A cluster randomised controlled trial to prevent injury due to falls in a residential aged care population

We undertook a cluster randomised controlled trial. All residential aged care facilities with at least 20 beds in the Hunter and Lower Mid North Coast areas of New South Wales were invited to participate in the study. Of the 92 eligible facilities, 88 took part. Consenting facilities were stratified according to mix of bed type (high-care, low-care and dementia-specific) and randomly allocated within strata into intervention or control groups by the statistician (R E G) using the procedure “surveyselect” in SAS statistical software, version 9.1 (SAS Institute Inc, Cary, NC, USA). Forty-six facilities were allocated to the intervention group and 42 to the control group. Facility staff were not blinded as to whether they were in the intervention or control group. Data collection commenced in July 2005 and concluded in June 2007. Ethics approval was obtained from the Hunter New England Area Health Service Human Research Ethics Committee.

The intervention involved employment of a project nurse to encourage a range of best-practice strategies at the facilities during the 17-month intervention period. The strategies promoted were: falls risk assessment; mobility assessment; use of hip protectors; calcium and vitamin D supplementation; continence management; exercise programs; appropriate footwear; medication review; and post-fall management review.

The project nurse provided link people from each intervention facility with information and resources to assist with preventing falls and fractures. An initial training session was held in November 2005. A set of resources was developed to support the 2005 falls prevention guidelines (known as the “Big Green Box”).4 The implementation of these resources was further encouraged during 3-monthly network meetings, held from February 2006 to June 2007 (six meetings in total), convened by the project nurse and attended by the link people. Intervention facility staff were also invited to attend a workshop to learn how to plan and run exercise programs. Support for the intervention was obtained from Divisions of General Practice.

Staff from facilities allocated to the control (usual care) group attended a workshop where data collection procedures were explained, and they were prompted to submit monthly data.

The main outcomes of interest were change in use of vitamin D supplements and hip protectors, and change in the rate of fall events. Measurement and comparison of outcomes were undertaken at two levels.

The first level assessed whether the intervention resulted in an overall increase in use of vitamin D supplements and hip protectors and a reduction in falls in intervention facilities when compared with control facilities (facility-level data). For 7 months before the intervention and for each intervention month, facilities conducted a record audit and completed a standard form providing data on the aggregate number of falls, falls resulting in fracture and/or hospitalisation, and deaths within 3 months of fractured neck of femur resulting from a fall. Data were also collected on the number of medication audits carried out by a pharmacist, the number of residents taking vitamin D supplements or receiving and using hip protectors, and the number of falls risk assessments completed on admission. Failure to produce monthly data was followed up by the project nurse.

The second level involved assessment of the effectiveness of the intervention for reducing falls risk among the cohort of residents who were in the selected facilities at the start of the intervention period (individual-level data). All participating facilities completed a “census” during January–February 2006 that recorded current residents’ date of birth, sex, resident classification scale (RCS) score, falls risk assessment score on admission, whether they were ambulant, and whether they were residing in a dementia-specific unit. Linkage to hospital separations data allowed ascertainment of fractured neck of femur and death.

Facility-level data were investigated to determine change in the rate of events per 100 beds. A two-piece mixed (both fixed and random effects) model was fitted to the data for each outcome of interest, adjusting for bed type (low-care, high-care and mixed low- and high-care). The first piece of the model estimated the baseline rate (intercept) and change (slope) in the 7 months before the intervention commenced (pre-intervention period), and the second piece estimated change (second slope) for the duration of the intervention.

Differences in the number of falls resulting in fractured neck of femur between the intervention and control groups for individual-level data were estimated using single-level survival analysis before fitting a multilevel Cox proportional hazards model. Covariates within facility level included age, mobility, dementia status, RCS score and falls risk assessment score. The between-facility part of the model accounted for group (intervention or control), bed type and facility size. Differences in the number of deaths between intervention and control groups were estimated using survival analysis. All facilities were analysed according to random allocation (intention to treat) using SAS version 9.1 and Mplus version 5.1 (Muthén and Muthén, Los Angeles, Calif, USA). Sample size was fixed by the number of available facilities.

At the time of the census, there were 5391 permanent residents. Box 1 shows that randomisation produced reasonably similar characteristics for residents in the control and intervention groups.

Monthly falls data collection forms were returned by 76%–97% of facilities for each month from July 2005 to June 2007. There was no systematic bias in returns according to type of facility or phase of the study (pre-intervention or intervention). Overall, six facilities withdrew from the project during the intervention. All withdrawing facilities provided sufficient data to allow retention in analyses (Box 2).

Mean use of vitamin D at baseline was 12.7 supplements per 100 beds (95% CI, 7.4 to 18.1) in the control group and was 6.7 per 100 beds (95% CI, 1.2 to 10.9) lower in the intervention group. Slope did not change during the pre-intervention period for either group (P = 0.4) but increased significantly during the intervention period, with mean slope of 2.0 supplements per 100 beds per month averaged over both groups (P < 0.001). A two-piece model with quadratic term, adjusted for bed type, showed that the intervention group had a significantly lower supplementation rate at baseline than the control group (P = 0.015). However, there were no differences in slopes, for either the first or second stage (pre-intervention and during intervention: P = 0.161 and P = 0.092, respectively), with respect to study group.

The baseline use of hip protectors was low, at only 5.1 per 100 beds (95% CI, 3.1 to 7.0). There was no difference between intervention and control groups, with both groups showing a small increase in use: first stage slope, 0.25 per 100 beds (95% CI, 0.06 to 0.43; P = 0.008); and second stage slope, 0.29 per 100 beds (95% CI, 0.17 to 0.41; P < 0.001).

There were no changes apparent for the use of risk assessments, medication reviews or calcium supplementation. Continence assessment was considered adequate at baseline in all but one facility. Similarly, only one facility did not have a regular exercise program, and nearly all facilities (87%) had regular footwear assessment programs in place.

Falls events recorded by facilities are shown in Box 3. For the entire cohort (intervention and control groups), there were 13.5 fractured neck of femur events per 1000 falls. Of these events, two occurred within the first 3 months of admission (15%). The risk of death within 3 months of a fall that resulted in a neck of femur fracture was 33%. Similar rates for all events were reported during the pre-intervention and intervention periods, after accounting for high variability of fall rates within facilities per month. The two-stage longitudinal regression model showed no evidence of change in the rate of falls from 16.0 per 100 beds (95% CI, 14.2 to 17.9) for either the pre-intervention stage (0.14 falls per 100 beds per month; 95% CI, − 0.17 to 0.45; P = 0.37) or after intervention commencement (− 0.023 falls per 100 beds per month; 95% CI, − 0.14 to 0.09; P = 0.686), when averaged over both groups. There were also no significant differences between intervention and control groups at commencement, with the intervention group being 2.40 falls per 100 beds (95% CI, − 1.25 to 6.24; P = 0.198) higher than the control group (12.91 falls per 100 beds; 95% CI, 6.89 to 18.93), nor over the first slope representing the pre-intervention period (0.18 more falls per 100 beds per month; 95% CI, − 0.39 to 0.76; P = 0.532) or over the second slope representing the intervention period (0.13 fewer falls per 100 beds per month; 95% CI, − 0.36 to 0.10; P = 0.259).

Of the 5391 people in the cohort, 215 were identified as having fractured neck of femur during the intervention period (106 from the control group and 109 from the intervention group). Rates of femoral neck fractures were similar in both groups (P = 0.8) and over time. There was no difference in rates of fractured neck of femur between intervention and control groups during the 17 months of intervention, and no difference in fracture rates in the entire cohort between the first 6 months of the intervention and the last 6 months. Resident factors found to be related to an increased risk of fracture were being ambulant (hazard ratio [HR], 0.61; 95% CI, 0.03 to 1.19; P = 0.04), having dementia (HR, 0.63; 95% CI, 0.27 to 0.999; P = 0.002), increasing age (HR, 0.04; 95% CI, 0.02 to 0.05; P < 0.001) and having a high falls risk assessment score. When compared with a high falls risk assessment score, those with a low score had a negative HR (− 0.56; 95% CI, − 0.94 to − 0.16; P = 0.006), as did those with a medium score (HR, − 0.72; 95% CI, − 1.08 to − 0.37; P < 0.001).

A sensitivity analysis excluding facilities that withdrew indicated no change in results.