This study has attempted to compare two quite different types of geriatric units in order to explore the issue of repeat fallers in two such different settings. Given that fall prevention strategies were already in place in both settings, this study may provide some insight into how to focus a secondary wave of fall prevention in different settings.
Factors associated with repeat falls differed by unit. Among fallers on the GRU, only age reached statistical significance for repeat falls but only at the 0.05 level, which given the large number of comparisons made, needs to be viewed with caution. The GRU may be seen as a very high risk environment, given that many of the patients have been admitted from an acute care hospital and most are deconditioned as well as carrying the burden of their recent illness. This is reflected in their high rate of first falls. For such a population the adoption of universal prevention measures is indicated. However, falls and repeat falls do occur and the risk of repeat falling appears to be greatest in the very old. Further fall prevention strategies may need to be specifically designed for, and more focused on, the older GRU patients, who are likely to be frailer and possibly sicker. A study of the factors operating in the very old may need to be undertaken to identify potentially modifiable factors to guide more specific interventions in these people.
Among fallers on the LSVU, none of the study variables were significantly associated with repeat falls. However, it is noteworthy that while falling on the LSVU was less common, repeat falling was more so. These patients are medically more stable and likely more independently mobile. Under these circumstances, residents who fall despite the presence of generic prevention programs, seem to identify themselves as being at high risk for repeat falling. One question raised by these results is whether a more generic falls prevention program is likely to be sufficient in such a setting or whether, in such a setting, individualized efforts should focus on first time fallers, with the goal being to determine the cause of the fall in order to reduce the risk of another fall. In a way one is relying on the patients to declare themselves at risk by falling once, and they can potentially be seen as people for whom the generic programs are insufficient. Of some reassurance is the observation that the risk of serious injury in the first fall was low.
The present study did not find any association between having a walker listed as a gait aid on the Incident Reporting Tool at time of first fall and repeat falling. However, it could not be determined if the fall was sustained while actually using a walker or if the fall event occurred in a patient who usually uses a walker but was not using their walker as prescribed. In a previous study it was shown that the use of a walker may be an added risk factor for falling in people living alone, the assumption being that if you live alone you may have to perform activities yourself for which a walker is a hazard . Whether similar situations exist in these care settings is not known. This uncertainty points to the need for further research on the use/non-use of the mobility aid at the time of the fall and the associated patient activities.
This study suggests that some repeat fall prevention strategies should be specific to the person (e.g., greater attention to older GRU patients), some strategies should be unit (population) specific (e.g., higher risks at certain times of day on the LSVU) and these are in addition to global strategies that are in place to enhance safety and reduce falls risk across the institutional setting (e.g. assessment and incident recording protocols). Strategies that include a comprehensive fall risk assessment which focused on the common and recognized extrinsic and intrinsic risk factors for all persons at time of admission, as suggested by the Canadian Task Force on Preventive Health Care, may reduce falls and perhaps also repeat falls. The impact of such a strategy has already been demonstrated in randomized control trials conducted in LTC settings  and in sub-acute rehabilitation settings .
There is emerging evidence that fall risk profiles and evidence-based approaches to intervention differ considerably among different geriatric care settings [14, 21, 22]. These differences in risk factors and risk profiles may be attributable to differences in the type of setting, the measurement tools used, the population demographics and characteristics with a few risk factors inherent for that particular setting .
Given that our study was conducted in units where fall prevention programs are in place, but falls still occur, this may speak to the need for a secondary phase of interventions at a more individual level directed to those for whom the global interventions are insufficient. The identification of factors which might predict further falling in those who have fallen once may allow more focused attention on those so identified in order to reduce future risk.
The present study has several limitations. Data from the pilot version of the Incident Reporting Tool were used to identify factors associated with repeat fall events. However, some of the data elements were not included or were incorrectly coded in this new tool. Level of incident (near-miss or actual fall event) was not consistently captured and gender of the patient/resident who fell, a factor known to be associated with repeat falls, was not included. Individual-specific dates of admission and discharge were also not available, thereby limiting assessment of exposure to fall risk (i.e. length of stay). Such data need to be added later. In addition, retrospective analysis of data using incident reports has been known to be confounded by issues of partial recording and under-reporting and by the fact that such tools were not designed specifically for research. However, they are a valuable source of information as they can be examined to identify population specific risk factors for falls, which can be utilized for targeted fall prevention strategies .
As well it was challenging to interpret some of the information. For example, it was not clear if the person was actually using the walker as a gait aid when s/he experienced the fall. While there is a team debrief following each fall event that likely captures a number of contextual factors, this information is not recorded in this database. Variables shown by the literature as being key to understanding repeat fall events such as mobility deficits, cognitive issues, and medications taken within the last 24 hours were not included in this version of the database thereby limiting a greater exploration of relationships among the predictor variables. Moreover, interpretation is difficult without control data drawn from the non-fallers on the units.
Finally, despite using two years of data, unit specific analyses lacked power.