Introduction
Aging is related to morphological, functional, biochemical, and psychological changes
that difficult the adaptation of a person to his/her environment, increasing the vulnerability
and the incidence of pathological processes that directly interfere in the quality
of life and in the mortality of elderly individuals.[1] In this population, proximal femoral fractures (PFFs) represent events of great
significance, both in frequency and severity, since they are associated with loss
of independence and with a reduction in life expectancy.[1]
It is known that the trauma resulting in PFFs in the elderly is mostly of low energy
and related to populational features, such as osteoporosis, malnutrition, decreased
visual acuity, impaired cognitive functions, and sarcopenia.[2] In elderly patients, PFFs are correlated with a mortality rate of ∼ 30% in the 1st year after the injury, being the main cause of death by trauma in individuals > 75
years old.[3] Some factors show a clear correlation with the increased mortality in PFF patients,
such as age, cognitive decline, time elapsed between the triggering event and the
surgical approach, prefracture mobility capacity, and previous comorbidities.[4]
[5] The early identification of patients with a higher predisposition to triggering
events and complications may help to reduce mortality in this scenario.[6]
Data on the population of the USA highlighted the significance of these fractures;
the average number of 250 thousand PFF cases in the 1990s is expected to duplicate
or triplicate until 2040.[7] It is believed that this exponential increase is closely related to a higher life
expectancy, since the associated risk factors become more prevalent as people grow
older.[8] In Brazil, Loures et al[9] found an average total cost of BRL 1,933.79 for patients submitted to surgical PFF
correction in the public health system in 2011 and in 2012. In the 2007/2008 period,
there were 34,284 intertrochanteric femoral (FIT) fractures in Brazil, and their treatment
costed a total of ∼ BRL 30.8 million in 2008.[10]
The present study aims to evaluate the factors related to the mortality of ≥ 70 years
old PFF patients submitted to surgical treatment and followed-up for 6 months.
Material and Methods
This is an observational, retrospective cohort study. The medical records of 141 PFF
patients who underwent surgical treatment from January 2009 to December 2015 by the
same senior surgeon were analyzed. Proximal femoral fractures included transtrochanteric
or FIT fractures with or without subtrochanteric trace, and femoral neck fractures
(FNFs). The treatment of FIT fractures was performed with intramedullary nails (IMNs)
or with dynamic hip screws (DHSs), whereas FNF treatment employed hip arthroplasty
or inverted pyramid compression (PCP); for statistical purposes, the method type and
the selected implant for the treatment of each case were not considered.
For the sample survey, the inclusion criteria were the existence of medical records
of PFF patients who underwent surgical treatment in a private hospital in Juiz de
Fora, state of Minas Gerais, Brazil, under uniform conditions, performed by the same
surgeon-researcher, and emphasizing the premise of surgical approach in the 1st 48 hours after the trauma, as recommended in the literature.[11] Uniformity conditions were defined as similar circumstances regarding structure,
classification, therapy, technique, material, and support. The exclusion criteria
were: pathological fractures due to neoplasia, fractures not classified as FIT or
FNF, patients < 70 years old, and lack of information about the studied variables
in the hospital documentation. Patient records that did not present a minimum follow-up
of 6 months were also excluded.
The primary outcome was death in the 1st 6 postoperative months. Secondary outcomes were gender, age, total hospitalization
days, number of previous comorbidities, lesion (fracture) topography, and hospitalization
at the intensive care unit (ICU). Regarding the variable comorbidities, a quantitative
evaluation was selected instead of a qualitative analysis. The variable ICU hospitalization
was solely based on the indication of the anesthesiology team, with no discussion
of its rationale.
The Ethics and Research Committee for research on human beings of our institution
approved the present study under the number 69173717.6.0000.5133 at the Certificate
of Submission for Ethical Appreciation (CAAE, in the Portuguese acronym) in the Brazilian
platform, with the original title Predictive Factors of Death After Surgery for Fixation of Proximal Femoral Fracture (Fatores preditivos de morte após cirurgia para fixação de fratura de fêmur proximal).
Statistical analysis
The statistical analysis was performed in three stages: univariate, bivariate and
multivariate. In the univariate analysis, descriptive statistics characterized the
sample through mean, standard deviation (SD), median, interquartile range (IQR), and
absolute and relative frequencies. In the bivariate analysis, the chi-squared (χ2) test or the Fisher exact test, when appropriate, were used to determine the association
between each of the independent variables (risk factors), and the dependent variable
(outcome: death); moreover, the death risk, as odds ratio (OR), with a 95% confidence
interval (CI), was calculated. The multivariate analysis employed a binary logistic
regression. The model was constructed with the Enter method with block input. The
logistic model was assessed using Likelihood Value (-2LL), Nagelkerke pseudo R2, and the Hosmer and Lemeshow test. The statistical significance of each coefficient
was analyzed by the Wald test. The predictive capacity of the model was evaluated
with a classification matrix, using the value of 0.3 as the cutoff point. Data were
analyzed using the statistical software IBM SPSS Statistics for Windows, Version 20.0
(IBM Corp., Armonk, NY, USA), with statistical significance defined as p < 0.05.
Results
[Table 1] presents the general demographics of the patients. The mean age was 84.4 years old
(±6.8 years old), ranging from 70 to 100 years old; most of the patients were female
(70.2%), and had at least 1 associated comorbidity (81.4%). Regarding the topography
of the lesion, most of the patients presented FIT (62.9%). The death rate in the total
sample was of 34.7% at the end of the 6-month follow-up period.
Table 1
|
Variables
|
n
|
%
|
|
Number of patients
|
124
|
100.0
|
|
Gender (female)
|
87
|
70.2
|
|
Age group (> 85 years old)
|
56
|
45.2
|
|
Hospitalization period (> 7 days)
|
54
|
43.5
|
|
Number of comorbidities
|
|
0
|
23
|
18.5
|
|
1
|
40
|
32.3
|
|
2
|
35
|
28.2
|
|
3
|
20
|
16.1
|
|
4
|
6
|
4.8
|
|
Topography
|
|
FIT
|
78
|
62.9
|
|
FNF
|
46
|
37.1
|
|
ICU Admission (yes)
|
51
|
41.1
|
|
Outcome (death)
|
43
|
34.7
|
[Table 2] shows the association between the categorical variables and the outcome (death)
using OR as a risk measure. In patients > 85 years old, the death risk is 2-fold higher
compared with those < 85 years old. The death risk is also 2.5 times higher in patients
hospitalized for > 7 days. Compared with patients with no comorbidities, the death
risk in patients with some comorbidity was four times higher. In addition, patients
who were admitted to the ICU were four times more likely to die than those who did
not. The death risk was similar between men and women and independent of the topography
of the lesion (p > 0.05). The data analysis also showed that FNF patients who died were older when
compared with the other patients (p = 0.027) ([Fig. 1]).
Fig. 1 Comparison of patients with transtrochanteric/intertrochanteric (FIT) or FNF and
their outcome in relation to age. *, statistically significant difference, p = 0.027.
Table 2
|
Variable/category
|
Deaths
|
p-value
|
OR
|
95%CI
|
|
n
|
%
|
|
Gender
|
|
Male
|
14
|
37.80
|
–
|
–
|
–
|
|
Female
|
29
|
33.30
|
0.630
|
0.82
|
0.37–1.83
|
|
Age group
|
|
≤ 85 years old
|
18
|
26.50
|
–
|
–
|
–
|
|
> 85 years old
|
25
|
44.60
|
0.030[*]
|
2.24
|
1.705–4.76
|
|
Hospitalization period
|
|
≤ 7 days
|
18
|
25.70
|
–
|
–
|
–
|
|
> 7 days
|
25
|
46.30
|
0.020[*]
|
2.49
|
1.17–5.31
|
|
Comorbidities
|
|
No
|
3
|
13.00
|
–
|
–
|
–
|
|
Yes
|
40
|
39.60
|
0.020[*]
|
4.37
|
1.22–15.68
|
|
ICU Admission
|
|
No
|
16
|
21.90
|
–
|
–
|
–
|
|
Yes
|
27
|
52.90
|
< 0.001[*]
|
4.01
|
1.84–8.75
|
|
Topography
|
|
FIT
|
24
|
30.80
|
–
|
–
|
–
|
|
FNF
|
19
|
41.30
|
0.230
|
1.58
|
0.74–3.38
|
[Table 3] summarizes the logistic regression coefficients and their significance in the model.
The model was valid for the mortality status classification. About 20% of variability
in the mortality status can be explained by the model. The model showed an accuracy
of 70.2% in death classification, with 81.4% sensitivity (death accuracy) and 64.2%
specificity (non-death). The probability of death is higher in older patients, those
hospitalized for > 7 days, in the ICU, and presenting with FNF. It is worth noting
that hospitalization for > 7 days and FNF were not statistically significant, but
they were kept in the model due to the biological plausibility and to improve its
final adjustment. The discriminatory power of the model can be deemed acceptable.
The area under the curve (AUC) was 0.75 (95%CI = 0.63–0.82; p = 0.19) ([Fig. 2]).
Fig. 2 Receiver operating characteristic (ROC) curve of the diagnostic capability of the
logistic model to discriminate the death outcome in patients ≥70 years-old surgically
treated for proximal femoral fracture. AUC, Area under the curve; 95%CI, 95% confidence
interval.
Table 3
|
Variable
|
Parameter estimation
|
Standard error
|
p-value
|
Odds Ratio (95%CI)
|
|
Age
|
0.068
|
0.033
|
0.039
|
1.07 (1.00–1.14)
|
|
Hospitalization > 7 days
|
0.631
|
0.427
|
0.140
|
1.88 (0.81–4.34)
|
|
FNF
|
0.502
|
0.426
|
0.239
|
1.65 (0.72–3.81)
|
|
ICU Admission
|
1.106
|
0.426
|
0.009
|
3.02 (1.31–6.96)
|
|
Intercept
|
- 7.421
|
2.856
|
0.009
|
0.001
|
Discussion
A 6-month follow-up period was proposed to compare the mortality rate of the sample
both with similar and with 1-year follow-up samples. Our findings were satisfactorily
comparable to those of Forster et al;[12] since we have obtained a mortality rate of 34.7% at 6 months postoperatively, whereas
these authors observed mortality rates of 50% after 6 months, and of 56% after 1 year
of follow-up. It should be noted that the study by Forster et al is based on a population > 100
years old (mean age of 101 years old), with an average hospital stay of 14 days, which
is higher than the data used in our statistical evaluation. As such, we believe that
the demographics of Forster et al[12] may justify the higher mortality rates in their study, since our investigation revealed
that older age and longer hospitalization periods increase the mortality rate.[13]
[14]
[15]
[16]
[17]
[18]
[19] On the other hand, compared to our study, Garcia et al[20] obtained a lower mortality rate at 6 months and at 1 year of follow-up (14% and
30%, respectively), as did Guerra et al,[13] with a 1-year mortality rate of 23.6%. Wood et al[21] and Parker et al[22] observed mortality rates of ∼ 14% at the 1-year follow-up. As such, the differences
in the literature regarding the mortality rate in the elderly population with PFF
are evident.
In the present study, the postoperative mortality rate was lower than the general
mortality rate of individuals > 60 years old in Brazil (34.7% versus 58.6%).[23]
[24] Although this subject remains controversial in the literature, the mortality rate
was not influenced by gender, corroborating the findings of van Laarhoven et al[25] and by Antes et al.[26]
We have opted for a quantitative, instead of a qualitative, evaluation of comorbidities,
because we understand that, in general terms, the patients with more severe comorbidities
also present a greater number of them. In addition, we assume that the presence of
comorbidities considered severe is often a consequence of prior, milder conditions.
These inferences are based on a study by Garcia et al,[20] who found an abrupt drop in the number of individuals with > 4 comorbidities when
dividing their sample according to the number of conditions presented by each patient;
these results are similar to those presented in [Table 1]. In addition, a review of the literature revealed studies showing that the higher
number of comorbidities is associated with a worse outcome regarding death, although
the severity of the disease was not taken into account.[27]
[28] Even though we have not evaluated the severity of the conditions nor found a statistically
significant difference in the absolute number of comorbidities (our results just allow
affirmations regarding the presence or absence of comorbidities), we believe that
the severity and the number of diseases distinctively interfere in the death of a
patient and are difficult to dissociate from it; as such, the solely quantitative
comorbidity evaluation does not invalidate a study. Thus, the finding that the risk
of death is four times higher in patients with at least one type of comorbidity compared
with those with no comorbidities is considered relevant, and it confirms data from
Guerra et al,[13] who observed that the absence of comorbidities is associated with their so-called
alive group, and that the presence of three comorbidities is associated with their so-called
death group. The same occurs with Shebubakar et al[29] and with Campos et al,[30] who demonstrated that the presence of ≥ 2 comorbidities is associated with an increase
in morbidity and mortality rates. Also regarding the choice for a purely quantitative
analysis, we also consider that one of our goals was to create a line of reasoning
that is reproducible and applicable to the general population, and not to specific
groups and to their different levels of pathological involvement (patients with heart,
coronary heart, liver disease, kidney or lung diseases, etc.).
Considering the findings presented on [Table 3], in which hospitalization for > 7 days and the presence of FNF did not reach statistical
significance, we believe that it is still important to take both variables into account.
The former due to the statistical significance and exuberant OR presented in [Table 1] (2.5-fold death risk), and the latter for its important OR shown in [Tables 1] and [3]. As such, we believe that a greater sample size may reveal a significant statistical
difference regarding the death outcome according to the topography of the lesion (FIT
versus FNF), especially considering the requirement of ICU admission.
Conclusion
The sample studied follows the epidemiological trend established in the literature
regarding mean age, gender, and fracture topography related to the mortality of elderly
patients with PFF submitted to surgical treatment. Advanced age, comorbidities, longer
hospital stay, and ICU admission are also already consolidated as associated to a
greater number of deaths in this population, and, similarly, these features were related
to a greater mortality in our study.
Regarding the death outcome, although we did not find a statistically significant
difference regarding the topography of the lesion and its behavior in its coexistence
with ICU hospitalization, we believe that further investigations under this perspective
are required in this population. This need is justified because the proposed classification
model shows a higher death probability in elderly patients hospitalized for > 7 days,
admitted to the ICU and presenting FNF; as such, we ask whether a bigger sample size
would result in a statistically significant increase in the death risk of patients > 70
years old with FNF and admitted to the ICU.
