In this study of older adults being discharged from the ED, we found that over 40% reported falling within the past year. This high percentage was reported in a cohort in which no patient presented with a fall-related complaint, and is consistent with rates reported in other studies of community-dwelling elders [21]. It demonstrates the importance of continued efforts to find effective and usable falls risk-stratification tools for older ED patients. Previous studies have largely concentrated on patient questionnaires and comprehensive geriatric assessment instruments [12, 13, 24, 25]. Many have used additional staff with geriatrics expertise, a resource not available in most EDs [24, 25]. These attempts have met with varying degrees of success. Those utilizing only ED personnel have generally been unsuccessful, likely due to failure of ED staff to follow the protocol suggestions [12, 13]. As a result, future efforts should concentrate on finding modalities acceptable to and adaptable by most EDs. These would ideally be rapidly and easily implemented. For example, the TUG test requires no additional equipment and can be performed by any trained ED personnel. The balance plate requires a modest initial investment, but could be adopted in EDs if purchased by them. The plate is mobile and can be kept on a small cart. It does not require recalibration with moving. The time to complete both tests in our study, although not specifically measured, was approximately 2-3 minutes.
Our goal was perform a pilot study analyzing the relationships between several potential falls risk-assessment modalities in the ED setting. Patient-supplied history of falls is only one of several potential risk factors for future falls and may provide an incomplete picture of risk of future falls [15]. As comprehensive review of all falls risk factors is unlikely to occur in the ED setting, identifying easily administered and interpretable testing modalities is crucial. The first steps in assessing such modalities include assessing their ability to be completed in the ED. In our study, both balance plate and TUG tests were obtainable in the ED as all patients were able to complete the TUG test and all but three were able to complete balance plate testing.
The second step is to understand the relationship between the modalities. If results differ between modalities, further study would be required of all of them. Conversely, if results do not vary, future studies could concentrate on only one. In our ED population, there was minimal correlation between TUG and balance plate results. This may be due to the different components of balance measured by the two modalities as TUG measures dynamic balance and the balance plate measures static balance. Other studies have noted only moderate association between dynamic and static balance in elders [26]. In fact, balance assessment modalities measuring different constructs may be complementary [17]. As a result, further study should clarify the advantages, if any, of complementary testing as compared to selecting a single modality in the ED.
Balance plates using limits of stability measurements have been used to predict fall risk in both institution-dwelling and community-dwelling elders [18, 19, 27, 28]. In addition to the lack of correlation between balance plate and TUG testing, there was no relationship between the balance plate testing and patient provided history of falls in univariate logistic regression analysis. The balance plate NSEO and NSEC measures did have an AUC of >0.60 in identifying falls in the week prior to ED visit. For these measures, cutoffs could be identified with a sensitivity >80% which were somewhat useful in ruling out a fall within the past week with a negative likelihood ratio of approximately 0.3. However, specificity was low and the confidence intervals for the ROC curves were wide, limiting the conclusions that may be drawn from them and indicating that few patients would be judged to be at low risk of falls.
An additional concern limiting conclusions to be drawn from our use of the balance plate was the decision to proceed with a single assessment of each balance plate test. Several authors have noted that multiple repeat sessions may be required to obtain the most reliable intra-session measurements and best correlation between measurements when performing balance plate testing [29, 30]. We chose a single measurement for two reasons. First, it is the recommended regimen from the balance plate manufacturer. Second, the test is most useful in the ED if it is short and easily accomplished. Repeat measurements would tend to decrease the usability of the test in the ED. However, given our results, it appears that a single session of COP measurements may not provide useful information in the ED setting.
The TUG test is recommended as a quick, routine falls-screening modality for older patients [15, 31, 32]. It is easy to perform, has demonstrated high intra-tester and inter-rater reliability [14], has shown construct validity [14, 16, 33], does not require specialized personnel, and is recommended by current guidelines [15]. In this ED study, TUG test results were related neither to balance plate testing nor to patient self-reported history of falls. In the regression models, the only near-significant relationship was between TUG testing and falls within the past 6 months. AUC for patient report falls was generally poor and with wide confidence intervals. The AUC was greatest for falls within the past month, 6 months, or year. For these time periods, TUG cutoffs could also be identified with a negative likelihood ratio sufficient to provide a small to moderate decrease in posttest likelihood of fall. The optimum TUG cutoffs of 12-15 seconds we found are consistent with those of other studies in community-dwelling elders [20]. Again, however, the results of the regression analysis and the wide CIs of the ROC curves indicate that there is generally poor agreement between TUG and patient reported falls history. In a study conducted among ED patients, Walker et al found that the TUG test was poorly predictive of ED revisit or admission, further supporting its lack of a proven role in ED patients [34].
The lack of association between TUG and falls history in the ED is different than previous reports in community-dwelling elders where TUG was able to discriminate between those with a history of falls and non-fallers, correctly classifying approximately 70% of patients [35]. In another study, TUG had a high sensitivity and specificity of 87% in predicting past falls [20]. It may be that in the acutely-ill ED setting, the TUG test has different test characteristics than in other community-dwelling elder populations. Based on out results and the results of Walker et al [34], the TUG test should not be adopted for ED use without validation in this specific population either alone or as part of a multifactorial risk assessment model.
Our study was limited by the fact that, although eligible, no patients presenting with a fall were included in the study cohort. Most previous studies of balance assessment have occurred in such patients, and this high-risk group is the recommended target for balance assessment [15]. ED patients who present with a fall have been shown to have worse performance on dynamic and static balance testing than non-fallers [7]. It may be that studying these modalities in elders presenting to the ED with a fall will improve the test characteristics.
We did not classify falls and did not focus on patients with known risk factors for falling. As a result, our cohort may have been at particularly low risk of falling and this may have affected our results. The possibility of a Type II error may also have occurred due to the size of the sample studied. Relying on patient recall may have resulted in missed episodes of falling. Similar self-reporting has previously been proven valid, with 80-89% sensitivity and 90-95% specificity for recall of a fall at 1 year in a review of 6 studies of falls recall [23]. However, these studies have not been conducted in acutely-ill ED patients, raising the possibility of even greater rates of misreporting. The possibility of at least 20% underreporting may have influenced the negative association in our study as noted by the example of TUG testing in the results section. In the worst case, assuming that those with the highest TUG scores had failed to report their falls, there was a substantial increase in AUC for the TUG test. Therefore, prospective evaluation of future falls would be the ideal method to identify an association between these tests and falling. We did examine prior falls at various time periods in our models given the acute nature of most ED visits as it is unclear if testing in acutely-ill ED patients will have similar characteristics to that conducted in stable outpatients. Additionally, we did not gather specific data on time taken to complete the tests which would be of interest prior to adoption in the ED. Most importantly, prior to applying these testing modalities in the ED, it will take further prospective trials to determine if these can reliably predict falls after the ED visit, and if acting on that information will be of benefit.