Prior nonhip limb fracture predicts subsequent hip fracture in institutionalized elderly people
- First Online:
- Cite this article as:
- Nakamura, K., Takahashi, S., Oyama, M. et al. Osteoporos Int (2010) 21: 1411. doi:10.1007/s00198-009-1081-0
- 76 Views
This 1-year cohort study of nursing home residents revealed that historical fractures of upper limbs or nonhip lower limbs were associated with hip fracture (hazard ratio = 2.14), independent of activities of daily living (ADL), mobility, dementia, weight, and type of nursing home. Prior nonhip fractures are useful for predicting of hip fracture in institutional settings.
The aim of this study was to evaluate the utility of fracture history for the prediction of hip fracture in nursing home residents.
This was a cohort study with a 1-year follow-up. Subjects were 8,905 residents of nursing homes in Niigata, Japan (mean age, 84.3 years). Fracture histories were obtained from nursing home medical records. ADL levels were assessed by caregivers. Hip fracture diagnosis was based on hospital medical records.
Subjects had fracture histories of upper limbs (5.0%), hip (14.0%), and nonhip lower limbs (4.6%). Among historical single fractures, only prior nonhip lower limbs significantly predicted subsequent fracture (adjusted hazard ratio, 2.43; 95% confidence interval (CI), 1.30–4.57). The stepwise method selected the best model, in which a combined historical fracture at upper limbs or nonhip lower limbs (adjusted hazard ratio, 2.14; 95% CI, 1.30–3.52), dependence, ADL levels, mobility, dementia, weight, and type of nursing home independently predicted subsequent hip fracture.
A fracture history at upper or nonhip lower limbs, in combination with other known risk factors, is useful for the prediction of future hip fracture in institutional settings.
KeywordsCohort studiesFracture historyHip fractureNursing homesRisk factors
Hip fractures affect many aspects of health in elderly populations, including the ability to conduct activities of daily living (ADL) and especially mobility . The cost of hip fracture treatment and aftercare is considerable. In Japan, it is estimated to be 600 billion yen (5 billion dollars) annually and is increasing . Although the incidence of hip fracture in East Asians is lower than that of Caucasians , it appears to be increasing more rapidly than the rate of aging . Thus, hip fracture prevention in elderly people is an important public health issue in Japan.
Nursing home residents are recognized as a high-risk group for hip fracture. Previous epidemiologic studies have shown that hip fracture incidence in nursing home residents is 1.5- to fourfold higher than the general population [5–7], suggesting that risk factors for hip fracture in nursing homes may differ from those of the general population . Factors associated with hip fracture in institutionalized elderly people have been studied; reduced mobility and ability to conduct ADL [7–9] were shown to be key risk factors, in addition to conventional factors, such as falls and low bone mass [10–12]. It is also known that a history of bone fracture predicts hip fracture in elderly people [13,14]. This has an important implication for the prediction of hip fracture in nursing home residents, because information on past history of fracture is readily available. Some epidemiologic studies conducted within institutional settings [11,12,15] have examined how fracture history influences hip fracture risk, and one study  showed that historical fracture influences subsequent fractures. However, no studies have clearly defined the strength of the association between historical fractures and subsequent hip fractures in such settings.
Until now, ability of fracture history to predict hip fracture in institutional settings has been unclear. The aim of this study was to evaluate the utility of fracture history for the prediction of hip fracture in a cohort of nursing home residents in Japan.
Subjects and methods
We targeted 201 nursing homes that provide long-term care service in Niigata Prefecture, Japan; the homes were registered with Niigata Council of Institutions for the Elderly. Most nursing homes providing long-term care service in Japan fall into one of two categories: nursing homes for frail elderly who require extensive care (“special nursing homes”) and nursing homes for elderly requiring only limited care. Of the 201 homes, 140 homes (95 special nursing homes) participated in this study. Subjects for this study included 8,905 elderly persons living in the 140 homes on 1 April 2007. The study protocol was approved by the Ethics Committee of Niigata University School of Medicine.
The present cohort study was initiated on 1 April 2007, and the follow-up concluded on 31 March 2008. Nurses or caregivers at each home were asked to provide subjects' information on 1 April 2007; age, gender, height, weight, and past fracture history were obtained from medical records. Historical fractures were classified as fractures at the upper limbs, proximal femur (hip), nonhip lower limbs, and other sites except for vertebrae. We did not analyze historical vertebral fractures. X-p-related information is very important for vertebral fracture diagnoses, but we did not obtain this information for historical vertebral fracture. Ability to conduct ADL was evaluated by caregivers in charge using the Barthel Index , which consists of the following ten items: feeding, transfers (bed to chair and back), grooming, toilet use, bathing, mobility on level surfaces, stairs, dressing, bowels, and bladder. The level of assistance (complete dependence to independence) required for each item was scored on a two- to four-point scale, with a maximum total score of 100. Residents with a Barthel Index score of 0 were judged to be totally dependent in ADL (the lower the score, the higher the level of dependence). Mobility (ambulatory or not) was assessed using the item “mobility on level surfaces” in the Barthel Index. Dementia was assessed by caregivers in charge using the dementia scale provided by the Ministry of Health, Labour, and Welfare of Japan [18,19], which is an observer-rating scale widely used in the long-term care insurance system to indicate the level of individual care needed. This scale includes five categories: (1) no dementia (normal); (2) some dementia, but almost independent in his/her daily living (I, slight); (3) dementia with some difficulty in communication, but independent in his/her daily living on condition of someone's observation (II, light); (4) dementia with some difficulty in communication and requiring care sometimes (III, moderate); (5) severe dementia with much difficulty in communication and requiring care all the time (IV or V, severe).
Each nursing home was asked to report incident fractures, including hip fractures (fractures at femoral neck, trochanter, and femoral head) during the 1-year follow-up period. Diagnosis of hip fracture was based on hospital medical records. Dates of discharge from nursing homes were also obtained to calculate person-years.
Cox's proportional hazards model was used to calculate the hazard ratio (HR) of hip fracture. In the multivariate model, HRs were adjusted for gender, age, weight, dependence (1, independent or partially dependent; 0, totally dependent), total Barthel Index score (1, 55–85 (suboptimal level); 0, other), mobility (1, ambulatory; 0, not ambulatory), dementia (1, light–severe; 0, normal or slight), and type of nursing home. The stepwise method was used to find the best multivariate model that included fracture history to predict hip fracture. Candidate predictor variables were previous fracture history, gender, age (1, 80–89 years; 0, other), weight (1, <39 kg; 0, ≥39 kg), dependence (1, independent or partially dependent; 0, totally dependent), total Barthel Index score (1, 55–85; 0, other), mobility (1, ambulatory; 0, not ambulatory), dementia (1, light–severe; 0, normal or slight), and type of nursing home. Categorization of dummy variables (age, weight, total Barthel Index score, and dementia) was based on preliminary analyses in a subgroup of the present population . SAS software (release 9.13, SAS Institute Inc., Cary, NC, USA) was used for statistical analysis. A P value less than 0.05 was considered significant.
Baseline characteristics of subjects
Mean (SD) or proportion
Mean (SD) or proportion
n = 6,841
n = 2,064
n = 6,544
n = 1,973
n = 6,828
n = 2,058
Body mass index (kg/m2)
n = 6,541
n = 1,972
Special nursing home residents
Totally dependent subjects (total Barthel Index score of 0)
Level of dementiab
IV or V (severe)
Number and proportion of subjects with a nonvertebral fracture history by fracture site
(n = 6,841)
(n = 2,064)
Nonhip lower limbs
Hazard ratios (HRs) for hip fracture in subjects with or without a nonvertebral fracture history
Past history of fracture
Incidence (per 1,000 person-years)
Unadjusted HR (95%CI)
Adjusted HR (95%CI)a
Absent (n = 8,461)
Present (n = 444)
Absent (n = 7,654)
Present (n = 1,251)
Nonhip lower limbs
Absent (n = 8,494)
Present (n = 411)
Absent (n = 8,631)
Present (n = 274)
Absent (n = 6,454)
Present (n = 2,451)
Hazard ratios (HRs) for hip fracture of subjects with or without a combination of fracture histories of upper limbs and nonhip lower limbs
Combination of past fracture histories
Incidence (per 1,000 person-years)
Unadjusted HR (95% CI)
Adjusted HR (95% CI)a
Upper or nonhip lower limbs
Absent (n = 8,086)
Present (n = 819)
Upper and nonhip lower limbs
Absent (n = 8,869)
Present (n = 36)
Independent predictors of hip fracture selected by the stepwise method of Cox's proportional hazards model
Adjusted hazard ratio
Fracture history of upper or nonhip lower limbs (1, present; 0, absent)
Dependence in activities of daily living (1, independent or partially dependent; 0, totally dependent)
Mobility (1, ambulatory; 0, not ambulatory)
Total Barthel Index scorea (1, 55–85 (suboptimal level); 0, other)
Weighta (1, <39 kg; 0, ≥39 kg)
Dementia (1, light-severe; 0, normal or slight)a
Type of nursing home (1, special nursing home; 0, other)
Numerous studies have reported associations between prior and subsequent fractures. The meta-analysis conducted by Klotzbuecher et al. in 2000  revealed that a prior fracture predicts subsequent hip fracture with a relative risk of approximately two. In peri- and postmenopausal women specifically, the relative risk of prior fractures at wrist, vertebral, and hip for subsequent hip fracture was 1.9, 2.3, and 2.3, respectively , suggesting that any prior fracture may be predictive of hip fracture. However, such data are sparse in institutional settings. Only recently have researchers demonstrated that previous hip and other osteoporotic fractures are predictors of subsequent fractures , and it is still unknown whether previous fracture predicts subsequent hip fracture; therefore, the present study attempted to address this issue. The final multivariate model in the present study showed that those with historical fractures at nonhip limbs had a 2.1-fold higher risk of future hip fracture, which is consistent with the results of Klotzbuecher et al.
The present study is the first to demonstrate the importance of lower limb fractures at sites other than the hip in predicting future hip fracture. One possible explanation for this result is that hip fracture and other leg bone fractures may share a same cause (i.e., low bone mass) . In addition, prior leg fracture may be associated with the tendency to fall and the subsequent increase in hip fracture risk. In contrast, the present study did not show that historical hip fracture predicted future hip fracture, which is inconsistent with current knowledge . We studied residents cared for by nurses or trained caregivers day and night in an institutional setting. In such a setting, residents with a previous hip fracture may be more carefully treated than those without it. Furthermore, among the 1,251 subjects with a previous hip fracture, 613 (49%) underwent total hip replacement surgery (data not shown). These factors may have decreased the risk of a second hip fracture in nursing home residents in this study.
Historical fracture at the upper limbs alone predicted future hip fracture with a HR of 1.8 (borderline significance), which was lower than that associated with lower limb fracture (HR, 2.4) in our study. Upper limb bones are not weight bearing, unlike lower limb bones, which may result in the lower predictive ability. The meta-analysis conducted by Klotzbuecher et al.  showed that the relative risk (RR) associated with prior wrist fracture for subsequent hip fracture was 1.9. More recently, two large cohort studies have evaluated the RR of prior wrist fracture for subsequent hip fracture. Although results from a clinical cohort study (mean age, 63) indicated that prior wrist fracture was not significantly associated with subsequent hip fractures (adjusted HR, 1.29) , a population-based cohort study (mean age, 72) showed a significant association (age-adjusted odds ratio, 1.43) . Results from these two studies and our study suggest that an association between prior upper limb fracture and subsequent hip fracture may be stronger in older populations.
A combination of historical fractures at upper limbs or nonhip lower limbs significantly predicted future hip fracture with HRs of approximately 2. This result is consistent with the meta-analysis . Moreover, a combination of fractures in upper and nonhip lower limbs produced significant and higher HRs (up to 5). Although cases of multiple historical fractures and incident hip fracture were not common in this study, this type of combination appears to be promising for hip fracture prediction. Larger cohort studies are needed to confirm this finding.
The impact of risk factors in the present population (other than fracture history), such as ADL, mobility, dementia, and type of nursing home, has been discussed elsewhere [7,19]. Low body weight has also been found to be an independent predictor of hip fracture. This association may be mediated by bone mineral density (BMD), because body weight is an important predictor of BMD among the elderly [22,23]. Furthermore, low BMD is a predictor of osteoporotic fracture among nursing home residents .
This study revealed that some prior bone fractures are able to predict subsequent hip fracture independent of other important factors (e.g., ADL, mobility, body weight, dementia, and nursing home type). As shown in Table 5, historical fractures of limbs other than the hip, lower mobility, suboptimal ADL levels (Barthel Index score, 55–85), low body weight (<39 kg), and dementia are associated with an up to 23-fold (2.14 × 2.19 × 1.80 × 1.51 × 1.79) higher risk for hip fracture. A prior fracture at nonhip limbs, in combination with other factors, may help identify persons at high risk for hip fracture; this predictive value should be retested in future studies.
This study has some limitations. First, it is possible that misclassification bias regarding determination of historical fractures occurred. Information on subject historical fracture was based on medical records, and thus, data on fractures occurring during the stay in nursing homes are considered to be reliable. However, data pertaining to fractures which occurred before admission to the homes may be less reliable. This was the major limitation of the present study. If the misclassification occurred, the association between prior fracture and subsequent fracture would be expected to be significantly weakened. Second, we did not obtain data regarding the time that prior fractures occurred. It has been suggested that the association between the first and second fractures wanes with increasing time after the first fracture [24,25]. Schousboe et al.  showed that prior nonspine nonhip fracture increased subsequent fracture risk significantly during the first 5 years of follow-up (adjusted HR, 1.7), but that it increased only modestly after 10 years (adjusted HR, 1.2). The elapsed time after the previous fracture was not ascertained in the present study, and thus, we estimated the average risk of prior fracture for subsequent hip fracture.
This is the first study to evaluate the impact of historical fracture on future hip fracture risk in institutionalized elderly people. We conclude that prior fracture at limbs other than the hip independently predicts subsequent hip fracture in this population. In particular, historical fracture at nonhip lower limbs is the best predictor of future hip fracture among historical single fractures, and the combination of historical fracture at upper or nonhip lower limbs is the best predictor of future hip fracture. These findings, in combination with other known risk factors, have practical utility for the prediction of hip fracture in institutional settings.
We wish to thank the Niigata Council of Institutions for the Elderly for their assistance in data collection. We also thank Ms. A. Kagoshima for help with data analysis. This study was supported in part by a grant from the Chiyoda Mutual Life Foundation, 2006.
Conflicts of interest