How early is early? When should rehabilitation begin in critical illness?

Critical care mortality has significantly declined over the last decade, resulting in a growing population of individuals who have survived a period of critical illness (Zimmerman et al. 2013). In the United Kingdom (UK) alone almost 140,000 people were discharged alive from critical care in the year to April 2018, in comparison with under 75,000 in the same period to 2009 (Intensive Care National Audit and Research Centre (ICNARC) 2010; ICNARC 2018). However, this survivorship is frequently characterised by long-term deficits in physical, cognitive and psychological health (Cuthbertson et al. 2010; Herridge et al. 2011; Pandharipande et al. 2013). Therefore, the National Institute for Health and Care Excellence (NICE) recommends that rehabilitation with the aim of optimising recovery commence ‘as early as clinically possible’ in those at risk of morbidity (NICE 2009). Despite this, the paucity of evidence in favour of any given rehabilitation strategy means that the optimal
type and timing of intervention remains an area of debate. The following review will explore when physical rehabilitation should begin in critical illness, focusing on some of the key factors that limit the current evidence base.

The physical sequelae of critical illness are most commonly loss of skeletal muscle strength and endurance, with associated loss of function and independence (Herridge et al. 2011; Batt et al. 2013). The term intensive care unit acquired weakness (ICU-AW) is used to refer to this presentation of weakness. ICU-AW is attributed to multiple physiological mechanisms relating to systemic inflammation and catabolic illness, resulting in varying proportions of polyneuropathy, polymyopathy, or both (Kress and Hall 2014). Those with ICU-AW take longer to wean from mechanical ventilation, and have poorer long-term functional outcomes than those without (De Jonghe et al. 2002; Dinglas et al. 2017; Kelmenson et al. 2017).

Early physical rehabilitation has been proposed as a potential treatment, or prophylaxis, for ICU-AW (Morris and Herridge 2007; Needham 2008). For the purposes of this discussion, ‘early rehabilitation’ is defined as active exercise of some form, commenced whilst an individual is on critical care. Passive exercises are not included, as there is no evidence for these increasing muscle strength or endurance (for example, Nafziger et al. 1992; Marshall et al. 2011). Multiple other definitions of early rehabilitation exist: some specifying begin- ning rehabilitation within a certain number of days or hours of critical care admission or commencement of mechanical ventilation; some include passive movements; while others simply fail to define early rehabilitation at all. A recent systematic review of early mobilisation in mechanically ventilated patients found that only 15 of the 76 included studies defined early mobilisation fully (Clarissa et al. 2019).

It might appear axiomatic that physical rehabilitation should be a treatment option for ICU-AW. Exercise-based interventions have strong evidence for improving strength and function in other clinical populations, such as chronic respiratory disease (McCarthy et al. 2015). However, evidence for early rehabilitation on critical care as a treatment for ICU-AW remains poor. Initial studies demonstrating benefit have been followed by multiple subsequent trials showing no effect of intervention (Castro-Avila et al. 2015; Hodgson and Cuth- bertson 2016; Tipping et al. 2016; Fuke et al. 2018). One problem with demonstrating effect may be due to a change in usual care on ICU, due to early indications of benefits of rehabilitation. For example, in Schweickert’s 2009 randomised control trial comparing early physical and occupational therapy on critical care with ‘usual care’ (data collected between 2005 and 2007). The control group was usual care in two North American hospitals at the time. They commented that ‘neither site routinely provides physical therapy for patients who are on mechanical ventilation for less than 2 weeks, nor has dedicated physiotherapists for such practice’ (Schweickert et al. 2009). Usual care therefore, was little or no physical rehabilitation on critical care. Subsequent studies, particularly those based in Australia or Europe, have compared ‘usual care’ – provision of some physical rehabilitation on critical care – with an intervention of either increased dose, earlier commencement, or more pro- longed delivery of rehabilitation (e.g. Denehy et al. 2013; Walsh et al. 2015). Although these interventions have been shown to be feasible and safe, they have demonstrated little or no improvement in functional outcomes (for reviews, see Castro-Avila et al. 2015; Hodgson and Cuthbertson 2016; Nydahl et al. 2017; Tipping et al. 2017).

The NICE Clinical Guideline ‘Rehabilitation after Critical Illness’ (NICE 2009), Quality Standard of the same name (NICE 2017), and the Intensive Care Society (ICS) Guidelines for the Provision of Intensive Care Services (ICS 2015), all recommend early rehabilitation to improve physical functioning. The NICE 2017 quality standard states, ‘[I]n the critical care unit, reduced mobility, weakness and fatigue will be the main problems, for which the overall goal will be early mobilisation’ (NICE 2017, p. 8), and that adherence to the quality standards are expected to contribute to improvements across multiple outcomes, including physical recovery and quality of life. However, both NICE and the ICS could be criticised for making recommendations in the absence of a strong evidence base to support them. For example, the statement above in the NICE quality standard is referenced as being supported by ‘expert opinion’ only, the weakest quality of graded evidence (Ryan 2016). This is precisely because there is still no higher quality evidence to support it, even eight years after the publication of the corresponding clinical guideline (NICE 2009).

The speed of onset of muscle wasting (Puthucheary et al. 2013; Parry et al. 2015) and resultant loss of strength and function that occurs in ICU, means that many clinicians feel rehabilitation should be commenced as soon as possible following admission. Proponents of early rehabilitation often cite studies that commenced their intervention within 72 hours of ICU admission as those with positive outcomes (Schweickert et al. 2009; Dong et al. 2016; Schaller et al. 2016). They then explain the failure of other studies to show benefit as being due to later initiation of the intervention (such as Denehy et al. 2013; Morris et al. 2016; Wright et al. 2018). This is a possibility, but downplays the range of other differences between these studies that could have affected the outcomes, such as differing inclusion and exclusion criteria, differences in usual care of the control group, varied interventions, and different primary outcomes.  For example, Schweickert and colleagues’ only included patients who were ‘functionally independent’ prior to admission, and Dong and colleagues’ only patients who were able to ‘stand up and walk’ at enrolment (Schweickert et al. 2009; Dong et al. 2016). Whereas, Denehy and Wright’s trials had no exclusion on the basis of pre-morbid or enrolment function (Denehy et al. 2013; Wright et al. 2018). As such, the evidence base does not clearly support the statement that physical rehabilitation improves function in all critical care patients provided it is commenced in the first two or three days, and not if it is not (Castro-Avila et al. 2015; Hodgson and Cuthbertson 2016; Tipping et al. 2016).

Another factor that may contribute to the inability of early rehabilitation trials to show improvements in morbidity or mortality is the heterogeneity of the patient group, both in their individual characteristics, and the way in which they respond to an intervention. Multiple authors have suggested this as a likely confounder when analysing their results (for example, Denehy et al. 2013; Wright et al. 2018). In fact, heterogeneity of treatment effect – the idea that not every critical care patient will respond in the same way to a given intervention – is thought to contribute to the difficulty in showing effect of intervention in other areas of critical care medicine too, not only in rehabilitation (Iwashyna et al. 2015). Many early rehabilitation studies have included all patients on one or more critical care units, therefore assessing the impact of a given intervention on a cohort of patients for whom acuity of illness and associated dependence are the only guaranteed common factors. Even cohorts with more narrowly defined inclusion criteria, such as patients with acute respiratory distress syndrome, or sepsis, have significant variation within groups, due to the syndromic nature of such diagnoses (Iwashyna et al. 2010; Herridge et al. 2011).

Critical care patients as a cohort have worse mortality and morbidity outcomes than hospital inpatients requiring ward-based level of care (Lone et al. 2016), but acuity of illness within a critical care population is a poor predictor of outcome (Cuthbertson 2016; Bagshaw et al. 2018). Therefore, although all critical care patients clearly have something in common (how sick they are), the discriminatory ability of this severity of illness is limited.

This makes trial design inherently difficult. There are few, if any, other areas of research into physical rehabilitation that select their patient cohort on the basis of acuity of illness, unrelated to specific diagnosis. This might be equivalent to an outpatient physiotherapy programme providing the same intervention to all patients with a certain severity of symptoms, unrelated to part of the body affected.

In light of this, various authors have suggested stratification of the ICU cohort by factors that correlate with levels of physical morbidity post-ICU (Iwashyna 2012; Hodgson and Cuthbertson 2016; Herridge et al. 2016). They aim to describe the differing recovery trajectories following ICU for specific groups of patients. In particular, allowing identification of those at risk of prolonged severe morbidity. Cuthbertson hypothesises three distinct group of ICU survivors – those with broadly preserved function, those with some functional decline, and those who either do not survive, or do so with severe disability (Cuthbertson 2016). He identifies pre-hospital morbidity, acuity of illness and overall hospital exposure as discriminatory factors to identify these distinct groups; citing work by Iwashyna and colleague (2010), who identified a link between pre-hospital functional trajectory and post- ICU function in a cohort of patients with sepsis, as well as that of Lone et al. (2016) and Ou et al. (2016). Iwashyna (2012) suggested three possible trajectories of recovery following critical illness: ‘The big hit’, an initial loss of function as a result of acute illness, followed by a period of recovery to near pre-morbid level; ‘the slow burn’, in which the initial impact of critical illness permanently alters a person’s functional trajectory to one of decline; and the ‘relapsing recurrences’, similar to that understood in populations with multiple sclerosis, or chronic obstructive pulmonary disease (COPD). He discusses these in the context of a study by Woon and colleagues (2012) examining cognitive outcomes following ICU, which demonstrates subgroups of patients following each of these trajectories. However, he argues that they hold true for physical functioning also, and emphasises the importance of an ‘appropriately defined patient population’ when designing controlled trials in critical care (Iwashyna 2012). This could be crucial when assessing physical rehabilitation interventions in those with critical illness, as the impact of a given intervention may depend on the recovery trajectory of each patient, and the extent to which this is modifiable.  If early physical rehabilitation does improve functional outcome, perhaps it is only in those on a certain trajectory – the ‘big hit’, or broadly preserved function, say. If we were able to identify these groups, then interventions could be assessed against their ability to modify these trajectories. It is worth recalling that both the trials led by Dong et al. (2016) and Schweickert et al. (2009) demonstrating positive impact of early rehabilitation interventions, excluded individuals with poor pre-morbid function from their study sample.

Herridge and colleagues have performed the most thorough stratification analysis of an ICU patient cohort, retrospectively examining outcomes of almost 400 ICU patients who had received a week or more of mechanical ventilatory support (Herridge et al. 2016). This was a diverse sample of medical and surgical patients, from five different ICUs, therefore should be generalisable to most ICU populations. They found that age and ICU length of stay (LOS) could be used to divide the sample into four distinct ‘disability risk groups’, where LOS is defined as short if less than two weeks, long if more than. These groups, in order of risk of physical morbidity, were: young (less than 42) with short LOS; mixed-age and variable LOS (either older than 42 with short LOS, or less than 45 with long LOS); older (46–66 years) with a long LOS; or oldest (greater than 66) with long LOS. Group membership predicted functional ability at one-week post ICU discharge, as measured by the functional independence measure (FIM). This score in turn, was predictive of FIM score at one-year post critical care. Therefore, age and LOS alone were able to predict physical functioning at one year following ICU discharge. This is a significant result, as it allows those designing future interventional trials to perform subgroup analysis by disability risk group, thereby assessing the impact of a given intervention on the predicted one-year recovery trajectory of physical functioning. This analysis differed in that it did not identify pre-ICU morbidity as a discriminatory factor explicitly. However, the lowest risk group had the lowest level of pre-ICU morbidity (estimated by comorbidity indices and work status), and the highest risk group the highest level. Indicating that age, LOS and pre-ICU morbidity are closely linked.

Accurately assessing the pre-admission morbidity burden of an ICU cohort is challenging. Ideally, all patients would be assessed over a period of time prior to ICU admission, to give an indication of functional trajectory prior to the onset of critical illness. However, the un- predictable nature of acute illness and therefore number of unplanned admissions to ICU renders this impractical, if not impossible. There may be some sub-populations of patients for whom these data are already collected as part of the management of a long-term condition; those with COPD, for example, for whom measures of functional ability and disease burden are routinely measured. These could give an indication of morbidity severity prior to ICU-admission for this group. For the majority of patients, surrogate measures must be used to retrospectively attempt to assess pre-ICU morbidity, such as work status, or an estimated functional measure based on proxy report (Schweickert et al. 2009; Herridge et al. 2016).

If factors such as pre-ICU morbidity, age and LOS combined were able to create a model with consistent predictive ability for trajectory of physical functioning post-ICU, then it may also identify groups of ‘responders’ versus ‘non-responders’ to early physical rehabilitation interventions. For example, perhaps older patients with poor pre-morbid function, who require a prolonged critical care stay, will not regain the physical function lost as a result of their critical illness, regardless of exercise-based intervention. It is important in this con- text to consider both appropriate utilisation of resources – targeting treatments at those most likely to benefit, and treatment burden for individual patients. Qualitative analysis performed by Corner (2017) examined the experience of early rehabilitation from the point of view of individuals who have survived ICU. This highlighted that physical rehabilitation on ICU is something that patients find fatiguing, and often dread. Therefore, it is imperative that the critical care community work to identify groups of potential ‘non-responders’ to this intervention. It is ethically problematic to ask patients to engage in a treatment that is difficult and tiring, for potentially little or no gain. If such groups were identified, then it could be argued that we are at a point of equipoise – should any physical rehabilitation be offered on critical care to those with risk factors making them likely to be non-responders to this intervention? Rather than asking ‘when’ we should commence early rehabilitation, a trial comparing current usual care of physical rehabilitation on critical care, with no physical rehabilitation may be warranted.

As the number of people surviving an ICU admission continues to grow, the population of individuals living with the residual effects of critical illness, including ICU-AW, grows also. Although it is well established that the pathophysiological mechanisms driving ICU-AW begin early in critical illness, there is currently insufficient evidence to recommend any specific treatment strategies to combat it. The question of timing – when early rehabilitation should commence on ICU – cannot be answered until one or more interventions, undertaken at any time, are found to consistently improve outcomes. An appraisal of the current evidence base with a focus solely on timing is problematic, because of the multiple other methodological differences between any given studies. Clarification, and perhaps standardisation, of usual care for control groups; as well as stratification of ICU populations according to factors that influence functional outcome trajectory, may aid future trials assessing the impact of rehabilitation interventions. Interestingly, NICE recommended ‘individualised’ rehabilitation plans in their guideline in 2009 (NICE 2009). Given the failure of trials to date to demonstrate benefit of interventions undertaken with heterogeneous cohorts of critical care patients, perhaps further individualisation is the answer. A focus on patients’ individual characteristics may be essential in designing an effective rehabilitation intervention. After which, optimisation of timing could be considered.