The number of patients admitted to intensive care units (ICUs) increased within the past years with an ongoing upward trend and an overproportion of the patients advanced in years (1, 2). Older patients are more likely frail, which is a multifaceted condition characterized by the loss of physiologic and cognitive reserves (3, 4). The observational studies suggest that the patients with frailty currently account for up to one-third of the critically ill (5, 6). Consequently, the patient outcome is determined not only by the acute critical illness but increasingly by the reduced functional reserve of pre-existing frailty resulting in higher 30-day mortality (5, 7, 8). In accordance, the likelihood to be discharged to a nursing home is greater in the patients with frailty (9), if the critical illness is survived. This might be caused by the higher odds of disability in the activities in daily living (10, 11) and increased functional dependence (12, 13). Despite this finding, the factors affecting the recovery of physical function after a critical illness remain poorly understood. The patient-level characteristics should be evaluated as the recovery trajectories differ between the cohorts in both the extent and speed of recovery of physical function. A functional trajectory is used to describe this complex process by measuring the changes in the functional status at different time points (14, 15).

While early mobilization might be an important element to maintain the autonomy and mobility in the prior functionally independent patients (16–18), little is known about the functional trajectory of the patients with frailty during the hospital stay (13, 19). Since information about the functional status and mobility of the patients before their ICU admission is typically missing (11, 20), it is unknown if the functional decline is caused by frailty itself, the critical illness, or the combination of both. This might have been important implications for the resource allocations in the acute care setting if the mortality is high and the functional decline cannot be prevented (21, 22).

This study aimed to describe the influence of pre-existing frailty on the functional trajectory of patients with a critical illness during their hospital stay. We hypothesized that the patients with frailty have a greater deterioration of function compared with the patients with non-frailty.

Between April 1, 2017, and May 31, 2019, we included 1,172 patients (Figure 1). The median age was 68 [56–77] years, of which 290 patients (25%) were assessed as frail (CFS levels 5–9). Further patient characteristics are presented in Table 1. Using the propensity score matching, 687 patients were selected, of which 232 were frail and 455 non-frail (Table 2 and Supplementary Table 1 in the Appendix). Applying the same criteria to the survivors, only lead to 393 patients of which 125 were frail (Supplementary Tables 2, 3 in the Appendix).

This prospective observational study refutes the assumption that pre-existing frailty deteriorates the functional status to a greater extent, i.e., the functional trajectory of critical care patients with frailty was not worse compared with the patients with non-frailty when adjusted for age, comorbidity, and the triggering reason for intensive care. Actually, the pre-existing differences in the functional status of patients with frailty and non-frailty converged at hospital discharge, indicating that intensive care is justified in patients with pre-existing frailty as well.

Approximately 25% of the patients were frail when admitted to our ICU for at least 24 h. They had more comorbidities as defined by the CCI, had more insufficient organ systems as indicated by a higher SOFA score, and the overall severity of their diseases was more profound as scored by the APACHE II. Accordingly, the mortality was higher in this patient sub-cohort. These data are in accordance with the German sub analysis of the VIP1 Trial, which included only patients >80 years of age (5). Accounting for more than 50% of patients with frailty indicated a strong relationship between age and reduction of the physiological reserve. Notwithstanding, LOS in the ICU was longer in our cohort when compared with the subcohort of the German VIP Trial (9 vs. 3 days) most likely due to our inclusion criteria of >24 h ICU stay as well as a higher portion of neurocritical care patients.

The patients with frailty expectedly had a reduced functional status during the complete trajectory, i.e., significantly lower MTB. This is in accordance with the other observational studies, which additionally demonstrated increased peri- and postoperative complication rates, morbidity, and mortality (10, 12, 27, 36). Furthermore, Bagshaw et al. suggested that the pre-existing frailty impaired the long-term outcome of ICU survivors, as one-third of their 421 patients reported a reduced health related quality of life with reduced mobility in the physical component score at 6 and 12 months after critical illness (20). In another study in more than 1,000 patients, frailty was again associated with an increased disability after critical illness (11). In both studies, the parameters of functionality were obtained before the onset of critical illness. Brummel et al. investigated the functional status at 3- and 12-month after critical illness, which did not evaluate the influence of intensive care on the clinical outcome. Similarly, in the study from Bagshaw et al., the outcome was assessed at 6 and 12 months. Furthermore, the authors mentioned a limitation that they were not able to integrate the baseline functional measures, such as mobility. Therefore, the conclusion that the functional status deteriorated during intensive care is not justified due to frailty based on those data.

More recently, Ferrante et al. also reported that the patients with frailty ≥70 years old had higher mortality when becoming critically ill (8). Those patients with frailty had an increased disability when compared with the patients with non-frailty. However, the trajectory showed no difference between the patients with frailty and non-frailty. This is similar to our observation in the patients with frailty, whose functional status was also decreased but their trajectory was at least not worse compared with the patients with non-frailty. This also accounts for the long-term outcome study of the same group (37).

The observation of partly improved or unaffected functional status in the patients with frailty who survived their critical illness demands critical consideration. It can be speculated that in older patients, depression, and social isolation often lead to reduced daily activities and accumulation of disability (37). In intensive care, those patients are exposed to stimuli by the caregivers and to early mobilization therapy resulting in the improvement of their functional status (19).

The strength of this study is its prospective approach and the high number of patients included, limited by in-hospital data only. Although this was a single-center study, our cohort showed a heterogeneous group of adult patients regarding diagnosis or prognosis of the disease suggesting generalizability. Assessing the prehospital functional status makes this study unique compared with others and offers new perspectives in understanding the trajectory of critically ill patients with frailty. However, the prehospital frailty status was assessed retrospectively which is a limitation. Although we implemented detailed and recurrent training of our study staff, the assessment depends on the ability to either correctly recall the prehospital status by the patients or to adequately know the status by proxy (memory and information bias). This important problem is not satisfactorily answered yet. The upcoming results of the ASTON study (NCT03785444) will likely improve our insight on assessing the prehospital functional status in patients with critical illness. Until then we must accept this as a limitation. Since the majority of our patients are surgical or trauma, the results should be validated in the medical ICU patients as well.

Addressing the functional outcome is a current focus after surviving the critical illness. Due to the scaling and granularity of the chosen MTB, subtle nuances of the functional outcome might be missed. Since there is no defined core outcome set for the functional outcomes in critical illness yet, we considered the subdomains of the Barthel-Index a suitable option, as it is easy to assess, reproducible for the caregivers, and relevant for the patients (24, 25).

Performing propensity score matching reduced the cohort of the primary analysis considerably. The factors leading to non-selection in propensity scoring were admission from a nursing home, an advanced age, a low GCS, a high APACHE II score, and a high level of comorbidities represented by CCI in the patients with frailty. In the patients with non-frailty, younger age and a low CCI (showing a healthy overall status) lead to non-selection. Nevertheless, this approach strengthens the conclusion that the effect is due to frailty itself and reduces the likelihood of bias. The assumed risk of limited generalizability is countervailed by the confirmation of all the results in the sensitivity and exploratory analysis performed in the complete cohort.

In conclusion, the patients with frailty have a reduced functional status. Their intrahospital functional trajectory, however, is no worse than those in the patients with non-frailty. Even more, our data suggests a significant rehabilitation potential of functional mobility in the patients with frailty if they survive.

Data can be obtained from the corresponding author on reasonable scientific request and as long as German data protection law can be complied with. Requests to access the datasets should be directed to Stefan Schaller, stefan.schaller@charite.de.

The studies involving human participants were reviewed and approved by Ethics Committee of the Faculty of Medicine, Technical University of Munich. The patients/participants provided their written informed consent to participate in this study.

SJS is the principal investigator and developed the protocol and BU is the study statistician. SJS and MB were involved in the ethical approval. KEF, JJG, BW, RM, BU, BJ, MB, and SJS were involved in the analysis and interpretation of the data. KEF, SF, ML, RB, MH, BK, SK, CL, JM, and SJS were involved in the data acquisition and quality assurance. All authors critically revised the manuscript and approved its final version.

BW reports personal fees from Orion Pharma Ltd and national (DAAD) and international grants (ESICM) outside the submitted work. BJ received honoraria for giving lectures from Pulsion Medical Systems SE (Feldkirchen, Germany). MB received research support from MSD (Haar, Germany) not related to this manuscript, received honoraria for giving lectures from GE Healthcare (Helsinki, Finland) and Grünenthal (Aachen, Germany). SS reports grants and non-financial support from ESICM (Brussels, Belgium), Fresenius (Germany), Liberate Medical LLC (Crestwood, USA), Reactive Robotics GmbH (Munich, Germany), STIMIT AG (Nidau, Switzerland) as well as from Technical University of Munich, Germany, from national (e.g. DGAI) and international (e.g. ESICM) medical societies (or their congress organizers) in the field of anesthesiology and intensive care, personal fees and non-financial support from Bavarian Medical Association, all outside the submitted work; SS holds stocks in small amounts from Alphabeth Inc., Bayer AG, Rhön-Klinikum AG, and Siemens AG. These did not have any influence on this study. The remaining authors declare that they have no competing interests.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.