Description of the condition
Frailty is a common geriatric syndrome, characterised by decline across multiple body systems, causing decreased reserve and increased vulnerability to adverse health outcomes (Fried 2004). It is estimated that seven per cent of the community-dwelling population over 65 years in America are frail (Fried 2004). Frailty is independently predictive of falls, worsening mobility, deteriorating functioning, impaired activities of daily living and death (Fried 2004; Pel-Littel 2009). Frailty is therefore costly to the individual, their families and society. The economic and healthcare burden will increase as the population ages (Howe 1997).
Reducing functional decline should be considered a key goal of intervention for frail older people. High rates of functional decline are evident in frail, community-dwelling older people (Gill 1995), with an estimated 27% of frail older people experiencing concurrent disability (Fried 2004). Interventions that slow functional decline in the frail population will impact upon morbidity and mortality (Gill 2004).
In the International Classification of Functioning, Disability and Health (ICF) (WHO 2001), 'functioning' is an umbrella term for function at the level of the tissue or organ, the individual and the individual in a social context, called body functions/structures, activity and participation respectively. The evidence to guide intervention for increasing functioning in frail older people is both sparse and conflicting. Many trials systematically exclude frail older adults, largely due to their co-morbidities and difficulties associated with their recruitment (Ridda 2008).
Description of the intervention
We will examine the effect of interventions that target improvement in mobility. In accordance with the ICF definition of mobility in the activity and participation domains, mobility is defined in this review as changing and maintaining a body position, walking and moving. Common interventions used to increase mobility include progressive resistance training and balance-challenging exercises. Multifactorial interventions will be included if the majority of the intervention aims to improve mobility, and interventions that involve performance of mobility tasks will be included if participants' performance is monitored and reviewed during the period of intervention delivery.
How the intervention might work
Studies have shown that frail older people can improve their degree of frailty (Gill 2006), functioning and health, and the benefits of exercise on health status in community-dwelling frail older people may be greatest in those who are more frail at baseline (Hubbard 2009). Although some interventions delivering exercise to frail older people have demonstrated improved mobility (Fiatarone 1994), physical functioning (Binder 2002; Worm 2001) and reduced rate of functional decline (Gill 2002), other trials have found no effect on functioning or disability (Chin 2001; Latham 2003; Rydwik 2004), and inconsistent effects on falls outcomes (Faber 2006; Wolf 1996). A systematic review of interventions targeting disability in community-dwelling frail older people indicated that the effect on performance of activities of daily living was inconclusive (Daniels 2008). It is unclear which parameters of a mobility training intervention (e.g. dose, degree of difficulty) are necessary to improve mobility and functioning in frail older people. In addition, the safety of such interventions is unclear, with reports of increased musculoskeletal injury in frail older people (Latham 2003).
Cochrane reviews describing studies of exercise delivered to older people have implications for the frail older population. Recent Cochrane reviews indicate exercise interventions can improve strength (Liu 2009) and balance (Howe 2011) and prevent falls in older people (Cameron 2012; Gillespie 2012). These outcomes have potential consequences for functioning and although the impact of exercise on functioning is less clear (Ashworth 2005; de Morton 2007), we do know that specific interventions, such as progressive resistance strength training (Liu 2009), can increase function in older people.
This Cochrane review differs from other Cochrane reviews of the effects of various exercise interventions in older people. While a proportion of participants in previous Cochrane reviews were frail (Ashworth 2005; de Morton 2007; Cameron 2012; Gillespie 2012; Howe 2011; Liu 2009), our review will be restricted to frail older people. We will focus on frail older people living in the community, rather than the hospital setting studied by de Morton and colleagues (de Morton 2007). We will build upon previous Cochrane reviews that evaluated the effect of exercise intervention on measures of functioning (Howe 2011; Liu 2009) by including by including a broader range of interventions that may affect mobility and functioning outcomes. Finally, the main outcome will be that of functioning, as opposed to narrower outcomes such as balance (Howe 2011) and falls (Cameron 2012; Gillespie 2012).
Why it is important to do this review
If the proposed systematic review and meta-analysis determine that mobility training interventions are effective in increasing functioning, the clinical implications for the frail population will be significant. An intervention that increases functioning in this population will have the potential to impact on hospitalisation and admission to residential care facilities, along with the associated costs to government and society.
A systematic review is needed to identify the trials in this field and a meta-analysis is required to summarise the evidence for researchers, professionals, policy makers and others with an interest in this area.
To summarise the evidence for the benefits and safety of mobility training on overall functioning and mobility in frail older people living in the community.
Criteria for considering studies for this review
Types of studies
We will include randomised controlled trials and controlled clinical trials that use quasi-randomisation methods (e.g. allocation by date of birth or alternation).
Types of participants
Men and women aged 65 and older, who live in the community and who are described as being frail.
We will include trials only when the participants are described as frail and the authors justify the use of the term 'frail' (for example, the Cardiovascular Health Study Criteria (Fried 2001), Frailty Index score (Mitnitski 2004), difficulty performing specified tasks, specified age or other measures of vulnerability).
Trials will be included if participants are described as older adults, aged, seniors, geriatric, elderly or all over the age of 60, and the mean or median age of participants is over 65 years.
We will only include trials where the majority of participants live, or plan to live, in the community. The community will be defined as residing either at home or in places that do not provide rehabilitative services or residential health-related care, for example retirement villages, sheltered housing or hostels will be included. In the event that a trial includes participants living both in the community and higher-dependency residences, for example nursing care facilities or hospitals, participants will be included in this review only if data are provided for the subgroup living in the community.
This review will focus on trials delivering interventions to frail people with a wide range of co-morbidities. We will exclude trials of a single diagnostic group (as listed in the International Classification of Diseases 10
Types of interventions
The intervention must target improvement in mobility, as evidenced by measurement of mobility as an outcome. Mobility will be defined according to the ICF: changing and maintaining a body position, walking and moving. Interventions used to improve mobility commonly include progressive resistance training, balance-challenging exercises, and mobility training that is prescribed based upon individualised assessment and ongoing reassessment.
If an intervention contains multiple elements, we will include only trials where there is an aim (explicit or implicit) to improve mobility for the majority of the intervention. Training that involves performance of a task must have a monitored component, with review of participants' performance during the period of intervention delivery. The mobility training intervention may be augmented by cognitive interventions such as goal setting, education and motivational interviewing.
We will include trials where the intervention is compared with a control group that receives no intervention, usual care, sham exercise (the exercise appeared to be of insufficient intensity and progression to have beneficial effects on balance and mobility) or a social visit. Trials comparing two or more interventions will also be included if the difference between the intervention and control groups is exercise.
Types of outcome measures
The following are the outcomes for the review:
- Adverse effects of the interventions
- Admission to hospital/nursing care facility
Outcomes to be included in the 'Summary of findings' table
- Adverse effects
- Admission to hospital/nursing care facility
The main outcome of interest will be functioning, measured in terms of the activity or participation levels of the ICF (WHO 2001). We will assess functioning as a continuous variable where possible.
Trials will be excluded if they report outcomes at the 'body structure/function' level of the ICF (such as proprioception or strength), but do not report outcomes at the level of activity or participation. Balance as an outcome will be treated as an activity and therefore included. We will classify instruments containing multiple ICF concepts, for example the EuroQol (EQ-5D) and SF-12, as health-status instruments and will not consider them measures of functioning.
The second outcome of interest will be mobility, defined in the ICF as changing and maintaining a body position, walking and moving (WHO 2001). We chose to report on measures of mobility that relate to ambulation (as opposed to rolling, standing, driving) as this is most relevant to everyday activity. We will assess mobility using physical performance measures.
If meta-analysis is deemed appropriate, we will select outcomes to facilitate quantitative pooling. Trials are likely to have measured functioning using many different instruments. To enable inclusion of as many trials as possible in the meta-analysis, we will use the method reported by Allen et al (Allen 2011) to pool results across multiple outcomes. If a trial reports results for more than one outcome measure, we will use the highest priority functioning measure and the highest priority mobility outcome in the analysis. The highest priority functioning measure will be that which encompasses the most activities and participation domains of the ICF. The highest priority mobility measure will be that which most reflects everyday mobility. The analysis will involve the pooling of the most comprehensive functioning measure from each trial and pooling of the highest priority mobility measure from each trial. The order of priority is based upon consensus of the review authors.
The order of priority follows:
- Functioning: Lawton Independent Activities of Daily Living Index, The Groningen Activity Restriction Scale, Barthel Index, Functional Independence Measure.
- Mobility: Timed Up and Go test, walking distance, walking speed, turning time, timed sit to stand, stair climbing, Functional Reach test, single leg stand time, sway.
We will contact authors directly if it is not clear whether a trial meets all the criteria for inclusion in this review.
Search methods for identification of studies
We will search the Cochrane Central Register of Controlled Trials (The Cochrane Library current issue), MEDLINE (OVID ONLINE) (from 1950 to present), EMBASE (1980 to present), AMED (1985 to present), PsycINFO (1967 to present), PEDro (1929 to present), ClinicalTrials.gov, and the WHO trials portal. There will be no language restrictions. The search strategy is in Appendix 1.
Searching other resources
We will check reference lists of all included studies and will identify further trials by contacting researchers in this field.
Data collection and analysis
Selection of studies
One review author (NF) will screen the titles and abstracts for inclusion and full papers will be retrieved for all trials that potentially meet the inclusion criteria. Two review authors (IC or CS = expert in the field, NF) will independently screen all full papers for compliance with the eligibility criteria, using a pre-tested form. Review authors will not be blinded to author and journal of publication. Disagreement will be resolved by consensus, or third party adjudication.
Data extraction and management
Two review authors will independently extract data using a pre-tested data extraction form.
Disagreement will be resolved by consensus, or third party adjudication.
We will extract the following study characteristics:
1. Methods: study design, total duration of study, details of any 'run in' period, number of study centres and location, study setting, withdrawals, and date of study.
2. Participants: N, mean age, age range, sex, disease duration, severity of condition, diagnostic criteria, important baseline data describing degree of frailty; inclusion criteria, and exclusion criteria.
3. Interventions: intervention: content of intervention, dose of intervention (intensity, frequency, duration, progression), supervision (self, individual, groups), supervisor (self, peer, healthcare professional, exercise specialist), number of supervisors, setting (home, community/gym, healthcare provider) and adherence rate. Comparison.
4. Outcomes: primary and secondary outcomes specified and collected, and time points reported. Number of events and number of participants per treatment group for dichotomous outcomes, and means and standard deviations and number of participants per treatment group for continuous outcomes.
5. Characteristics of the design of the trial as outlined below in the 'Assessment of risk of bias in included trials' section.
6. Notes: funding for trial, and notable declarations of interest of trial authors.
7. The primary time period of outcome assessment will be completion of the intervention programme. We will also collect time periods less than 12 months after intervention ceases and more than 12 months after intervention ceases.
Main planned comparisons
1. Mobility training intervention versus control (either no intervention, usual care, sham exercise, social visit)
2. Mobility training intervention versus other intervention not including exercise.
Assessment of risk of bias in included studies
Two review authors will independently assess risk of bias will using the Cochrane 'Risk of bias' tool (Higgins 2011b). We will assess the five domains in the 'Risk of bias' tool (sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessors, incomplete outcome data and selective outcome reporting). We will also assess trials using the PEDro rating scale (Maher 2003). Authors will not assess their own trials. Review authors will not be blinded to author and source institution. Disagreement will be resolved by consensus, or third party adjudication.
We will consider blinding separately for self-reported outcomes (e.g. function) and outcomes assessed by another outcome assessor.
Measures of treatment effect
Data relating to functioning
We expect most of the functioning scales used to be continuous.
We will use standardised mean difference (SMD) as a summary statistic as we anticipate different outcome measures will be reported across trials. We will extract means and standard deviations of pre-therapy scores, post-therapy scores and change scores and their respective standard deviations, where available. If the standard deviation of change score can be reliably obtained from other statistics (such as confidence interval (CI) or P value) then we will use this. Where the study sample size exceeds 70, and the median and inter-quartile range are reported, we will use the median as an approximation of the mean and estimate the inter-quartile range as 1.35 standard deviations (Higgins 2011a; Hozo 2005). We will use unadjusted values in the meta-analysis, unless the adjustment was for clustering. In the case that only adjusted values are reported, we will enter these values. If scales decrease with increasing function, we will multiply the mean values by -1 to correct for differences in the directions of scales. When available, we will include baseline and follow-up measurements of the outcome as an analysis of covariance (ANCOVA) and enter in the meta-analysis using the generic inverse-variance method. When only follow-up values are available, we will use these. We will combine studies in the meta-analysis, to calculate the standardised mean difference.
SMD will be back-translated to clinically meaningful units (such as walking speed and minimum chair height from which a person can stand up) by multiplying the SMD by a typical among-person standard deviation for each particular scale (derived from a representative observational study). If no representative observational study exists, we will use the control group baseline SD from the most representative trial (Schünemann 2011b).
Data relating to mobility
We expect most of the mobility scales used to be continuous.
We will use the standardised mean difference to summarise the size and direction of the effect (using the method described above).
Data relating to death, adverse outcomes, falls, hospitalisation and movement to nursing care facilities
We will treat these outcomes as dichotomous.
For each trial with data available, we will calculate the risk ratio (a comparison of the number of subjects in each group with one or more event) and 95% CI using the generic inverse variance option. We will use unadjusted values in the meta-analysis, unless the adjustment was for clustering or unadjusted values are available. We will enter the number of subjects contributing data to each group in the analysis; if this is not known,we will use the number randomised to each group.
Data relating to costs
We will report costs and cost-effectiveness analysis associated with individual trials.
We will conduct all analyses using the Cochrane Review Manager software (RevMan 5) (RevMan 2012), except meta-regression, which we will conduct using the 'metareg' command in Stata v. 11 (College Station, Texas).
Unit of analysis issues
We will use outcomes from all randomised participants in the meta-analysis where possible. If intention-to-treat is not reported, we will use per-protocol results.
In the presence of cluster-randomised trials, we will use an intraclass correlation coefficient (ICC) to enable analysis at the individual level while allowing for clustering in the data. If one of the included trials reports an ICC, we will use this in the analysis. If an ICC is not reported, we will use the ICC of 0.01 (Elley 2005).
In the event of multiple intervention groups, we will combine the groups relevant to the review and containing similar intervention to create a single pair-wise comparison. If the interventions are markedly different, we will include each pair-wise comparison in the meta-analysis separately, with the control group divided evenly among the intervention groups.
Dealing with missing data
Only data from subjects whose outcomes are available will be included in the analysis. We will address the potential influence of missing data in the 'Risk of bias' assessment.
If the standard deviation of change score can be reliably obtained from other statistics (such as CI or P value) then we will use this. Where the study sample size exceeds 70, and the median and inter-quartile range are reported, we will use the median as an approximation of the mean and estimate the inter-quartile range as 1.35 standard deviations (Higgins 2011a; Hozo 2005)
Assessment of heterogeneity
For studies judged as clinically homogeneous, we will assess and quantify the possible magnitude of inconsistency (i.e. heterogeneity) across studies, using the I
Assessment of reporting biases
We will assess reporting bias using Egger's test and visual inspection of the symmetry of the funnel plot.
We will perform meta-analysis using random-effects modelling and will do sensitivity analysis to explore whether fixed-effect modelling gives similar results, to assess the possible presence of small sample bias in the published literature (i.e. in which the intervention effect is more beneficial in smaller studies). In the presence of small sample bias, the random-effects estimate of the intervention is more beneficial than the fixed-effect estimate (Sterne 2011).
If appropriate, we will then pool results of studies with comparable participant characteristics and interventions using the generic inverse variance method in Review Manager 5. We will use the random-effects model to calculate SMD with 95% CI.
Following analysis, we will convert outcomes back to clinically meaningful units, such as walking speed and minimum chair height from which a person can stand up. We will multiply the SMD by a typical among-person standard deviation for each particular scale (derived from a representative observational study). If no representative observational study exists, we will use the control group baseline SD from the most representative trial (Schünemann 2011b).
Summary of findings table
We will present the main results of the review in a 'Summary of findings' (SoF) table which provides key information concerning the quality of evidence, the magnitude of effect of the interventions examined, and the sum of available data on the outcomes functioning, mobility, adverse events, admission to hospital, falls and death, as recommended by The Cochrane Collaboration (Schünemann 2011a). The 'Summary of findings' table includes an overall grading of the evidence related to each of the main outcomes, using the GRADE approach (Schünemann 2011b). In addition to the absolute and relative magnitude of effect provided in the 'Summary of findings' table, for dichotomous outcomes, we will calculate number needed to treat to benefit (NNTB) or the number needed to treat to harm (NNTH) from the control group event rate (unless the population event rate is known) and the risk ratio using the Visual Rx NNT calculator (Cates 2008). For continuous outcomes, we will calculate the NNT using the Wells calculator software available at the Cochrane Musculoskeletal Group (CMSG) editorial office. We will determine the minimal clinically important difference (MCID) for each outcome for input into the calculator. Where direct measurements of important change are absent, we will estimate the MCID as half the baseline standard deviation of raw scores, using the methodology of Norman et al (Norman 2003).
Subgroup analysis and investigation of heterogeneity
We will conduct meta-regression analyses to explore any significant heterogeneity. Based upon a maximum of one meta-regression analysis per 10 trials, we will decide which of the following analyses to conduct.
- Dose of the intervention.
- Interventions specifically prescribed to challenge the participant versus interventions that are not prescribed with the intent to challenge the participant. We will define intent to challenge balance as prescription of at least two of the following: minimising upper limb support, movement of the centre of mass and narrowing of the base of support (Sherrington 2008). Intent to challenge muscle strength is defined as addition of resistance to muscle contractions.
- Frailty defined using a validated clinical tool versus frailty defined using an impression such as age, difficulty performing a task at home, recent hospitalisation.
- Adherence to intervention. Measurement will be dichotomous (yes/no) or in quartiles (< 25%, 26% to 50%, 51% to 75%, > 76%), depending on how it is measured in the included trials.
- Quality indicators, as appropriate. Examples are concealed allocation, blinding of assessors and intention-to-treat analysis.
We will undertake sensitivity analyses to examine the robustness of our results. We will carry out sensitivity analyses for the inclusion of trials with inadequate or unclear allocation concealment (including quasi-randomised trials), trials using self-reported outcomes (rather than performance-based outcomes) and using an ICC of 0.05 for cluster-randomised trials. We will evaluate whether the pooled effect differs upon omitting each single trial from the analysis.
The authors would like to thank Renea Johnston for her support at the editorial base and Louise Falzon for assistance in developing the search strategy.
Appendix 1. Search strategy
Cochrane Central Register of Controlled Trial (Ovid SP)
1. MeSH descriptor Frail Elderly explode all trees
3. (1 OR 2)
* indicates truncation
ti,ab denotes word in the title or abstract
Ovid MEDLINE (1950 to present)
1. exp Frail Elderly/
4. randomized controlled trial.pt.
5. controlled clinical trial.pt.
Ovid EMBASE (1988 to present)
1. Frail Elderly/
4. (random$ or placebo$).ti,ab.
5. ((single$ or double$ or triple$ or treble$) and (blind$ or mask$)).ti,ab.
6. controlled clinical trial$.ti,ab.
7. RETRACTED ARTICLE/
9. (animal$ not human$).sh,hw.
10. 8 not 9
Footnote for OVID
.pt. denotes a Publication Type term
.ab. denotes a word in the abstract
.sh. or / denotes a Medical Subject Heading (MeSH) term
.ti. denotes a word in the title
Last assessed as up-to-date: 16 January 2013.
Protocol first published: Issue 5, 2013
Contributions of authors
The protocol was developed by all authors. Nicola Fairhall is responsible for co-ordinating the review, carrying out the searches and locating studies. Both she and Catherine Sherrington will decide independently and then by consensus which studies meet the inclusion criteria. All three authors will assess quality and extract data from included studies. Nicola Fairhall prepared the protocol draft, will draft the review and perform primary data entry and analysis into RevMan. Catherine Sherrington and Ian Cameron will provide guidance with this process. All authors will be involved in re-analyses and revisions at all stages. Nicola Fairhall and Catherine Sherrington are the guarantors of the review.
Declarations of interest
Sources of support
- Rehabilitation Studies Unit, School of Medicine, The University of Sydney, Australia.
- The George Institute for Global Health, Sydney, Australia.
- Australian National Health and Medical Research Council, Australia.Salary (NF, CS, IC)