Methods

Setting and participants

TNH is an acute hospital located in Melbourne, Australia. The hospital’s bed capacity increased from 225 to 323 during the study. A fall prevention programme was introduced in 2002. Fall data from inpatients admitted between 1999 and 2007 were included in this retrospective study. Ethics approval was obtained from the Northern Hospital Medical Research Ethics committee. Patient informed consent was not obtained as this project was considered an evaluation of usual care practice.

Definitions

A fall was defined as an event resulting in a person coming to rest inadvertently on the ground, floor or other lower level. A fall injury was defined as any physical damage that resulted from a fall (including bruising, abrasions, lacerations and fractures).

Interventions: the falls prevention programme

In 2002, a fall prevention committee was established, which included nursing, medical and allied health staff. The committee objectives were to (1) design an evidence-based fall prevention programme, sensitive to the local population at TNH, (2) work closely with clinical staff in the development and implementation of the programme, (3) develop tools and strategies that were feasible within current staffing and (4) integrate the programme into the existing care plan documentation, to ensure that strategies were implemented promptly and updated daily.

An audit of inpatient falls occurring in 2001 found that most falls occurred on the general medical and surgical wards, in association with toileting and while patients were alone. Fallers were more likely to have been admitted as a result of a fall, to be confused, impulsive or agitated and to require supervision or assistance with mobility. The STRATIFY risk assessment is a reliable and valid tool for identifying patients most at risk of falling.22 The findings of the local falls audit led to some modification of the tool to improve its sensitivity to the local falls risk factors at TNH (table 1).16 18

The fall prevention programme was introduced into the high-risk wards identified by the audit. The programme consisted of risk assessment of all patients and targeted interventions for patients classified as high risk. Staff selected one or more interventions as shown in table 2. The patient’s fall risk score and interventions were recorded each shift on the nursing care plan. Education for all hospital nursing and allied health staff was provided before the introduction of the programme in the form of in-services and placement of a fall prevention manual on each ward. All new hospital clinical staff received an overview of the programme as part of the hospital orientation. All hospital nursing and allied health staff attended a fall prevention educational session including competency assessment annually. Further details of the interventions and staff training are available on the on-line appendix (Appendix 1 http://qshc.bmj.com).

Data collection

It is TNH policy that all patient incidents including falls are reported. Before 2002, the hospital incident reporting system was paper based, but after this time, Riskman (Software Design & Enhancement, Victoria, Australia), a computer-based reporting database was introduced. Fall incident reports were reviewed and collated monthly by the injury prevention coordinator. Data on all patient-reported incidents were examined to investigate changes in incident reporting over the 9-year period. Data on the age, sex and length of stay of the studied patients were extracted from the hospital-maintained data sets. Bimonthly audits of nursing staff compliance with documentation of fall risk score and interventions on the care plan were completed by the injury prevention representatives on each ward.

Statistical analysis

The primary outcomes were hospital-wide and high-risk wards fall and fall injury event rates per 1000 occupied bed-days (OBD). The fall data were analysed using a generalised additive model (GAM), and the autocorrelated fall injury data using a generalised additive mixed model (GAMM) using R (R V.2.4, The R Core Development Team, http://cran.r-project.org/). Log OBD was used as an offset, and a first-order autoregressive (AR1) object was employed in the GAMM analysis. A cubic regression spline was selected. An approximate two sigma equivalent upper control limit was included in the charts derived from the GAM and GAMM analyses.23 These were at the 97.725th percentile of the Poisson distribution relative to the fitted values.

Results

The number of admissions and OBD increased over the study period. Mean length of stay decreased between 2002 and 2007 when compared with 1999 (p<0.05). The mean age of patients from 2001 to 2007 was lower than in 1999 (p<0.05). Hospital-wide fall rates increased in the preprogramme years of 1999 to 2001. Fall rates remained stable in the 3 years after implementation of the programme (2002–2005), increased moderately in 2006, then in 2007 decreased to just below the rate in 2002–2005 (table 3). Based on the GAM model, there was an initial rate rise between January 2000 and December 2001 from 2.54 (95% CI 2.15 to 3.00) to 3.84 (95% CI 3.32 to 4.44). Between January 2002 and September 2006, the rate remained stable (range: 3.41 to 4.11). Between October 2006 and December 2007, the fitted rate decreased from 4.13 (95% CI 3.65 to 4.67) to 2.83 (95% CI 2.24 to 3.59; p = 0.005) (fig 1). Hospital-wide fall injury rates decreased throughout the study period, with a more marked decrease between 2002 and 2004 and a relatively stable rate between 2004 and 2007 (table 3). The GAMM-fitted fall injury rate in January 1999 was 1.66 (95% CI 1.24 to 2.21), and in December 2007, it was 0.61 (95% CI 0.43 to 0.88). The reduction was significant (p<0.001) (fig 2). The rate of all reported patient incidents increased from 5.50 in 1999 to 32.39 in 2006 (table 3).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

GAM plot of hospital-wide falls data from January 1999 to December 2007.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

GAMM plot of hospital-wide falls injury data from January 1999 to December 2007.

High-risk ward fall rates increased in the preprogramme years of 1999 to 2001. Fall rates remained stable after implementation of the programme (2002–2007) (table 4). Based on the GAM model, there was an initial rate rise between January 1999 and December 2001 from 4.80 (95% CI 3.54 to 6.52) to 6.33 (95% CI 5.53 to 7.36). Between January 2002 and August 2006, the rate was variable (range: 5.56 to 7.12). Between September 2006 and December 2007, the fitted rate fell from 7.12 (95% CI 6.33 to 8.02) to 4.44 (3.49 to 5.66; p = 0.001) (fig 3). The GAMM-fitted fall injury rate in January 1999 was 3.27 (95% CI 2.74 to 3.90) and December 2007 was 0.94 (95% CI 0.76 to 1.17). The reduction was significant (p<0.001) (fig 4). Staff compliance with documentation of fall risk scores and interventions was high (>70%) (table 4).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

GAM plot of high-risk ward falls data from January 1999 to December 2007.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

GAMM plot of high risk ward falls injury data from January 1999 to December 2007.

Discussion

This study demonstrates that fall injuries, a common adverse event in older inpatients, can be reduced through implementation of a targeted fall prevention programme in the acute hospital setting. However, fall rates varied throughout the observation period, and no significant change in the rate from preprogramme to postprogramme implementation was observed in the follow-up period. Finding a significant and sustained reduction in injuries is promising, as this is a primary goal of fall prevention programmes. Further, the reduction in fall injuries achieved is superior to that reported in both randomised controlled trials (RCTs)9 19 and systematic reviews.11 Fall-related injuries increase resource utilisation, length of stay and the chance of unplanned re-admission or discharge to nursing home care.24 Thus, these results reflect significant improvements in patient outcomes, hospital efficiency and efficacy.

Sustainability of positive outcomes such as reduced fall injuries is an important outcome. Even if an intervention is found to be effective in an RCT, if it is not accepted by clinical staff and patients, or is not feasible to integrate into daily clinical practice, the intervention is of little benefit. The results of this research show that the fall prevention programme implemented at TNH appears to be effective, acceptable and feasible.

The finding of no significant reduction in the rate of falls in rate from preprogramme to postprogramme implementation in the follow-up period may be explained by several factors. The number of falls may have reduced more substantially, but the reduction may have been masked by an increase in the reporting of non-injurious falls after the programme implementation. The late reduction in falls could be consistent with a plateauing of reporting late in 2006, demonstrated by fewer reported patient incidents in 2007 compared with 2006. The rate of all patient incidents reported increased throughout the study period (table 3). This may have been fostered by implementation of computer-based reporting, increased feedback of patient incident data to staff and introduction of a coronial falls investigation standard in 2003. Increased staff awareness of the definition of a fall, and that all falls must be reported, as communicated in the training, may also have increased reporting. Historically, nurses reported injurious falls, but not those that were unwitnessed or non-injurious. The phenomenon of increased fall reporting after implementation of a fall prevention programme has been identified by past researchers.1 25 Alternatively, the modest reduction in falls may be attributable to a low fall rate before the programme implementation. Past studies in acute hospitals have reported fall rates per 1000 OBD of 9.26,14 12.526 and 10.6,27 while the rate on TNH high-risk wards was 4.63 in 1999 before the programme implementation (table 4). Therefore, there may have been a “floor” effect in the ability to decrease the fall rate significantly. However, the sustained reduction in fall injuries is novel.

Past studies that have found no significant reduction in falls or injuries have indicated that the results may have been reflective of poor adherence to the fall prevention protocol rather than an ineffective protocol.12 One study reported a 43% non-adherence with the fall prevention protocol;28 in contrast, over 70% compliance with fall prevention documentation was achieved at TNH. The differences in efficacy of previously reported studies may also be because of differences in the interventions included, programme intensity and resources dedicated to the design, implementation and monitoring of the programme.

There are several additional factors that may have contributed to the programme’s success. The fall risk assessment was sensitive to the local population facilitating direction of resources to patients who would benefit most from them and thus did not overwhelm nursing workload. The fall risk assessment could be completed in minutes and was not dependent on the input of multiple disciplines. Previous programmes have required input from nursing, allied health, medical and pharmacy staff to complete the fall risk assessment process. In this setting, where the length of stay is short, completion of a multidisciplinary assessment would result in delay in the identification of high-risk patients and subsequent delivery of interventions, especially after hours or on weekends. The use of a multidisciplinary fall risk assessment may have contributed to failure of similar programmes in the acute care setting.12 The inclusion of the risk assessment and interventions on the patient care plan created a formal process for the assessment and management of fall risk within each nursing shift.

This programme included only a small number of interventions from the myriad promoted in the literature, and the interventions were achievable within current staffing structures. One previous study required nursing staff to choose from 18 possible interventions and found no effect on the fall rate.13 It is not known whether one particular component of the programme was primarily responsible for the reduced rate of falls and injuries or whether it was the combination of interventions that was most effective. High–low beds when lowered reduce the height from which a person falls and, consequently, the likelihood of an injury. As such, use of high–low beds may reduce injuries without reducing falls. It is plausible that the results of this study of greater reductions in injuries than falls may be partially attributable to the use of high–low beds. The comparable price of high–low beds with other electronic beds appears to offer a cost-effective fall injury prevention strategy. The relationship between use of high–low beds and incidence of fall injuries, and the relative efficacy of the other interventions used in this programme, warrants further investigation.

The methodology used was not an RCT, and thus, some observed effects may have been because of factors other than the intervention; for example, changes in the hospital population, policy and environment. Although the magnitude of these other changes cannot be quantified, this research employed robust analyses on a large sample and with a long observation period. The chosen methodology addressed a specific question: can the application of RCT findings and practice guidelines in a “real-world” setting show demonstrable improvements? Our results suggest it can.

Conclusion

The implementation of an evidence-based targeted fall prevention programme reduced fall injuries but not falls in this acute hospital. The reduction in fall injuries was substantial and sustained. A number of factors were identified that drove this improved outcome. They were the identification of a local problem, involvement of clinical staff at all stages, the development and use of a falls risk assessment to guide the delivery of simple interventions, integration of processes into daily clinical practice and creating systems that demand accountability of staff.