Background
Patients who have a hip fracture are at considerable risk for premature death [
1,
2]. A recent report of osteoporosis in the European Union estimated that mortality related to low-impact trauma hip fracture is greater than road traffic accidents and equivalent to breast cancer [
3]. This mortality burden will increase over the next few decades commensurate with the aging of the population [
3]. Targeted interventions among at-risk groups may contribute to mortality reductions [
1], thus, a contemporary epidemiology of hip fracture mortality would be useful in developing risk profiles and estimates of potential lives saved. To this end, country-specific mortality data should be collected to refine estimates of the longitudinal burden of hip fracture and conduct economic evaluations of hip fracture prevention and treatment measures [
4].
The data describing excess mortality after hip fracture are well established in developed Western European and North American countries [
1,
5]. In a systematic analysis, the documented cumulative excess risk (i.e., exceeding mortality rates among non-hip fracture or community control populations) during the first year after hip fracture varied widely from 8.4% to 36% [
1]. Studies of short- vs. long-term mortality almost always note increased mortality soon after the fracture, within the first 3–6 months [
1,
2,
5‐
9]. Although the relative risk decreases in subsequent years, it does not return to that of age- and sex-matched reference groups even 10 years post-fracture [
5]. The excess risk increases with advancing age, although this differential becomes less pronounced over the following years due to increased mortality, unrelated to hip fracture, in the reference populations [
1,
5]. The excess risk is also higher among men than women [
1,
5], notably among the oldest age categories (≥80 years) in the first months and years after fracture [
5].
Estimates of hip fracture mortality in Eastern Europe are scarce [
10,
11]. However, the data in this region suggests a sex-specific difference in the incidence of hip fractures between Eastern and Western Europe [
12‐
14], and the age and sex-specific all-cause mortality rates in Eastern Europe differ from those in Western countries [
15]. Longitudinal data is needed to quantify the change in excess mortality after hip fracture by temporal, clinical and geographic characteristics. This study estimates the impact of hip fracture on 10-year all-cause mortality among Estonian men and women ≥50 years of age.
Discussion
Our study assessed the impact of hip fracture on the all-cause risk of death over 10 years among people 50 years of age or older in Estonia. We observed a pronounced and sustained excess mortality risk after a hip fracture that was highest within 6–12 months after the fracture and persisted for the full 10-year follow-up period. After adjustment for age and pre-fracture comorbidities (CCI), hip fracture was associated with a 21.1% (95% CI 20.0–22.5%) 10-year cumulative excess risk of death (RR 1.37, 95% CI 1.35–1.40). Even after 10 years following the hip fracture, more than 1 in 4 deaths among hip fracture patients was attributable to the initial injury. The 1-year average relative all-cause excess mortality (4.1 in men, 3.4 in women) after hip fracture was comparable to that of dementia, cancer, and heart failure [
21], mental disorders [
33], and higher than that for diabetes [
34].
The excess risk of death differed by the duration of follow-up time after the hip fracture, by sex, and according to age at the time of fracture. We found a high immediate excess risk of death in older age groups (≥80 years) and a gradually increasing excess risk in younger age groups (50–79 years) that was more pronounced in men than in women. In the elderly, hip fracture has an immediate devastating impact on mortality that lasts for years. For example, in the group of men ≥90 years old, the excess risk of death 3 months after fracture was 45%, and the risk of dying was 8 times higher than in men who had not had a fracture. Consequently, over half of the men in that group had died within 3 months of the fracture, and within 12 months over two-thirds of the men had died. It is important to note that the excess risk among the elderly (aged 80 years or more) persisted throughout the 10-year follow-up period and did not disappear in any age- or sex-specific group. In contrast, the excess risk in younger age groups increased in a linear fashion over the follow-up period. For example, in the 60–69-year group the excess risk increased with time to a maximum of 30%. Thus, 6 out of 10 men with a hip fracture and 1 in 2 women in that age group died during the 10-year follow-up period. The risk of death was 2–3 times higher in the fracture group than in the reference group even 10 years after the fracture.
Previous studies have demonstrated an immediate elevated risk of mortality after hip fracture [
1,
2,
5,
6,
35‐
39], however, the evidence of persistence is not universal [
1,
7,
9,
40,
41]. Overall, our results are in line with the meta-analysis suggesting that the excess mortality after hip fracture in patients over 50 years is extensive already in the first months after fracture and persists for at least 10 years [
5]. Notably, the relative risk of all-cause mortality within the 3 months after hip fracture was as high as 9.5 in men and 8.3 in women, comparable to that in the meta-analysis [
5]. However, compared to the pooled estimates [
5] the excess risk of death in Estonian men and women in younger (50–79 years) age groups was rather high, particularly in the first months and years after fracture. For example, in the 70–79-year-old men the excess risk in our study reached as high as 18% within 1 year, and 30% within 5 years, whereas in the meta-analysis the respective estimates were lower (11% and 20%). Likewise, in women of the same age we found the excess risk to be 14% in 1 year and 24% in 5 years, versus 5% and 13% in the meta-analysis. It is difficult to explain the reasons for increased mortality in these groups, but insufficient case management upon discharge and low utilization of rehabilitation, nursing care, and social care [
42] could be potential contributors. In addition, it is possible that not all comorbid conditions were diagnosed and recorded in these patients (especially men), and their impact on mortality therefore could not be adjusted in the models. Finally, excess mortality study results are difficult to compare due to differences in study design and sources of data, ascertainment of cases and controls, determination of death, differences in follow-up time, presentation of results, and adjustment for confounding [
1,
5].
A number of confounding factors, such as advanced age, male sex, poor preoperative health status and multiple comorbidities have been associated with excess mortality following hip fractures [
43]. Our results are in line with the collective evidence confirming that excess mortality increases with age, and is higher in men than in women in all age groups [
1,
5]. The average excess risk among men in our study did not exceed that among women during 3 years following the fracture; this may be explained by the different age distribution of fractures in men and women in our study. We know that most hip fractures in Estonian men occur at a younger age (50–79 years) due to the elevated incidence of hip fractures in that age [
14] and low life expectancy in men (15.1 years at age 65, 2014 data) [
16], whereas over half of fractures in women occur among those ≥80 years [
14]. Due to the considerable age difference between sexes in our study (8.2 years) women experienced an elevated crude cumulative 10-year risk of all-cause death among both the hip fracture and reference groups (see Fig.
1), and the weighted average excess risk in both study groups (see Fig.
3) was influenced by the higher-weighted age groups, with younger groups in men and older groups in women.
Numerous studies have reported that the presence of pre-fracture concomitant medical conditions are negative predictors for survival [
35,
36,
43‐
46], whereas the extent to which underlying conditions contribute to the excess mortality associated with hip fracture is still unclear [
1]. In our study the hip fracture patients had higher CCI score than the age- and sex-matched reference subjects. However, because the sample in our study was matched for age and sex, and the risks were adjusted for major confounders, we believe that the results strongly suggest that hip fracture is an independent and attributable risk factor for death. This implies that preventative efforts and post fracture rehabilitation and social care are essential to reduce the excess risk of death.
The possible reasons for the greater mortality risk in men than in women following hip fracture are still poorly understood [
1]. Previously described risk factors in older men include multi-morbidity, smoking, lower dietary protein, greater height combined with the use of antidepressants leading to a greater impact upon falling, whereas the traditional risk factors in women (rheumatoid arthritis, use of benzodiazepines and corticosteroids) were not related to hip fractures in men [
47]. It has also been suggested that men have higher rates of pneumonia and septicemia than women [
48], or more severe medical comorbidities prior to the hip fracture [
49,
50]. However, in our study the CCI score was lower in men than in women in both study groups, suggesting that men were healthier than women. It is possible that the lower CCI score in men was related to their younger age compared to women. Our study adjusted for CCI, yet the excess risk was higher in men than in women.
We acknowledge the limitations of our study. Our objective was strictly aligned with assessing the impact of hip fracture on mortality, thus, we did not assess the impact of hip fracture complications on excess mortality, or the causes of death. Nursing home or facility residence, poor preoperative walking capacity, poor activities of daily living and poor mental state have been identified as strongly predictive factors for the excess mortality [
43] suggesting that frail and disabled elderly are at higher immediate risk of death after hip fracture [
7,
46]. It is possible that the level of functional impairment in our study was higher among hip fracture patients. We used data from the (administrative) health insurance database that covers the overwhelming majority of the Estonia’s population. However, we are not aware of any data documenting the completeness of the database. Furthermore, the data on additional useful indicators (sociodemographic factors such as income, education, occupation, social deprivation, and other health/lifestyle indicators (BMI, smoking, alcohol consumption)) [
6,
43,
46,
51] were not available. Finally, we did not account for changes in hip fracture mortality in the population over time.
We also did not analyze in detail the impact of comorbidities on excess mortality in detail. We used the CCI as a well-accepted comorbidity burden index for adjusting for concomitant diseases [
22,
23,
44,
45,
52,
53]. We chose the CCI because of its adaptability to large population databases using diagnostic codes from the ICD-10 [
44]. It has also been documented that excess deaths among hip fracture patients can mainly be explained by the conditions predominantly responsible for mortality in the general population, i.e. those represented in the CCI [
54]. However, we know that as a composite index it does not discriminate well between diseases, i.e. it equates the entities. Models incorporating comorbidities as individual variables perform better in predicting mortality than the weighted index [
22]. The CCI ignores most of the disorders known to cause secondary osteoporosis [
54], and it does not include hyperthyroidism or Parkinson’s disease which are known to increase the propensity to falls [
47]. Furthermore, it does not allow for risk stratification based on disease severity [
22,
44]. Further research is needed to identify the specific diseases most responsible for the excess mortality.
It is possible that the excess risk of death in our analysis was slightly overestimated due to measurement bias. Our data collection started from 2004, and some subjects with unascertained fractures before 2004 might have been misclassified as incident cases (fracture group) or non-fracture patients (reference group). This misclassification might have resulted in slightly overestimating mortality in both groups. However, as the risk for further hip fracture after previous hip fracture is over 2-fold [
3,
55] and a subsequent fracture is associated with increased mortality risk [
56], the overestimation would have been higher in the fracture group, resulting in a slightly overestimated excess mortality.
The comorbidity data for both groups were collected at the time of, and for 1 year before, the index dates of a hip fracture cases. We are aware that by using hospitalization data as a part of index case definition we might have introduced differential misclassification into assessing co-morbidity. However, we believe that people with severe life-threatening conditions would have received health care, and in including data from hospitalization episodes (including primary and secondary diagnoses) within the 12-months recall period into CCI for individuals in the reference group might mitigate some of this bias. Further, we speculate that potential differential misclassification described above might lead to overestimating the effect of comorbid conditions on mortality and thus support our main finding of hip fracture as a major independent risk factor for death.
To our knowledge, this is the first population-based observational study to estimate the impact of hip fracture on mortality in Eastern Europe. The strength of our analysis lies in the use of a data source with nationwide coverage (EHIF data). We had a large sample size of a representative population (given the >94% population coverage of the EHIF), long follow-up, and standardized recording of health events across the period of observation, which avoids problems related to imperfect recall and incomplete records. The large sample size provided a high number of events (deaths) to derive precise estimates over the long follow-up period, and the high frequency of observations allowed for assessment of rapid and extensive changes during the first months after fracture. We believe that our study provides informative results allowing inferences to other Eastern European hip fracture populations ≥50 years of age.