survival curves with inverse prob- org/10.1001/jama.2017.13889​ ability weights. Comput Methods Programs Biomed 75:45–49 69 40. Davies J, Whitlock J, Gutmanis I, Kane SL (2018) Inter-rater reliability of the 42. Flaatten H, Oeyen S, de Lange DW (2018) Predicting outcomes in very old retrospectively assigned clinical frailty scale score in a geriatric outreach ICU patients: time to focus on the past? Intensive Care Med 44:1344– population. Can Geriatr J. 21:1–5. https​://doi.org/10.5770/cgj.21.263 1345. https​://doi.org/10.1007/s0013​4-018-5262-1 41. Shears M, Takaoka A, Rochwerg B et al (2018) Assessing frailty in the intensive care unit: a reliability and validity study. J Crit Care 45:197–203. https​://doi.org/10.1016/j.jcrc.2018.02.004