Study design

We conducted a single center, randomized, controlled trial among older people who were frail in Sydney, Australia. The trial protocol has been previously reported [8]. The Northern Sydney & Central Coast Health Human Research Ethics Committee has approved the study protocol - Research Protocol Number 0709-191M - and the trial is registered with the Australian New Zealand Clinical Trials Registry: ACTRN12608000250336. Trial registration was delayed and 8% of the participants had been randomized at the time of registration. Participants provided written informed consent before randomization.

Study population, screening and randomization

Eligible participants were identified from older people treated by clinicians working within the Division of Rehabilitation and Aged Care Services (DRACS) at Hornsby Ku-ring-gai Health Service (Sydney, Australia). DRACS is a large clinical service that has programs operating both in the community and hospital settings. Eligible participants first completed their usual treatment program before being approached to enter the study. Following participation agreement, informed consent was sought, often in conjunction with relatives. If granted, the study nurse screened for inclusion criteria. These were: adults aged 70 years or older with three or more of the CHS frailty criteria [1]; not usually living in a residential aged care facility; residing in the Hornsby or Ku-ring-gai local government areas; without moderate or severe cognitive impairment (defined as a Mini Mental State Examination score of ≤18) [9]; not an ongoing client of DRACS; without an illness likely to be associated with a life expectancy of <12 months, estimated by a score of ≤3 on a modified version of the Implicit Illness Severity Scale [10]; and not participating in another physical intervention research project. All randomized participants met the eligibility criteria.

Interventions

Participants in the intervention group received a multifactorial, interdisciplinary treatment program intended to target frailty for a 12-month period following randomization. The interventions were individually tailored to each participant based on their frailty characteristics as assessed at baseline, and additional problems as identified during a detailed assessment by the two experienced physiotherapists providing the intervention program. Geriatric evaluation and management principles underpinned both the assessment and intervention [11].

Details of the approach to intervention are described in the protocol paper [8]. To summarize, the CHS frailty components that were present in each participant were specifically targeted [12]. If the participant met the weight loss criterion, a dietician evaluated nutritional intake. Home-delivered meals were recommended if appropriate clinical criteria applied. In addition, if the participant's body mass index was <18.5 kg/m², or mid-upper arm circumference was <the 10th percentile (using Australian age and gender specific norms), nutritional supplementation was offered using commercially available, high energy, high protein supplements.

If the exhaustion criterion was met and the Geriatric Depression Scale [13] score was high, the study team considered referral to a psychiatrist or psychologist. Where the participant was socially isolated, options to encourage greater social engagement were identified, such as participation in day activity groups and telephone contact with a volunteer.

Participants who met the weakness, slowness or low energy expenditure criteria received up to 10 home-based physiotherapy sessions and performed a home exercise program, over the course of 12 months. The Weight-bearing for Better Balance (WEBB) program [14], designed to improve mobility, increase physical activity and prevent falls, was tailored to individual's physical impairments, prescribed three to five times per week, and reviewed regularly. Two physiotherapy sessions targeted the participant's mobility goal. Equipment was also recommended as necessary.

Case management by the physiotherapist, and regular case conferences involving the physiotherapist, geriatrician, rehabilitation physician, nurse and dietician, facilitated coordination of the delivery of the intervention. Reassessment was ongoing throughout the intervention phase. The physiotherapist was the co-ordinator of the intervention. Home visits usually involved several intervention components and included not only the WEBB exercise program, but other identified interventions that were relevant to the frail person at that particular time.

For all participants, additional interventions were then provided or recommended based on a comprehensive geriatric evaluation, for example review by the study geriatrician or rehabilitation physician, follow-up of chronic diseases, treatment of pain, and management of other identified conditions such as urinary incontinence.

The major limiting factor in implementing the intervention as planned was the limited ability of the participants to be able to adhere with the recommended intervention plan. The physiotherapists coordinating the intervention were careful not to overwhelm the participants with complex treatment plans and, in addition, participants often declined specific interventions.

There was a standardized approach to interventions in the study based on the study protocol and regular case discussions of each participant. The physiotherapist primarily responsible for each participant documented adherence to the study protocol and estimated a global level of adherence (in five categories) over the 12-month intervention period.

Usual care, as received by the control group in this study, consisted of those health and aged care services that would normally be available to older people. These include general practitioner and medical specialist consultations, and nursing and allied health interventions as appropriate. Australia has a system of universal health insurance so that all of its population has access to health care without significant cost. Aged care services include assistance with housekeeping and personal care and these types of services are also heavily subsidized by the Australian government for older people with care needs. Northern Sydney has a well-developed system of health and aged care that has been operating in its current form for more than 20 years.

Outcomes

The primary outcomes measured were frailty and mobility. Frailty was assessed using the CHS definition of the frailty syndrome [1]. See Additional file 1 for details of frailty assessment criteria. Mobility was measured using the SPPB [7], which assesses the ability to stand (for 10 s) with the feet together in the side-by-side, semi-tandem and tandem positions; time to walk 4 meters; and time to rise from a chair and return to the seated position five times. The SPPB score and the lower extremity continuous summary performance score were calculated [15].

Secondary outcomes included hospitalizations and admissions to nursing care facilities that were reported on monthly calendars and confirmed from either hospital records or the relevant facility; disability was measured with the Barthel Index [16], and health-related quality of life measured by the EuroQol-5D (EQ-5D) [17]. Psychological status was assessed using the Geriatric Depression Scale (short form) [13] and deaths were recorded and verified by hospital records.

We initially chose the Timed Up and Go measure as a co-primary outcome and included this on the Clinical Trials Registry. However, early in the study we recognized that it was not feasible to collect data for this outcome and hence determined that the SPPB and Frailty would be the co-primary outcome measures. These are the co-primary outcomes recorded in the study protocol paper [8].

The data were collected in the participants' homes by experienced research nurses trained specifically for the trial. Training in the assessment tools was provided, and joint sessions conducted to standardize the administration of the assessment tools, ensuring consistent interpretation of the data recorded. Inter-rater reliability checks were conducted at commencement and mid-way through the trial.

The study aimed to recruit approximately 230 participants, to detect a clinically and statistically significant 15% difference in mobility as assessed by the lower extremity continuous summary performance score between the two groups (power = 80%, P = 0.05, dropouts = 15%, non-compliance = 15%, standard deviation (SD) = 0.7).

A permuted block randomization approach was used to achieve balanced treatment allocation. There were two strata (frail with three CHS frailty criteria and very frail with four or five CHS frailty criteria). A random number sequence was generated for the order of treatment allocation within the blocks using SPSS v15 RV.UNIFORM function (SPSS, Inc., Chicago, IL, USA). Block sizes of four and six were used and these blocks were randomly arranged within blocks of ten.

Project personnel not involved in assessing participants or in providing the intervention managed the randomized group allocation. The treatment allocation tables for both strata were stored centrally off site.

Staff members performing the outcome assessment and data analysis were masked to group allocation. It was not possible to blind participants and staff administering the interventions to group allocation. The number of participants inadvertently unblinding the outcome assessors was recorded.

Harms

Adverse events were monitored and recorded by the treating clinicians. Deaths were monitored by the study statistician. A data monitoring procedure was implemented but, after consultation with the independent epidemiologist monitoring the study, interim analyses were not performed because of the limited number of adverse events.

Statistical analysis

The study design assessed study participants at baseline, at 3 months, and at 12 months after randomization. Data were coded to permit blinding to group allocation in the statistical analysis. The primary analyses were undertaken in accordance with the intention-to-treat principle. Frailty was treated as both a dichotomous (that is, more than three CHS criteria met or not) and continuous variable, and other study outcomes as continuous variables. The chi-square test was used for frailty as a dichotomous variable, and linear regression models with baseline values as a covariate were used for continuous outcomes. We report between-group differences in percentages, or mean, with 95% confidence intervals (CIs) at the 3- and 12-month follow-ups. We tested whether the pattern of changes in frailty and mobility were modified by frailty severity at baseline, by including an interaction term of study groups with frailty severity at baseline in the regression analyses. Secondary analyses were also carried out to explore the effect of different rates of adherence (as a category variable: <25%, 25% to 49%, 50% to 74% and ≥75%) on the outcomes in the intervention group at the 12-month follow-up. The Cox regression model was used to assess time to admission for those permanently admitted to aged care facilities.

Characteristics of the study population

Participant flow is shown in Figure 1. Of the eligible people approached, 75% (241 out of 322) agreed to participate in the trial. Recruitment commenced in January 2008 and concluded in April 2010. Follow-up concluded in June 2011.

At 3 months, follow-up data from 109 out of 120 of the intervention group and 117 out of 121 of the control group were available for analysis (Figure 1). At 12 months, the availability of these follow-up data was 107 out of 120 and 109 out of 121 respectively. The follow-up attrition rate was 10.4%, with 22 of the 25 losses due to death.

Interventions:The intervention and control treatments were implemented as planned. Adherence was 0% for 16 participants (13%), including three people who died before the intervention could commence. The median adherence overall was in the category of 26% to 50%.

There was a median of 10 face-to-face sessions with a physiotherapist for each participant in the intervention group, including a median of eight sessions to teach the WEBB program. In addition, there was a median of four telephone calls to each participant and a median of four telephone calls to other parties. The WEBB program was delivered to 93% of intervention group participants. A dietetic assessment and intervention was provided to 50% of participants (resulting in 29% being recommended nutritional supplements). A medical specialist consultation (with a geriatrician or rehabilitation physician) was arranged for 24% of participants. Referral to a psychologist or psychiatrist was arranged for 3% of participants. Further details of the intervention as implemented are available from the authors on request.

Primary outcomes

Mobility remained relatively stable in the intervention group, whereas it declined substantially in the control group. At 12 months, there was an average decline on the 12-point SPPB scale of 0.98 (SD = 2.30) in the control group and an average increase of 0.52 points (SD = 2.47) in the intervention group (between-group difference 1.44 points; 95% CI: 0.80, 2.07; P <0.001). A similar result was seen in mobility when assessed using the lower extremity continuous summary performance score (Table 2).

Secondary outcomes

There were no major differences between the groups with respect to the secondary outcomes of the trial (see Table 2). The number of deaths in the intervention group was 12 (10.0%), with 10 in the control group (8.26%) (P = 0.64). A large number of hospital admissions occurred, (intervention 74, control 67) with no significant differences between the groups (P = 0.32). There were similar numbers of permanent admissions to nursing care facilities in both groups (intervention 16, control 21) and, compared with the control group, the hazard ratio of time to admission for the intervention group was 0.69 (95% CI: 0.35, 1.33; P = 0.27).

A number of participants (169 out of 241, 70%) inadvertently disclosed their treatment status to the assessors. At the 12-month follow-up, disclosure of treatment status occurred in 51% of cases. We analyzed SPPB scores by whether the rater was unblinded or not, and there was no significant difference in the scores (data not shown).

Harms

No major adverse events specifically attributable to the intervention were evident. Two participants experienced back pain that met the adverse event criterion reported in a previous study [18] and required modification of their exercise program.

Strengths and limitations of study

The trial was completed in accordance with the published protocol. It was a pragmatic randomized trial that included participants meeting a widely accepted definition of frailty and had few exclusion criteria. In keeping with studies of this type, it was not possible to blind participants and treating clinicians to the intervention. Outcome assessors were blinded but many participants inadvertently disclosed their treatment status. The components of the frailty definition are partly self-reported and partly performance based, but the co-primary outcome (the SPPB) is a performance-based measure that should reduce observer bias.

The lack of a control group providing a sham intervention could be seen as a limitation and our trial does not tell us whether the nature of the contact with the staff providing the intervention was important. It is unlikely that such contacts could have specific effects on frailty and mobility, but could have had an effect on mood that may influence the exhaustion criterion in the CHS frailty phenotype [21]. However, we consider our trial to be comparative effectiveness research in which we are comparing the two options of usual care versus potential care through this program.

Adverse events were minor and responded to a change in the prescribed exercise intervention. Skilled physiotherapists tailored and delivered the intervention. Dropouts were not due to adverse events but were rather related to participants' beliefs or to major changes in their health conditions.

Generalizability:The study was conducted in a relatively affluent country with well-developed health and care services for older people. It used a specific clinical team with considerable experience and knowledge of the locally available aged care services. It should be feasible to generalize the intervention to other situations with similar health and care services. It is a moderate intensity intervention that could be provided as a program through an aged care health service. The intervention costs about a quarter of the amount of the program currently administered by the Australian government, designed to assist a similar group of older people in the transition from hospital to home or nursing care facility [22].

Comparison with other studies

Trials involving older people with reduced functioning, labeled as 'frail', have shown variable improvements in disability and its components. A recent systematic review of exercise interventions in frail populations concluded that multicomponent exercise treatments for frail people are likely to be effective if undertaken on a regular basis over a prolonged period [6]. The geriatric evaluation and management literature has shown mixed effects with respect to overall benefit. This study adds to previous research by demonstrating that it is possible to identify frail older people for inclusion in a randomized trial, and that intervention can reduce the degree of frailty and improve mobility outcomes in this population.

In the group studied, it is difficult to disentangle 'frailty' from 'disability' because these two states coexisted in almost all participants. Although improvement in frailty was seen there was also improvement in mobility disability. The changes in frailty and mobility are similar in magnitude and represent medium effect sizes. In this study, changes in frailty and mobility appear closely linked. The measure of activities of daily living disability that was used, the Barthel Index, was close to its maximum and ceiling effects may have limited the ability to detect changes in disability as defined by this measure.