Falls are a major cause of disability and death in older adults.1–4 A practical and efficient means of identifying people at an increased risk is needed. In 2001, the American Geriatrics Society (AGS), the British Geriatrics Society (BGS) and the American College of Orthopaedic Surgeons (AAOS) jointly published an evidence-based clinical practice guideline for fall prevention, which will be referred to as the AGS/BGS/AAOS guidelines.2 The guidelines contain a screening algorithm to assist health care professionals in the assessment of fall risk and intervention recommendations.
The components of the AGS/BGS/AAOS algorithm are evidence based but the algorithm itself has only been evaluated in one study. Lamb et al5 found the algorithm did not achieve thresholds indicating strong evidence of performance, though their results suggest that modifications could improve the estimation of fall probability. Work documenting the implementation of fall assessment guidelines by community and emergency department physicians has shown both an underdetection of falls and an inadequate evaluation of detected falls.6,7 Further evaluation of the performance of the screening algorithm is warranted in different populations to evaluate its ability to identify and stratify fall risk. The American Physical Therapy Association was represented on the committee reviewing the original guidelines and physiotherapists have an interest in the ongoing evaluation of the guidelines.2
The purpose of the present study was to evaluate the AGS/BGS/AAOS Guidelines' algorithm for fall risk screening in a group of community-dwelling older adults. As the algorithm can be applied in several settings, this evaluation takes the perspective of screening as part of routine care, not individuals presenting after a fall to a health care professional. Specific study objectives were (1) to evaluate the ability of the algorithm to identify and stratify fall risk on the outcomes of any fall and any injurious fall and (2) to calculate measures of prognostic accuracy (sensitivity, specificity, likelihood ratios, and diagnostic odds ratios) for the algorithm.
This study used prospective falls occurrence data that were collected during a field trial of a falls prevention program, ‘The Project to Prevent Falls in Veterans' (PPFV). The PPFV was carried out at the University of Western Ontario, funded jointly by Veterans Affairs Canada and Health Canada. Initial eligibility was determined by Veterans Affairs Canada, who generated the simple random samples from their client lists of veterans of WWII and the Korean War living independently in the community in Southwest Ontario. Two questionnaires were sent to each household, one for the veteran and the other for the veteran's caregiver (ie, spouse or other family member residing in the household). Additional details of the sampling and data collection procedures have been described in detail elsewhere.8
Phase II of the PPFV was a fall risk factor modification trial with 1 year follow-up. Participants from 2 regions in Phase I who consented to be recontacted and who self-reported at least one modifiable risk factor for falling were stratified into 5 groups representing the number of risk factors identified (lower extremity weakness; greater than or equal to 4 prescription medications; and balance, foot, and vision problems) (n = 348). Participants were randomized to 1 of 2 treatment groups within each stratum, the Specialized Geriatric Services (SGS) group or the Family Physician group. The SGS group received a comprehensive fall assessment performed by a geriatrician or physical therapist and was provided with individual recommendations for fall risk factor reduction. The assessment and recommendations was the intervention, there was no provision of any other treatment and participants were not recontacted to monitor or encourage adherence to recommendations. The Family Physician group and the participant's family physician were sent a letter summarizing the risk factors identified on the mailed questionnaire. Any treatment was left at the discretion of the family physician representing usual care. People with no reported modifiable risk factors formed an open arm in the study, who received the same comprehensive geriatric assessment, as the SGS group and educational materials on fall prevention and healthy living.
There were no statistically or clinically significant differences between the study arms in the proportion falling, the number of falls, or the time to first fall allowing the data to be used for the prognostic evaluation of the algorithm to identify individuals at risk for falling.9 For the present study, only data from participants in the SGS group were analyzed.
One hundred thirty participants in the SGS group received the comprehensive geriatric assessment, 129 had complete information on history of falls in the 12 months prior to the assessment. One hundred seventeen (91.4%) people had complete follow-up data of falls over the duration of the study. There were 23 paired observations of husbands and wives residing in the same household who contributed data analyzed in the present study, the statistical methodology outlined in the following text took into account the clustering of observations from people residing in the same household. Participant characteristics are presented in Table 1. People with incomplete falls data over the study period (n = 10) were excluded from this analysis. These individuals were not statistically significantly different from those who had complete follow-up information on their baseline characteristics, except more women were lost to follow-up. Reasons for missing falls data were: no longer interested in participating in the study (3 people), and did not submit any falls calendars after the baseline assessment (7 people).
Informed consent was obtained from all study participants. The project was approved by the University of Western Ontario Research Ethics Board for Health Sciences Research Involving Human Subjects.
The baseline comprehensive fall assessment utilized the basic version of the interRAI Community Health Assessment, a subset of the Minimum Data Set for Home Care (MDS-HC) version 2.0.10,11 Reliability and validity of all items of the MDS-HC in community settings have been reported.12,13 The interRAI asks each participant about a fall history in the 3 months before the assessment and an additional question about fall history in the previous 12 months was included. A study supplement included the Berg Balance Scale (BBS), a valid and reliable scale for the assessment of balance in older adults.14–16 The BBS consists of 14 balance tasks scored on a scale of 0 to 4, where 0 indicates the inability to perform a task and 4 indicates that the task is performed independently. The maximum possible score of 56 indicates no identified balance difficulties.
The AGS/BGS/AAOS screening algorithm indicates people with a history of one fall are to receive an assessment for balance and gait impairment. There are no specific prescriptive recommendations for measurement scales, though the guidelines state a person should be observed as they stand up from a chair without using their arms, walk several paces and return, an evaluation consistent with the Get Up and Go Test (GUGT).2 The guidelines state people with difficulty or who demonstrate any unsteadiness during the performance of this task require further assessment.2
The GUGT was not performed during the geriatric assessment of the PPFV. The GUGT is a composite test of the 3 functions of standing up and sitting down, walking, and turning.5 Information on the component functional tasks (sit to stand, stand to sit, ambulation, and turning) was available in the data set and were combined to produce a single variable used as a surrogate for the GUGT. An observational gait assessment was performed during the geriatric assessment with an abnormality defined as the presence of any one of the following: antalgic, ataxic, spastic, steppage, leg length discrepancy, waddling, frontal lobe, vestibular, and Parkinsonian that produced an unsteady gait and was recorded as a dichotomous outcome (present/absent). The use of any ambulatory aid was also scored as a mobility problem. The performance of the algorithm was also evaluated using a more detailed balance assessment; the Berg Balance Scale (BBS) with a score of less than 50 denoted impairment. The BBS score was combined with the observational gait assessment for the presence of impairment among people who reported a single fall in the previous 12 months.
Prospective information on daily falls was collected for 12 months by each participant's monthly submission of a falls log calendar. A “fall” was defined as the person coming to rest unintentionally on the floor or ground. An “injurious fall” was defined as a fall resulting in an injury that required the person to see a doctor. Participants who indicated falling in the previous month were interviewed by telephone to obtain detailed information about the specifics of the fall. Data collection for the baseline comprehensive geriatric assessment was started in May 2002, and collection of 1-year follow-up information on prospective falls was completed in January 2004.
The essential elements of the AGS/BGS/AAOS algorithm2 are reproduced in Figure 1. The key steps have been labeled to facilitate the following description of how the algorithm was applied to the data: all participants entered the algorithm in the “Any Falls in the Past Year?” box at the upper left. Participants' responses to the question on the number of falls in the previous year were used to determine subsequent flow through the algorithm: zero falls (Box A); has one fall; and 2 or more fallen (Box C). Participants reporting one fall were linked to data from their clinical assessment for the presence of gait and balance problems. Those with no identified problem were assigned to Box B, and those with an identified problem were assigned to Box C.
Prospective risk for any fall (# participants with history of falls/N) and any injurious fall (# injured participants with history of falls/N) were then calculated for each arm of the algorithm (boxes A to D) using the fall calendar data. A Mantel-Haenszel chi-square test for trend with one degree of freedom was performed across the 4 risk categories for each fall outcome. The presence of an increasing monotonic trend across the categories was evaluated by the calculation of incremental risk ratios using a multivariable Cox proportional hazard regression model for correlated data controlling for age and sex.17,18 Goodness of fit testing was performed through the calculation of Schoenfeld residuals for each predictor variable (P < .05) in the regression model.19
The 4 risk categories were then collapsed into the 2 recommended intervention categories as described in the algorithm: Boxes A and B into the “no treatment required” or low-risk group and Boxes C and D into the “detailed fall evaluation required” or high-risk group. The intervention categories were used to calculate the relative risk for falling and the diagnostic accuracy of the overall performance of the algorithm for each fall outcome. Success for the algorithm was defined as people identified as requiring further detailed assessment who then subsequently fell during follow-up.
Sensitivity and specificity and 95% confidence intervals were calculated using a generalized estimating equation for correlated values.20 The positive likelihood ratio (LR+) was calculated as [sensitivity/(1-specificity)], representing how much more likely a person who falls in follow-up had a positive test; negative likelihood ratio (LR-) as [(1-sensitivity)/ specificity], how much more likely a person who does not fall has a negative test; and the diagnostic odds ratio (DOR) as (LR+/LR-), a summary measure of LR+ and LR- providing a ratio of odds for a positive result in people who fall compared with the odds of positive test result in people who do not fall during the follow-up period of the study.21,22 The presence of LR+ more than 5 and LR- less than 0.2 demonstrates strong diagnostic evidence for the measurement scale. A DOR value of 1 indicates the test does not differentiate between people with and without the outcome, and values less than 1 indicate more negative test results among people with the outcome, whereas better discriminatory ability corresponds to higher values.22,23 All analyses were performed using SAS, version 8.2 (SAS Institute, Inc, Cary, NC).
Fifty-two (44.4%) people fell at least once, of whom 26 (22.2%) individuals experienced a single fall, 26 (22.2%) sustained multiple falls (≥2), and 36 (30.8%) sustained an injurious fall. Participants in the group reporting no history of falls in the previous 12 months possessed known and potentially modifiable risk factors for falls on the clinical assessment (Table 1).
The algorithm produced 4 distinct fall risk groups for each fall outcome (Figure 2). The observed increasing risk of falls across the 4 categories was statistically significant using the Mantel-Haenszel chi-square test for trend (P < .001). The adjusted regression analysis conformed to an increasing monotonic trend for each fall outcome with all incremental risk ratio estimates greater than 1.0, though the values were not statistically significant (data not presented). This finding supports that the average risk for any fall and any injurious fall increased hierarchically from Box A (no history of falls) to Box D (multiple falls).
Evaluation of the intervention groups for the outcome any fall found fall risk was 33% in the low risk group and 68% in the high-risk group recommended for a fall evaluation. The relative risk estimate was 2.08 (95% CI 1.42-3.05) indicating a statistically significant fall risk for people with a history of recurrent falls or a single fall with balance/ gait problems. The risk difference of 35% (95% CI 15.5-55.5%) indicated the amount of fall risk that is explained by the presence of a history of recurrent falls or a single fall with balance/gait problems. For the outcome any injurious fall, the fall risk was 20% in the low-risk group and 53% in the high-risk group. The relative risk estimate was 2.60 (95%CI: 1.53, 4.42) for people with history of recurrent falls or single fall with balance/gait problems. The risk difference of 32% (95% CI 12.2-52.5%) indicated the amount of excess fall risk that is explained by the presence of the history of recurrent falls or single fall with balance/gait problems.
The use of the complete Berg Balance Scale along with the observational gait assessment produced the same discriminatory ability as the functional components of the GUGT to identify future fall risk for people with a history of a single fall for either outcome of any fall or any injurious fall.
The overall prognostic accuracy of the algorithm for the outcome any fall was sensitivity of 0.50 (95% CI 0.36-0.64), specificity of 0.82 (95% CI 0.70-0.90) for a LR+ of 2.71, a LR- of 0.61 and DOR of 4.44 for correctly identifying people who will fall in the subsequent 12 months as needing further evaluation. For the outcome any injurious fall, values of prognostic accuracy were a sensitivity of 0.56 (95% CI 0.38-0.72), specificity 0.78 (95% CI 0.67-0.86), LR+ of 2.5, LR- of 0.57 and DOR of 4.29 for correctly identifying future participants with history of falls as needing further investigation.
This study demonstrated the AGS/BGS/AAOS algorithm is able to identify and stratify fall risk in a sample of community-dwelling older adults. The relative risk for falls between the 2 treatment recommendation groups is both statistically and clinically relevant for the outcomes of any fall and any injurious fall indicating significant public health relevance. The overall prognostic accuracy of the algorithm though is moderate with values just below the threshold levels set to identify superior clinical utility. The study by Lamb et al5 found lower values of prognostic accuracy for AGS/BGS/ AAOS algorithm, though their study was limited to a representative sample of older women with disability which may impact the relative importance of the screening items to identify future fall risk.
The inability of the algorithm to achieve threshold values of prognostic accuracy could reflect the 2 proposed intervention options may not adequately address the 4 risk categories that are collapsed to produce the 2 treatment recommended groups. An association of a monotonic increasing trend for fall risk was demonstrated across the 4 risk categories created by the 2 screening steps in the algorithm. It is clear that the 2 risk categories who would receive a comprehensive falls assessment possess a marked risk for any future fall at 60% to 74% and merit further comprehensive assessment. This section of the algorithm appears supported as being able to identify a high-risk group of individuals for falling and it further supports previous findings that the risk for falling increases as the number of risk factors increases.24–28 The success of falls prevention has been stated in linking treatment to the identified deficits on assessment.2,29
Although the trend association with regression modeling was not statistically significant it does suggest that the groups collapsed to the “no treatment” intervention group may be 2 distinct groups. Future fall risk in the 2 fall risk groups that are collapsed to the no intervention recommended group varied from 29% to 43.9%, representing a common outcome. Half of all future falls occurred in these groups due to the larger number of participants streamed to these groups. It is unclear in the fall prevention literature what value of fall risk denotes “low-risk.” Specifically, what level of future risk are we willing to accept, as not meeting a threshold to initiate further intervention? This particular point influences whether the algorithm can be stated to possess clinical utility as there needs to be consensus that fall risk in excess of 60% defines the threshold separating high from low risk.
People with a single fall but no balance or gait problems on assessment may need to be placed in a separate intervention category. The self-report of a recalled fall has been identified as an important independent predictor of future falls, which would support that individuals with a history of a single fall but no problems with balance and gait on clinical assessment require more study to best meet their needs to prevent future falls.30
The baseline assessment indicates that the absence of a history of falling does not preclude the presence of other risk factors for falling including risk factors that are potentially modifiable. Lamb et al5 found that the self-report of balance problems while walking was an independent predictor for people with no history of falling in their evaluation of the AGS/BGS/AAOS algorithm. The combination of self-report and the clinical assessment of functional performance measures improved overall performance of the algorithm. Other research has demonstrated the use of a questionnaire to elicit information on perceived balance disorders has been found to be an independent predictor of fall risk.31 Self-report and clinical assessment provide different information and when used together identify gradients of risk for mortality and nursing home admission.32 The systematic review by Ganz et al33 recommends that people without a history of falling in the previous 12 months should be assessed for balance and gait impairment both by self-report and clinical assessment. The results of these studies further support the importance of a formal evaluation of the components of the algorithm to refine its performance using evidence-based research of the algorithm directly.
Prospective studies of falls in healthy community-dwelling older adults without a history of falls in the previous 12 months have found the risk of falling in the next 12 months varies from 16% to 49% which is consistent with the present studies findings.34,35 In this population, nonthreatening activities account for the majority of fall-related activities.34 Considering it takes only one fall to produce deleterious effects, it may not be appropriate to recommend “no intervention” as an option when addressing the public health importance of preventing falls in the older adults. The most current Cochrane Collaboration review on interventions to prevent falls reports people with the presence of modifiable risk factors without a history of falls will benefit from treatment.29 Recent recommendations from the American College of Sports Medicine and the American Heart Association also recommend regular physical activity in older adults citing the specific benefits of a reduction in falls and injurious falls, and the prevention or mitigation of functional limitations.36 Use of the screening process does not have to be limited to fall prevention, but to the maintenance of autonomy, healthy aging and prevention of functional decline in general.
In this study, though gait and balance problems were obtained from a specialized geriatric assessment, the implementation of tests to screen is within the capacity of and lend themselves easily to the primary care setting. Furthermore, research on the form of clinical measurement used for gait evaluation, that is, timed tests versus the observation of abnormalities, could further refine the discriminative ability of the algorithm. Importantly, the evaluation of balance with a more detailed measurement tool did not increase the discriminating ability of the algorithm supporting the use of a simple screening measure like the GUGT.
No separate substudy was performed to directly compare performance between the formal administration of the GUGT and the composite variable of functional tasks that would be executed during the performance of a GUGT, which may be seen as a limitation. The composite variable was felt to evaluate the same essential tasks as the GUGT. The analysis using the composite variable produced the same results as the BBS, a comprehensive multi-item balance assessment scale. The AGS/BGS/AAOS guidelines2 define the presence of any difficulty in the task, as sufficient to merit further detailed evaluation and the composite variable would have achieved this requirement lending support to the study's findings.
The subsample of the PPFV analyzed in this study included only people who had reported modifiable risk factors on a mailed questionnaire. The time frame between the questionnaire and the comprehensive geriatric assessment was on average a year, therefore, the questionnaire responses do not necessarily reflect the person's status or concerns at the time of the geriatric assessment. The group analyzed may still not be fully representative of people who would be screened in the community, a critique that is also relevant to the study by Lamb et al,5 and the results may overestimate prognostic accuracy as a result. Information on a history of falls in the 12 months prior to the geriatric assessment was not available for the open arm of the PPFV Phase II, precluding their inclusion in the present study.
The sample size for this study was small, but considering the prominence of the guidelines and limited research evaluating the algorithm this study is an important contribution to the literature. The fall risk values are consistent with the study by Lamb et al,5 so the sample size appears to have had an effect on the precision of the estimates rather than the magnitude of the risk estimates. Repeat evaluation of the algorithm in a larger sample and a population consistent with the general community-dwelling older adults is merited. The authors are aware that the AGS/BGS/AAOS guidelines have undergone a revision but those results have not been released as a formal update, this study therefore represents evaluation of the current guidelines available to clinicians.37
The strengths of the study include the prospective design and the use of a reliable method to collect falls data in a large sample of community-dwelling older adults.38 All participants received a standardized comprehensive geriatric assessment, which minimized any bias that may result, if only some of the participants received a detailed evaluation as the AGS/BGS/AAOS algorithm2 sets out.
The algorithm can identify and stratify fall risk in community-dwelling older adults. The 2 intervention groups created by the algorithm have distinct fall risks with risk doubled in people with history of multiple falls or history of a single fall with balance/gait problems. The prognostic accuracy of the algorithm to maximize the identification of people who will fall as needing further detailed assessment falls below thresholds for clinical utility. Further research is required to improve the section of the algorithm for people having no history of falls and a single fall but no balance or gait impairment to identify factors that are associated with falling and establishing a consensus statement of the level of risk that defines the threshold between low and high risk.
1. Public Health Agency of Canada. report on seniors' falls in Canada. Ottawa: Public Health Agency of Canada; 2005.
2. American Geriatrics Society; British Geriatrics Society and American Academy of Orthopaedic Surgeons Panel on Falls Prevention. Guideline for the prevention of falls in older persons. J Am Geriatr Soc. 2001; 43:664–672.
3. Clemson L, Cusick A, Fozzard C. Managing risk and exerting control: determining follow through with falls prevention. Disabil Rehabil. 1999; 21:531–541.
4. Perell KL, Nleson A, Goldman RL, et al. Fall risk assessment measures: an analytic review. J Gerontol A Biol Sci Med Sci. 2001; 56A:M761–M766.
5. Lamb SE, McCabe C, Becker C, Fried LP, Guralnik JM. The optimal sequence and selection of screening test items to predict fall risk in older disabled women: the women's health and aging study. J Gerontol A Biol Sci Med Sci. 2008; 63A:1082–1088.
6. Rubenstein LZ, Solomon DH, Roth CP, et al. Detection and management of falls and instability in vulnerable elders by community physicians. J Am Geriatr Soc. 2004; 52:1527–1531.
7. Salter AE, Khan KM, Donaldson MG, et al. Community-dwelling seniors who present to the emergency department with a fall do not receive Guideline care and their fall risk profile worsens significantly: a 6-month prospective study. Osteoporos Int. 2006; 17:672–683.
8. Speechley M, Belfry S, Borrie MJ, et al. Risk factors for falling among community-dwelling veterans and their caregivers. Can J Aging. 2005; 24:261–274.
9. Moons KG, Royston P, Vergouwe Y, Grobbee DE, Altman DG. Prognosis and prognostic research: what, why, and how? BMJ 2009; 338: b375.
10. Institute of Gerontology U of M. interRAI. 2005; www.interrai.org
. Accessed May 6, 2007.
11. Morris JN, Bernabei R, Ikegami N, et al. RAI-HOME CARE (RAI-HC) Assessment Manual for Version 2.0: Primer on Use of the Minimum Data Set-Home Care (MDS-HC) Version 2.0 and the Client Assessment Protocols (CAPS): for use in Canada and the United States. interRAI Corporation, Washington, DC; 2002.
12. Landi F, Tua E, Onder G, et al. Minimum data set for home care: a valid instrument to assess frail older people living in the community. Med Care 2000; 38: 1184–1190.
13. Morris JN, Fries BE, Steel K, et al. Comprehensive clinical assessment in community setting: applicability of the MDS-HC. J Am Geriatr Soc 1997; 45:1017–1024.
14. Berg KO, Wood-Dauphinee SL, Willimas JI, Gayton D. Measuring balance n elderly: preliminary development of an instrument. Physiother Can. 1989; 41: 304–311.
15. Berg KO, Maki BE, Williams JI, et al. Clinical and laboratory measures of postural balance in an elderly population. Arch Phys Med Rehabil. 1992; 73: 1073–1080.
16. Berg KO, Wood-Dauphinee SL, Willimas JI, Maki B. Measuring balance in the elderly: validation of an instrument. Can J Public Health. 1992; S2:S7–S11.
17. Rothman KJ, Greenland S. Modern Epidemiology. 2nd ed. Philadelphia: Lippincott Williams & Wilkins; 1998.
18. Maclure M, Greenland S. Tests for trend and dose response: misinterpretations and alternatives. Am J Epidemiol. 1992; 135:96–104.
19. Kleinbaum DG, Klein M. Survival Analysis: A Self-learning Text. 2nd ed. New York, NY: Springer Science & Business Media; 2005.
20. Smith PJ, Hadgu A. Sensitivity and specificity for correlated observations. Stat Med. 1992; 11:1503–1509.
21. Deeks JJ. Using evaluations of diagnostic tests: understanding their limitations and making the most of available evidence. Ann Oncol. 1999; 10:761–768.
22. Glas AS, Lijmer JG, Prins MH, Bonsel GJ, Bossuyt PMM. The diagnostic odds ratio: a single indicator of test performance. J Clin Epidemiol. 2003; 56:1129–1135.
23. Deeks JJ. Systematic reviews in health care: systematic reviews of evaluation of diagnostic and screening tests. BMJ. 2001; 323:157–162.
24. Tinetti ME, Speechley M, Ginter SF. Risk factors for falls among elderly persons living in the community. New Engl J Med. 1988; 319:1701–1707.
25. vanBemmel T, Vandenbroucke JP, Westendorp RGJ, Gussekloo J. In an observational study elderly patients had an increased risk of falling due to home hazards. J Clin Epidemiol. 2005; 58:63–67.
26. Tinetti ME, Williams TF, Mayewski R. Fall risk index for elderly patients based on number of chronic disabilities. Am J Med. 1986; 80:429–434.
27. Tinetti ME, Doucette J, Claus E, Morottoli R. Risk factors for serious injury during falls by older persons in the community. J Am Geriatr Soc. 1995; 43:1214–1221.
28. Nevitt MC, Cummings SR, Kidd S, Black D. Risk factors for recurrent nonsyncopal falls: a prospective study. JAMA. 1989; 261(18):2663–2668.
29. Gillespie LD, Gillespie WJ, Robertson MC, Lamb SE, Cumming RG, Rowe BH. Interventions for preventing falls in elderly people. Cochrane Database Syst Review. 2003;(4):CD00034.
30. Gill TM, Williams CS, Tinetti ME. Assessing risk for the onset of functional dependence among older adults: the role of physical performance. J Am Geriatr Soc. 1995; 43:603–609.
31. Gerdham P, Ringsberg KAM, Akesson K, Obrant KJ. Clinical history and biologic age predicted falls better than objective functional tests. J Clin Epidemiol. 2005; 58:226–232.
32. Guralnik JM, Simonsick EM, Ferrucci L, et al. A short physical performance battery assessing lower extremity function: association with self-reported disability and prediction of mortality and nursing home admission. J Gerontol A Biol Sci Med Sci. 1994; 49:M85–M94.
33. Ganz DA, Bao Y, Shekelle PG, Rubenstein LZ. Will my patient fall? JAMA. 2007; 97:77–86.
34. Hill K, Schwarz J, Flicker L, Carroll S. Falls among healthy, community-dwelling older women: a prospective study of frequency, circumstances, consequences and prediction accuracy. Aust NZ J Public Health. 1999; 23:41–48.
35. Gabell A, Simons MA, Nayak USL. Falls in the healthy elderly: predisposing causes. Ergonomics. 1985; 28:965–975.
36. Nelson ME, Rejeski J, Bair SN, et al. Physical activity and public health in older adults: recommendation from the American College of Sports Medicine and the American Heart Association. Circulation. 2007; 116:1094–1105.
38. Ganz DA, Higashi T, Rubenstein LZ. Monitoring falls in cohort studies of community-dwelling older people: effect of the recall interval. J Am Geriatr Soc. 2005; 53:2190–2194.