Design and setting
An investigator and assessor-blinded, parallel group multisite RCT will compare a progressive balance and strengthening exercise program (Staying UpRight) with a low intensity seated exercise program (Flex and Stretch), provided to older people living in LTC facilities located in Auckland and Hamilton, New Zealand. The study design is outlined in Fig. 1. Reporting of results will conform to the recommendations of the Consolidated Standards of Reporting Trials (CONSORT) statement . Any protocol amendments will be submitted to NZ Health and Disability Ethics Committee for review and will be updated on the trial registry.
Residents aged 65 years or older within participating LTC facilities will be invited to take part. Residents who are in psychogeriatric care, respite or palliative care, acutely unwell or immobile (bed-bound) will be excluded. Study investigators will work with the LTC clinical lead to identify eligible residents. Participants will provide consent before enrolment. For residents unable to provide informed consent due to cognitive impairment, written consent will be sought from the LTC clinical manager, in accordance with New Zealand HDEC requirements when undertaking research among people with cognitive impairment.
To control for LTC facility factors, randomization will be stratified by facility and by level of care (high dependency, low dependency and dementia-level care). Levels of care in New Zealand are determined by health authority-appointed needs assessment service coordination (NASC) agencies. After baseline assessment, participants will be individually randomized to intervention (Staying UpRight) or control (Flex and Stretch) by a researcher distant from recruitment using a computer-generated random sequence.
The primary outcome measure is fall rate (per 1000 person years).
Using the LTC facility’s incident report records, fall registers will be audited for the 6 months prior to study commencement (to provide baseline data) and for the 12 months of the study. Estimates of complete accuracy in falls ascertainment is impossible as most falls are unwitnessed , but the possibility of increased falls being reported as programs are implemented  will be avoided by using already accepted and implemented reporting systems (RiskMan™ or VCare™) that are standardised throughout all LTC facilities in the trial.
Secondary outcome measures are:
- Risk of falling, expressed as proportion of fallers per group i.e. those who sustained at least one fall during follow-up.
- Rates of hospitalisation for fall-related injury (fracture, intracranial or extracranial haemorrhage), expressed as number of fall hospitalisations per 1000 person-years of follow-up. Hospitalisation data will be collected from NHI matched Ministry of Health data over 12 months.
- Fall rate relative to activity exposure, expressed as number of falls per 100,000 steps. Participants will wear a tri-axial accelerometer (Axivity AX3, York, UK), secured on the lower back at the fifth lumbar vertebrae (L5) using a hydrogel adhesive (PALStickies, PAL Technologies, Glasgow, UK), covered with an adhesive dressing (OPSITE Flexifix™ and Hypafix™, Smith+Nephew Ltd, England). The accelerometer is programmed to sample at a frequency of 100 Hz (range ±8 g).
- Gait volume, pattern and variability of ambulatory activity, measured using accelerometery as described above. Algorithms are valid for gait spatio-temporal features for macro gait i.e. volume (total steps, time spent walking, bout number), pattern of activity (bout length and bout distribution), variability of ambulatory bouts and micro gait i.e. pace, rhythm, variability, asymmetry and postural control domains of gait [28-30].
- Physical performance, measured using the Short Performance Physical Battery (SPPB)  and the Timed Up and Go (TUG) [32, 33]. The SPPB comprises ability to stand for 10 secs (with feet side by side, semi tandem and tandem stance), timed chair stand (5 times chair rise) and gait speed (measured over 3 metres). The maximum score is 12, with higher scores indicating better function. The TUG requires the participant to rise from a chair, walk 3 metres quickly but safely to a mark on the floor, turn, walk back and sit down. A lower time indicates better function.
- Cost-effectiveness of Staying UpRight versus seated exercise will be estimated as both an additional cost per fall prevented and per fall requiring hospitalisation.
Demographic and health information (health conditions, medications, independence in activities of daily living) will be collected using a standardised Minimum Data Set (MDS 2.0; interRAI Corporation 1999) at baseline prior to randomisation. The interRAI™ (International Resident Assessment Instrument) Long Term Care Facilities (LTCF) assessment is a standardised comprehensive observational assessment, completed on admission and 6-monthly thereafter by nursing staff for all residents in LTC.
Cognitive function will be measured at baseline, 6 and 12 months using the Montreal Cognitive Assessment (MoCA) . The MoCA comprises 16 items to assess multiple cognitive domains. A score of ≤10/30 is considered severe cognitive impairment, 11-18 moderate impairment, 19-23 mild and >23/30 normal cognitive function .
Ambulatory activity over 7 days will be collected using the tri-axial accelerometer and data uploaded to an encrypted, secure platform (eScience Central online platform, Newcastle University, UK)  for storage and blinded processing.
Outcome measures will be completed at baseline, 6 and 12 months (Fig.2). Assessors will be trained in all aspects of participant assessment, activity monitoring and digital data management to ensure the assessments are standardized. Assessors will watch a standard assessment and each rate the responses. Responses will be compared and discussed.
Both intervention and control group participants will participate in any usual activities provided in the LTC facility.
Assessors will be blinded to group allocation for all assessments. Group assignment will only be available to the project coordinator (LT), data manager (SM) and the intervention coordinator (EB) and will be accessed online via a password protected site. Participants and the staff providing the intervention cannot be blinded.
Participants randomized to the intervention group will attend the Staying UpRight program. Staying UpRight is a supervised balance and strength group exercise program (up to 8 per group) delivered for 1 hour twice weekly over 1 year. Classes will be led by a physiotherapist trained in program delivery, supported by an assistant. Balance exercises comprise static and dynamic activities progressed by reducing hand support, reducing the base of support, reducing visual input or adding a cognitive task. The strength exercises use body weight resistance, low repetitions (2 x 10 repetitions at 5-7/10 effort) in weight-bearing positions where possible (Table 2). Exercises are progressed by increasing the number of sets, the speed or amplitude of the movement, or the complexity of the task (Table 3). The program is manualised, with exercises and progressions selected based on participants’ abilities. Exercises completed in each session, and exercise progressions for the group will be recorded on a spreadsheet and reviewed as part of fidelity monitoring.
To test sustainability, the LTC facility staff (physiotherapist or physiotherapy assistant) will take over delivery of the classes for the second 6 months, ensuring continuity of the program for 1 hour twice weekly over the 12-month period. If the on-going provision of classes is compromised within a facility, alternative funding streams will be sought.
Control group participants will attend a seated group activity program (Flex and Stretch) delivered for 1 hour twice weekly over 1 year. Classes are led by the LTC activities staff or a volunteer trained in program delivery. Activities comprise lower limb, trunk, upper limb, head and neck movements without resistance or progressions e.g. seated swimming, boxing, seated marching, heel and toe tapping; seated stretches; as well as activities e.g. balloon catch and throw, pass the parcel. The program is manualised, with class duration for each session recorded.
Class attendance, class duration and reasons for nonattendance will be documented for both intervention and control groups.
Intervention fidelity monitoring
To ensure fidelity of the intervention, an exercise class at each facility will be observed by a research investigator within the first 2 months of physiotherapist delivery and within the second 6 months of LTC facility delivery. A fidelity checklist which includes the number of participants, exercises completed, total time spent in standing and total class duration will be used to identify any deviations from the exercise protocol, with feedback given to the class facilitator. The research intervention coordinator will also audit all class exercise spreadsheets monthly to identify any deviations from the protocol.
Contamination between intervention and control groups is best controlled by a cluster randomised design, but that also introduces considerable heterogeneity given differences in length of stay, spatial design of the LTC, and differences in staffing ratios. Fall rates in facilities in a previous trial varied from 0.68 falls per resident year to 7.67 falls per resident year . LTC facility factors found to be associated with falls included the level of care (low level dependency had highest rate of falls) and staffing ratios . Individualised randomisation addresses these issues, although introduces the possibility of contamination. Contamination between intervention and control groups will be managed by separating delivery of the intervention and control classes, maintaining attendance registers for intervention and control classes and using different facilitators for intervention and control groups.
Risk management and safety monitoring
In the event of a fall or medical event, standard LTC facility procedures will be followed. Falls, mortality and unplanned re-admissions to an acute hospital service sustained during the trial period will be reported to an independent Data Monitoring Committee (see project governance) and reported to the relevant ethics committee.
All data will be stored in a confidential manner. Participants will be assigned a code, which will be used for all data management and analyses. Paper data will be stored at the local site in a locked filing cabinet. Coded electronic data will be stored on a computer server at the host organisation (The University of Auckland) for the duration of the study. Access to the computer files will be password-protected and accessible only to the research team. Data quality will be monitored by the data manager on a regular basis and reported to the Data Monitoring Committee. All project investigators will have access to the final, de-identified data set.
Sample size estimation
The sample size is estimated on the primary outcome, falls rate. To detect a 25% reduction in falls, assuming a control rate of 2.6 falls per resident per year (based on pilot data), we estimate a required sample size of 264 in each group (n= 528; two-tailed test, a=0.05, power= 90%). The anticipated drop-out rate is 35%, which will be replaced by recruitment at participating facilities throughout the trial. Final recruitment will stop 12 weeks prior to completion of the 12-month intervention in each facility.
The primary analysis will be conducted on an intention-to-treat basis using data from all randomized participants, although a ‘per protocol’ analysis of the primary outcome will also be reported.
Total number of falls, number of fallers, people sustaining a fall-related fracture or brain injury; fall rate (falls per 1000 person years); multiple fallers and number in each analysis will be reported. Prior fall-incidence rates will be calculated as number of falls/resident/year using the audited 6 months prior to enrolment.
Negative binomial regression models will be fitted to determine the Incident Rate Ratio (IRR) for differences in fall rates between groups both for overall follow-up time as the primary endpoint and secondarily activity adjusted. Similar negative binomial models will be built for fall injury rates; fractures and head injuries both combined and separately. Logistic regression models for fallen during follow-up or not fallen and for having fallen multiple times during follow-up or not fallen will be compared between intervention and control groups.
All models will control for prior falls rate, level of dependency and cognition (MoCA) as these are confounding factors. Baseline data will be compared between the two groups, with any strongly imbalanced factors further adjusted for in the analysis.
Per protocol analyses will be performed including those with higher attendance and pre-planned subgroup analyses will include those with moderate and high levels of cognitive impairment.
Fall rate relative to activity exposure will be calculated and compared between groups. Within group change will be examined using repeated measures. Generalized linear mixed-effects regression models will be used for volume, pattern and variability of ambulatory activity and cognition.
The cost effectiveness model will look at the difference in total costs of hospitalisation due to fracture or head injury and overall between intervention and control groups, compared to difference in the cost of two exercise programs. Unit costs will be assigned using weighted inlier equivalent separations New Zealand (WIESNZ) by type of hospitalisation, indicative District Health Board costs for outpatients, and literature/expert judgement for anything not already covered. Incremental cost effectiveness for intervention versus comparator activity will be assessed (i) per fall prevented and (ii) per injury fall (resulting in hospitalisation) prevented. The cost-effectiveness of Staying UpRight as an adjunct to usual activity will be explored for these same outcomes by removing the costs of the comparator activity (retaining any effect). Sensitivity analyses will be performed for all analyses to indicate the uncertainty in our estimation of the costs and consequences using bootstrapping. Decision uncertainty will be assessed using a probabilistic sensitivity analysis using standard diagrams (cost-effectiveness acceptability curve/ cost-effectiveness frontier) and varying assumptions around LTC facility size.
A range of further sensitivity analyses will be considered, including modifying the potential size of attendance within the intervention or control groups (where not fully subscribed), fall-related versus all-cause hospitalisation costs, and the likely cost of other healthcare utilisation in LTC.
The Principal Investigator (PI) is responsible for overseeing all aspects of the project. The University of Auckland’s research processes oversee the trial and provide financial integrity.
Ongoing monitoring of the study for futility, data integrity, and safety will be conducted by an external independent Data Monitoring Committee (DMC). The DMC will meet every 6 months. One of these meetings will involve formal review of interim statistical analyses. A nominated member of the DMC will be provided immediate access on an on-going basis to patient-specific information on Suspected Unexpected Serious Adverse Reactions.
Internal monitoring is through the Steering Committee and the Operations Committee (PI and key researchers), responsible for project management including recruitment, assessment and intervention delivery.
The results of the trial will be presented in international peer-reviewed journals and presented at conferences. Decisions about publications arising from the data set will be ratified by the Steering Committee and all named investigators will be eligible for authorship. Trial staff will also provide workshops to clinicians to assist the translation of findings to clinical practice.