Incidence of and risk factors for falls following hip fracture in community-dwelling older adults.

Hip fractures in older adults are a major medical problem leading to increased morbidity and mortality. (1-7) Research has shown that hip fractures in elderly people can lead to decreased strength (force-generating capacity of muscle), impaired balance and mobility, and long-term loss of independence. Less than 50% of patients will regain their prior level of function following hip fracture. (2-5,7-10) Following a fracture, older adults are 3 times more likely to be functionally dependent, and 4 times less likely than those who have not fallen to return to walking in the community. (6) Up to 38% of older adults who previously lived independently in the community and who survive a hip fracture will require long-term care. (5,9)

Several studies have identified predictors of morbidity and functional outcomes following hip fracture. Prefracture predictors of functional outcome include age, (3,4,11,12) comorbidities, (3) history of falls, (11) perceived risk of falling, (11) previous fracture and hospitalization, (8) decreased mobility, (1,11-13) and decreased independence in activities of daily living (ADL). (8) Ingemarsson et al (1) found that Timed "Up & Go" (TUG) test scores at hospital discharge, prefracture walking habits outdoors, and activity level are strong predictors of walking ability and activity level 1 year after fracture. Individuals are more likely to require nursing home placement if they have decreased balance or poor gait (13) or if they are confused or delirious on admission to the hospital. (14) Patients who require readmission to the hospital following hip fracture are more likely to require total assistance for ambulation, to require nursing home placement, and to die. (15)

There is limited information on the incidence of postfracture falls and the factors associated with those falls among older adults. et al (12) followed 57 patients for 2 months and reported that 17.5% of patients with hip fractures had subsequent falls. Colon-Emeric et al (16) reported that 19% of community-dwelling men and male veterans sustained a second hip or pelvic fracture within 1 year of their initial hip fracture. Decreased quadriceps femoris muscle strength (measured with a strain gauge) and increased postural sway while standing have been reported as indicators of increased risk for falls following hip fracture. (17) Fox et al, (13) however, concluded that mobility status 2 months after fracture did not predict future falls.

Identifying elderly patients with hip fracture who have an increased risk for postfracture falls is critical in light of research documenting that participation in a fall prevention program can improve balance and mobility and reduce the rate of injurious falls in community-dwelling older adults. (18,19) In addition, identifying factors that predict the likelihood for a poor outcome following a fall-related hip fracture may facilitate discharge planning by identifying patients who will require additional rehabilitation resources to improve functional recovery and reduce the likelihood for future hospitalizations associated with postdischarge falls. Thus, the goals of this study were to examine incidence of falls in older adults with a fall-related hip fracture and to identify factors that would predict falls in the 6 months following hospital discharge.

Method

Subjects

The first 100 older adults admitted to Northwest Hospital from May 1994 to August 1995 for care of a fall-related hip fracture who met the inclusion criteria and agreed to participate were included in this study. Of the 100 participants admitted into the study, 90 were available for the 6-month follow-up, 3 had died, 3 had moved from the area, and 4 refused the follow-up contact.

Criteria for inclusion in the study were that the subjects: were aged over 65 years and living in the community, either in their own home or in a retirement center; had sufficient cognitive ability to provide informed consent; and had experienced a fall-related hip fracture. Subjects were excluded if they had a prior history of neurological or musculoskeletal impairment that would increase risk for falls, such as cerebrovascular accident or Parkinson disease, or if they had a history of lower-extremity joint replacement. People with unstable cardiac conditions or impaired cognition and dementia (as reported in the medical records) also were excluded.

The mean age of study participants was 83.4 years (SD=6.5, range=68-98). The majority of the study participants were women (83%) and people who were living alone in their own homes (80%). Subjects reported a high level of independence in ADL function ([bar.x]=47.8, SD=1.9) in the month prior to the fall-related hip fracture. Their prefracture and baseline characteristics are presented in Table 1. The number of subjects' comorbidities ranged from 0 to 9 ([bar.x]=3.8, SD=1.9). Sixty percent of the individuals used no ambulation device prior to the hip fracture, whereas 34.4% used a cane and 5.6% used a walker. Of all participants, 36.7% had a diagnosis-related group (DRG) code of "Major Joint and Limb Replacement," 38.9% had a DRG code of "Hip/Femur Procedures With Complications," and 17.8% had a DRG code of "Hip/Femur Procedures Without Complications." (20) Length of hospital stay ranged from 2 to 14 days ([bar.x]=5.6, SD=1.9). Ninety percent of the participants were discharged to a transitional care unit, 9% to their own home, and 1% to a nursing home.

Procedure

Subjects were interviewed within 48 hours of admission for a fall-related hip fracture. After informed consent was obtained, subjects were interviewed to obtain information on demographics and premorbid health and functional status. Demographic factors included age, sex, and residential and marital status. Premorbid functional status included a self-report history of 1 or more falls in the previous 6 months (not including the fall resulting in the current fracture), the type of ambulation device used in the previous 6 months, and the level of independence during the month before hospital admission when performing 7 basic forms of ADL: walking, bathing, dressing, eating, transferring, toileting, and grooming. Although the Functional Independence Measure (21) (FIM) was not administered to the participants, we determined the level of self-reported independence in ADL using a similar scoring method (1 = "total assistance or not testable" to 7 = "complete independence"). The total score for ADL was calculated by summing the scores of all individual items forming the scale, and could achieve values from 7 to 49. Change in ADL was calculated as the difference between the ADL total score at 6-month follow-up and the prefracture ADL score. A negative difference was considered a decline in ADL. Assistive device used for gait prior to the fall related fracture and at 6 months also was determined, and a hierarchical ordering of device types was established: no device, cane, walker, and wheelchair.

Medical factors were determined from the patient's medical record and included the primary DRG, (20) including "Major Joint and Limb Replacement" (DRG 209), "Hip/Femur Procedures, >17 Years of Age With Complications (DRG 210)," and "Hip/Femur Procedures, >17 Years of Age, Without Complications" (DRG 211); coexisting medical conditions; and number and type of prescription medications. Length of hospital stay (in days) also was recorded.

Six months following discharge from the hospital, participants were interviewed in their homes regarding their current health status, number of falls in the previous 6 months, and independence in ADL function. In addition, a 1-hour evaluation of balance and gait was conducted. Level of dependence performing ADL was determined by self-report using the same FIM scoring method. Balance was tested using the Berg Balance Scale, which rates balance during the performance of 14 tasks, including sitting, standing, reaching, leaning over, turning, and stepping. (22) Gait speed was calculated by timing participants as they walked at a comfortable pace using the assistive device most often used when walking in the community. Time (in seconds) taken to walk the middle 2.4 m (8 ft) of a total distance of 3.7 m (12 ft) was converted to velocity (in meters per second). The evaluations were conducted by 1 of 2 licensed physical therapists. Interrater reliability between the 2 physical therapists on the clinical evaluation protocol was good (intraclass correlation coefficient=.98). Subjects were asked to report the number of falls that had occurred in the 6 months following discharge. A fall was defined as any event that led to an unplanned, unexpected contact with a supporting surface. The number of hospital admissions for fall-related injuries was determined from medical records.

To study whether the type of change in assistive device was different in the 2 groups of subjects (those who had fallen and those who had not fallen), we defined the following categories: no change (indicating the same type of device was used before the fracture and after the 6-month follow-up); higher degree of assistive device (indicating a change to more dependence on a device, such as a change from no device to a cane or from a walker to a wheelchair); and lower degree (indicating less dependence on a device, such as a change from a cane to no device).

Data Analysis

The data were analyzed using the SPSS software program, version 11.5.* All variables were analyzed using descriptive methods. Differences in outcomes between the 2 groups of subjects (those who fell and those who did not fall during the time period leading up to the 6-month follow-up) were assessed by the Mann-Whitney test (23) for continuous variables and the Fisher exact test (23) for categorical data. The nonparametric Mann-Whitney test was chosen because of the lack of normality of some of the outcome data. Logistic regression was used to study potential predictors of a fall during the 6-month period after fracture. Initial variables chosen as potential explanatory variables included: sex, age, premorbid ADL function, premorbid use of an ambulation device, DRG, length of hospital stay, and premorbid fall history. These variables were selected because we hypothesized that they would be associated with postdischarge falls. Age and sex were entered first into the model. Then all other variables were entered using a forward stepwise method. An alpha level of .10 was used for all tests. This level was chosen because this study was exploratory in nature and because using a less restrictive significance level would avoid discarding potential risk factors.

Results

Fall Status at 6 Months

At the time of the 6-month follow-up, 53.3% (48/90) of the patients reported falls, with 62.5% (30/48) of them reporting 2 or more falls. Of the patients who had fallen, 18% reported that they had been readmitted to the hospital with fall-related injuries.

Functional Status

Balance and mobility. Table 2 compares balance and mobility measures at the 6-month follow-up in the 2 groups of older adults (those who reported falls during the 6-month period versus those who did not report falls). Older adults who reported falls had lower Berg Balance Scale scores and slower gait speed compared with those who did not fall (P<.001 for both measures). Prior to their hip fracture, 23.8% (10/42) of the participants who did not fall used assistive devices (cane or walker). In contrast, 54.2% (26/48) of the participants who fell used assistive devices. At 6 months, 71.4% (30/42) of the participants who did not fall used assistive devices, and 93.8% (45/48) of the participants who fell used assistive devices. Thus, both groups showed a substantial increase in the use of an assistive device for walking at 6 months following hip fracture. The change in the level of assistive device was not different between the 2 groups (P=.33), although the group of people who had fallen had a higher proportion of individuals using a higher degree of assistive device at the 6-month follow-up (Tab. 2).

ADL function. Results for ADL function are shown in Table 2. The follow-up ADL scores of the participants who had fallen were lower ([bar.x]=43.0, SD=7.0) compared with those of participants who had not fallen ([bar.x]=47.0, SD=3.1; P<.001). The mean changes in ADL scores between prefracture and 6-month postdischarge measurements were -4.2 (SD=6.4, median=-2, range=-27-3) for the participants who had fallen and -1.4 (SD=2.7, median=0, range=-13-1) for those who had not fallen (Mann-Whitney test, P=.001). Negative change in the total ADL score was considered to be a decline. Overall, 60% (54/90) of the subjects demonstrated a decline in ADL function, and 40% (36/90) had returned to prefracture ADL status. Of the participants who had fallen, 77% (37/48) demonstrated a decline in ADL function compared with only 40.5% (17/42) of those who had not fallen (P<.001). One person among the 42 who did not fall (2.4%) moved to an assisted-living type of facility, and 7 out of 48 of those who had fallen (14.6%) made such a move (P=-.06).

Factors Predicting Postfracture Falls

A logistic regression analysis was used to determine factors that predicted the probability of falling in the 6-month period following hospital discharge. The resultant model is shown in Table 3, including odds ratios, 95% confidence bounds, and P values for the test of the null hypothesis that the odds ratio was 1. Age and sex were not statistically significant (P=.42 and .57, respectively). In addition, premorbid ADL function was dropped from the model because there was too little variation in scores, thus limiting its usefulness as an explanatory variable. After adjusting for age and sex, premorbid use of a gait assistive device and a history of at least 1 fall in the 6-month period prior to fracture were significant predictors of postdischarge falls (P=.02 and .002, respectively). All other factors being equal, an older adult who used an assistive device was 3.15 times more likely to fall by the 6-month follow-up than a person who did not use an assistive device. Similarly, an older adult with a history of premorbid falls was 8.77 times more likely to fall by the 6-month follow-up than a person who had no history of premorbid falls.

The final logistic regression equation can be used to calculate an estimate of the probability that a person will fall in the 6 months following discharge based on the values observed at baseline (Appendix). For example, using the formula and the appropriate variable values, an 82-year-old woman who used an assistive device prior to the fracture and who had fallen once in the 6-month period prior to fracture would have an estimated probability of falling in the 6 months after discharge of 0.923. Similarly, a 70-year-old man who did not use any assistive device and did not have any falls prior to fracture would have a probability of falling in the 6 months after discharge of 0.172.

Fixing a cut point and classifying the subjects in terms of whether they will experience a fall or not if their probability is greater or smaller than that value gives us measures of sensitivity and specificity of the model. Using the logistic regression model and a cut point of 0.5, the overall proportion of older adults correctly classified was 73%, sensitivity was 70.8%, and specificity was 76.2%. Sensitivity could be increased to 91.7% with a cut point of 0.29; however, specificity would decrease to 35.7%. These levels of sensitivity and specificity mean that, if we adopted the cut point of 0.29, we would identify correctly those who will experience a fall 91.7% of the time, but we could expect to have a large number of false positive classifications.

Discussion

This study showed that about half (53%) of the older adults who sustained a fall-related hip fracture experienced another fall in the 6 months following hospital discharge. This rate of falls is higher than that reported by either McKee et al, (12) who reported that 17.5% of older adults fell again within 2 months of their hip fracture, or Colon-Emeric et al, (16) who reported a fall rate of 19% in community-dwelling men and male veterans in the year following fracture. Differences may be due to variations in methodology. The study by Colon-Emeric et al involved only men, and McKee et al followed older adults for only 2 months. Consistent with higher fall rates, hospital readmissions (18%) in our study also were higher than readmission rates reported by Boockvar et al (15) and Colon-Emeric et al. (16)

Fall rates in this study were higher than reported fall rates among older adults hospitalized for any reason. Mahoney and colleagues (24) reported that 14% of patients over the age of 70 years who were hospitalized for medical illness fell within 1 month of discharge, indicating a risk for falling in older adults who are hospitalized independent of hip fracture. Similar to results from our study, Mahoney and colleagues reported that decreased walking function and use of an assistive device were predictive of falls among this group of older adults.

Several studies (2-5,7-11) have shown that many older adults do not recover their baseline level of function following hip fracture. Only 40% to 60% of older adults will regain their prefracture mobility status within 6 months following a hip fracture, and less than 50% will regain their ADL status. (2-5,7-9,11) These studies have suggested that functional decline in older adults following hip fracture is related to the fracture itself, not aging or the presence of comorbidities. Magaziner et al (25) compared 594 older adults with hip fracture with 3 groups without hip fracture from different geographic regions and matched for age and comorbidities. They concluded that 25% of the group with hip fracture developed lower-extremity disability that otherwise would not have occurred. Norton et al (6) compared 911 patients with hip fracture with 910 people without hip fracture and found a decline in physical function (as measured by self reported mobility, dependence in ADL, and level of physical activity) in the group with hip fractures at 2 years, independent of the effects of increasing age or pre-existing medical conditions and disabilities. Although the design of our study does not allow us to determine whether the changes in ADL in our participants were a cause or effect of the falls during the 6-month postdischarge period, our results suggest that declines in ADL function are greatest in older adults who experience postfracture falls. In addition, postdischarge declines in ADL function and falls were associated with impaired balance and mobility at 6 months, suggesting that such declines may be related more to the coimpairments of balance and mobility than to the presence of a hip fracture.

The importance of balance impairments to postfracture outcomes is supported by results from the logistic regression analysis. Factors predicting falls in the 6 months following a fall-related hip fracture included use of an ambulation device and a history of falls prior to the one causing hip fracture. Thus, premorbid factors gathered on admission were useful in identifying older adults who were at risk for postdischarge falls. These findings have important clinical implications for postfracture care in older adults. Studies (18,19) have shown that falls in community-dwelling older adults can be reduced using a multidimensional approach that targets changing underlying risk factors. Exercises designed to increase strength, improve balance, and enhance mobility function have been shown to reduce the risk for falls in elderly people. (19) Identification of older adults at risk for poor outcomes early in their hip fracture care pathway could result in improved discharge planning. Specifically, older adults who are determined to be at risk for falling could be referred for outpatient physical therapy for exercises designed to improve strength, balance, and mobility function. More research is needed to determine whether a change in care pathway to include postdischarge physical therapy results in improved outcomes in older adults with a history of fall-related hip fracture. The sociodemographic characteristics of the older adults with hip fractures included in this study are consistent with those reported in other studies that examined health outcomes following hip fracture, suggesting the generalizability of the results of this study.

Limitations

There were a number of limitations in the current study. Prefracture ADL and fall history were determined by subject report and, therefore, were subject to error. In addition, falls in the 6 months following hospital discharge were reported retrospectively by self-report and, therefore, were subject to reporting error. Accuracy in reporting falls would have been improved by asking subjects to report falls on a monthly basis. The model developed to identify older adults who are at high risk for falling needs to be tested prospectively with a larger sample of older adults. Finally, we could not determine whether the ADL decline identified at 6 months occurred prior to or following the fall(s) reported during that period.

Conclusions

Older adults admitted for care of a fall-related hip fracture can be evaluated early in their hospital stay to determine risk for falls following discharge. Indicators may include a previous history of falls and prefracture use of an ambulation device. Elderly patients who fall within the 6 months following a hip fracture are likely to demonstrate poorer balance, slower gait speed, and greater decline in ADL from the prefracture level than those who do not fall. Restoring elderly patients to independent function following a hip fracture may be facilitated by expanding the focus beyond fracture healing to the identification and management of balance and mobility deficits, which may be associated with postfracture falls and loss of functional autonomy.

Appendix.

Logistic Regression Model

logit (p) = -3.30 + 0.03 x age - 0.37 x sex + 1.15 x device + 2.17 x prefall

where:

p = probability that a person will fall in the 6 months after discharge;

logit (p) = logarithm of (p/(1-p));

age = age at time of hip fracture (in years);

sex = 0 if female, 1 if male;

device = 0 if the person did not use assistive device prior to fracture, 1 if the person did use any type of device;

prefall = 0 if no fails in the 6 months prior to fracture, 1 if 1 or more falls during that period.

For a new patient, we can substitute the observed values in the equation, find the estimate of logit(p), and calculate an estimate of the probability p by calculating:

p = exp (logit(p))

1 + exp (logit(p))

References

(1) Ingemarsson AH, Franden K, Mellstrom D, et al. Walking ability and activity level after hip fracture in the elderly: a follow-up. J Rehabil Med. 2003;35:76-83.

(2) Roder F, Schwab M, Aleker T, et al. Proximal femur fracture in older patients: rehabilitation and clinical outcome. Age Ageing. 2003;32: 74-80.

(3) Michel JP, Hoffmeyer P, Klopfenstein C, et al. Prognosis of functional recovery 1 year after hip fracture: typical profiles through cluster analysis. J Gerontol A Biol Sci Med Sci. 2000;55:M508-M515.

(4) Eisler J, Cornwall R, Strauss E, et al. Outcomes of elderly patients with non-displaced femoral neck fractures. Clin Orthop. 2002;399: 52-58.

(5) Eastwood EA, Magaziner J, Wang J, et al. Patients with hip fracture: subgroups and their outcomes. J Am Geriatr Soc. 2002;50:1240-1249.

(6) Norton R, Butler M, Robinson E, et al. Declines in physical functioning attributable to hip fracture among older people: a follow-up study of case-control participants. Disabil Rehabil. 2000;22:345-351.

(7) Magaziner J, Hawkes W, Hebel JR, et al. Recovery from hip fracture in eight areas of function. J Gerontol A Biol Sci Med. 2000;55: M498-M507.

(8) Wolinsky FD, Fitzgerald JF, Stump TE. The effect of hip fracture on mortality, hospitalization and functional status: a prospective study. Am J Public Health. 1997;87:398-403.

(9) VanBalen R, Steyerberg EW, Polder JJ, et al. Hip fracture in elderly patients: outcome for function, quality of life, and type of residence. Clin Orthop. 2001;390:232-243.

(10) Kuisma R. A randomized, controlled comparison of home versus institutional rehabilitation of patients with hip fracture. Clin Rehabil. 2002;16:553-561.

(11) Hannan EL, Magaziner J, Wang JJ, et al. Mortality and locomotion (6) months after hospitalization for hip fracture: risk factors and risk-adjusted hospital outcomes. JAMA. 2001;285:2736-2742.

(12) McKee KJ, Orbell S, Austin CA, et al. Fear of falling, falls efficacy, and health outcomes in older people following hip fracture. Disabil Rehabil. 2002;24:327-333.

(13) Fox KM, Hawkes WG, Hebel JR, et al. Mobility after hip fractures predicts health outcomes. J Am Geriatr Soc. 1998;46:169-173.

(14) Dolan MM, Hawkes WG, Zimmerman SI, et al. Delirium on hospital admission in aged hip fracture patients: prediction of mortality and 2-year functional outcomes. J Gerontol A Biol Sci Med Sci. 2000;55: M527-M534.

(15) Boockvar KS, Halm EA, Litke A, et al. Hospital readmission after hospital discharge for hip fracture: surgical and non-surgical causes and effect on outcome. J Am Geriatr Soc. 2003;51:399-403.

(16) Colon-Emetic CS, Sloane R, Hawkes WG, et al. The risk of subsequent hip fracture in community-dwelling men and male veterans with hip fracture. Am J Med. 2000;109:324-326.

(17) Sherrington C, Lord SR. Increased prevalence of fall risk factors in older people following hip fracture. Gerontology. 1998;44:340-344.

(18) Tinetti ME. Preventing falls in elderly persons. N Engl J Med. 2003;348:42-49.

(19) Shumway-Cook A, Gruber W, Baldwin M, et al. The effect of multidimensional exercises on balance, mobility, and fall risk in community-dwelling older adults. Phys Ther. 1997;77:46-57.

(20) International Classification of Diseases. 9th rev. Los Angeles, Calif: Practice Management Information Corp; 2002.

(21) Keith RA, Granger CV, Hamilton BB, et al. The functional independence measure: a new tool for rehabilitation. In: Eisentberg MG, Grzesiak RC, eds. Advances in Clinical Rehabilitation. Vol 1. New York, NY: Springer Verlag; 1987:6-18.

(22) Berg K, Wood-Dauphinee SL, Willimans JT. Measuring balance in the elderly: validation of an instrument. Can J Public Health. 1992;83: $9-S11.

(23) Fisher LD, vanBelle G. Biostatistics: A Methodology for the Health Sciences. New York, NY: John Wiley & Sons Inc; 1993.

(24) Mahoney J, Sager M, Dunham NC, et al. Risk of falls after hospital discharge. J Am Geriatr Soc. 1994;42:269-274.

(25) Magaziner J, Fredman L, Hawkes W, et al. Changes in functional status attributable to hip fracture: a comparison of hip fracture patients to community-dwelling elderly. Am J Epidemiol. 2003;157:1023-1031.

A Shumway-Cook, PT, PhD, is Associate Professor, Department of Rehabilitation Medicine, University of Washington, Box 356490, Seattle, WA 98195 (USA) ([e-mail]). Address all correspondence to Dr Shumway-Cook.

MA Ciol, PhD, is Research Assistant Professor, Department of Rehabilitation Medicine, University of Washington.

W Gruber, MD, was Medical Director, SAGE Program, Northwest Hospital, Seattle, Wash, at the time of the study.

CA Robinson, PT, MS, is Clinical Instructor, Department of Rehabilitation Medicine, University of Washington.

All authors provided concept/idea/research design and consultation (including review of manuscript before submission). Dr Ciol, Dr Gruber, and Ms Robinson provided writing and data analysis. Dr Shumway-Cook provided data collection, project management, writing, and fund procurement. Dr Shumway-Cook and Dr Gruber provided subjects. Dr Gruber provided institutional liaisons.

Study was approved by the institutional review board of Northwest Hospital.

This investigation was supported by a grant from Northwest Hospital Foundation, Seattle, Wash.

This article was received August 23, 2004, and was accepted December 21, 2004.

Table 1.
Patient Characteristics at Baseline
Characteristic (a)

Age
 [bar.x]                                  83.4
 SD                                        6.5
 Range                                    68-98

Sex (%)
 Female                                   83.3
 Male                                     16.7

Prefracture residence type (%)
 Home                                     80
 Retirement center-independent            20

Prefracture ADL
 [bar.x]                                  47.8
 SD                                        1.9
 Range                                    40-49

No. of comorbidities
 [bar.x]                                   3.8
 SD                                        1.9
 Range                                     0-9

Premorbid use of ambulation device (%)
 None                                     60
 Cane                                     34.4
 Walker                                    5.6

DRG code for fracture type (%)
 Major Joint and Limb Replacement (209)   36.7
 Hip/Femur Procedures (B) (210)           38.9
 Hip/Femur Procedures (C) (211)           17.8
 Missing                                   6.6

Acute care length of stay
 [bar.x]                                   5.6
 SD                                        1.9
 Range                                     2-14

Acute care discharge disposition (%)
 Home                                      8.9
 Transitional care unit                   90.0
 Nursing home                              1.1

(a) ADL=activities of daily living, DRG=diagnosis-related group.

(b) > 17 years of age, with complications.

(c) > 17 years of age, without complications.

Table 2.
Comparison of Outcomes at 6-month Follow-up

                                       Fall        No Fall
Outcome                                (n=48)      (n=42)         P

Berg Balance Scale score (a)                                   < .001
 [bar.x]                               31.0        44.7
 SD                                    10.8         8.2
 Median                                31.5        45.8
 Range                                  7-49       23-56

Gait speed (m/s) (a)                                           < .001
 [bar.x]                               23.3        41.7
 SD                                    15.6        18.6
 Median                                18.3        40.8
 Range                                  0-66.0      6.1-86.4

Gait device, % (n) (b,c)                                         .013
 None                                   6 (3)      29 (12)
 Cane                                  33 (16)     14 (6)
 Walker                                44 (21)     36 (15)
 Wheelchair                            17 (8)      21 (9)

Change in assistive device,                                      .33
 % (b,c)
 No change                             33.3 (16)   45.1 (19)
 Higher degree of assistive            64.6 (31)   54.8 (23)
  device
 Lower degree of assistive device       2.1 (1)     0.0 (0)

Activities of daily living (ADL) (a)                           < .001
 [bar.x]                               43.0        47.0
 SD                                     7.0         3.1
 Median                                45          48
 Range                                 20-49       34-49

Change in ADL (a,d)                                              .001
 [bar.x]                               -4.2        -1.4
 SD                                     6.4         2.7
 Median                                -2           0
 Range                                -27-3       -13-1

Subjects with decline in ADL,          77.1 (37)   40.5 (17)   < .001
 % (b,e)

Residence status change, % (n) (b,f)                             .06
 Same status (independent living)      85.4 (41)   97.6 (41)
 Change to assisted living             14.6 (7)     2.4 (1)

(a) P value from Mann-Whitney test of difference of medians.

(b) P value from Fisher exact test.

(c) No change indicates same type of device before fracture and after
6-month follow-up. Higher degree indicates a change to more dependence
on the device (eg, from no device to cane, from walker to wheelchair).
Lower degree indicates less dependence on a device (eg,from cane to no
device).

(d) Change calculated from ADL prefracture to ADL at Crmonth follow-up
(negative values mean a decline in ADL).

(e) Decline in ADL is defined as any change smaller than zero.

(f) Same status indicates that subject went back to independent living
(either home or retirement center/independent living). Change to
assisted living indicates moving to assisted living or nursing
facility.

Table 3.
Potential Predictors of a Fall by the Time of the 6-Month Follow-up:
Logistic Regression Results

                                                     95% Confidence
                                         Estimated   Bounds for Odds
                                         Odds        Ratio
Explanatory Variable               p     Ratio       Lower   Upper

Age                               .42    1.03        0.96     1.11
Sex                               .57
 Female (reference)                      1.00
 Male                                    0.69        0.19     2.50
Premorbid ambulation device       .02
 None (reference)                        1.00
 Used assistive device                   3.15        1.16     8.55
Falls in 6 mo prior to fracture   .002
 None (reference)                        1.00
 1 or more falls                         8.77        2.28    33.72