Hip Fracture and Increased Short-term but Not Long-term Mortality in Healthy Older Women FREE

Hip fractures are a major public health problem, with nearly 300 000 occurring annually in the United States.¹ Hip fractures cause substantial short- and long-term disability¹ as well as increased mortality.^2- 19

In a recent meta-analysis,¹⁶ women had nearly 3-fold increased mortality risk in the year after hip fracture. Long-term (10-year) mortality data are mixed.^{2,4,7,9,12,14,20- 21} In the meta-analysis,¹⁶ excess mortality risk from hip fracture decreased during the first 2 years after fracture but did not return to the rate of age-matched control participants during long-term (10-year) follow-up.

Previous mortality risk estimates have some important limitations. Some studies^{3,6,9- 10,12,14,21} used hospitalized patients or registries to identify samples while drawing controls from the general community. Such studies are likely biased by differences in the health status of hip fracture cases compared with nonfracture controls. The use of hospitalized patients or registries also does not allow for examination of health status or other important covariates, such as bone mineral density (BMD), before fracture occurrence. Other studies¹⁶ have combined nursing home residents and community-dwelling women, populations that would be expected to have very different health status.

Such methodological limitations have made it difficult to determine whether the noted increase in mortality after hip fracture is the result of underlying poor health or the hip fracture itself. Some studies have found that healthy participants without comorbidities have no increased mortality risk after hip fracture^3- 4,15 and have concluded that comorbidities explain much of the increased mortality risk after hip fracture^2,6,19,22; others have noted higher mortality in individuals without comorbidities compared with those with comorbidities.¹⁴

Previous studies exploring the influence of age on mortality after hip fracture have conflicting results, and limitations include insufficient information about health status^13,23 and the use of registries to retrospectively identify cases.^9,13 Some studies^6,12,24- 26 have found that mortality after hip fracture increases with age. However, in several studies from Europe,^9,13,23 the relative risk of death after hip fracture compared with the risk of death in the general population was lower in older vs younger patients with hip fracture.

Among 5580 community-dwelling ambulatory women 65 years or older enrolled in our prospective cohort (1116 incident hip fracture cases and 4464 age-matched controls), our aim was to determine the short-term (≤1 year), intermediate-term (>1 to ≤5 years), and long-term (>5 to ≤10 years) mortality associated with hip fracture. Because expected mortality rates increase significantly with advanced age combined with poor health, a secondary aim was to determine whether healthy women 80 years or older would have increased mortality associated with hip fracture compared with healthy age-matched controls.

In 1986-1988, the Study of Osteoporotic Fractures (SOF) recruited 9704 community-dwelling women 65 years or older (>99% non-Hispanic white) in 4 US regions: Baltimore County, Maryland; Minneapolis, Minnesota; Portland, Oregon; and the Monongahela Valley near Pittsburgh, Pennsylvania.²⁷ Women who were unable to walk without assistance and those with bilateral hip replacements were excluded. All women provided written consent, and the SOF was approved by each site's institutional review board.

We prospectively observed 1116 women who had an adjudicated incident hip fracture (cases) during almost 20 years of follow-up until December 2005 (mean, 14.4 years; Figure).²⁷ Each fracture case was matched by exact year of age at baseline with 4 controls (n = 4464) without hip fracture; controls had to similarly survive at least to the time of the case hip fracture. To confirm that the selected controls were representative of the SOF population, we compared frequencies and means of selected baseline characteristics (age, body mass index, hip BMD, pack-years smoked, modified Mini-Mental State (3MS) examination score,²⁸ health status, standing from a chair, and medical conditions) between nonfracture participants who were selected and not selected for this analysis. The 2 groups were relatively similar except that those who were not selected for the study were younger (69 vs 73 years [age was a matching criteria for the fracture cases]). We observed women for up to 19 years after hip fracture (mean, 4.8 years).

To determine whether healthy women 80 years or older have increased mortality associated with hip fracture compared with healthy age-matched controls, we also analyzed a healthy older subset of women who were chosen on the basis of their status at the year 10 SOF examination (Figure). Inclusion criteria for the healthy subset were (1) survival to age 80 years or older at postfracture year 10, (2) ability to return to the clinic for the year 10 examination (ie, the examination did not need to be performed at the participant's home), and (3) reporting of good or excellent health at year 10. Of this healthy subset, 192 had an adjudicated incident fracture during a mean of 6.7 years' follow-up after the year 10 examination (defined as cases for this analysis). These cases were matched by age with 4 controls (n = 768) from the healthy subset who had survived to the time of the case's fracture.

Weight was measured in light clothing without shoes by a balance-beam scale, and height was measured by stadiometer. These measures were then used to compute body mass index.²⁷ Cognitive function was assessed using a modified version of the Mini-Mental State Examination (the 3MS examination).²⁸ Neuromuscular function tests included a participant's ability to stand from a chair without using her arms.²⁹ Age, smoking and alcohol consumption, medical history, and self-reported health were determined by questionnaire and interview at baseline, and current medications were recorded.

Dual-energy x-ray absorptiometry (DXA) was first available and measured by a whole-body bone densitometer (Hologic QDR 1000; Hologic, Inc, Bedford, Massachusetts) in 1988-1990 at the 2-year examination. The DXA BMD measurement standards and precision have been previously detailed,³⁰ and quality control procedures were rigorous.²⁷ Complete data for DXA total hip BMD was available for 86.1% of the cohort at the year 2 examination.

Participants were contacted every 4 months by postcard (with telephone follow-up for nonresponders) to ascertain incident hip fractures or death; more than 95% of these contacts were completed. Verification of incident hip fractures was obtained from radiology reports. Cause of death was determined from death certificates and, if available, other supporting documentation (medical records or hospital discharge data) at each SOF clinic site. For consistency in coding across SOF sites, all final outcomes (hip fracture and cause of death) were centrally adjudicated by a physician at the SOF coordinating center. The final adjudicated cause of death was determined using International Classification of Diseases, Ninth Revision codes, and the cause of death was categorized at the coordinating center.³¹

We identified death within the first year following the fracture for each woman within a matched case and control quintuplet (1 case plus 4 controls) according to the date of fracture for the case. When no death occurred for any woman within the quintuplet, we also evaluated mortality during the time beyond postfracture year 1 through postfracture year 5; quintuplets without a death within 5 years after the case fracture were further evaluated for mortality beyond year 5 through year 10 after the case fracture date. Thus, beyond year 5 through year 10, all 5 women in the quintuplet had to survive at least 5 years.

We used conditional logistic regression to evaluate the associations between hip fracture and mortality during several postfracture periods. Based on previous studies of the association between fracture and mortality,^2,32 models were adjusted for age, BMD, smoking status, self-reported health status, SOF clinical site, 3MS examination score, the ability to stand from a chair at the baseline examination, and history of diabetes mellitus, hypertension, stroke, or Parkinson disease. Total hip BMD was then added to the model to evaluate its effect. We tested for the presence of interaction by stratifying for age group (<70, 70-79, and ≥80 years). We also examined the associations between intertrochanteric and femoral neck fractures and mortality. We repeated the analyses in the healthy subset of older women as defined at the year 10 examination. We performed the analysis with SAS statistical software (version 8.2; SAS Institute, Inc, Cary, North Carolina) and considered P < .05 to be significant.

Hip fracture cases had lower body mass index, lower total hip BMD, and more cigarette exposure than did age-matched nonfracture controls (Table 1). A slightly greater proportion of cases had Parkinson disease and needed to use their arms to stand from a chair compared with controls, but the mean 3MS examination scores were similar in the 2 groups. In those younger than 70 years (but not in the older age groups), more cases than controls had diabetes mellitus (P = .002). Self-reported health status was worse in those 80 years or older who had a fracture compared with nonfracture controls (P = .04), but the health status was not different between younger cases and controls (data not shown).

Overall, 48.2% (n = 2690) of the sample died during a mean (SD) 14.4 (4.1) years of follow-up. Mortality risk was highest in the first year after hip fracture (Table 2). During the first year after an index hip fracture, 189 cases (16.9%) died compared with 374 nonfracture controls (8.4%) (odds ratio [OR], 2.3; 95% CI, 1.9-2.8). This increase in mortality persisted when the analysis was adjusted for other risk factors for hip fracture, including total hip BMD (OR, 2.4; 95% CI, 1.9-3.1). Although deaths in the control group were evenly spread across the first year of observation (approximately 25% every 3 months), 72.5% of year 1 deaths (137 of 189) in the cases occurred within 6 months of a hip fracture. In addition, more than half the deaths (99 of 189 [52.4%]) in the first year following hip fracture occurred within the first 3 months for the cases.

Mortality risks for participants who did and did not experience a fracture were generally similar after the first year (Table 2). In the 3340 participants who remained in the analysis after the first year (because no one in their quintuplet had died), 164 (24.6%) of the hip fracture cases and 585 (21.9%) of the controls died by year 5 (OR, 1.2; 95% CI, 1.0-1.5); the association was no longer significant after adjusting for potential confounders (1.2; 0.9-1.5). Among the 1265 remaining in the analysis between postfracture years 5 and 10, 36 (14.2%) of the hip fracture cases compared with 101 (10.0%) of the nonfracture controls had died by the end of follow-up (OR, 1.6; 95% CI, 1.0-2.6); however, this increased risk was no longer present in the multivariable models (1.4; 0.8-2.6). The associations were similar when stratified by type of fracture (femoral neck and intertrochanteric fracture data are available in eTable 1 and eTable 2).

We noted a significant interaction between age and risk of death after hip fracture (P = .002). When stratified into 3 age categories (aged <70, 70-79, and ≥80 years at the baseline examination), the risk of death from hip fracture decreased as age increased (Table 3). Those aged 65 to 69 years who experienced a hip fracture had a 5-fold increased risk of death in the first year (16.3% vs 3.7%; OR, 5.0; 95% CI, 2.6-9.5). In contrast, there was no increased risk of death in the first year after a hip fracture in those 80 years or older (20.3% vs 16.8%; OR, 1.1; 95% CI, 0.6-2.1). Those aged 70 to 79 years had a risk of death in the first year that was intermediate between the youngest and oldest SOF participants (16.5% vs 8.9%; OR, 2.4; 95% CI, 1.8-3.3). Examining longer-term follow-up, those aged 65 to 69 years had an increased risk of mortality after hip fracture during postfracture years 1 through 5 (17.2% vs 11.3%; OR 1.9; 95% CI, 1.1-3.2) and after year 5 through year 10 (12.1% vs 5.9%; OR 3.2; 95% CI, 1.0-10.2) follow-up periods, although risks were attenuated compared with the first year. The older age groups did not have an increased risk of death during long-term follow-up.

Healthy women who were 80 years or older did not have overall increased mortality after hip fracture (Table 4). However, short-term mortality was increased among these women: hip fracture cases had a nearly 3-fold increased risk of death in the first year after a fracture, even after adjustment for other risk factors, including BMD (OR, 2.8; 95% CI, 1.5-5.2). In the 600 women who remained in the analysis after the first year (because no one in their quintuplet had died), the risk of death was not increased. Indeed, there may even have been a decreased risk of death in hip fracture cases from 1 through 5 years after hip fracture, although the results were statistically significant only in the multivariable models adjusted for total hip BMD (OR, 0.4; 95% CI, 0.2-0.8).

In the overall sample, the leading causes of death were coronary heart disease, cancer, and stroke in both the hip fracture case and control groups. There were no differences in death rate from coronary heart disease between the hip fracture cases and controls (24.9% vs 26.0%), for stroke (9.8% vs 11.7%), or for sepsis (1.8% vs 2.1%). However, a greater proportion of hip fracture cases than controls died of pneumonia (10.5% vs 5.6%; P < .001), cognitive disorders (9.2% vs 6.7%; P = .03), and osteoporotic fracture (2% vs 0%; P < .001); in contrast, more controls died of cancer (11.0% cases vs 18.2% controls; P < .001).

We found that women 65 years or older who experience a hip fracture have a 2-fold increased risk of death in the first year after the fracture compared with similarly aged women who do not have a hip fracture. After the first year, mortality rates were similar when we examined the entire cohort of SOF women. Our overall findings of increased short-term mortality are consistent with a recent meta-analysis¹⁶ that found an almost 3-fold increased risk of death in the first year after a hip fracture. Previous data on long-term mortality are mixed,^{2,4,7,9,12,14,16,20- 21} with some,¹⁶ but not all,^7,12 showing increased mortality risk for up to 10 years after hip fracture, although excess mortality risk was lower during later years of follow-up compared with the first year or two after hip fracture.¹⁶

Limitations in previous study designs have made it difficult to determine whether excess mortality was driven by the hip fracture itself or by differences in prefracture health status, age, or both. Because of the large size and prospective extended follow-up of the SOF cohort, we were able to examine the effects of both health status and age on the association between hip fracture and mortality and found that they may influence mortality risk after fracture. Those aged 65 to 69 years had the greatest risk of short-term death (5-fold) after a hip fracture. In contrast, the oldest participants (≥80 years) overall had no increased short-term mortality risk after a hip fracture. However, when we separately examined women 80 years or older who reported good or excellent health status, we found a more than 2-fold increased risk of death in the first year after a fracture. Similarly, when we examined long-term mortality by age and health status, we found that those aged 65 to 69 years had a persistent increase in intermediate and long-term (5 to 10–year) mortality, although the risks were attenuated compared with short-term risk. In contrast, the oldest participants (≥80 years) had no increased intermediate or long-term mortality risk after a hip fracture regardless of their health status. Our data suggest that previous mixed results with regard to mortality, especially long-term mortality, after hip fracture may have been the result of differences in the underlying age and health status of the population being studied.

Our finding that short-term mortality risk was elevated only in the exceptionally healthy participants who were 80 years or older—and not the whole population 80 years or older—suggests that hip fracture itself may contribute to increased mortality in these older women. Most,^{2- 4,6,15,19,22} but not all,^9,14 previous studies have concluded that increased mortality risk after hip fracture was the result of underlying comorbidities, such as cardiovascular disease, stroke, respiratory disease, diabetes, or cancer. An earlier study² conducted in the SOF cohort concluded that a 2-fold increased mortality risk in hip or pelvic fracture cases compared with nonfracture controls during 5.9 years of study (64 deaths) was a consequence of increased comorbid conditions in the hip fracture cases.

Because the background mortality rate is very high in women 80 years or older, especially those with poor health, it would be more difficult to distinguish an additional independent risk of hip fracture on mortality. Our purpose in evaluating the exceptionally healthy subset was to better isolate the independent effects of hip fracture on mortality from those of comorbidities, and indeed we found an increased risk in the first year after hip fracture compared with similarly healthy age-matched controls.

Consistent with hip fracture being a contributor to death, we also found among the whole cohort that the highest increase in mortality risk was in the first 3 months after hip fracture. Although, in our study, less than 15% of the deaths were due to infection or osteoporosis (the most likely causes of death to be directly attributed to the fracture itself), hip fracture could have been a contributing cause in many of the remaining deaths, including those attributed to coronary heart disease and stroke.

Our finding that younger women have the highest mortality risk is consistent with some,^9,13,23 but not all,^6,12,24- 26 previous studies. However, previous studies were unable to disentangle the effects of age from health status. Our identification of women before their fracture, the prospective study design, and the extensive data on comorbidities allowed us to better examine how age and health status each influences mortality after hip fracture. We hypothesize that age influences the risk of death after hip fracture by affecting the baseline death rate in the population. Those who are younger and/or healthier have a low risk of dying from other causes. Therefore, experiencing a hip fracture may increase their mortality risk compared with nonfracture controls. In contrast, octogenarians generally have a relatively high risk of dying from other causes; therefore, experiencing a hip fracture does not result in an increased risk of death during the next year compared with other women their age, unless they are exceptionally healthy. This hypothesis is supported by the recent meta-analysis¹⁶ on the association between hip fracture and mortality; in that analysis, mortality from other causes increased rapidly with age, reducing the absolute excess mortality after hip fracture in the oldest group of patients.

Our study has important strengths. It was set within a large community-based prospective study of older women, had rigorous quality control of BMD measurements, and had study physician–adjudicated hip fracture data. Retention of survivors was excellent, with well over 95% completion of the follow-up information—including women who became too frail to attend subsequent visits. In addition, we were able to adjust for several important covariates.

Our study also has several limitations. It was conducted in postmenopausal white women 65 years or older and may not be generalizable to men, other ethnic groups, and younger women. However, it provides important information on the group most likely to experience a hip fracture—older white women. One of our study's strengths—its long observation period—is also a limitation because standard care for treatment of hip fractures has changed over time.³³ However, mortality rates after hip fracture have not decreased much, if at all, during the past 10 to 40 years^7,34- 35; therefore, our results are likely still relevant to considerations for prevention of hip fracture and its associated mortality.

Despite our selection of a healthy subset and our adjustment for multiple risk factors, we cannot exclude residual confounding by other diseases or frailty. We often measured covariates months to years before the time of fracture and, during that period, women's health status may have changed. In the healthy subcohort, women may have developed new comorbidities that might have changed their health status between the year 10 visit (when the healthy subcohort was defined) and fracture occurrence. It is also important to note that, in age-stratified analyses and because of decreased overall survival, the numbers of cases and events in the oldest population were lower than those in the younger age groups; thus, we had decreased power to detect a significant association if it were present.

Because hip fracture risk increases with age, hip fractures are expected to become an even larger public health issue as the population ages; some experts predict that there will be 6.3 million to 8.2 million hip fractures annually worldwide in 2050.^1,36 Overall, older women who experience a fracture have a 2-fold increased mortality in the first year after hip fracture. Women who are younger than 80 years or who are 80 years or older and healthy have the highest risk of dying in the first year compared with women of similar age and health status. If our findings are replicated, they would suggest that research should focus on hip fracture prevention and interventions in these groups that could decrease mortality during that high-risk period. Women who are 65 to 69 years of age continue to have an increased risk of mortality for up to 5 to 10 years; therefore, prevention of hip fractures in these women should be of high priority.