Mortality after osteoporotic hip fracture: incidence, trends, and associated factors

Background It is known that mortality after hip fracture increases compared to the general population; the trend in mortality is a controversial issue. The objective of this study is to examine incidence, trends, and factors associated with mortality in patients with osteoporotic hip fractures. Methods This is a retrospective cohort study that uses the Registry for Hospital Discharges of the National Health System of our hospital. Patients older than 45 having an osteoporotic hip fracture between 1999 and 2015 were identified. Demographic data and comorbidities were obtained. A survival analysis was performed (Cox regression and Kaplan-Meier). Incidence rate, standardized death rate (SDR), trend (Poisson regression), and risk (hazard ratio) were calculated. Results During 1999–2015, in our hospital, there were a total of 3992 patients admitted due to osteoporotic hip fracture. Out of these 3992 patients, 3109 patients (77.9%) were women with an average age of 84.47 years (SD 8.45) and 803 (22.1%) were men with an average age of 81.64 years (SD 10.08). The cumulative incidence of mortality was 69.38%. The cumulative mortality rate for 12 months was 33%. The annual mortality was 144.9/1000 patients/year. The 1-year mortality rate increased significantly by 2% per year (IRR 1.020, CI95% 1.008–1.033). The median overall survival was 886 days (CI95% 836–951). The probability of mortality density for a period of 10 years following a hip fracture was 16% for women and 25% for men (first 90 days). The SDR was 8.3 (CI95% 7.98–8.59). Variables that showed statistically significant association with mortality were aged over 75, masculine, institutionalization, mild to severe liver disease, chronic kidney disease, COPD, dementia, heart failure, diabetes, the Charlson Index > 2 , presence of vision disorders and hearing impairment, incontinence, and Downton scale. Conclusions For the last 17 years, an increase of mortality for patients with hip fracture and a higher mortality rate in men than in women were observed. Institutionalization combined with comorbidities is associated with a higher mortality. Electronic supplementary material The online version of this article (10.1186/s13018-019-1226-6) contains supplementary material, which is available to authorized users.


Introduction
Osteoporotic hip fracture is a major health problem due to increased mortality, morbidity, and functional impact in these patients (only 30-40% of these patients recover their previous functional status) [1,2], as well as the economic cost for the National Health System. In addition, it is expected to increase as life expectancy increases.
Increased mortality after hip fracture has been widely reported. The cumulative mortality after 1 year of a hip fracture occurrence ranges between 20 and 40% [3][4][5][6][7] with higher mortality rates in men than in women [8][9][10]. After having a hip fracture, the mortality risk from any other cause increases between five and eight times, and although it decreases over the first years, excess mortality stays higher than mortality for the general population for a period of at least 10 years [10,11].
Literature describes different factors associated with mortality, among them were age, masculine gender, dementia, heart pathology, institutionalization, comorbidity, and change of residence or rehabilitation programs carried out after hip fracture [12][13][14].
One aspect of mortality associated with hip fracture, and seldom researched, is the trend of mortality rates over a long period of time.
The purpose of this study is to analyze mortality, as well as factors associated and trends in mortality over time in patients with osteoporotic hip fracture in an urban municipality of Spain.

Material and methods
Retrospective observational cohort study for the area covered by the Hospital Universitario Fundación Alcorcon from January 1, 1999, until December 31, 2015.

Introduction
Alcorcon is in the metropolitan area of Madrid in Spain. It is a clear example of a dormitory town since half of its population works in Madrid. Its population pyramid is typical of an aging population. The Hospital of Alcorcon is the only public hospital servicing this district and all the hip fractures occurring in the area. Until 2011, the sphere of influence of the Hospital included 18 municipalities (17 were rural and 1 urban), among them was Alcorcon city, with a total population of 273,703 inhabitants. However, from 2012, the hospital only attended one municipality: Alcorcon city with 167,136 inhabitants.

Data source and definitions
Records on hip fractures (HFx) were obtained from the admissions' database of this hospital. All hospitals in Spain use this database which collects the minimum basic data set (MBDS) of each patient admitted to any Spanish hospital: reasons for admission, procedures, and diagnosis codified in accordance with International Classification of Diseases 9 (ICD-9). We ascertained all hip fracture cases occurred in persons 45 years old or older from ICD-9 codes 820.0 to 820.9 in the study years. All cases that occurred to subjects aged 45 years or older from ICD-9 820.0 to 820.9 codes during that period of study were identified, including both first and second instances of fracture. Thanks to a recent work published by our team and where the same database was used, we know that 7.5% of all fractures included in that period correspond to second fractures [15]. Only osteoporotic fractures were taken into consideration. As an example, pathologic fractures (ICD-9 733.1-733. 19), metastasis, metabolic diseases such as Paget's disease (M89.9), femur fractures of different location (821.0-821.9), or acetabular pelvis fractures (808.0-808.9) as well as fractures due to traffic accident (V87.9) were excluded. Data derived from readmissions were removed. An admission in the same center within the first 30 days after discharge due to the same medical reason was considered a readmission. Patients having a fracture within that period had a follow-up until the end of the study (December 31, 2015). The MBDS was queried to check whether the patient had died or had been readmitted for a different reason after the fracture. Losses to follow-up or deceased patients were crosschecked with death certificates obtained from the National Death Index or INDEF, by its Spanish acronym (https:// indef.msssi.es/indefWeb/loginAction.do). Second fractures were excluded from the survival analysis, and only the date of the first fracture was considered for analysis. In this study, with the objective of facilitating the comparison with previous studies, only population over 45 years of age were included.
The following data were collected at the moment of the first fracture: sex, age, place of residence (nursing home or own home), type of fracture (intra-or extracapsular), time elapsed until surgery, and time of hospitalization. Additionally, the following data on comorbidity (identified by ICD code) were recorded: dementia, diabetes mellitus, obesity, neoplasm, mild to severe liver disease (defined as evidence of portal hypertension: ascites, esophageal varices, or encephalopathy), chronic obstructive pulmonary disease, skin sores, cardiovascular disease (ischemic cardiopathology, brain stroke, congestive heart failure, and peripheral artery disease). Charlson Index calculation was performed based on comorbidity data. Additional data were collected from the Nursing Assessment Form on admission. This questionnaire started to be used in our center in 2010, so only data from the last 5 years of study were collected. During this period of 5 years, 1154 patients were included, and out of this total, the form was completed for 810 patients. Data collected by this questionnaire consisted of vision disorder and hearing impairment, urinary and/or fecal incontinence, all the Norton scale items plus its final score that quantifies the risk of bed sores, the Downton scale, technical aid to the patient's evolution, and habitual treatment.
To calculate standardized mortality ratios (SMRs), the population and mortality in Madrid were used as a reference. Data were obtained from the Statistical Institute of the Community of Madrid (http://www.madrid.org/iestadis).

Statistical analysis
The follow-up period covers the time between the dates of hospital admission, due to hip fracture, and patient's death. Three 4-year initial periods and one 5-year final period were defined for the analysis.
The annual mortality rate was calculated after the occurrence of an osteoporotic hip fracture per 100,000 inhabitants in people older than 45 and by sex, the numerator being the cases of death after first hip fracture registered in the MBDS and the denominator being the population at risk (those who had a first hip fracture and were still alive).
The cumulative rate of mortality was calculated for 1, 3, 6, 12, and 36 months over the time of study (17 years), global, and by sex. The probability of mortality density was also analyzed during a period of 10 years following a hip fracture. Specific mortality rates by age and sex, based on the time between fracture and death or end of study, were compared to rates of expected mortality by age and sex in the Madrid population (standardized mortality ratios or SMRs). Assuming a Poisson distribution, statistical significance and the respective 95% confidence intervals (CI95%) were calculated.
Kaplan-Meier survival analysis was used to compare survival time between the different periods and by age and sex. Bivariate and multivariate Cox regression analyses were performed to analyze differences between patients and factors associated with mortality. Results were expressed as hazard ratios (HR) with their corresponding CI95% values.
Change in the annual rates of death after hip fracture was analyzed by generalized linear models (GLM) with Poisson distribution or negative binomial distribution in case of over-dispersed data. Models were used to estimate the incidence rate ratio of the variables at 1, 3, and 12 months with CI95%; the p value of association was noted as a result of the trend test. The significance level was established at 1% to account for the use of aggregated data. To allow for a clearance period, year 1999 was excluded from the analysis.
Patients' clinical and demographic characteristics were described. Quantitative data were described by their median and standard deviation or by median and interquartile range (IQR). Qualitative data were described by counts and percentage. The relative frequency of the different variables for each period was also analyzed in order to find out if they have changed over the study period.
The statistical analysis was carried out with IBM SPSS 24.0 and Stata 14 software.
The majority of patients (73.9%) were living in their own homes and not institutionalized before the fracture occurrence. Among institutionalized patients, 81.95% were women. The average Charlson Index was 1.08 (SD 1.50), and in women, it was 0.70 (SD 1.14). Out of the total amount, 7.4% of patients had a Charlson Index above 2 (men 12.1% out of the total of men and women 6.1% out of the total of women).

Trend in mortality
Through the use of GLM, it is observed that 1-year mortality rate increased 2% per year in a significant way (IRR 1.020, CI95% 1.008-1.033). One-month mortality rate remained stable without increasing or decreasing (IRR 1.020, CI95% 0.967-1.050) and the same happened for 6-month mortality rate (IRR 1.016, CI95% 0.997-1.035) (Fig. 1). Figure 2 shows probability of mortality density for a period of 10 years following the hip fracture occurrence. It is around 16% of women and 25% of men in the first 90 days, falling rapidly in the first year and stabilizing around 2% every 90 days in the following years.

Factors associated to mortality after hip fracture occurrence
The bivariate regression Cox analysis (Table 3) shows association between mortality and the following items: In the survival analysis, men die more frequently and earlier than women after hip fracture (Additional file 1: Figure S1). In the analysis by periods, the trend is that mortality is also earlier (Fig. 3).

Discussion
Numerous studies proved that the mortality risk after hip fracture is increasing [10,11,16,17]. This increase in mortality is not exclusive to hip fracture, and it is also associated with vertebral fractures [18,19] and practically all major osteoporotic fractures, such as humerus fracture [20,21], pelvis fracture [22,26], distal end of femur fracture [22], and rib fracture [22,26], and at all ages [18,19]. Moreover, a second fracture results in an additional increase in mortality [18].
In this study, mortality in patients with hip fracture has been analyzed over a long period of time (1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015). This allowed for research into trends in mortality rates, which is currently a controversial topic. The main finding of this study is that the trend in 1-year mortality rate has increased 2% per year during the period of study. Other findings corroborate well-known results obtained from other studies: in the first place, mortality in patients with hip fracture increases a lot compared to the general population; secondly, mortality rate is higher in men than in women; thirdly, death risk is very high right after the fracture occurrence (it drops over the first year and remains stable afterwards); fourthly, institutionalization combined with comorbidity is associated with a higher mortality.
A few studies presented results on trends in mortality after osteoporotic hip fracture. Some of them, from different nationalities, show that mortality remained stable over the years [16,23,24]. Other studies described small increases in excess mortality due to hip fracture in specific population groups (elderly women) [10], and others described a variable reduction, although moderate in general, in mortality rates over time [25][26][27][28]. Being a study that covers a long period of time, this study allows us to analyze trends in mortality in patients with hip fracture. The study observes a growing trend of mortality of 2% per year at 1 year, while short-term mortality remains stable (1 to 6 months). We believe that the observed increase in mortality rate is due to a growing proportion of men over women occurred over the years of the study period, and men, as is explained below, have a higher mortality rate. While in the first period of the study (1999)(2000)(2001)(2002)   In accordance with different studies, cumulative 1-year mortality after hip fracture varies between 20 and 40% [5,29,30] with higher rates of mortality observed in men than women [8,31]. In previous studies done in Spain, 1-year mortality has been placed around 21-23% [12,13] and varies widely depending on the regions [32]. In this study, a cumulative 1-year mortality of 30% in women and 43% in men has been obtained, which is clearly higher than data obtained from previous studies. The difference between results obtained in this study and previous studies can possibly be found in the methodology used. In this study, we have searched for patients' deaths through the national database of death certificates, while in the rest of studies this was not done. In this study, the standardized mortality index for the whole period of study (17 years) has been analyzed in comparison with the global mortality in Madrid based on data obtained from the National Institute of Statistics (INE, by its Spanish acronym). We obtained an 8 SMR. This finding is similar to those of other authors working with series such as the Forsen study [33], the Johnell study [34], or the Fransen study [35]. Forsen et al. [33] found a 1-year mortality risk of 9 in subjects aged between 50 and 74 years, 5.1 for subjects aged between 75 and 85 years, and 5.7 for subjects older than 85. Johnel et al. [34] found a 1-year mortality risk of 10.2 in men and 9.1 in women. Fransen et al. [35,40] found a 2-year mortality risk of 7.18 in women.
Epidemiological studies have consistently proved that men present a higher risk of mortality [8,9]. This study, as other studies did, found a higher risk of mortality in men than in women. Masculine sex presents a 30% increase in death risk compared to women. In contrast to other studies that do not find that excess mortality in men compared to women is due to comorbidities [36], this study observed that men present higher comorbidity, measured by the Charlson Index which would justify this increase of mortality.
Likewise, the study observes, as other studies do [7], the fact that death risk is very high immediately after hip fracture and that this risk rapidly drops over the first year, stabilizing with a slight trend downwards over the following years (Fig. 2).
Lastly, the role of comorbidities as a factor increasing risk mortality is controversial [37]. In our study, the risk factors most strongly associated with mortality (HR > 2) were moderate-severe liver disease, to be older than 75 years, and heart failure. Other risk factors also associated with mortality (HR between 1 and 2) were male, Charlson index greater than 2, diabetes mellitus, living in a nursing home, COPD, myocardial infarction, dementia, and renal failure. It has been hypothesized that comorbidities play a major role as a cause of mortality after fracture. In fact, several studies [12,38], as this study did, found an association between the severity of comorbidities and mortality risk. In a study based on Medicare USA data, after adjustment by comorbidities, the risk of mortality dropped significantly over the early post-fracture period (first 6 months) and disappeared completely over the later post fracture period of follow-up (over 6 months) [37]. Due to this dramatic drop of excess mortality with adjustment of state of health, the authors concluded that the majority of early deaths and all late deaths were not attributable to hip fracture. In a different Spanish study, the severity of comorbidities was also found to be associated with an increase of mortality after vertebral fracture [37].
Nevertheless, not all studies found an association between comorbidity and post-fracture mortality. In the Study Osteoporotic Fractures (SOF), the adjustment by comorbidity did not affect the relationship between mortality risk and any of the examined fractures [39]. Similar findings were obtained in other studies [40]. This study is groundbreaking as we had the chance to analyze the data collected from a Nursing Assessment Form on admission for a subgroup of 810 patients. This form collects specific data on the patient's quality of life before admission, such as vision and hearing capacity, presence of urinary incontinence, fecal incontinence, Downton scale (measuring risk of falling), and Norton scale (measuring risk of developing bedsores). In this study, it was found that, with the exception of Norton scale total score, all these factors are associated with risk of death. It seems that we can conclude that a previous bad quality of life is associated with a higher mortality after hip fracture. Unfortunately, the absence of data on established treatments does not allow us to assess with certainty whether different treatments have been taken or avoided in these groups with higher risk. It is striking that among the associated factors are institutionalized and older people, that is, the more fragile patients who sometimes have to stabilize before proceeding with surgical treatment, delaying one of the known factors that increase mortality. In relation to the increase in mortality in males, the associated comorbidities that sometimes require prior stabilization is also one of the main causes of surgical delays. In daily practice, knowing the most vulnerable subgroups forces us to maximize therapeutic care in these subgroups with a higher risk of mortality. This study also presents some limitations, among them were those typical of a retrospective study that uses administrative databases as main source, and these limitations may cause a loss of data. Additionally, since it is a largescale population and a long period of study, some variables could not be obtained from all patients (for example, date of surgery or data collected on the Nursing Assessment Form on admission which was only implemented in our center for the last few years of study). Time delay between the fracture and surgery could not be observed. Important variables for this profile of patient, i.e., the use of different treatments such as bisphosphonates and other therapeutic methods, could not be examined since our database lacks this type of data. Another limitation of our work, it is that which refers to the use of HR in retrospective cohort studies. The use of the HR for causal inference is not straightforward even in the absence of unmeasured confounding, measurement error, and model misspecification. Endowing a HR with a causal interpretation is risky for 2 key reasons: the HR may change over time, and the HR has a built-in selection bias. Nevertheless, this study presents some strength. It is a series with a large number of patients (3992) that includes all osteoporotic hip fractures treated in a tertiary hospital. For this reason, we think that this study is very representative of the urban population of the center of Spain, and it was carried out over a long period of time (17 years). Additionally, numerous variables were collected, and some of them, related to mortality after hip fracture, have not been studied until now, such as Downton or Norton scales. In conclusion, in this study, we found that 1-year mortality has increased over the last 17 years, and we corroborate specific aspects of mortality that were established in other previous epidemiologic studies, i.e., that mortality in patients with hip fracture increased highly compared to the general population, that mortality risk is higher in men than in women, that death risk is very high immediately after fracture and drops over the first year remaining stable afterwards, and that institutionalization combined with comorbidity are associated with higher mortality. Additionally, some factors related to quality of life before hip fracture are identified, i.e., vision disorders, hearing impairment, incontinence, or Downton scale, which are associated with an increase in mortality for these patients.