The authors state that they have no conflicts of interest.
It is assumed that nursing homes are the setting with the highest incidence of hip fractures. This observation is, however, based on very little data. The aim of this study was to analyze hip fracture rates and the associated excess mortality in a large nursing home population. A cohort of >69,000 women and men newly admitted to German nursing homes were used to calculate sex- and age-specific incidence rates of hip fractures. To calculate excess mortality, a retrospective cohort study was conducted. To each patient with a hip fracture (n = 4342), four residents without hip fracture (n = 17,368) were matched by sex, age, and level of care (measure for the need of care). Hazard regression models were applied. During 91,850 person-years, 4342 hip fractures were observed. The crude incidence rates for hip fractures were 50.8/1000 person-years in women and 32.7/1000 person-years in men. The incidence rates increased with increasing age categories and were highest in the first months after admission to the nursing home. Increasing care need reduced the risk of hip fracture. Mortality in patients with a hip fracture was increased (women: hazard rate ratio for the first 3 mo after fracture, 1.72; 95% CI, 1.59–1.86; men: hazard ratio, 2.14; 95% CI, 1.80–2.53), but excess mortality was limited to the first months after injury. Data are presented for hip fracture rates and excess mortality after a hip fracture. Our results have implications on the timing and the allocation of specific measures for hip fracture prevention.
The incidence and secular trends of hip fractures have been analyzed extensively in many countries. The majority of studies have been of community-dwelling samples of older people where hip fracture is observed to rise exponentially with increasing age. The prevalence of risk factors for osteoporosis and falls are especially high in residents of nursing homes, and it is assumed that nursing homes are a setting with an especially high incidence of hip fractures. It is therefore surprising that nearly no data exist about incidence rates of hip fractures in residents of nursing homes. The few studies reporting incidence rates of hip fractures in nursing homes have either small case numbers or fail to report the time that each resident is observed for, making accurate estimation of incidence rates impossible.
In western countries, the vast majority of hip fractures are treated by surgery, which is associated with intra- and postoperative mortality risk. In addition, hip fractures are often followed by a considerable deterioration of health status. Several studies have estimated excess mortality caused by hip fractures in community-dwelling people. Excess mortality was found to be time dependent, with the highest risk within the first 6 mo and to differ by sex such that men were more vulnerable than women. Compared with community-dwelling people, residents of nursing homes may have higher mortality rates after a hip fracture because of a reduced general health status. The degree of excess mortality, however, depends on the baseline mortality, which is especially high in nursing home populations. To our knowledge, there exists only a subgroup analysis by Leibson et al., evaluating excess mortality caused by hip fractures in nursing homes. Their case numbers, however, were low and did not allow sex-specific analyses.
The aim of our study was (1) to estimate incidence rates of hip fractures in nursing homes, (2) to analyze the time period between nursing home admission and hip fractures, (3) to examine the association between functional impairment and hip fracture incidence, and (4) to evaluate excess mortality caused by a hip fracture during institutionalization.
MATERIALS AND METHODS
The dataset consisted of all people ≥65 yr of age, insured with the Allgemeine Ortskrankenkasse (AOK) and newly admitted to a nursing home between January 1, 2000 and December 31, 2005 in Baden-Württemberg, a federal state with 10.7 million inhabitants in southwest Germany. People resident in the nursing homes on January 1, 2000 were not eligible, because it was not known if they had suffered a first hip fracture during their stay at the home before January 1, 2000.
Health insurance, including cover for nursing home care, is statutory in Germany. The AOK is Germany's largest statutory health insurance company and covers ∼60% of all residents living in nursing homes in Baden-Württemberg. Fourteen percent of the 385,000 citizens ≥80 yr of age in Baden-Württemberg are residents of 1 of ∼1200 nursing homes. The insured persons are free to live in a nursing home of their own choice.
Since October 2003, a fall prevention program was introduced step by step in ∼300 selected nursing homes by the AOK. Each year, ∼100 nursing homes started at the same time with the program. Because the fall prevention program may have influenced fracture rates, residents newly admitted in these nursing homes after the start of the program were not considered. For residents already living in these fall prevention intervention homes, the time under observation ended at the date of the start of the intervention program. Homes were included in the program according to a waiting list. Before the start of the program (October 2003), baseline characteristics were very similar between residents of the intervention homes and of the remaining homes (mean age, 84.2 ± 7.2 and 83.8 ± 7.3 [SD] yr; level of care I [52.83% and 51.66%], level of care II [40.40% and 41.19%], level of care III [6.77% and 7.15%]; women [78.0% and 76.4%, respectively]). The final study cohort consisted of 69,692 people admitted to a nursing home between January 2, 2000 and December 31, 2005.
We used the routine data collection systems of the health insurance company to gain data on sex, age, date of admission to the home, level of care (see below), and, if appropriate, hip fractures and date of death for each individual. All data are held by the same company. Therefore, no linkage procedure between data from different sources was necessary.
Hospital discharge diagnoses were used to identify hip fractures (S72 in the 10th revision of the International Classification of Diseases [ICD-10]). Information about a second or third hip fracture during the observation time was also available. However, it was not always clear if a person had actually suffered from two different fractures or if one fracture had been coded twice. Therefore, only the information about a first hip fracture during institutionalization was used.
Level of care
In 1995, a long-term care insurance was introduced in the German social insurance system that is compulsory for all citizens. To claim for long-term care benefits, people must need a minimum of 45 min of assistance with basic care per day. Depending on the extent of care required, recipients are categorized into three levels after an assessment by a physician (levels I, II, and III requiring basic care such as washing, eating, or dressing for at least 0.75, 2, and 4 h/d, respectively). The level of care is therefore a measure for the need of care and the degree of functional impairment. All residents of nursing homes are assigned to one of the three levels of care.
Person-years at risk were accumulated between date of admission to the nursing home and date of discharge, the end of the study (December 31, 2005), introduction of a fall prevention program, date of death (residents without hip fracture), or date of first hip fracture. The crude incidence rate of hip fractures was calculated by dividing the number of fractures by the total number of person-years. Age standardization was performed according to the distribution of the official German standard population (men and women combined), and 95% CIs were calculated.
To show the association between age and hip fracture rates, seven age categories (each containing 5 yr) were created. Person-times at risk for age-specific fracture rates were assigned to the age interval in which they were accumulated.
To calculate excess mortality, a retrospective cohort study using an exposed cohort (those who had experienced hip fracture) and an unexposed cohort (those who had not experienced hip fracture) was conducted. Because personal characteristics may be associated both with fracture risk and mortality, a matching procedure was used to make both groups as comparable as possible. Residents in the comparison group had to be alive and institutionalized at the date of the matched patient's hip fracture. They were matched by sex, age (in 5-yr intervals), level of care, and date of admission to the nursing home. To each patient with a hip fracture (n = 4342), four residents without hip fracture (n = 17,368) were matched. Survival time for patients and their matched residents started at the date of the injury (admission date to the hospital). In subsequent analyses, no adjustment was performed for these matching factors.
Excess mortality in this cohort is shown by time-dependent mortality rates (30 days, 3 mo, 6 mo, and 12 mo) and by proportional hazard regression models. Because survival curves did not meet the proportional hazard assumption, proportional hazard regression models were performed stratified by four time intervals (0–3; 3–6; 6–12; and >12 mo).
Analyses were performed separately for women and men. All calculations were carried out with SAS version 9.1 (SAS Institute, Cary, NC, USA).
The study cohort admitted to nursing homes in Baden-Württemberg between 2000 and 2005 was made up of 52,946 women (76.0%) and 16,746 men (24.0%) ≥65 yr of age. At the time of institutionalization, the median age was 85.2 yr in women and 81.5 yr in men. Female residents were categorized into levels of care I, II, and III in 54.8%, 39.5%, and 5.7% and male residents in 48.7%, 43.4%, and 7.9%, respectively. During a total observation time of >91,000 person-years, 4342 hip fractures were observed (Table 1). The incidence rate (IR) for hip fractures was higher in women than in men (IR, 50.8/1000 versus 32.7/1000 person-years, respectively).
Table Table 1.. Crude, Age-Standardized, and Age-Stratified Hip Fracture Rates* in the Cohort of Residents Admitted to Nursing Homes in Baden-Württemberg Between 2000 and 2005
The IR ratio between women and men was 1.55 and remained essentially unchanged when calculating age-standardized instead of crude incidence rates.
To be able to follow persons newly admitted to the nursing home for some minimum amount of time, sensitivity analyses were performed excluding residents newly admitted to nursing homes during the last month and the last 3 mo before the end of the study. In both analyses, no change in hip fracture rates was observed.
The IR of hip fractures increased with increasing age categories up to ≥95 yr in men and up to 90–94 yr in women. In each age category, the IR of women exceeded that of men (Table 1).
A near linear association was observed between the level of care at nursing home admission and the risk of an incident hip fracture. Higher levels of care were associated with a reduced risk of hip fracture both in women and men (Fig. 1).
The IR was highest during the first months after admission and declined thereafter (Fig. 2). This pattern was more pronounced in women than in men. Thus, a substantial proportion of all hip fractures occurred during the first 3 mo after admission (23.4% in women and 28.2% in men).
In the subgroup in which excess mortality was analyzed, the mean age was 85.9 yr (SD, 6.7 yr), and 86.6% were women. The distribution of sex and level of care between residents with and without a hip fracture was identical, and the mean age between both groups differed only slightly (86.0 and 85.9 yr). Absolute and relative excess mortalities were increased during the first months after occurrence of the hip fracture but attenuated considerably during the first year (Table 2). Mortality and excess mortality were higher in men than in women during the whole observation period (Table 2). The risk of dying was significantly increased during the first 3 mo in women and the first 6 mo in men (women: HR in the first 3 mo, 1.72; 95% CI: 1.59, 1.86; men: HR in the first 3 mo, 2.14; 95% CI: 1.80, 2.53; Table 3). However, residents suffering a hip fracture who survived the first half year had even a more reduced risk of dying during the second half of the first year (women: HR, 0.81; 95% CI: 0.73, 0.91; men: HR, 0.77; 95% CI: 0.58, 1.02).
Table Table 2.. Mortality in the Hip Fracture Group (n = 4342) and in the Comparison Group* (n = 17,368) 30 Days, 3 mo, 6 mo, and 1 yr After the Time of Injury in Residents Admitted to Nursing Homes in Baden-Württemberg Between 2000 and 2005
Table Table 3.. Risk of Mortality in the Hip Fracture Group (n = 4342) and in the Comparison Group* (n = 17,368) in Different Periods After the Time of Injury in Residents Admitted to Nursing Homes in Baden-Württemberg Between 2000 and 2005
Our findings add substantial information to the literature on hip fracture epidemiology: To our knowledge, we present, for the first time, incidence rates of hip fractures analyzed in a large sample of residents of nursing homes and estimates of excess mortality after hip fracture. We found the risk of hip fractures to be highest in the first months after admission to the nursing home and observed hip fracture rates to decrease with increasing levels of care needs.
We observed a crude IR of 50.8 and 32.7 hip fractures/1000 person-years in institutionalized women and men. Not unsurprisingly, the sex- and age-specific IR of hip fractures in residents of nursing homes exceeded that in the German population substantially. This was especially pronounced in younger age categories. In the age category 65–69 yr, for example, IR was 13-fold higher in male residents and 20-fold higher in female residents than in men and women of the German population in 2003. This difference decreased with increasing age, but even in the highest age categories, IRs were about doubled compared with those in the German population. When comparing IRs between the nursing home population and the German population, one has to keep in mind that hip fracture rates of the nursing home population are included in the fracture rates of the German population. Therefore, the real difference is likely to be higher.
Not only women but also men have to be regarded as highly vulnerable to hip fracture in nursing home settings. The observed ratio between female and male fracture rates was lower than risk ratios from national and international populations independently of whether crude or age-standardized ratios are compared. For example, age-standardized sex ratios of 2 and more have been observed in populations of highly industrialized European countries. In the German population of 2003, the age-specific ratios between female and male fracture rates varied between 1.3 and 2.1 and were, for each age category, higher than in the nursing home population (exception: 65–69 yr).
To our knowledge, there exists no data about the time between nursing home admission and the risk of hip fractures. We found the risk of hip fractures to be highest in the first months after admission to the nursing home, with a considerable percentage of all hip fractures occurring during the first 3 mo. The pattern of decreasing hip fracture rates over time was more pronounced in women than in men. The interpretation of our finding remains speculative. However, it seems to be plausible that reorientation in a new environment (toilet, furniture, lighting, etc.) may be associated with an increased risk of falling. This is supported by a publication from Friedman et al., who found the fall rate of residents to be increased during the first 3 mo after their relocation to a new facility. In addition, nursing home admissions are frequently preceded by hospital admission. Therefore, a morbidity-related weakness may also contribute to the increased fracture rate immediately after transition.
Because hip fracture rates decrease with increasing living time in a nursing home, it is crucial to differentiate between fracture rates of a cohort (as we did) and fracture rates of a nursing home population at a specific calendar year comprising both newly admitted residents with high fracture risks and survivors with lower fracture risks. During 2000, for example, hip fracture rates in women already institutionalized at January 1, 2000 were considerably lower (42.3/1000 person-years) than those in women newly admitted during the year (66.8/1000 person-years).
Although the assessment of care needs used in this study was based on a qualitative judgment made by a physician, the method has been shown to have good levels of inter-rater reliability. The need for care is strongly associated with functional impairment, in particular mobility and comorbidity. We observed that the risk of hip fractures decreased with increasing degrees of care needs. Residents categorized at level III had less than one half the IR for hip fractures than those in level I. The most likely explanation for this observation is that people with lower care needs are more mobile, and by default, expose themselves to more risky situations that result in falls. This hypothesis is supported by Chandler et al., who found independence in transfer to be a significant predictor of osteoporotic fractures in female nursing home residents. A different intensity or quality of care in different levels of care may further modify the observed association.
Our results have implications on preventive measures in nursing homes. Basically, fall and fracture prevention in nursing homes have to be intensified because fracture rates in this setting are especially high. Prevention should be focused on residents with a lower degree of care need, and measures should be implemented immediately after institutionalization, because the risk of a hip fracture is highest during the first months. Measures such as an adjusted bed height, antislip stockings, appropriate lighting, pressure mats, and hip protectors are available to reduce the risk of falls or fractures straight after admission to the home. Sex and age remain independent risk factors, but they are much less important in residents of nursing homes as they are in community-dwelling people.
We observed an excess of mortality during the first months after occurrence of the hip fracture. The risk of dying was reduced during the second half of the first year. This may be because of a selection of relatively healthy survivors after hip fracture. To our knowledge, excess mortality attributable to hip fractures in residents of nursing homes has been examined only in a small subgroup analysis in residents from Olmsted County, MN. Similar to our results, they found excess mortality to be limited to the first months after the injury. In community-dwelling people, a similar limited temporal effect was observed in studies of the oldest old or in subgroups with functional impairments or comorbidity. In our data, excess mortality after a hip fracture was higher in men than in women. This is consistent with results in population-based studies.
A common problem in studies has been to disentangle two components of excess mortality: one caused by the hip fracture itself and one caused by morbidity associated both with an increased risk of falls and fractures and an elevated baseline risk of mortality. It is therefore essential to have an appropriate comparison group with a similar health status and function. In our study, the matching procedure considered, among others, a measure for functional status (level of care), resulting in a more robust comparison between residents with and without hip fracture.
Major strengths of the study are its large number of study participants and of hip fractures and the exact documentation of the time under risk at an individual level. An accurate estimation of the denominator is important to not underestimate fracture rates because survival time of residents in nursing homes is short and does usually not exceed several months to a few years. The case ascertainment is systematic, and the outcome misclassification should be low because nearly all hip fractures are brought to hospital. Therefore, we believe that the certainty of the presented fracture rates and of the fracture-related mortality is high.
Several limitations have to be considered. The legal situation for being eligible for a nursing home is the same in all Federal States in Germany, but our data were derived from only one health insurance company and may not be representative for the whole German nursing home population. Compared with insurants of other statutory or private insurances, persons insured at the AOK represent lower rather than higher socioeconomic levels. However, ∼60% of all residents of nursing homes in the federal state of Baden-Württemberg were covered by our analysis. The composition of the persons in nursing homes may be different in different countries and may be influenced by different social and political determinants. This could influence the external validity of our results. Furthermore, variables like nursing home size, environmental factors, or structure and education of the staff, which could have influenced hip fracture rates, were not available. In addition, we cannot exclude the possibility of some misclassification in the coding of the diagnosis of hip fracture at the hospitals.
In summary, we examined hip fracture rates and excess mortality after hip fractures in a large cohort of newly admitted residents to nursing homes. Hip fracture rates were substantially greater than those of community dwelling people and were highest in the first months after admission to the nursing home. Excess mortality caused by hip fractures was limited to the first months after the injury.
The authors thank Luzia Erhardt-Beer and Heinrich Wölfle from the Allgemeine Ortskrankenkasse (AOK) for the admission to the data and for the support of our analyses. This analysis was supported by a grant of the Forschungskolleg Geriatrie of the Robert Bosch Foundation.