Chandler JM, Zimmerman SI, Girman CJ, Martin AR, Hawkes W, Hebel JR, Sloane PD, Holder L, Magaziner J. Low Bone Mineral Density and Risk of Fracture in White Female Nursing Home Residents. JAMA. 2000;284(8):972–977. doi:10.1001/jama.284.8.972
Author Affiliations: Department of Epidemiology, Merck Research Laboratories, Blue Bell, Pa (Drs Chandler and Girman and Ms Martin); Department of Epidemiology and Preventive Medicine, University of Maryland at Baltimore (Drs Zimmerman, Hawkes, Hebel, and Magaziner); Department of Diagnostic Radiology, University of Maryland Medical Center, Baltimore (Dr Holder); Cecil G. Sheps Center for Health Services Research (Drs Zimmerman and Sloane), School of Social Work (Dr Zimmerman), and Department of Family Medicine (Dr Sloane), University of North Carolina, Chapel Hill.
Context Low bone mineral density (BMD) is a strong risk factor for fracture
in community-dwelling white women, but the relationship in white female nursing
home residents, for whom fracture rates are highest, is less clear.
Objective To assess the relative contribution of low BMD to fracture risk in nursing
Design Prospective cohort study with baseline data collected April 1995 to
June 1997, with 18 months of follow-up.
Setting Forty-seven randomly selected nursing homes in Maryland.
Patients A total of 1427 white female nursing home residents aged 65 years or
Main Outcome Measure Documented osteoporotic fracture occurring during follow-up as a function
of baseline BMD measurements higher vs lower than the median, and after controlling
for demographic, functional, cognitive, psychosocial, and medical factors.
Results A total of 223 osteoporotic fractures occurred among 180 women. Low
BMD and transfer independence were significant independent risk factors for
fracture in this nursing home sample (P<.001)
and the 2 factors acted synergistically (P = .06)
to further increase fracture risk. Compared with women whose BMD was higher
than the median (0.296 g/cm2), those whose BMD was lower than the
median had an unadjusted hazard ratio for risk of fracture of 2.1 (95% confidence
interval [CI], 1.5-2.8); women who were independent in transfer had a hazard
ratio of 1.6 (95% CI, 1.2-2.2) compared with women dependent in transfer.
Among residents independent in transfer, those with BMD below the median had
a more than 3-fold increase in fracture risk compared with those with higher
BMD (unadjusted hazard ratio, 3.1; 95% CI, 2.2-4.4). Among residents dependent
in transfer, those with BMD below the median had a 60% increase in fracture
risk (unadjusted hazard ratio, 1.6; 95% CI, 1.1-2.3). Adjustment for covariates
did not alter the BMD-fracture relationship.
Conclusions Our data indicate that low BMD and independence in transfer are significant
predictors of osteoporotic fracture in white female nursing home residents.
The incidence of osteoporotic fracture, especially at the hip, increases
dramatically with age and is highest among white women.1
Fractures in elderly, community-dwelling white women are associated with high
health care costs, long recovery periods, and high mortality rates.2- 4 Low bone mineral density
(BMD) is a strong predictor of fracture in postmenopausal, community-dwelling
white women.5- 7
However, the relationship is less clear among white female nursing home residents,
despite the high prevalence of osteoporosis and the nearly 10-fold higher
hip fracture rate than in community-dwelling elderly persons.8
Risk factors for fracture in other studies have included old age, impaired
mobility, poor cognition, psychotropic medication use, and high risk of falls.8- 13
Recent evidence suggests that bone loss continues into the 9th and 10th decades
and may accelerate,14- 16
which may combine with other factors to increase fracture risk in this population.
We conducted a prospective cohort study of white women in a representative
sample of nursing homes in Maryland to determine whether low BMD predicts
osteoporotic fractures independent of other risk factors. Knowing whether
BMD contributes to fracture risk in this population may be useful for designing
effective risk assessment and fracture prevention strategies in long-term
Residents were recruited from facilities participating in the Maryland
Long-Term Care Project, which is dedicated to the study of quality of life
among long-term care residents. All 221 licensed long-term nursing facilities
in Maryland were categorized by location and number of beds, and a stratified
random sample of 47 homes was selected.17 The
cohort followed up longitudinally included white women aged 65 years or older.
Residents were excluded if they were in a coma; had bone metastases, terminal
cancer, a prosthetic implant in both wrists/forearms, significant open skin
lesions on both hands/arms; or were admitted for rehabilitation only. Residents
were randomly selected within a facility if eligible women exceeded a quota
set to maintain representation across regional and bed strata. Institutional
review board approval was obtained prior to study initiation. Written informed
consent was obtained from residents or from family members if residents were
unable to provide consent.
Two trained teams of evaluators collected baseline data between April
1995 and June 1997. Distal radius BMD (bone mineral content/area) of the dominant
arm was assessed with a DTX-100 single x-ray absorptiometer (Osteometer A/S,
Rødovre, Denmark). Reliability of distal radius BMD was excellent.18 The project radiologist (L.H.) reviewed the quality
of all scans, discarding only 2% because of poor quality.
Demographic characteristics were collected from residents, family members,
nursing home staff, and medical records. The most recent Minimum Data Set
(MDS), a valid and reliable comprehensive nurse-completed assessment of functional,
cognitive, psychosocial, and medical status,19- 22
was obtained from each participant's chart. Variables derived from the MDS
included functional status (dependency in transfer, dressing, eating, bathing,
bed mobility, hygiene, and toileting), ambulatory status, vision, and history
of falls in the past 180 days. Transfer and ambulatory independence appeared
to capture similar information among those with both variables. Because 30%
of the sample was missing information on ambulatory status, only transfer
independence was used in analyses.
Cognition was assessed using the Mini-Mental State Examination (MMSE)23 and depression using the Cornell Depression Scale.24 Current medications, history of fracture, chronic
disease information (for calculating a Charlson comorbidity score25), and height and weight were obtained from the resident's
medical chart. Grip strength of the dominant side was measured using a Jamar
hand dynamometer (Sammons Preston Inc, Bolingbrook, Ill).
Trained medical record abstractors reviewed charts of all participants
for new fractures during the 18 months following baseline examination. Fractures
were counted only if documented by a radiographic report or written physician's
note (suspected fractures were not included), and only osteoporotic fractures
were included in these analyses. Osteoporotic fractures are those shown to
be associated with low bone mass in elderly community-dwelling white women,
and include vertebrae, hip, humerus, wrist, rib, clavicle, pelvis, leg, foot,
and toes.26 Excluded from analyses as nonosteoporotic
were fractures of the face (n = 6), fingers (n = 8), ankle (n = 9), and patella
(n = 4). Residents who died or were lost to follow-up were included in the
analysis up to time of death or loss to follow-up, after which they were censored
Bone mineral density was standardized using the mean (SD) of the full
cohort of white female nursing home residents. The BMD-to-fracture relationship
was also assessed by comparing residents with values above and below the median
BMD. The BMD-to-fracture risk relationship was illustrated graphically using
Fracture rate for any osteoporotic fracture was calculated by dividing
the number of first fractures occurring during follow-up by the number of
person-years of follow-up (calculated as time from initial examination to
the first fracture, end of study, death, or loss to follow-up). A similar
rate was calculated for hip fractures only. The relationship between BMD and
time to first osteoporotic fracture during follow-up was analyzed using Cox
proportional hazards models. A similar analysis was attempted for hip fracture,
although the study was powered only for osteoporotic fractures. When multiple
fractures occurred at the first time point, only the first listed fracture
diagnosis was included in analyses. Results were summarized using hazard ratios
with 95% confidence intervals (CIs).
Proportional hazards assumptions were assessed by the interaction of
time and BMD and by inspecting parallelism of estimated survival functions.
Linearity of the relationship between standardized BMD and fracture risk was
supported by nonsignificance of squared and cubic terms of BMD (P>.30). Effect modification by key variables (age, fracture history,
transfer status, and body mass index) was evaluated by including interaction
terms in the model. Potential confounding was investigated by assessing whether
the regression coefficient for BMD changed by more than 20% when individually
adding variables (shown previously to be related to fracture and BMD) to a
model with BMD only.27
Of 1953 eligible residents approached, 1456 (74.6%) agreed to participate.
Twenty-nine women (2%) were lost to follow-up because of missing data or lack
of access to medical records. Baseline characteristics of women lost to follow-up
differed little from the 1427 women who were followed up (Table 1). At baseline, 71 residents (5%) were taking calcium and/or
cholecalciferol; 28 (2%) were taking a prescription osteoporosis medication
(such as estrogen); and 10 (0.7%) were taking calcium, cholecalciferol, and
prescription osteoporosis medication. Overall, 89 residents (6.2%) were taking
any osteoporosis treatment. Forty-three residents (2.9%) had a history of
fracture in the past 6 months.
Mean (SD) and median BMD for this sample (0.302 [0.07] and 0.296 g/cm2) corresponds to a value of 3.5 SDs below the normal mean for young
women (manufacturer's young-normal mean is 0.504 [0.058] g/cm2).
Eighty-two percent of the sample had BMD values 2.5 SDs below or lower and
54% had values 3.5 SDs below or lower than the normal mean (Figure 1).
Four hundred forty-one (31%) of the 1427 residents died during the 18-month
follow-up, with 45 (10%) of the deaths occurring following a fracture. Nineteen
residents were transferred to another facility and 22 were discharged home
During the 18 month follow-up period (1762 person-years), 223 osteoporotic
fractures occurred among 180 women (Table
2). Twelve residents had more than 1 fracture on the same date and
38 patients had more than 1 fracture at different times during the 18 month
follow-up. One hundred eighty women had at least 1 new osteoporotic fracture
during follow-up, for an incidence rate of 109 per 1000 women per year (95%
A total of 84 hip fractures occurred during follow-up, accounting for
38% of total osteoporotic fractures. Nine residents had a second hip fracture
during follow-up. Hence, 75 women had at least 1 new hip fracture during follow-up,
for a rate of 44 hip fractures per 1000 women per year (95% CI, 34-54).
When assessed univariately (without BMD in the model), age 85 years
or older, history of fracture, and transfer independence were associated with
a 30% to 60% increase in fracture (Table
3). By contrast, 3 or more comorbid conditions, history of falls
in the past 6 months, and impaired vision were not associated with increased
fracture. Dependence in 3 or more basic activities of daily living (bathing,
dressing, eating, hygiene, bed mobility, or toileting), a variable strongly
associated with dependence in transfer, was associated with an approximately
33% lower risk of fracture.
Bone mineral density was a strong predictor of osteoporotic fracture
(P<.001); fracture risk increased with lower BMD
levels (Table 3). Nursing home
residents with values below the cohort median BMD (0.296 g/cm2,
corresponding to 3.6 SDs below young peak normal) had a more than 2-fold increase
in risk compared with women with values above the median (Table 3). Consistent with this finding, the cumulative incidence
of fracture was higher in nursing home residents with BMD in the lower 2 quartiles
of BMD than in those in the upper 2 quartiles (Figure 2). Using standardized BMD, fracture risk increased 40% for
every SD decrease in BMD (Table 4).
When age, comorbidity, cognition, grip strength, depression, body mass
index, falls in the past 6 months, fracture history, and psychotropic medication
use at baseline were added to the model individually, the hazard ratio (HR)
of fracture associated with BMD did not change more than 4%. Transfer independence,
however, was significant when added to the BMD-fracture model (P<.001) and there was suggestion of an interaction (P = .06).
Nursing home residents independent in transfer had a more than 3-fold
increase in risk of fracture (HR, 3.1; 95% CI, 2.2-4.4) if their BMD was below
the median and a 90% increase in risk for every SD decrease in BMD (HR, 1.9;
95% CI, 1.5-2.4). Nursing home residents who were dependent in transfer had
a 60% increase in risk of fracture if their BMD was below the median (HR,
1.6; 95% CI, 1.1-2.3) and a modest, nonsignificant elevation in fracture risk
for every SD decrease in BMD (HR, 1.2; 95% CI, 0.98-1.4). Adjustment for age
and other covariates separately or simultaneously did not change the fracture
risk estimates for BMD by more than 8% in either transfer-independent or transfer-dependent
residents (Table 4).
The rate of hip fracture was numerically higher in those with BMD below
the median than in those with BMD above the median (53/1000 person-years [95%
CI, 37-68] vs 35/1000 person-years [95% CI, 22-47], respectively). Risk of
hip fracture was 2.5 times higher in those with BMD below than above the median
among nursing home residents who were independent in transfer (HR, 2.5; 95%
CI, 1.5-4.2), but not among residents dependent in transfer. Adjustment for
age and other covariates did not change the risk of hip fracture associated
with BMD by more than 12% in transfer-independent or transfer-dependent residents.
We report the findings of the first large prospective study of the relationship
between BMD and fracture in a sample of white women residing in a representative
statewide sample of nursing homes. The findings demonstrate a strong relationship
between low BMD and fracture in this population, even after accounting for
other known risk factors. Importantly, although more than 80% of the sample
had BMD values more than 2.5 SDs below young adult normal mean, and mean BMD
levels in this population were lower than in community-dwelling white women
of similar age,28 a gradient with fracture
rate greatest in the lower quartiles of BMD was still observed. Studies of
elderly community-dwelling white women support a strong link between low BMD
and increased fracture risk, with at least a 50% increase in risk for every
SD decrease in BMD, regardless of the site of BMD measurement.5- 7
Even in community-dwelling white women aged 80 years or older, modest reductions
in BMD substantially increase fracture risk.28
We expressed the HR for fracture in terms of SD changes in BMD to facilitate
comparison of these findings with those previously reported. However, the
risk of fracture may not increase linearly with decreasing BMD across the
entire range of BMD in this population. Nevertheless, it is clear that low
BMD is a risk factor, even in the nursing home population, and the HR per
SD change may be considered a reasonable estimate of average effect size.
This nursing home cohort showed a substantially higher rate of fractures,
especially of the hip, than elderly community-dwelling white women.9,28 The hip fracture rate of 44/1000 person-years
in our cohort compares with 14.2/1000 person-years in community-dwelling women
aged 80 years or older and 4.2/1000 person-years in women aged 65 to 79 years.27 Wrist and humerus fracture rates in our cohort (14.4
and 13.2/1000 person-years, respectively) compare with rates of 10.2 and 6.4
in community-dwelling women aged 80 years or older and 7.4 and 3.4, respectively,
in community-dwelling women aged 65 to 79 years.27
The observed incidence rate of 109 osteoporotic fractures per 1000 person-years
of follow-up is also higher than rates of 55 to 82 per 1000 person-years reported
previously for nursing home populations.8 It
is possible that women in our sample were more impaired than nursing home
residents in earlier studies because of the current shift toward alternate
long-term care options for the less impaired. On the other hand, the fracture
rates in this study may be underestimated due to exclusion of suspected fractures
not documented by radiography or physician note. In addition, previous studies
have been based primarily on single homogeneous nursing homes and may not
be representative of nursing homes in general.
The most common fracture site was the hip, accounting for nearly 40%
of the fractures in this cohort. This finding is slightly less than the 50%
observed in prior nursing home studies,8,10
possibly because of more complete ascertainment of less severe fracture types
in this sample. The hip fracture rate in this study is 4 times higher than
the rate found in elderly community-dwelling women aged 75 years or older.9
Transfer independence emerged as a strong independent predictor of fracture
risk. Studies in smaller nursing home samples also have found higher fracture
rates among ambulatory residents.10,29
Intuitively, mobile nursing home residents have greater opportunity for falls
(and fracture) than less mobile residents.22
In a study of fall-related fractures among nursing home patients, more than
a third of fractures occurred during a transfer activity.8
Unlike in studies of community-dwelling populations, older age,7
psychotropic medication use,10 fracture history,30 low body mass index,31
and poor cognition12 were not independent predictors
of fracture when added to the model containing BMD. Perhaps these factors
are so prevalent in nursing home residents that they do not distinguish fracture
risk as well as in healthier community-dwelling populations.
Independence in transfer and low BMD appear to be the strongest predictors
of fracture in the nursing home, and the 2 operate synergistically to increase
the risk of fracture. When residents who can transfer independently also have
very low BMD, their risk of fracture more than triples compared with residents
with relatively higher BMD levels. Even among residents requiring human assistance
with transfer, those with the lowest BMD levels have a 60% higher risk of
fractures than those with higher BMD levels.
Approximately 70% of fractures in this sample were the result of a fall.
Information on the number of residents who fell but did not incur a fracture
during follow-up was not available. Previous studies in community-dwelling
populations have shown that, although most fractures result from falls, only
a small percentage of falls (1%-6%) result in fracture.32
Since low BMD contributes to the incidence of fracture among nursing home
residents independent in transfer, relatively higher BMD levels may offer
some protection against fracture.
We studied white women because of the high prevalence of osteoporosis
and fracture in this group,33,34
and because white women comprise more than two thirds of the US nursing home
population.35 Osteoporosis and fracture, though
less prevalent, are important considerations among male and nonwhite nursing
home residents17 and should be evaluated in
future studies. Few residents were being treated at baseline with medications
that influence BMD or fracture risk; information about such medication use
during follow-up was unavailable. It is possible that a small percentage of
women initiated therapy after BMD results became known to their physicians.
Hence, our BMD-fracture risk estimates may be conservative since including
treated residents would tend to dampen the BMD-fracture risk estimate.
This study was conducted in nursing homes in Maryland; it is unknown
how these results generalize to other regions in the United States or worldwide.
However, the nursing homes were drawn from a stratified representative sample
of nursing homes in Maryland, and participants were comparable to the US nursing
home population in demographic and functional characteristics.17
Low BMD and independence in transfer are significant predictors of osteoporotic
fracture in nursing home residents. Whether interventions to increase BMD
will lower the fracture rate among nursing home residents will require further
study. Effective programs to modify these factors are needed to reduce overall
nursing home fracture rates and the high costs of associated morbidity and
health care utilization.