Does a functional activity programme improve function, quality of life, and falls for residents in long term care? Cluster randomised controlled trial

Participants and recruitment

Further recruitment details are reported elsewhere.23 Residents in low level dependency residential care in two cities of New Zealand were eligible for this study. Elderly people living in low level dependency residential care homes need assistance with most instrumental activities of daily living and at least two activities of daily living but can usually ambulate to some degree and feed themselves.

The New Zealand Ministry of Health supplied a listing of all residential care homes (subsequently termed “homes”) in the two centres. We excluded homes caring exclusively for young disabled patients or delivering only palliative care. We invited homes to participate in random order by using computer generated random numbers. The owners and managers of the homes and all eligible residents, their family members, or guardians gave written informed consent.

Eligible residents were aged 65 years and over, able to engage in a conversation about a goal, remember the goal, and participate in a programme to achieve the goal (a proxy for cognitive state). The clinical nurse in charge of the resident at the time of recruitment judged ability. We excluded residents who were unable to communicate to complete the study measures, had anxiety as their main diagnosis, were acutely unwell, or were in a terminal state. We recruited residents between February and November 2004 and followed each resident for 12 months.

Measures

We ascertained staffing levels—registered nurses’ hours per day and healthcare assistants’ hours per day—by a structured interview of management staff. Trained independent research nurses recruited residents and, using standardised techniques, collected demographic data, health information, and drug use data from the medical and nursing record and did standardised face to face interviews. We estimated socioeconomic status by using the main occupation of the person or their spouse and by establishing whether the resident was publicly funded by a means tested entitlement. We used Hodgkinson’s abbreviated mental test score to establish cognitive function.24

The primary outcomes were function, self reported and observed; quality of life; and falls over 12 months of follow-up. We measured self reported function with the late life function and disability instrument validated for frail elderly people.25 The two main components, the functional and disability components, were administered at baseline and 12 months’ follow-up. Sub-domains included upper extremity function, basic lower extremity function, and advanced lower extremity function. The function component was also administered at six months.

We assessed observed basic mobility and functional tasks at baseline, six months, and 12 months with the timed up and go,26 the elderly mobility scale,27 and the FICSIT-4 balance test.28 We measured quality of life at all three time points with the EuroQol instrument and the life satisfaction index.29 30 We defined falls as “an unexpected event in which the participants come to rest on the ground, floor, or other lower level.”31 Independent researchers used this definition to audit all medical and nursing records of participating residents for three months before the trial started and then every two months over the year of the trial.

Secondary outcome measures were depressive symptoms assessed by the geriatric depression scale,32 fear of falling measured by the modified fear of falling scale,33 and hospital admissions over the 12 months of the trial from the national minimum dataset, which records all discharges from New Zealand’s public hospitals and is held by the New Zealand Health Information Service.

We ascertained adverse events, including muscle pain, fatigue, and general aches and pains, by self report at baseline, six months, and 12 months. We established mortality and permanent admission to a high dependency level of long term care (nursing home care) by direct enquiry at each follow-up visit.

Randomisation and blinding

After recruitment of all homes and residents and collection of baseline data, a biostatistician not involved in recruitment randomised homes to the intervention or control group by using computer generated random numbers. We used randomisation by home (cluster) to avoid contamination between groups resulting from the intervention design involving staff training and thus potentially affecting all residents in the home.34 Research nurses blinded to the group allocation of the homes used standardised methods to assess outcomes. All intervention materials were removed from each resident’s room before follow-up reassessment visits.

Interventions

Two trained gerontology nurses, one at each site, delivered the promoting independence in residential care (PIRC) intervention on the basis of a successful efficacy trial.21 To standardise intervention delivery they were trained by KP and did dual assessments and plan development for five residents at the start, after six months, and towards the end of the enrolment period at both sites. The gerontology nurses were not involved in any of the outcome assessments and had a well designed study protocol to guide the intervention phases.

Goal setting—The resident, assisted by the gerontology nurse, set a mutually agreed goal that had to meet two criteria: it had to be relevant and meaningful to the resident, and it had to promote progressive increases in physical activity. The goal setting often required one to two visits, depending on the resident’s abilities

Functional assessment and activity programme design—The gerontology nurse then completed a functional assessment and designed an individualised programme of physical activities based on repetitions of activities of daily living, such as rising from a chair, additional walking, or repeated transfers, aiming to improve the physical functions needed to achieve the goal. Exercise activities were planned to be done daily or several times a day in short doses as part of the resident’s usual activities. A physiotherapist and occupational therapist were available to assist in assessment and programme design when asked by the gerontology nurse. A prescriptive plan, the promoting independence plan, was constructed and placed on the wall in the resident’s room and in the resident’s file. The plan was developed in part as a template, such that if rising from a chair was challenging (and was needed as part of achieving the goal) the plan incorporated initially five sit to stands twice daily with increasing repetitions as the individual resident gained greater lower leg strength.

Staff implementation—The gerontology nurse trained the healthcare assistants in implementing the promoting independence plan on a one to one basis for each resident. The healthcare assistants then implemented the plan under the supervision of the usual nursing staff of the facility. For the more independent residents, the healthcare assistants provided minimal supervision. The more dependent residents needed up to 15 minutes of close supervision of the plan twice daily.

Ongoing support—The gerontology nurse provided support to the home staff every week for the first month and monthly for the next six months. Support included visits; discussions about residents, exercises, and goals; and reviewing progress and renegotiating a new goal with enrolled residents when the first goal was met. The intervention lasted six months in total. After that time the facility staff were expected to continue encouraging engagement in activity.

Control group—Residents in control group homes received usual care and were offered two social visits by a social science researcher to control for the attention received by the resident from the gerontology nurse visits.

Uptake—Gerontology intervention nurses following enrolled residents reported compliance of residents with the activity programme. Staff were asked to complete checklists of residents’ episodes of exercises. A record was made of all social visits.

Sample size calculation

We used previous data on the main outcome variables to complete power calculations. Based on a baseline average score of 15 (SD 6) of the elderly mobility scale,35 we needed 132 participants in each group to detect a true difference of 2 units (90% power, 0.05 level of significance, two sided test). Inflating this according to an estimated design effect of 1.44 to account for the cluster randomised design (intraclass correlation of 0.02 from the efficacy study21), we needed 190 residents in each group. To detect a 20% reduction in falls and a 15% reduction in injuries with 90% power at the 0.05 level of significance (two sided test), we needed 329 residents for falls and 235 residents for injuries, based on a baseline proportion of 46% falling and 30% sustaining injury over a year and inflating raw estimates to allow a design effect of 3.2 for falls and 1.8 for injuries.16 36 Allowing a response rate of 93%,16 and taking the largest of these estimates, we needed a total of 707 elderly people to participate. These were to be recruited from between 40 and 44 homes, assuming an average cluster size of 20 based on half the residents in homes being eligible and the average home size being 40 people.21

Statistical analysis

We sought to examine the effect of the intervention at an individual level, adjusting for the impact on variance of the clustering inherent in the sampling design process. We investigated differences in changes over time in the two groups by testing the effect of the interaction of time with the randomisation group on the main outcomes. We did planned tests for interactions between depression, randomisation group, and time and between mental status, intervention group, and time. If appropriate, we did subgroup analyses.

We described baseline characteristics for intervention and control groups. We used generalised linear mixed models, with rest home as a random factor, city as a fixed factor, and time as a repeated measure, to investigate whether the intervention was associated with a change in function (late life function and disability index, timed up and go, elderly mobility scale, balance) or quality of life (life satisfaction index, EuroQol) over time and in particular whether this change differed between the social and intervention groups. We used an autoregressive structure to model the correlations between repeated measures over time. Where outcome measures were not normally distributed, we log transformed them or converted them to categorical variables for analysis. Where appropriate, we used a logit link for binary data. We used a generalised linear mixed model with a logit link for the secondary outcome of depression. We investigated time to fall, allowing for multiple falls per resident, by using a Cox proportional hazards model with rest home included as a random effect.

We included mental status (abbreviated mental test score), depression (geriatric depression scale score), age, sex, previous falls, total number of drugs (as a proxy for health status), time in the low level dependency home, and socioeconomic status in the models for the primary outcomes. We used negative binomial regression models allowing for the clustering of participants within rest homes to investigate hospital admissions and adverse effects (aches and pains). We used the χ² test to test for differences in the proportion of participants who were admitted to high level dependency care or died. We used SAS version 9.1 for all analyses.

The quality of the trial met all criteria outlined for randomised controlled trials in the User’s Guide to the Medical Literature II,37 except that double blinding was not possible.