The Frailty Intervention Trial, a prospective, parallel-group, assessor-blind, randomised, controlled, single centre trial, was undertaken from January 2008 to June 2011. The protocol for this study was registered with the Australian New Zealand Clinical Trials Registry (ANZCTRN 12608000250336) and has been published elsewhere 14. This paper reports on the mobility-related disability outcomes of the Frailty Intervention Trial, which were registered separately (ANZCTRN 12608000507381). The primary outcomes of the trial are reported elsewhere (manuscript under review).

The protocols for the Frailty Intervention Trial and the mobility-related disability component were approved by Northern Sydney Central Coast Health Human Research Ethics Committee (1 November 2007 and 20 August 2008, respectively). Written informed consent was obtained from all participants.

Potential participants were identified from community-dwelling patients being discharged from the hospital or community arm of the Division of Rehabilitation and Aged Care Services at Hornsby Ku-ring-gai Health Service in urban Sydney, Australia. All potential participants had completed their treatment before they were approached. Participants were eligible if they were 70 years or older; did not usually reside in a residential aged care facility; were defined as frail according to the Cardiovascular Health Study (CHS) Frailty Phenotype 1; did not have severe cognitive impairment (defined as a Mini Mental State Examination 16 score of 18 or less); had a life expectancy exceeding 12 months (estimated by a score of three or less on a modified Implicit Illness Severity Scale) 17; and resided in the Hornsby or Ku-ring-gai local government areas.

Community-dwelling older people were screened for eligibility by a research nurse (KL). Participants who gave informed consent underwent baseline assessment prior to randomisation into groups. A data analyst (NM) not involved in recruitment or assessment developed the group allocation schedule using a computer generated random number sequence and stored the list off site. Sets of permuted blocks were generated for each of two strata (three frailty criteria versus four or five frailty criteria). Block sizes of four and six were randomly arranged within blocks of 10. After baseline assessment was completed, KL was unblinded to group assignment.

Participants in the intervention group received a multifactorial interdisciplinary intervention targeting the CHS frailty phenotype for one year. Intervention was coordinated by two physiotherapists with extensive relevant experience and delivered by an interdisciplinary team comprising the physiotherapists, a dietician, a geriatrician, a rehabilitation physician and a nurse. Intervention was delivered primarily in the participants' homes, with additional outpatient appointments (for example, dietician, continence clinic), occupational therapy and community exercise programs offered as indicated.

The intervention is described in detail elsewhere 14. In brief, the intervention was tailored to each participant based upon the CHS frailty criteria present in the individual at baseline (three or more of slow gait speed, weak grip strength, exhaustion, low energy expenditure and weight loss); problems identified during comprehensive geriatric evaluation; and ongoing reassessment during the 12-month intervention period. The interdisciplinary intervention was coordinated by regular case conferences and case management by the treating physiotherapist, who liaised with the participant, family, health professionals, service providers and coordinated services as indicated.

The component delivered to the intervention group that most directly targeted mobility was the 10 home-based physiotherapy sessions of 45 to 60 minutes duration. There were five sessions in the first three months after randomisation, and five sessions over the following nine months. Two of the physiotherapy sessions specifically targeted a participant-centred mobility goal. Initially developed by the participant and an assessor blinded to group, the goal was then discussed between the participant and treating physiotherapist and modified if necessary. The treating physiotherapist made a clinical assessment of the barriers to goal attainment, and then delivered interventions to target the potentially remediable barriers, such as walking capacity, social support, anxiety and community environment. The components of the goal were practised in isolation then in the target physical environment, and the degree of assistance and/or support was decreased in a safe and appropriate manner. The participation intervention protocol is available at http://www.webb.org.au.

Eight physiotherapy intervention sessions addressed the weakness, slowness and low energy expenditure CHS frailty criteria by teaching a home exercise program designed to improve mobility, increase physical activity and prevent falls (the Weight-bearing Exercise for Better Balance (WEBB) program, http://www.webb.org.au). The program was tailored to the individuals' physical impairments, prescribed three to five times per week, and reviewed and modified regularly. Appropriate equipment items such as mobility aids were also recommended.

The physiotherapist responsible for each participant recorded adherence to the study protocol and estimated a global level of adherence (in five categories) during the 12-month intervention.

Participants assigned to the control group received the usual care available to older residents of the Hornsby Ku-ring-gai area from their general practitioner and community services, which may include medical management of health conditions, allied health input, assessment of care needs and provision of care.

The outcome of interest was mobility-related disability, measured at the levels of participation restriction and activity limitation. Participation was evaluated in terms of satisfaction and performance. Satisfaction with level of community access was measured using the question 'do you get out of the house as much as you would like?'13. This question has a dichotomous yes or no response and has demonstrated reliability and responsiveness 13. Mobility during the preceding month was quantified in terms of distance and frequency of travel and degree of independence using the University of Alabama at Birmingham Life Space Assessment 18. Scored on a continuous scale from 0 to 120, a higher score illustrates greater life space.

Achievement of individualised mobility-related participation goals was also evaluated using the Goal Attainment Scale (GAS) 19. Recommended as a measure of relevant person-centred outcomes in frail people 2021 and in the evaluation of complex interventions 22, the GAS has demonstrated responsiveness, adequate inter-rater reliability and concurrent validity with other rehabilitation outcome measures in older people 2123. Using an established methodology 24, a blinded research physiotherapist guided the participant in setting a mobility-related goal in the home environment (for example, able to weed the garden) or the community (for example, return to volunteer work) based on problems identified by the Reintegration to Normal Living Index 25. The participant and physiotherapist established how the goal would be measured, and then the physiotherapist predicted the treatment outcomes on a five-point scale (see Figure 1). Because deterioration was plausible, baseline function was scored at the second lowest point on the scale 26. Those who were unable to set a goal were allocated the lowest score at post-test.

Participation across multiple areas of life was measured with the Reintegration to Normal Living Index 25. Nine of the eleven original items measure participation, covering eight dimensions of life and five International Classification of Functioning domains 27. The nine items were included and the question phrasing was modified to increase comprehension, as validated by Daneski and colleagues 28. The nine declarative statements were scored using a five-point scoring system (strongly disagree, disagree, neither agree nor disagree, agree, strongly agree), yielding a total score from 0 to 36, with a higher score reflective of greater participation. The tool has demonstrated reliability, content validity and construct validity 26.

Mobility performance at the activity level was measured by walking speed, measured using the 4-metre Walk Test. Self-report measures of activity were the mobility components of the Activity Measure for Post Acute Care (AMPAC) 29, which measures activity level using Item Response Theory and computer-adaptive testing, and the Nottingham Extended Activities of Daily Living Index 30.

An adverse event was defined as a medical event or injury that arose as a consequence of the trial and resulted in medical attention or restricted activities of daily living for more than two days 31.

Outcomes were assessed in participants' homes by independent, blinded nurse assessors at baseline (before randomisation) and at 3 months and 12 months after randomisation. The AMPAC was assessed via telephone at baseline and 12 months. Baseline goal setting and AMPAC assessment occurred after randomisation, by a physiotherapist blinded to group allocation.

Researchers who collected outcome measures and recorded data were blinded to group allocation. To ensure blinding, participants were asked not to disclose group allocation to the assessors. The participants and treating staff could not be blinded to intervention group allocation.

A sample size of 240 participants was chosen as it would provide 80% power to detect a 15% between-group difference in a primary outcome of the Frailty Intervention Trial - the lower extremity continuous summary performance score 32.

This sample size also provided sufficient power to detect a clinically meaningful 20% between-group difference in goal attainment and satisfaction with getting out of the house, a between-group difference of 0.1 m/s in walking speed (assuming a within-group standard deviation of 0.23) 8, and a between-group difference of 10 points in the Life Space Assessment (standard deviation of 24.7) 18. For these calculations, we assumed an α of 0.05, non-compliance of 15% and a dropout rate of 15%.

The primary analyses were by intention to treat 33 using Stata version 11 (College Station, TX, USA). Missing data for individual variables were imputed using multiple imputation. There was a maximum of 2% of cases missing for any variable.

To determine the effect of group allocation on continuously scored outcome measures, we examined between-group differences using linear regression models. For dichotomous outcomes, between-group differences were compared using logistic regression models. Baseline scores were entered into the linear and logistic regression models as covariates. Statistical significance was set at P < 0.05 and we reported the differences in percentage or mean (95% confidence interval (CI)) between the two groups at the 3-month and 12-month follow-ups.

Goal attainment was treated as an ordinal outcome; between-group difference in the distribution of goal attainment was compared using ordinal logistic regression models. The reported odds ratios (ORs) express the odds of an improved distribution of outcome in association with the intervention. To aid interpretation of the GAS, the scores were also dichotomised (goal met versus goal not met), and ORs were calculated.

We undertook a preplanned analysis to determine whether there was evidence of a differential response to the intervention for people with a cognitive impairment using a test for statistical interaction 34 (grouped by a Mini Mental State Examination score ≤ 24 or > 24)16. Post-hoc analyses were also conducted to test for evidence of an interaction between group and frailty severity (number of CHS criteria met, three versus more than three), and to explore the relationship between different levels of adherence (as a category variable: < 25%, 25% to 49%, 50% to 74% and ≥ 75%) and outcomes at 12-month follow-up in the intervention group.

The participant flow is summarised in Figure 2. We recruited 241 participants (68% female) with an average age of 83.3 years (SD: 5.9 years) between January 2008 and April 2010. Table 1 shows the participant characteristics at baseline. Of the 241 people randomised, 226 (94%) completed the 3-month assessment and 216 (90%) completed the 12-month assessment. The majority (22 out of 25) of losses to follow-up were due to death.

The median number of face-to-face intervention physiotherapy sessions in the home was 10 (range 0 to 24), and there was a median of four telephone calls to the participant. The physiotherapist delivered the WEBB program to 112 intervention participants (93%), with a median of eight sessions (range 0 to 22). The study dietician provided assessment and intervention to 60 participants (50%); 29 participants (24%) were reviewed by a geriatrician or rehabilitation physician and 4 (3%) were referred to a psychologist or psychiatrist. The physiotherapist delivered or coordinated a broad range of additional interventions, including training targeting self-selected mobility goals (61 out of 120, 51%); provision, modification or advice about equipment (48 out of 120, 40%); referral to specialist aged care services such as community transport, meal delivery, day centres, nursing, social work, occupational therapy, or teams that assess and advise on care needs (49 out of 120, 41%); and referral to medical services, such as the general practitioner, or medication-related intervention (36 out of 120, 30%).

The median global level of adherence to the intervention program, as estimated by the treating physiotherapists blinded to outcome, was 25% to 50%. Of 220 participants in the intervention group, 16 (13%) did not complete any of the intervention, 34 (29%) completed 1% to 25% of the intervention, 19 (16%) completed 26% to 50%, 25 (21%) completed 51% to 75%, and 25 (21%) completed 76% to 100%. There were 61 participants who did not complete the goal focused aspect of the intervention and the reasons for this were: inability to set a goal (17 out of 120, 14%); death before intervention commenced (3 out of 120, 3%); goal became inappropriate due to deterioration in health (35 out of 120, 30%) or the environment (for example, unsupportive family or new accommodation, 6 out of 120, 5%).

Outcome assessors remained blinded to group allocation in 97% of baseline AMPAC and GAS assessments, 100% of other baseline measures and 49% of 12-month follow-up assessments.

Two intervention group participants with pre-existing musculoskeletal conditions experienced back pain severe enough to meet the study definition of an adverse event. Both resumed exercising once the prescribed exercises had been modified.