A 12-month multifactorial intervention targeting frailty was more effective than usual care in reducing mobility-related disability in community-dwelling frail older people. At the participation level, gains were present at both 3 and 12 months after intervention commenced. The intervention also increased mobility outcomes at the activity level at 12 months and was associated with minor adverse events. However, significant improvements were not detected using several outcome measures, and some statistically significant improvements may not have been clinically meaningful. To our knowledge, this is the first randomised trial to evaluate the effect of an intervention targeting frailty on mobility-related disability in older people who met specified frailty criteria.
Our sample of frail older people had poor mobility at baseline. They had an average of seven medical conditions and walked at one quarter the speed of healthy older people . Most had recently been discharged from an aged care and rehabilitation service and almost half did not get out of the house as much as they wanted to. At 12 months, we found the distribution of goal attainment on the GAS was significantly more favourable in the intervention group (OR 2.1, 95% CI 1.3 to 3, P = 0.004). We acknowledge that the GAS needs to be interpreted with caution however, as the non-linearity of ordinal scores at the margins of the score range (37% of scores at 12 months) can generate exaggerated change scores . The intervention as delivered also significantly increased life space; how often people mobilised in the home and community, how far they went, and their degree of independence. The improvement reached statistical, but not clinical, significance . Despite an increase in the extent of mobility, the intervention group showed a non-significant trend toward less satisfaction with their ability to get out of the house.
Interestingly, although participation in the mobility domain increased, there was no effect on the global measure of participation. This may be because the intervention focused primarily on the components of frailty, principally mobility, and targeted mobility-related participation goals. There are few trials measuring participation in frail older people with which to compare our findings. Previous systematic reviews of the effect of exercise and trials of geriatric evaluation and management on functional outcomes in frail older adults [10, 37, 38] have reported function primarily in terms of body structure and/or function and activity, whilst participation outcomes have been largely unreported.
There were statistically significant between-group differences for two of the three measures of the activity aspect of disability. At 12 months, the intervention group performed better on the AMPAC and walking speed. The effect of the intervention on activity outcomes is consistent with previous studies in frail older people, which indicate gait speed and composite activity measures may improve with regular multicomponent training over a prolonged period . The mean increase in walking speed of 0.05 m/s was at the suggested cut-off for a small meaningful change in a sample of older people with a higher level of functioning , so may be clinically significant in this comparatively frail group. Also, given the association between gait speed and survival , the significant improvement is noteworthy in this vulnerable population. In the absence of consensus on the clinically meaningful difference in AMPAC score in this population, we applied Norman and colleagues' criteria that the minimally important difference can be estimated as half the baseline standard deviation of raw scores . Although the mean between-group difference in AMPAC score was statistically significant, it was less than the minimally important difference. The relatively high degree of difficulty of the mobility tasks in the Nottingham Extended Activities of Daily Living Index mobility outcome may account for the lack of between-group difference in this outcome.
There appears to be a significant differential effect of the intervention on the walking speed and life space aspects of disability according to degree of frailty at baseline. The intervention had a lesser effect on walking speed and a greater effect on life space in participants who met three frailty criteria compared with the participants who met four or five criteria. This effect should, however, be interpreted with caution and requires further investigation . Participants with higher levels of adherence to the intervention had better outcomes after adjusting for possible confounders. This may indicate a dose-response effect of the intervention, but we acknowledge the potential biases associated with analysis of such adherence data .
It is not possible to determine which aspect of the multifactorial intervention increased participation in the mobility domain. There was a significant between-group difference in goal attainment, despite only half of intervention participants receiving the two intervention sessions specifically targeting participation goals. Participation is associated with multiple factors, including degree of frailty, mood, strength and walking speed [7, 42], and it is feasible that improvement in such elements contributed to gains in participation in life roles. Further research is required to understand which components of an intervention are required to improve participation and how participation can be improved across multiple areas of life.
The strengths of our study were the use of a validated definition of frailty, broad generalisability to recently hospitalised community-dwelling frail older people, the small losses to follow-up that were similar in both groups, and adherence to sound trial design and methodology. Also, the intervention - delivered in the setting of an existing health service by an interdisciplinary team experienced in aged care - resembles that deliverable in clinical practice. We acknowledge, however, that the study had limitations. First, participants could not be blinded to group allocation, which is a potential source of bias due to possible differential reporting of self-report outcomes such as goal attainment. Second, adherence with the program was variable; however, this is likely to be the case with treatment programs delivered to frail populations in the clinical setting, where health, physical and social needs fluctuate. Third, as there was no frequency-matched social intervention for the control group, we cannot exclude the impact of social aspects of the program on the difference between groups. Finally, although we attempted blinding of outcome assessors, 123 participants (51%) inadvertently disclosed their group status (that is, mentioned their exercise program) to research personnel at the 12-month follow-up.
Understanding disability in frail older people is hindered by the infrequent use of validated diagnoses, the systematic exclusion of frail older people from trials  and the often narrow conceptualisation of disability. This paper adds to the evidence that community-dwelling frail older people have the potential for functional improvement in both the participation and activity domains, through multifactorial intervention. Interventions that reduce disability in the frail population have the potential to impact on morbidity, hospitalisation and admission to residential care facilities, along with the associated costs to government and society.