Examining Differences in Recovery Outcomes between Male and Female Hip Fracture Patients: Design and Baseline Results of a Prospective Cohort Study from the Baltimore Hip Studies

Abstract

Background:

Incidence of hip fractures in men is expected to increase, yet little is known about consequences of hip fracture in men compared to women. It is important to investigate differences at time of fracture using the newest technologies and methodology regarding metabolic, physiologic, neuromuscular, functional, and clinical outcomes, with attention to design issues for recruiting frail older adults across numerous settings.

Objectives:

To determine whether at least moderately-sized sex differences exist across several key outcomes after a hip fracture.

Design, Setting, & Participants:

This prospective cohort study (Baltimore Hip Studies 7^th cohort [BHS-7]) was designed to include equal numbers of male and female hip fracture patients to assess sex differences across various outcomes post-hip fracture. Participants were recruited from eight hospitals in the Baltimore metropolitan area within 15 days of admission and were assessed at baseline, 2, 6 and 12 months post-admission.

Measurements:

Assessments included questionnaire, functional performance evaluation, cognitive testing, measures of body composition, and phlebotomy.

Results:

Of 1709 hip fracture patients screened from May 2006 through June 2011, 917 (54%) were eligible and 39% (n=362) provided informed consent. The final analytic sample was 339 (168 men and 171 women). At time of fracture, men were sicker (mean Charlson score= 2.4 vs. 1.6; p<0.001) and had worse cognition (3MS score= 82.3 vs. 86.2; p<0.05), and prior to fracture were less likely to be on bisphosphonates (8% vs. 39%; p<0.001) and less physically active (2426 kilocalories/week vs. 3625; p<0.001).

Conclusions:

This paper provides the study design and methodology for recruiting and assessing hip fracture patients and evidence of baseline and pre-injury sex differences which may affect eventual recovery one year later.

INTRODUCTION

Hip fracture, a very disabling and costly event, is typically thought of as a health problem of older women. However, 25–30% of the 300,000 annual hip fractures in the U.S. occur in men (1, 2). By 2030 the number of hip fractures is projected to decrease by 3.5% in women, but it is expected to increase in men by 51.8% (1); thus, making this a public health concern for older men and women, their families, and the healthcare system.

Because of the historically higher number of hip fractures in women compared to men and a subsequently higher number of women included in studies of hip fracture, much of the knowledge on treatment and outcomes following fracture is specific to women. However, studies examining men who fracture a hip have found that men are younger (3) and have more comorbidities than women at the time of fracture (4–7). Men have approximately twice the risk of mortality compared to women following fracture that can persist for at least 2 years after the event (4, 8–10).

There are biological differences in aging processes in men and women with particular relevance to hip fracture recovery including sex-related differences in muscle tissue, bone density, sex steroids, and insulin-like growth factor (11–13). Many disease processes and clinical events affect men and women differently (14). However, findings are inconsistent regarding sex differences in functional recovery in older patient populations and in hip fracture (15–19). Therefore, this study was designed to include equal numbers of male and female hip fracture patients using a recruitment methodology from acute care settings to determine whether sex differences exist in functional, physiologic, neuromuscular, metabolic, and clinical outcomes over the year after hip fracture. Thus, the overall goal of the proposed study was to determine whether at least moderately-sized (0.3–0.5 standard deviation units) sex differences exist across several key outcomes after a hip fracture.

Our working hypothesis was that men and women will have different trajectories of loss and recovery after hip fracture. More specifically, it was hypothesized that men will experience greater declines in bone mineral density (BMD) and bone strength, muscle mass and strength, and will gain more fat than women who sustain a hip fracture, and men will experience less recovery and will recover more slowly than women in multiple functional domains. Men also will reduce their level of physical activity and be less active post-fracture than women and will be less motivated to engage in rehabilitative strategies to improve their chances of recovering. In addition, it is anticipated that bone turnover, hormone effects, and cytokine regulation following fracture will differ by sex. In evaluating these differences, we will test the hypothesis that at least moderately-sized sex difference in outcomes exists in many if not all areas of functional, physiological, and metabolic change.

This manuscript puts forth the essential study design elements and data collection methods developed by the Baltimore Hip Studies (BHS) to meet the challenges of working with frail older adults within acute care settings and during the post-discharge period, while ensuring equal numbers of males and females, frequency matched (1:1) on calendar time of fracture and hospital. It also compares baseline characteristics of male and female hip fracture patients enrolled in the study.

METHODS

Recruitment from acute care sites

This prospective cohort study (the BHS seventh cohort (BHS-7)) was carried out in eight hospitals in the 25-hospital BHS network. The choice of hospitals was made using the number of hip fracture cases per hospital in 2003, according to data obtained from the Maryland Health Services Cost Review Commission. Hospitals were chosen to: 1) maximize number of male hip fracture patients expected for recruitment per hospital per year, 2) minimize geographic proximity between hospitals to minimize staff travel time and costs, and 3) maximize collection of baseline bone mineral density (BMD) measurements within 15 days of admission.

Recruitment occurred in acute care hospitals since most hip fracture patients receive initial medical care in that setting, and the target sample could be defined through admission logs and gaining access to the acute care patient population allowed study staff to follow patients to numerous post-discharge locations. Finally, the number of Institutional Review Board (IRB) and other administrative reviews was reduced compared to accessing patients from the large number of post-acute care settings.

Support was obtained from the chiefs of orthopedics who requested ‘blanket permission’ from all orthopedic surgeons in their departments to approach their patients. The study was also introduced to nursing and rehabilitation staff on orthopedic floors and to staff in admissions and medical records departments. In-service sessions, grand rounds, and appreciation gifts strengthened cooperation and assured ongoing knowledge of the study while ensuring “buy-in” from hospital staff.

Participant screening and recruitment

This study utilized trained research nurses to conduct recruitment activities in study hospitals. This assured control over standardized screening across hospitals, minimized the likelihood of missing newly admitted patients, and permitted sustained recruitment over several years with minimal burden on hospital staff. Research nurses were able to perform hands-on clinical assessments, function optimally in hospital settings, and communicate with families while having a clinical understanding of the event.

Patient sampling.

Our recruitment strategy ensured an adequate number of men in the sample with a comparison group of women who fractured during the same time period and were treated in the same hospital, thereby minimizing the effect of secular changes in care and hospital practice differences (e.g., care pathways). Since only 25–30% of hip fractures occur in men, recruitment for men was ongoing while recruitment of women was frequency-matched with men within each hospital. After a male enrolled, the study nurse began screening females for that day; all eligible women were approached and consented if eligible, even if it resulted in more than one woman per man. When this happened, the additional female(s) enrolled was used to match the next available male(s) to be enrolled, which may have resulted in a slight excess of eligible females.

Study sample.

The final cohort comprised eligible patients aged 65 or older admitted with diagnosis of hip fracture (ICD-9 codes 820.00–820.9) who consented (or whose proxy provided consent) to participate within 15 days of admission.

Patients were excluded if they had a pathologic fracture (e.g., Paget’s disease or bone metastases), were not community-dwelling at time of fracture, non-English-speaking, bedbound for 6 months prior to fracture, resided > 70 miles from the hospital, weighed > 300 pounds (weight restrictions on BMD scanner table), did not undergo surgery, or had hardware in the contralateral hip leaving no unaffected hip for BMD measurement. The protocol was approved by the IRB at the University of Maryland Baltimore, as well as by each study hospital’s IRB.

Data collection

In-person assessments were conducted at baseline (within 22 days of hospital admission), and 2, 6, and 12 months post-admission. Assessments included blood collection, cognitive testing, and interview- and physical performance-based measurements conducted at person’s place of residence. A complete list of measures and collection times is provided (see Appendix 1). Participants had BMD measurements using dual-energy x-ray absorptiometry (DXA) at one of seven study DXA centers located at or near recruitment hospitals. Transportation was provided and a research staff member accompanied participants in transport and throughout the visit. Monthly telephone calls obtained data on general health, symptoms of infections, and hospitalizations between study visits. A proxy was asked to provide appropriate data for those participants with a score <36 on the Modified Mini-Mental State Examination (3MS) (20) at their most recent visit.

Study measures

Body composition.

BMD of the non-fractured hip was measured using DXA. A whole body DXA scan provided data on muscle and fat mass. Of the seven DXA facilities, three machines were manufactured by Hologic (Waltham, MA, USA) and four by GE Lunar Prodigy (Madison, WI, USA).

The BHS used seven DXA centers in the community, rather than a central DXA site, to acquire the BMD scans. This maximized the ability to obtain the baseline BMD scan within 15 days of the admission for hip fracture at a center location either within the admitting hospital or at a location that was familiar to the participant and closer to the place of residence after discharge, thereby increasing the ability to get follow-up scans conducted on the same machine for each visit.

Standardized methods were used for quality control, certification of DXA operators, and scanning procedures to guarantee the reproducibility of results. Daily quality control (QC) plots were obtained from each site and annual cross-calibration of machines was conducted using a single spine phantom and a variable composition phantom from Bio-Imaging Technologies (Newtown, PA, USA.). Random study scans were selected across sites monthly and reviewed for quality by two expert investigators. To account for any inter-site and machine differences, statistical models include a time-varying indicator to capture the different DXA sites and machines, an approach that has been used previously in the study of BMD changes in patients with osteoporosis (21).

Anthropometric measures collected at baseline included weight, knee height using a sliding caliper to estimate height (22, 23), and calf circumference using a measuring tape.

Physical Function.

Survey-based items and grip strength were measured at all study visits, while performance-based items and actigraphy were obtained only at follow-up visits. Physical activities of daily living (ADL) assessment captured self-reported function of both upper and lower extremity activities. Lower extremity function was assessed using a modified form of the Functional Status Index (24), revised to specifically address functional issues relevant to hip fracture patients. Instrumental ADL were obtained using a modified version of the Older Americans Resources and Services Instrument (25), which asked about performance of seven tasks of daily living during the preceding two weeks. The Yale Physical Activity Scale provided data on five categories of activities performed during a typical week (26). Both hours spent in activities and kilocalories expended in a week were calculated.

Lower extremity performance was measured by the Lower Extremity Gain Scale, a performance-based measure that focuses on clinically relevant aspects of functioning for hip fracture patients (27–29). The Short Physical Performance Battery assessed timed physical activities of balance, gait, and strength (30). Grip strength was measured using the JAMAR Hydraulic Hand Dynamometer (31, 32), and an objective measure of activity was obtained with ActiGraphs (The Computer Science Applications Inc.) over a 48-hour period (33). The pager-size device was placed on the waist of participants at the completion of each follow-up visit and collected by study staff 2 days later.

Cognition.

Cognitive status was assessed using three standard tests. The 3MS (34) was administered first. If a participant scored ≥36, the interviewer attempted to administer the other cognitive measures and the survey (20). The Trail Making Test measured visual tracking, complex and divided attention, mental flexibility, and executive function (35–38). Subjects were allowed 3 minutes to complete each part. After preliminary evaluation of the initial 7 participants, 77% of the participants could not complete the Trail Making Test Part B in 3 minutes, and 10% could not complete the Trail Making Test Part A. Consistent with studies of older persons (39, 40), timing was increased to 5 minutes for each part and progress at the 3-minute mark was noted. The Hooper Visual Organization Test (41) measured ability to organize visual stimuli by mentally rotating parts of objects in space (42–44).

Psychosocial measures.

Depressive symptoms were measured using the Center for Epidemiological Studies Depression scale (45). Social Interaction was self-reported number of social activities engaged in during the prior two weeks using the social activity measure developed by House (46). The Life Orientation Test-Revised (47) assesses individual differences in generalized optimism versus pessimism and measures the participant’s own expectations of their future. Resilience was measured using the 15-item Resilience Scale (48, 49), which assessed the person’s ability to successfully cope with change or misfortune.

Metabolic Measures.

Blood specimens were collected between 7 a.m. and 10:00 a.m. and transported to the university campus no later than 1:00 p.m. for processing. Samples were processed within 6 hours of collection; serum and DNA were stored at −80 degrees Centigrade until assayed.

Demographics and health status.

Demographic and other descriptive information about health, physical status, factors related to surgery, and hospital events were obtained from medical chart abstraction or interview with participants or surrogates. Additional information was collected on pain experience during the past week (50), self-rated health and number of falls and resultant injuries via self-report, prescription medications, and healthcare utilization. Frailty was only measured at follow-up time points using the definition by Fried and colleagues (51) using the gender-specific cutoffs developed within the CHS cohort and sarcopenia was also only measured as follow-up time points.

Vital status and time and cause of death were ascertained from patients’ representatives throughout the first year post-fracture and by National Death Index search up to nine years post fracture for first patient enrolled.

Statistical Analysis

Baseline analyses.

Baseline descriptive statistics (means and standard deviations or counts and percentages) are presented for 339 enrolled participants with significance testing of sex differences (t-tests for continuous measures, chi-square for categorical measures). For all analyses, p values less than 0.05 were considered statistically significant. All statistical tests were 2-sided and analyses presented in this paper were performed using SAS version 9.3.

Longitudinal analyses.

A series of generalized estimating equations (GEEs) will be used for the trajectory and longitudinal sex difference analyses for each of the main outcomes (BMD, body composition, physical activity, neuromuscular function, functional performance, depression, cognition, and metabolic parameters). Using GEE, we are able to relate outcomes, at specific post-fracture time points, to both fixed factors (e.g. sex) and time-dependent covariates (e.g. use of bone-strengthening medication).

The GEE models will assess the effect of time on the various trajectories of outcomes during the post-fracture recovery period using the natural logarithm of the outcome measure and time measured in months. The trajectory described by the equation is an exponential curve which approaches a specific level asymptotically with increasing time. The time to reach the point on the curve where no appreciable change occurs thereafter is thought to be a meaningful indication of the time needed to recover from the fracture, i.e. the recovery time. The addition of sex and a sex by time interaction term to the model provides an assessment of the difference in recovery times between male and female hip fracture cases.

Time-specific sex differences in outcomes will be addressed using a longitudinal model where the fitted coefficients will provide estimates of male vs. female differences in the outcome at any of the specific time points. Covariates, such age, comorbidity, baseline measures of BMD, and other factors when seen as potential confounders, will be added to the model to enable the determination of adjusted differences.

Death certificate data will be used to determine male vs. female relative risk for cause-specific mortality and extend the analysis to sex differences in mortality. The effect modification influence of frailty, comorbidity and functional impairment, on the sex differences in metabolic, physiologic and functional changes, and mortality that follow hip fracture will be investigated by the inclusion of interaction terms in the models discussed above.

Sample Size

The target sample size was 404 (202 men and 202 women). We estimated a conservative 50% attrition rate (30% one-year mortality, 20% loss to follow-up) over 12 months providing 202 subjects at the end of the study (101 men and 101 women). These numbers would enable the detection of sex differences with effect sizes within the small to medium range (0.2 to 0.5 standard deviation (SD)) as defined by Cohen (52), with 80% power and 0.05 type I two-tailed error rate using within-subject correlations mainly in the range from 0.5 to 0.7 seen in a prior BHS cohort (53).

Interim power calculations were performed when study performance reports showed a lower projected loss to follow-up (47%) and higher than expected within-subject correlations (0.7–0.9) for key variables in each category of measurement. It was determined that there was excellent power (>85%) to detect the proposed sex differences across all categories of measures with a smaller sample size. In consultation with the Data and Safety Monitoring Board, recruitment ended on June 30, 2011 with a total sample of 362 enrolled.

RESULTS

During the recruitment period May 2006 – June 2011, 1709 hip fracture patients were screened, and 917 (54%) were eligible (405 males, 512 females) (Figure 1). A total of 180 men and 182 women consented to participate in the study. Twenty-three participants were withdrawn [five participants failed to provide data at the baseline and 2-month follow-up visit and another 18 were removed from the analysis sample as a result of an IRB-requested post procedure audit (6 participants were found to be ineligible and 12 participants were ineligible secondary to failures in the informed consent process)]. The final analytic sample was 339 (168 men, 171 women).

Figure 1.

Participant Recruitment Flowchart (May 2006-June 2011)

^* Only the first exclusion criterion is shown for each person

^† Participants withdrawn either due to predetermined criteria for no data provided at baseline and 2-month follow-up visits or as a result of an IRB-requested post procedure audit

Selected baseline demographic and medical characteristics are presented in Table 1 by sex. Proxy responses were provided for 18 men and 13 women at baseline. Men and women did not differ in age, race, education, current smoking, or depressive symptoms. Men were more often married (53% vs. 28%, p < 0.001), less often lived alone (25% vs. 49%, p < 0.001), and had fewer social contact activities (9.3 (SD=8.7) vs.11.5 (SD=9.3), p = 0.03). Men consumed more alcoholic drinks per week (mean number of drinks 1.3 (SD=1.5) vs. 0.9 (SD=1.2), p = 0.02). Women had a higher 3MS score (mean 86.2 (SD=16.4) vs. 82.3 (SD=16.4), p = 0.03).

Table 1.

Sex Differences in Demographic and Medical Characteristics of Hip Fracture Patients at Baseline; BHS-7 2006-2011

	Women	Men
Characteristic	n = 171	n = 168	P value
Self-Report
Age (years) mean (SD)	81.4 (7.9)	80.4 (7.8)	0.20
Non-white (%)	12 (7.3)	16 (9.8)	0.42
Education (years) mean (SD)	13.0 (3.0)	13.2 (3.8)	0.74
Currently married (%)	47 (28.5)	87 (53.4)	<0.001
Lived alone (%)	81 (49.4)	39 (24.7)	<0.001
Social contact (activities) mean (SD)	11.5 (9.3)	9.3 (8.7)	0.03
Current smoker (%)	11 (6.8)	9 (5.8)	0.72
Alcohol consumption (drinks/week) mean (SD)	0.9 (1.2)	1.3 (1.5)	0.02
Depression (CES-D) mean (SD)	17.7 (11.4)	17.4 (10.4)	0.82
3MS Score mean (SD)	86.2 (16.4)	82.3 (16.4)	0.03
Medical Chart
Height (m)	1.6 (.08)	1.76 (.08)	<0.001
Weight (kg)	63.65 (14.21)	79.04 (14.0)	<0.001
Body mass index (kg/m2)	24.97 (5.58)	25.56 (4.53)	<0.29
Comorbidities mean (SD)	4.0 (2.4)	4.5 (2.4)	0.09
Charlson Comorbidity Index mean (SD)	1.6 (1.6)	2.5 (1.9)	<0.001
Dementia/Alzheimer’s diagnosis (%)	18 (10.5)	28 (16.8)	0.09
Diagnosis of osteoporosis (%)	67 (39.2)	25 (15.0)	<0.001
History of fractures (%)	36 (21.1)	27 (16.2)	0.25
Length of stay mean (SD)	5.2 (2.8)	5.5 (2.4)	0.41
Fracture type			0.20
Femoral neck (%)	93 (54.4)	74 (44.6)
Intertrochanteric (%)	62 (36.3)	73 (44.0)
Other (%)	16 (9.4)	19 (11.5)
Surgical type			0.17
Internal fixation (%)	86 (50.3)	101 (60.5)
Arthroplasty (%)	79 (46.2)	61 (36.5)
Other (%)	6 (3.5)	5 (3.0)
Self-Report Medication Use–Prior to Fracture
Bisphosphonate (prior 6 months) (%)	64 (39)	13 (8)	<0.001
Calcium supplement (%)	112 (69)	43 (27)	<0.001
Vitamin D supplement (%)	126 (77)	97 (61)	0.003

3MS = Modified Mini-Mental State Examination; CES-D = Center for Epidemiological Studies Depression; SD = standard deviation

While total number of comorbidities did not differ by sex, men had a higher Charlson comorbidity score (2.5 (SD=1.9) vs.1.6 (SD=1.6), p < 0.001). Women more often had a diagnosis of osteoporosis in the medical chart (39% vs. 15%, p < 0.001), and used bisphosphonates in prior 6 months (39% vs. 8%, p < 0.001). Sixty-nine percent of women reported using calcium and 77% used vitamin D supplements at the time of fracture compared to 27% and 61% of males, respectively.

Sex differences in pre-fracture functional status are presented in Table 2. There were no differences in lower and upper extremity ADLs; however, men reported more need for assistance with overall Instrumental ADLs (mean 2.2 (SD=1.5) vs. 1.8 (SD=1.5) for women, p = 0.01). Additionally, men were less physically active prior to fracture as measured by kilocalories expended in a week on the Yale Physical Activity Scale (mean 2426 (SD=2366) kilocalories vs. 3625 (SD=2600), p < 0.001) and number of hours/week spent doing activities (mean 11.6 (SD=10.9) hours vs. 19.7 (SD=13.3), p < 0.001). Men had significantly higher grip strength (mean 29.3 (SD=9.7) kg vs. 18.6 (SD=6.7) kg, p < 0.001).

Table 2.

Sex Differences in Baseline Functional Status and Activity Level at Baseline; BHS-7 2006-2011

	Women	Men
Variable	Mean (SD), n = 164	Mean (SD), n = 162	P value
Lower extremity PADL summary score	2.7 (2.6)	2.4 (2.6)	0.27
Upper extremity PADL summary score	0.1 (0.6)	0.1 (0.5)	0.70
IADL summary score	1.8 (1.5)	2.2 (1.5)	0.01
YPAS (kilocalories/week)	3625 (2600)	2426 (2366)	<0.001
YPAS (hours/week)	19.7 (13.3)	11.6 (10.9)	<0.001
Grip strength (kg)	18.6 (6.7)	29.3 (9.7)	<0.001

IADL = instrumental activities of daily living; PADL = physical activities of daily living; SD = standard deviation; YPAS = Yale Physical Activity Scale

Measures of body composition are shown in Table 3. Women had a significantly lower total hip BMD irrespective of DXA machine type, resulting in a significantly lower sex-specific T-score at the total hip compared to men (−2.23 (SD=1.06) vs. −1.94 (SD=0.97), p = 0.03); interestingly, there was no significant difference in the sex-specific T-score at the femoral neck. Women had significantly lower lean body mass and higher fat mass, although the latter was not statistically significant in those with DXA measurements performed on Lunar machines.

Table 3.

Sex Differences in Body Composition at Baseline (Within 15 Days of Admission) by Machine Type; BHS-7 2006-2011

	Women	Men
	Mean (SD), n = 130	Mean (SD), n = 123	P value
Femoral neck BMD, T-score	−2.38 (0.91)	−2.32 (0.95)	0.58
Total hip BMD, T-score	−2.23 (1.06)	−1.94 (0.97)	0.03
Lunar	n = 101	n = 97
Total hip BMD (gm/cm2)	0.72 (0.13)	0.82 (0.14)	<0.001
Total body fat (g) *	23188.15 (9557.88)	21895.52 (9193.43)	0.34
Appendicular lean mass (g) †	15387.7 (3140.5)	21860.9 (4058.8)	<0.001
Hologic	n = 29	n = 26
Total hip BMD (gm/cm2)	0.69 (0.12)	0.78 (0.14)	0.019
Total body fat (g)*	24129.73 (11139.84)	17764.14 (4985.48)	0.008
Appendicular lean mass (g) †	13331.2 (3159.7)	20020.5 (4016.1)	<0.001

BMD = body mass index; SD = standard deviation

^*Due to edema on the fractured side at baseline, fat mass of the fractured leg was not included and fat mass of the non-fractured leg was doubled for the measure of total body fat mass

^†Due to edema on the fractured side at baseline, lean mass of the fractured leg was not included and lean mass of the non-fractured leg was doubled for the measure of total body lean mass and appendicular lean mass

DISCUSSION

At the time of hip fracture, men and women were different in several ways that could be important predictors of eventual recovery and choice of post-fracture treatment over the first 12 months. Although men were on average one year younger than women, this difference was not significantly different in this sample as had been reported by others (3, 54–57); however, men did suffer from more serious comorbidities than women (54–56, 58–60). Managing concomitant comorbidities could impact not only the potential for recovery, but also the level of recovery expected. It also could influence the type of medical management and timing of potential function-promoting therapies.

A greater percentage of men compared to women were married and were not living alone. These findings have been observed by others (16, 61) and may be explained by higher percentages of widowhood and differences in life-expectancy in women (58). These differences may have profound impact on discharge location and the ability to return to home and the recovery trajectory. Repeatedly, it has been reported that the presence of a spouse or familial support improves hip fracture recovery (58), and the impact may vary by sex. Arinzon et al., found that family support had a significant impact on ambulation and transfer ability among men post fracture but not among women (60). Thus, men may be more likely to return home and have more support for activities of daily living.

At the time of fracture, men have significantly lower 3MS scores (86 vs. 82). It is well known that dementia, delirium, and overall cognitive status impact recovery from hip fracture (62). While other hip fracture studies have shown lower cognition and more delirium in men after hip fracture (63), this finding is not consistent (16, 64). In community-dwelling elders examined on the 3MS, among those with at least a high school education, men do score lower than women (65). In addition, the men in this sample had slightly more dementia which also would contribute to the lower 3MS scores.

Many older adults at high risk of hip fracture are not being diagnosed and treated for osteoporosis (66, 67), particularly among men which is consistent with this study. History of prior fractures was not different by sex, yet only 15% of men compared to 40% of women were diagnosed with osteoporosis, and fewer than 10% of men had used bisphosphonates in prior 6 months compared to almost 40% of women. Similar differentials were also seen for use of calcium but not vitamin D supplements. While the under-treatment of osteoporosis in both men and women is not consistent with current practice guidelines, men may have increased risk of losing more BMD and re-fracturing as they are less likely to receive treatment after a hip fracture has occurred.

While both sexes are relatively sedentary prior to the fracture, men are significantly more sedentary, spending 8 fewer hours being active and expending approximately 1100 fewer kilocalories per week. While older women in the community have been found to be less physically active in terms of aerobic exercise compared to men (68), men are less physically active after a hip fracture. In the early post-hip-fracture period, participation in physical activity is predictive of functional recovery (69); thus, men may be at greater risk for poor recovery. Furthermore, individuals who remain sedentary after hip fracture are at increased risk for a second hip fracture and further functional decline (70).

Women had lower absolute BMD and sex-specific T-scores than men at the total hip; however, sex-specific T-scores did not differ at the femoral neck. The proportion of men and women with osteoporosis, defined as sex-specific BMD T-score at the femoral neck, was less than 50% in both sexes. This has been reported previously by Wainwright in women from the Study of Osteoporotic Fractures (71).

Body composition also differed by sex with men having significantly higher lean mass and lower fat mass at the time of fracture. These factors may play a key role in aspects of recovery. While lean mass has not been shown to affect functional performance in women post fracture (7, 72), it did predict functional outcomes in men (73, 74). Given the losses of lean muscle mass and increases in fat mass seen in women post fracture (75), it is unclear if the men will experience a similar magnitude of change in body composition and how it will impact recovery.

The design of the BHS infrastructure and methodology employed in this study provided an effective system for identifying older adults from acute care settings and recruiting from post discharge settings within 15 days of hospitalization. While early identification of hip fracture patients facilitated building a rapport and trust with them and their families as partners in this research study, screening for study eligibility was challenging given the short length of stay (3–4 days). A partial privacy waiver for recruitment facilitated the timely identification of eligible hip fracture admissions from medical records and eliminated the need for individual Health Insurance Portability and Accountability Act authorization prior to screening for eligibility. This reduced staffing costs and patient burden since patients who were ineligible through a chart review were not approached about the study.

This study provides critically important descriptive data on how men and women differ at the time of hip fracture and provides some insight into areas for potential sex-specific interventions. Most importantly, it allows for systematic examination of the multiplicity of factors that are most relevant to men who have hip fractures to inform future research in this understudied group. Although there may be limits to generalizability based on conducting this study in a single metropolitan area, it gains generalizability through recruitment in eight distinct hospitals. We cannot generalize to the group of patients we chose to exclude (i.e., nursing home residents; those with hardware in the contralateral hip or pathological fracture; or those who were bedbound prior to the fracture).

In conclusion, findings from this study to date provide evidence that older male and female hip fracture patients can be recruited during the early acute care period into research studies, and using a methodology that over-samples males to maximize statistical power for detecting between sex differences. Results from the longitudinal study should provide important information about the unique consequences of hip fracture in men and provide a greater understanding of the differences in recovery trajectories between men and women over the year after hip fracture. This information, in turn, should inform treatment decisions and enable design and testing of novel interventions to improve outcomes for men and women who fracture a hip.