The impact of an exercise intervention on frailty levels in hospitalised older adults: secondary analysis of a randomised controlled trial

Abstract

Key Points

• Frailty is reduced by an individualised exercise intervention in older adults admitted to an acute care ward.

• Baseline levels of frailty moderate the effect of the intervention for some health outcomes.

• There is a potential to reverse frailty even for those in the worst health state.

Introduction

Bed rest when recovering in acute care plays a major role in deconditioning, resulting in significant muscle wasting [1] and the development of other symptoms such as joint stiffness, pain, fatigue, etc. [2]. This has informed an increasing awareness of the need to replace prolonged periods of bed rest with physical activity [3] or exercise, since there is evidence that this therapy during hospitalisation is effective; even in those in poor health [4, 5]. Moreover, exercise can diminish damage from acute disease and hospitalisation [6]. Many older people admitted to hospital who appear to benefit from interventions to prevent deconditioning—notably including early mobilisation—have attributes that suggest underlying frailty: mean age 80 years, a female preponderance, coming from nursing homes or other facilities and a high prevalence of impaired function or mobility [7]. Even so, whether exercise treatment impacts the degree of frailty or frailty impacts the response to treatment has not been determined.

Although frailty is not a disease, it affects the expression of many age-related illnesses as well as the response to treatment [8, 9]. For example, for some frailer individuals, interventions that may seem harmless could be considered physiologic stressors. One such instance is that after a short period of strength exercise in frail older adults, autophagy of muscle cells is activated, impacting the overall health of muscle and therefore mobility [10]. Similarly, increased inflammatory responses have been described following an acute bout of aerobic exercise [11]. It is also less clear whether individuals with greater degrees of frailty have any response to an intervention [12, 13].

The objective of this study is to examine the impact of an exercise intervention on frailty levels in older adults admitted to an acute care ward, and to determine the impact of baseline frailty levels on the effectiveness of this intervention on changes in frailty, physical function (Barthel Index, BI), physical performance (Short Physical Performance Battery, SPPB) and other health outcomes.

Methods

Design, sample, setting and procedures

This is a secondary analysis of a randomised controlled clinical trial conducted in a hospital setting. The objectives and procedures of the main trial are detailed elsewhere [14]. In brief, it was completed in an acute care ward of the geriatrics department of a tertiary hospital, which recruited individuals 75 years and older from February 2015 to August 2017. A total of 370 older adults were randomised to receive the intervention (n = 185) or usual care (n = 185). Participants were not blinded but were instructed not to discuss their assigned group. Here, only those with complete data to calculate a Frailty Index (FI) score are included.

Intervention

Each session was performed in a room equipped ad hoc in the geriatric care unit by research staff; exercises were adapted from the ‘VIVIFRAIL’ multicomponent physical exercise program [15]. Morning sessions included individualised supervised progressive resistance, balance and walking exercises. Participants performed three exercises at a high speed—to optimise muscle power output—involving mainly lower-limb muscles (squats rising from a chair, leg press and bilateral knee extension), and one exercise involving the upper-body musculature (seated bench press). The resistance exercises were tailored to the individual’s functional capacity using variable resistance training machines aiming at 2–3 sets of 8–10 repetitions with a load equivalent to 30–60% of the one-repetition maximum. Balance and gait retraining exercises progressed in difficulty and included: semi-tandem foot standing, line walking, stepping practice, walking with small obstacles, proprioceptive exercises on unstable surfaces (foam pads sequence), altering the base of support and weight transferring from one leg to the other. The evening sessions consisted of functional unsupervised exercises using light loads (e.g. knee extension and flexion, hip abduction), and walking along the corridor. Adherence to the intervention for the sample included in the secondary analysis was like the adherence of the full sample that was included in the original study (95.8% morning sessions and 83.4% evening sessions) [14].

Outcomes

Assessing the impact of the exercise intervention on frailty, measured with the FI, was the primary outcome. The next step involved assessing the impact of the exercise intervention by admission frailty levels on changes in frailty, physical function (BI) and physical performance (SPPB) at hospital discharge. Finally, the impact of the intervention stratified by frailty levels on secondary outcomes was also assessed (see below).

Frailty index

Variables available both at admission and discharge were screened to create an FI, according to standard procedures [16]. To have an optimal balance between number of items and representation of different health domains in the FI, one criterion from the standard procedure was modified: ≤20% missing values for individual variables instead of ≤5%. This modification has been reported elsewhere [17].

Out of 75 potential items, 63 were included: Supplementary Table 1 details the questions, coding and frequency of each included item. Each item was coded according to conventional cut-off points—if available—with 0 representing no deficit and 1 representing deficit present (with the possibility of intermediate values). For those variables with no available cut-off value, terciles of the variable’s distribution were used. With these coded variables, the FI was calculated as the number of deficits present for any given individual divided by the total number of deficits, giving a score from 0 (lowest possible frailty burden) to 1 (highest possible frailty burden). We did not calculate FI scores for individuals with 10 or more missing items (>20%). Age, sex or number of people in the intervention group were not significantly different between those with and without an FI score; only cause of admission and quality of life score were significantly different in people without an FI score (Supplementary Table 2, Supplementary data are available in Age and Ageing online).

In addition to the continuous FI score change, we also examined whether people changed FI scores above the clinically meaningful level: better = FI score reduced by at least 0.03, no meaningful change = FI change of −0.031 to 0.29 and worse = FI score increased by at least 0.03 [18]. Moreover, to examine the impact of admission frailty levels on the effectiveness of the intervention, three groups were formed according to FI scores: <0.2, 0.2–0.29 and ≥0.3. When physical function and physical performance were assessed as outcomes, their respective variables were excluded from the FI, resulting in a 53-item FI for BI and a 59-item FI for SPPB. This same approach was used for other outcomes that were initially included in the FI. For example, when examining delirium as the outcome, this condition was removed from the FI.

BI and SPPB

Changes in the BI and the SPPB were categorised as in the original work [14]: SPPB much better = improvement of >3 points, better = improvement of 2 points, somewhat better = improvement of 1 point, unchanged = no difference and worse = <0; and for the BI much better = improvement of >10 points, better = improvement of 10 or less points, unchanged = no difference, worse = decline 10 or less points, much worse = decline of >10 points.

Secondary outcomes

Secondary outcomes were extracted from participant medical records including cognition, depression, quality of life, delirium, handgrip strength, length of stay, incident in-hospital falls, place of discharge, hospital readmission, emergency room visit after discharge and mortality (the last three outcomes were assessed up to 58 months post-discharge).

Cognitive status was assessed with the Mini-Mental State Examination, depression with the 15-item Geriatric Depression Scale (GDS), quality of life with the EuroQoL 5D visual analogue scale, delirium with the Confusion Assessment Method and handgrip strength with a pneumatic dynamometer; these variables were available at admission and discharge.

Statistical analysis

Descriptive statistics (mean, standard deviation [SD], median, maximum and minimum) for the FI were calculated. A kernel density plot was included to visualise the distribution of the FI. In addition, the logarithmic transformation of the FI was plotted against age and further stratified by sex. Survival analysis was performed comparing the three levels of frailty (<0.2, 0.2–0.29 and ≥0.3).

We performed an intention-to-treat analysis in participants with a complete FI at admission (n = 323), carrying forward the admission observation when discharge FI data were missing. Descriptive statistics were provided for the whole sample and by intervention group using means and SD for continuous variables and absolute/relative frequencies for discrete ones. Student’s t-test for comparison between groups and admission/discharge scores, along with paired t-test were used to assess statistical significance. Descriptive analyses were conducted both at admission and at discharge. To visually assess the difference for primary outcomes (i.e. FI, SPPB and BI) by baseline frailty level and by intervention assignment, bar graphs were used.

Linear mixed effect models were used to explore the mean differences in outcomes (continuous variables for both primary and secondary outcomes) using fixed effects for sex and time (length of stay in days) by intervention interaction, and random effects for the individuals. Logistic regressions were also used to examine the impact of the intervention on FI (worse versus unchanged/better), BI (worse/much worse versus much better/better/unchanged) and SPPB (worse versus much better/somewhat better/better/unchanged). Except for mortality in which Cox regression was used, for all other secondary outcomes logistic regression models were used.

Sensitivity analyses

Sensitivity analyses were performed to account for missing data at discharge. For this purpose, regression models with different approaches were performed: (i) per protocol: only participants with FI data at both admission and discharge (n = 279), (ii) FI missing data at discharge imputed with a negative outcome (worse FI scores at discharge compared with admission) and (iii) FI missing data at discharge imputed with a random outcome (better, worse or same FI scores at discharge compared with admission). Moreover, since the FI contained both the SPPB and the BI items and we already knew that these two measures improved over time [14], an FI without these variables was also constructed for sensitivity analysis: a 50-item FI without SPPB and BI items.

Results

FI characteristics

We were able to calculate admission FI scores for 323 out of 370 individuals (87.2%). The average age of the included sample was 87.1 years (± 4.8 SD) and 56.3% were female. Most individuals had primary or secondary education (87.6%) and the main cause of admission was a cardiovascular problem (38.4%; Supplementary Table 2, Supplementary data are available in Age and Ageing online).

The overall mean FI score at admission was 0.26 (±SD 0.1); range 0.05–0.57. The FI was higher in those >90 years and in females (Supplementary Table 3, Supplementary data are available in Age and Ageing online). The FI distribution was normal for both males and females (Supplementary Figure 1, Supplementary data are available in Age and Ageing online) and FI scores increased with age (Supplementary Figure 2, Supplementary data are available in Age and Ageing online). Individuals with an FI score ≥ 0.3 had higher mortality risk compared to those with lower scores (HR of 2.13, 95% CI 1.25-3.63, P = 0.005; Supplementary Figure 3 and Supplementary Table 4, Supplementary data are available in Age and Ageing online).

Impact of the exercise intervention on frailty

Participants in the intervention group improved their FI scores from 0.26 (± 0.1 SD) to 0.2 (± 0.1 SD), whereas the control group’s FI scores at admission and discharge worsened from 0.25 (± 0.1 SD) to 0.27 (± 0.1 SD; Table 1). More participants in the intervention group experienced a clinically meaningful improvement in their FI score during hospitalisation compared with the control group (Figure 1).

Figure 1

Clinically meaningful change in frailty index score for the intervention and control groups. Note: better = FI score reduced by at least 0.03, no meaningful change = FI change of −0.031 to 0.29 and worse = FI score increased by at least 0.03.

Open in new tabDownload slide

In the linear mixed effects model, FI change was greater in the intervention group compared to the control group with a beta coefficient of −0.06 (95% confidence interval [CI] −0.08 to −0.03, P < 0.001; Supplementary Table 5, Supplementary data are available in Age and Ageing online). The intervention group was also less likely to have a worse FI at discharge compared with the control group (adjusted OR 0.06, 95% CI 0.01–0.23, P < 0.001; Supplementary Table 6, Supplementary data are available in Age and Ageing online).

More people maintained or improved their FI scores among the FI ≥ 0.3 group compared with the lower FI groups at admission (Figure 2a). Logistic regression analyses also showed that among people with FI ≥ 0.3 the intervention group was less likely to worsen their FI compared with the control group (adjusted OR 0.01, 95% CI 0.01–0.37, P = 0.012); the difference was not statistically significant for the FI < 0.3 groups (Table 2). The linear mixed model effects also showed that the frailty change was higher for the intervention group across all levels of frailty (Supplementary Table 7, Supplementary data are available in Age and Ageing online).

Figure 2

Clinically meaningful changes in (a) Frailty; (b) Barthel Index and (c) Short Physical Performance Battery during hospitalisation for the intervention and control group by level of frailty. Note: Clinically meaningful changes for frailty: better = FI score reduced by at least 0.03, no meaningful change = FI change of −0.031 to 0.29 and worse = FI score increased by at least 0.03; (b) BI: much better = improvement of >10 points, better = improvement of 10 or less points, unchanged = no difference, worse = decline 10 or less points, much worse = decline of >10 points; (c) SPPB: much better = improvement of >3 points, better = improvement of 2 points, somewhat better = improvement of 1 point, unchanged = no difference and worse = <0.

Open in new tabDownload slide

Changes in BI and SPPB by admission frailty level

Over 80% of the people in the intervention group improved or maintained their BI and SPPB scores during hospitalisation (Figure 2b and c). Among people with FI ≥ 0.2, those who participated in the intervention group were more likely to have a significant change in BI compared with the control group; for people with FI < 0.2 the difference between the intervention and control groups was not statistically significant (Table 2). The linear mixed models showed that there was no significant change in BI scores in the intervention compared with the control group for all levels of frailty (Supplementary Table 8, Supplementary data are available in Age and Ageing online). There was no difference between the intervention and control groups regarding their impact on SPPB worsening irrespective of baseline frailty (Table 2). However, the linear mixed models showed that there was a significant improvement in SPPB scores in the intervention compared with the control group across all levels of frailty (Supplementary Table 8, Supplementary data are available in Age and Ageing online).

Changes in secondary outcomes stratified by baseline frailty level

The intervention group had lower GDS and higher EuroQoL 5D visual analogue scale scores at discharge compared with the control group (Table 1). In addition, the linear mixed model showed that these lower GDS scores were sustained across all FI categories. The change in the EuroQoL 5D visual analogue scale scores was significantly different between intervention and control groups for the frailest (FI ≥ 0.3) participants (Table 3). Other secondary outcomes had no significant differences in the change scores between the two groups.

Sensitivity analysis

When using an FI without the SPPB and BI items (50-item FI), FI change was still greater in the intervention group with a beta coefficient of −0.06 (95% CI −0.09 to −0.03, P < 0.001; Supplementary Table 9, Supplementary data are available in Age and Ageing online). Also, similar estimates were obtained when not using intention to treat analysis (per protocol, worst outcome imputation and random outcome imputation; Supplementary Table 10, Supplementary data are available in Age and Ageing online) and when stratifying by admission FI levels (Supplementary Table 11, Supplementary data are available in Age and Ageing online).

Discussion

The present study shows that a multicomponent individualised physical exercise intervention is effective for reversing frailty in acutely hospitalised older patients, which builds on previous evidence that showed positive results of this exercise program [14, 19–21]. Our data are in line with implementing an approach to measure finer degrees of frailty and indicates that a frailty cut-off point to initiate intervention is not needed. Indeed, limiting an intervention to a frailty level could exclude individuals who could still benefit from its implementation [22, 23].

Frailty is a common condition in older adults; however, there is no consensus on how to operationalise it in clinical practice [24]. Notwithstanding, a focus on the continuous deterioration of multiple systems in the ageing adult (i.e. the FI as a measure of multisystem deficit accumulation) captures the benefit of a complex intervention (e.g. multicomponent exercise). Moreover, past reports evidenced that the FI is sensitive to change when exercise is used as treatment [25]. Other outcomes, such as function may not be sensitive enough to change, specifically in individuals with poor baseline health status; this could explain why we found no significant change in BI scores in the intervention compared with the control group for all levels of frailty.

Evidence on the benefit of exercise on frailty keeps growing. For example, a recent systematic review showed a significant impact of physical activity on reducing frailty [26]. The pathway from which exercise leads to an improved health is believed to be multifactorial [27, 28], having the potential to benefit multiple domains. Moreover, physical activity combined with other lifestyle behaviours has shown to have an additive effect [29–33].

Our study has limitations. This is a secondary analysis of a randomised trial not intended to measure changes in frailty. We found no differences on some outcomes, which could be related to that the study was not originally designed to examine these outcomes and that a bigger sample size may be needed for this analysis. Another limitation is the difficulty of collecting at least 30 variables in routine care to construct an FI [34]. However, when routine hospital/administrative data are available, it becomes more feasible to use an FI, even in critical care [35]. It is always challenging to minimise participant drop-out in addition to addressing missing data, which is an important consideration when calculating an FI. To address this problem, a sensitivity analysis was performed, along with an FI calculation that was more flexible to data missingness to maximise inclusion of participants. In addition, individuals with FI (the sub sample of this secondary analysis study), had similar characteristics with the original sample [14]. However, our results might be affected by the characteristics of our population: very old people, fairly educated and very frail.

The results from this study should be explored further, since there is no certainty that improving physical performance and function during hospitalisation will translate into better long-term outcomes. On the contrary, frailty improvement in those with low admission levels could be associated with better long-term outcomes—or vice versa. Nevertheless, these results certainly pose intriguing questions about how interventions could be tailored to the individual based on frailty levels, to optimise the effect of that given intervention. Meanwhile, as physical exercise plays an essential role in preventing the functional and cognitive decline associated with hospitalisation in older adults, exercise and early rehabilitation programs are among the interventions through which functional decline is likely to be best prevented in hospitalised frail older adults [36]. Implementing an exercise program in routine care could increase health care costs if additional health care staff are needed. Pragmatic trials should examine the efficacy of these programs, especially across health care systems with different characteristics. Not only the exercise program needs to be individualised, but the implementation of this program should also be tailored to the specific characteristics of the health care system. It is also still to be determined what would be the impact of this exercise intervention on other older adults such as centenarians, supercentenarians, or younger older adults.

The convergence of frailty as a chronic steady state with the sudden insult of acute illness poses challenges to clinicians, since it will require additional skills that go beyond the classical paradigm of one-cause illness. It is well-known that hospitalised patients spend most of their time in bed during their hospital stay and most health care staff do not have the skills and training to prescribe exercise [37]. This is one of the biggest problems in implementing these interventions outside research settings. Future research should shift to pragmatic trials and educational programs that move from experimental settings to the real-world action. It is widely known that physical activity improves health across the life course. The time has come to start implementing these strategies in a systematic way.

Acknowledgements

We thank the participating patients and their families for their confidence in the research team. Data sharing statement—the data underlying the results presented in this study can be made available upon request from the Complejo Hospitalario de Navarra, by contacting Antonio García Hermoso Antonio.garciah@unavarra. Public sharing of individual participant data was not included in the informed consent form of the study, but data can be made available to interested researchers who meet criteria for access to de-identified data.

Declaration of Conflicts of Interest

MUPZ, NMV, SK, MI and OT have no competing interests.

KR has asserted copyright of the Clinical Frailty Scale through Dalhousie University’s Industry, Liaison and Innovation Office. Use is free for education, research and not-for-profit health care. Users agree not to change or commercialise the scale. In addition to academic and hospital appointments, KR is cofounder of Ardea Outcomes, which (as DGI Clinical) in the last 3 years has contracts with pharma and device manufacturers (Biogen, Hollister, Novartis, Nutricia, Roche and Takeda) on individualised outcome measurement. In 2019, KR was paid an honorarium for an interview with Biogen. In 2020, he attended an advisory board meeting with Nutricia on dementia, and chaired a scientific workshop and technical review panel on frailty for the Singapore National Research Foundation. Otherwise, any personal fees were for invited guest lectures, rounds and academic symposia, received directly from event organisers, for presentations on frailty. KR is associate director of the Canadian Consortium on Neurodegeneration in Aging, which is funded by the Canadian Institutes for Health Research, the Alzheimer Society of Canada and several other charities.

Declaration of Sources of Funding

MI and NMV obtained the funding for the original study by: Gobierno de Navarra project Resolución (grant 2186/2014); ‘Beca Ortiz de Landázuri’ award as the best research clinical project in 2014; ‘la Caixa’ Foundation (ID 100010434), under agreement LCF/PR/PR15/51100006. This work was supported by Research Nova Scotia (MED-EST-2017-1241) and by a University Internal Medicine Research Foundation Fellowship awarded to MUPZ.

References