Long-term Sex Differences in All-cause and Infection-Specific Mortality Post Hip Fracture

Abstract

Background:

Mortality rates among men are double that of women in the first two years after hip fracture and may be related to more infections. Research has only examined differences in short-term mortality after hip fracture. Thus, the objective was to determine if long-term all-cause mortality and infection-specific mortality rates are higher in men compared to women.

Methods:

Data come from a prospective cohort study (Baltimore Hip Studies 7th [BHS-7]) with up to 10.2 years of follow-up (2006–2018). The participants were selected from eight acute care hospitals in the 25-hospital BHS network. Enrolled women were frequency-matched (1:1) to men on timing of admission for hip fracture that yielded an analytic sample size of 300 participants (155 women, 145 men). Associations between sex and mortality were analyzed using Cox proportional hazard models and cause-specific Cox models adjusted for age, cognition, body mass index, pre-fracture lower extremity activities of daily living limitation, depressive symptoms, and comorbidity.

Results:

Participants had a mean age of 80 years, 48% (n=145) were men and the median follow-up was 4.9 (Interquartile Range=2.3–8.7) years. Over the follow-up period after hospital admission for hip fracture, 237 (79.0%) participants died of all causes (132 men and 105 women) and 38 (12.7%) died of infection-specific causes (25 men and 13 women). Men had significantly higher rates of all-cause mortality [Hazard Ratio (HR)=2.31(95% confidence interval (CI) 2.02–2.59)] and infection-specific mortality (HR=4.43, CI 2.07–9.51) compared to women.

Conclusions:

Men had a two-fold higher rate of all-cause mortality and four-fold higher rate of infection-specific mortality compared to women over a follow-up period of up to 10.2 years. Findings suggest that interventions to prevent and treat infections, tailored by sex, may be needed to narrow significant differences in long-term mortality rates between men and women after hip fracture.

INTRODUCTION

Hip fracture is a disabling medical event that represents the most significant consequence of osteoporosis and often results in death among older adults.¹ Since three out of four hip fractures occur in women, women comprise the majority of participants in extant research, which is therefore not generalizable to men.² Nonetheless, hip fracture rates among men are projected to increase by 51% from 2010 to 2030, while the number of hip fractures among women is expected to decrease by 3.5%.³ Men with hip fracture tend to be younger than women at the time of fracture, yet they are frailer and have more comorbid conditions and die at a higher rate in the first two years after fracture compared to their female counterparts.⁴

Even after adjustment for age, dementia, and other comorbidities that might impact mortality post-hip fracture, men are twice as likely to die within two years post-hip fracture compared to women.^4–6 There are unclear findings among the few studies that have assessed sex differences in all-cause mortality beyond two years post-hip fracture. Men had a higher risk of all-cause mortality in a 3-year and 5-year follow-up study but it was not always statistically significant.^7,8 Scant data are available regarding the changing relationship of sex with all-cause mortality over longer periods of follow-up after hip fracture.

Infection is one of the top causes of death over the year post hip fracture, where men are 20 times and women are 10 times more likely to die of infectious causes in the year after a hip fracture compared to age- and sex-matched counterparts who did not sustain a hip fracture.⁴ Additionally, infection is a preventable and treatable condition, and it is important to study its consequences. However, past studies have not examined long-term (>2 years) sex differences in infection-specific mortality within hip a fracture population.

The objective of this study was to determine whether male sex is associated with greater all-cause and infection-specific mortality after hip fracture over a 10.2-year follow-up period. We hypothesize that the cumulative incidence of all-cause mortality and infection-specific mortality are consistently greater for men compared to women over a follow-up period up to 10.2 years post-hip fracture.

METHODS

Study Design

Data come from the Baltimore Hip Studies 7th (BHS-7) cohort. Participants (≥ 65 years old) were enrolled from one of eight acute care hospitals in the 25-hospital BHS network and were admitted with a hip fracture diagnosis (ICD-9 codes 820.00–820.9). The enrollment period was from May 2006 through June 2011, and in-person assessments were done at baseline (within 22 days of hospital admission), and 2, 6, and 12 months post admission. All participants were consented (or proxy provided consent) within 15 days of admission. BHS-7 was designed to assess sex differences in the sequelae of hip fracture, and female patients were frequency-matched (1:1) to males on the timing of admission within each hospital to ensure equal numbers of men and women. Further details regarding the study design and methodology are available elsewhere.² The BHS-7 cohort enrolled 339 participants (168 men and 171 women) and was restricted to an analytic cohort of 300 participants (145 men, 155 women) with complete data on salient variables measured at baseline.

Mortality Outcomes

All-cause mortality and infection-specific mortality were the study outcomes, and death data were obtained from 2006–2018 for a maximum possible follow-up period of 10.2 years. All deaths and corresponding causes were verified through the National Death Index (NDI) by linking participant information such as first and last name, date of birth, sex, and social security number with the NDI records. If death dates were identified from NDI, then the participant was coded as deceased. Deaths were defined as infection-specific if one or more contributing conditions or the underlying causes of death as indicated by clinicians in the death certificate were any of the following ICD-10-CM codes: A04.7, A04.71–2, A41.9, A41.90, A49.0, A49.9, J15.20, J15.211–2, J15.29, J18.0–2, J18.8–9, J44.0, or N39.0.¹¹

Measures

The primary predictor variable of interest was biological sex defined as male or female. Potential confounders were selected a priori based on prior research suggesting their association with mortality difference between men and women who experienced a hip fracture. The Modified Mini-Mental State Examination (3MS) was used to assess cognitive function (score range 0–100).¹² Comorbid conditions were evaluated using the Charlson Comorbidity Index (CCI), a weighted index of comorbidities that takes into consideration the number and seriousness of chronic conditions.¹³ Higher CCI scores indicate greater comorbidity and are associated with greater mortality.¹³ In the current study, CCI score ranged 0–8 in which only moderate or severe (not mild) liver disease was recorded. The 20-item Center for Epidemiological Studies Depression scale (CES-D) measured depressive symptoms. Questions ask about symptomology occurring in the last week, and responses are scored on a Likert scale that yields a summary score ranging from 0–60.¹⁴ Lower-extremity Physical Activities of Daily Living (LPADL) function was assessed in the week prior to fracture using a modified form of the Functional Status Index, which measures the ability to perform an activity and level of difficulty.¹⁵ Body Mass Index (BMI) was measured as kilograms per meters squared (kg/m²).

Statistical Analysis

Separated by sex, percentages and frequencies were estimated for categorical variables, and mean and standard deviations were calculated for continuous variables. Unadjusted associations between sex and all-cause mortality were evaluated using Kaplan-Meier survival curves and log-rank tests.¹⁶ Unadjusted associations between sex and infection-specific mortality were evaluated using cumulative incidence for competing risk method and Gray’s test.¹⁷

Unadjusted and adjusted associations between sex and all-cause mortality were estimated using Cox proportional hazards models. Unadjusted and adjusted associations of sex and infection-specific mortality were assessed using cause-specific Cox models to account for competing causes of death.¹⁸ For both mortality outcomes, adjusted models included age, cognition, comorbidity, depressive symptoms, BMI, and pre-fracture LPADL limitations. Secondary analyses assessed whether the relationship between sex and mortality changes over time during hip fracture recovery by including an interaction term between sex and time (i.e., sex*time). Model goodness-of-fit was assessed using Akaike information criterion (AIC).¹⁹ Lastly, sensitivity analyses were conducted to assess differences in participant characteristics between those with and without missing data. Statistical significance was defined as p < 0.05 or 95% confidence intervals excluding the null, and all analyses were performed using SAS Version 9.4 (SAS Institute, Inc., Cary, NC).²⁰

RESULTS

Sample Characteristics and Mortality Rates

The final analytic sample included 145 (48.3%) men and 155 (51.6%) women (Table 1; Supplementary Figure S1). After a maximum 10.2 years of follow-up (median=4.9 years; IQR, 2.3–8.7), 90.9% (132/145) of men died, and 67.7% (105/155) of women had died. There were 38 infection-specific deaths with 25 (17.2%) men and 13 (8.3%) women dying due to infection. Pneumonia was the most common cause of infection-specific death (Supplementary Table S1). Men had significantly higher cumulative incidence of both all-cause death (p-value < 0.001) (Figure 1A) and infection-specific death (p-value= 0.018) (Figure 1B).

Table 1:

Baseline characteristics of enrolled participants and death by the end of ten years for men and women

Measures	Total (n=300)	Male (n=145)	Female (n=155)	Missing Frequency No. (%)
Death (by the end of 10 y), No. (%)
Infection related death	38 (12.6)	25 (17.2)	13 (8.3)
Non-infection related death	199 (66.3)	107 (73.7)	92 (59.3)
Alive	63 (21.0)	13 (8.9)	50 (32.2)
Survival time from the date of admission, median (IQR), y	4.9 (6.4)	3.6 (5.1)	6.3 (6.6)
Age, mean (SD), y	80.7 (7.7)	80.3 (7.7)	81.1 (7.7.)
3MS, mean (SD)a	85.7 (13.4)	83.7 (13.5)	87.6 (13.1)	7 (2.3)
CCI, mean (SD)b	2.0 (1.7)	2.4 (1.8)	1.6 (1.5)	1 (0.3)
CES-D, mean (SD)c	17.6 (10.9)	17.6 (10.3)	17.6 (11.4)	36 (12)
BMI (kg/m2), mean (SD)d	25.2 (5.1)	25.6 (4.5)	24.8 (5.6)	16 (5.3)
Pre-fracture LPADL, mean (SD)e	2.4 (2.4)	2.2 (2.3)	2.6 (2.5)	13 (4.3)

^aModified Mini-Mental Status ranges from 0 to 100. Higher score indicates better cognition.

^bCharlson Comorbidity Index ranges from 0 (no comorbidity) to 8.

^cCenter for Epidemiological Studies Depression scale ranges from 0 to 60. Higher score indicates greater depressive symptoms.

^dBody Mass Index, 21–25 kg/m² is normal range per World Health Organization.

^eLower Extremity Physical Activities of Daily Living ranges from 0–12 with higher score indicating more disability

Figure 1:

Figures showing the association of all-cause mortality and infection-specific mortality by sex over 10 years. Figure A: Cumulative incidence of all-cause mortality (failure probability = Survival-1) post hip fracture by sex. Figure B: Cumulative incidence of infection-specific mortality post hip fracture by sex.

Sex-Differences in Post-Fracture Mortality

After adjustment for covariates, men had a 2.31 times higher all-cause mortality rate (95% confidence interval [CI]= 2.02–2.59) and 4.43 times higher infection-specific mortality rate (95% CI=2.07–9.51) (Table 2). The interaction term between sex and time (years) was not significant for all-cause (p-value=0.84) and infection-specific (p-value=0.31) mortality, providing little evidence that the association between sex and mortality changed over time. Men also had higher mortality rates than women from non-infectious causes (P<0.01), but the hazard ratios were closer to the null than those for infectious causes. A sensitivity analysis treating comorbidity as an unweighted count showed similar results.

Table 2:

Unadjusted and adjusted association of sex with all-cause mortality and infection-specific mortality

	Model 1a		Model 2b		Model 3b
		P value		P value		P value
All-cause mortality
Male (HR, 95% CI)	1.06 (1.04–1.07)	<.001	2.31 (2.02–2.59)	<.001	2.21 (2.71–1.71)	<.001
Male*time (ratio of HR, 95% CI)					1.01 (0.91–1.10)	0.84
Infection-specific mortality
Male (HR, 95% CI)	2.20 (1.13–4.27)	0.02	4.43 (2.07–9.51)	<.001	7.09 (2.07–24.21)	<.001
Male*time (ratio of HR, 95% CI)					1.15 (0.87–1.53)	0.31
Non-infection related mortality
Male (HR, 95% CI)	1.59 (1.20–2.10)	<.01	2.11 (1.55–2.88)	<.001	1.72 (0.99–3.01)	0.054
Male*time (ratio of HR, 95% CI)					0.95 (0.85–1.06)	0.38

^aUnadjusted

^bAdjusted for age, Modified Mini-Mental Status (3MS), Charlson Comorbidity Index (CCI), Center for Epidemiological Studies Depression scale (CES-D), Body Mass Index (BMI), and Lower Extremity Physical Activities of Daily Living (LPADL)

DISCUSSION

Consistent with prior research, results from this study demonstrated that 79% of participants died by 10.2 years post hip fracture.²¹ The hazard of all-cause deaths and non-infection-specific deaths were two-fold higher in men whereas infection-specific deaths were four-fold higher compared to women over ten years of follow-up with negligible evidence of a change in the association over time after hip fracture. Collectively, these results indicate that older men have greater long-term post-fracture mortality rates compared to women that (in part) may be explained by a greater likelihood of infection-specific causes.

Men were twice as likely to die of all-causes and non-infection-specific causes compared to women even after controlling for known risk factors for mortality in hip-fracture patients such as age, pre-fracture activities of daily living limitations, comorbidity and cognition. Consistent with prior findings, men were at a survival disadvantage, despite being younger than women and having higher pre-fracture functional ability.⁴ More baseline comorbidity among men compared to women may explain greater all-cause mortality in men because comorbid conditions prior to fracture have been consistently demonstrated to predict post-fracture mortality.^22,23 Lower cognitive function at hospital admission among men than women could also explain their higher all-cause mortality rate since it has been shown that impaired cognition is associated with all-cause mortality at 6-months after hip fracture.⁶ However, both comorbidity and cognitive function were adjusted for in the models, so they are not likely to explain men’s increased mortality. While more comorbidity and poor cognition are associated with greater mortality, there may be specific re-occurring conditions like infections that increase risk of death and develop more frequently among men than women after hip fracture.

Men had a four times higher rate of infection-specific death compared to women. A higher rate of infection-specific mortality in men compared to women may be explained by sex-based differences in the immune system.²⁴ Immunosenescence occurs with normal aging; however, research has shown a greater male-specific decline of B cells than in women after 65 years of age.²⁴ B cells are a component of adaptive immunity and protect against infection by producing antibodies that are explicitly targeted to individual pathogens. Thus, a male-specific decline of B cells could make them more vulnerable to adverse outcomes associated with infectious disease. Alternatively, women may have greater protection against acute infection as studies imply the majority of the genes responsible for immune cell division after activation of cytokines are on the X-chromosome.^25,26 Additionally, other prognostic factors that are associated with infectious diseases may be related to sex differences in long-term mortality rates after hip fracture.

Strengths and Limitations

First, there were missing data on some baseline variables, which impacted inclusion into the analytic sample and hence may reduce generalizability to a general population of patients with hip fracture. However, descriptive data showed that participants with missing data did not significantly differ from those with complete data, which mitigates concerns about generalizability. Second, the BHS-7 sample was predominantly white and participants were all from Baltimore-area hospitals; hence, findings may not be generalizable to other populations. Third, the cause of death data obtained from death certificates in the National Death Index database provide little detail about the processes leading to death and other details relating to fatal illness that could be obtained from other sources, such as hospital records. Misclassification of causes of death on the death certificates is possible when autopsy is not performed and even then, there can be questions about underlying causes that are not detected. However, there is literature showing that death certificates are reasonably accurate in determining cause of death.²⁷ Despite these shortcomings, NDI-based mortality data is one of the most used sources in the surveillance of long-term mortality. Additionally, this study could not determine whether the higher risk of long-term mortality in men compared to women with hip fracture could explicitly be attributed to hip fracture, or if it is because, on average, men have a shorter lifespan than women as there were no matched comparators without hip fracture. Although there was no external comparison group, national data from a representative population of older adults indicates that men only have 1.7 times higher risk of death than women over 13 years.²⁸ By contrast, older men in this study had over 2.3 times higher rate of mortality over 10 years following the occurrence of a hip fracture compared to women which is a 35% greater than expected increase in all-cause mortality. This is the first study to evaluate sex differences in all-cause and infection-specific mortality beyond 2 years of follow-up after hip fracture. Moreover, the BHS-7 cohort study was specifically powered to examine sex difference in the sequelae of hip fracture, which has been a limitation of past studies among these patients.

CONCLUSION

Male sex was associated with a higher risk of all-cause mortality and infection-specific mortality throughout ten years of follow-up after hospital admission for hip fracture. More specifically, men had a four-fold higher rate of infection-specific mortality compared to women, and a two-fold higher rate of all-cause and non-infection-specific mortality. Findings imply that interventions during hip fracture recovery to prevent and treat infections, tailored by sex, may be needed to narrow significant differences in long-term mortality rates between men and women. Pneumonia was the leading cause of infection-specific deaths after a hip fracture and past evidence shows that men have more severe symptoms and are also more likely to die of pneumonia, especially in older age.²⁹ Future studies should explore sex difference regarding pathogens causing infection among older adults post hip fracture so that tailored screening and treatment regimens can be developed. Since typical signs of infection, like fever, may be blunted in older adults, clinicians should be vigilant of other signs like unexplained change in physical abilities or cognitive status compared to baseline.³⁰

Supplementary Material

supinfo

Supplementary Figure S1: Flow diagram of study screening, enrollment, and complete case analysis sample, stratified by sex.

Key Messages:

• Previously unknown if sex relates to all-cause mortality and infection-specific mortality beyond two-years post hip fracture.

• Men had a two-fold higher rate of all-cause mortality and four-fold higher rate of infection-specific mortality compared to women over a follow-up period of up to 10.2 years post hip fracture.

• Interventions to prevent/treat infection, tailored by sex, may be needed to narrow significant differences in long-term mortality rates between men and women.

Why Does This Matter?

Hip fracture is a traumatic event, and the number of hip fractures among men is expected to increase over the next 30 years with potential differences in long-term trajectories of recovery compared to women. This study showed that men die at twice the rate that women do from all causes and four times the rate of women due to infectious causes, and this sex-based disparity in mortality persists for at least 10.2 years post hip fracture. Therefore, research is needed to further ascertain potential mechanisms that explain these sex difference and evaluate possible interventions that can mitigate these disparities in mortality.