=Paper=
{{Paper
|id=Vol-2488/paper29
|storemode=property
|title="Analysis of the Early Posttraumatic Period Pathophysiology in Case of the Severe Combined Thoracic Trauma Using Multivariate Logistic Regression"
|pdfUrl=https://ceur-ws.org/Vol-2488/paper29.pdf
|volume=Vol-2488
|authors=Myroslav Stupnytskyi,Viktor Zhukov,Tatyana Gorbach,Oleksii Biletskii,Hakan Kutucu
|dblpUrl=https://dblp.org/rec/conf/iddm/StupnytskyiZGBK19
}}
=="Analysis of the Early Posttraumatic Period Pathophysiology in Case of the Severe Combined Thoracic Trauma Using Multivariate Logistic Regression"
==
https://ceur-ws.org/Vol-2488/paper29.pdf
Analysis of the Early Posttraumatic Period
Pathophysiology in Case of the Severe Combined
Thoracic Trauma Using Multivariate Logistic Regression
Myroslav Stupnytskyi1 [0000-0003-2960-1806], Viktor Zhukov2 [0000-0003-2086-0115],
Tatyana Gorbach2 [0000-0003-4819-7220], Oleksii Biletskii3,4 [0000-0001-8638-4823]
Hakan Kutucu5 [0000-0001-7144-7246]
1
Military Medical Clinical Center of the Western Region, Lviv, Ukraine
2
Kharkiv National Medical University, Kharkiv, Ukraine
3
Kharkiv Municipal Clinical Emergency Hospital named by prof. O.I. Meshchaninov, Kharkiv,
Ukraine
4
Kharkiv Medical Academy of Postgraduate Education, Kharkiv, Ukraine
stupnytskyima@gmail.com
5
Karabuk University, Karabuk, Turkey
Abstract. Severely injured patients are always challenging, even more so when
they have suffered critical trauma to the chest. The aim of this study is to create
a prognostic tool for outcome prediction for patients with combined thoracic
trauma based on the determination of main homeostasis parameters on the 1st
and 2nd day after injury. Multivariate logistic regression analysis with forward
elimination of the variables was used for modeling the dependence of outcome
on clinical and laboratory parameters that reflects main pathophysiological
mechanisms developed on the 1st and 2nd day after combined thoracic trauma.
73 Male patients with combined thoracic trauma were included in the study.
The results of fitting a logistic regression model show the relationship between
mortality and six independent variables: transferrin saturation, percentage of
eosinophils, TNF-a concentration, total iron binding capacity, inspiratory frac-
tion of oxygen and albumin concentration. Besides that, forward elimination of
the variables into the logistic regression equation helps to recognize relatively
independent pathophysiological mechanisms involved to progression of wound
dystrophy. The likelihood ratio tests can reflect the contribution degree of each
pathogenesis rout responsible for the negative outcomes of the severe combined
thoracic trauma. The study contributes to our understanding of interaction be-
tween pathophysiological mechanisms that make harmful effects and are in-
volved in the progression of wound dystrophy and compensatory reactions di-
rected on stabilization of vital function disturbances and maintenance of home-
ostasis during this type of wound dystrophy.
Keywords: Combined thoracic trauma, Multivariate logistic regression, Out-
come prediction, Pathophysiologic mechanisms of polytrauma.
Copyright © 2019 for this paper by its authors. Use permitted under Creative Commons License Attribution
4.0 International (CC BY 4.0)
2019 IDDM Workshops.
2
1 Introduction
Logistic regression provides a useful means for modeling the dependence of a binary
response variable on one or more explanatory variables, where the latter can be either
categorical or continuous. The fit of the resulting model can be assessed using a num-
ber of methods [1–3].
Polytrauma has great social and economic value as 80% polytrauma victims are
able-bodied aged [4, 5]. Thoracic trauma is the leading cause of death nearly 25% of
polytrauma patients and, when especially associated with other injuries, it may causes
death in additional 50% of polytrauma patients [6]. Severely injured patients are al-
ways challenging, even more so when they have suffered critical trauma to the chest
[7]. Blunt chest trauma is understood to be an injury to the thoracic cage affecting the
rib cage itself, the lung parenchyma, the heart, great vessels and/or mediastinal struc-
tures, although the bony structures are the ones that are usually the most damaged.
Such trauma is potentially life threatening, with a direct mortality rate of around 25%
and it is related indirectly to mortality in polytrauma patients in another 25% of cases
[8]. Management of chest injuries requires multidisciplinary approach [9] and in-
volves different specialists: emergency physicians both out of hospital and in hospital
settings, anesthesiologists, intensivists, radiologists, and surgeons [10].
The pathophysiology of polytrauma is complex, involves mostly all systems and
organs and is still being elucidated [11, 12]. The injury rapidly activates the immune
defense, which includes some protease cascades (coagulation and complement) and
the cellular innate and adaptive immune response. Central part of the pathophysiolog-
ical changes is the trauma-induced coagulopathy, hypothermia, and acidosis. The
complex pathophysiological interactions of damaged and dysfunctional molecules,
cells, and organs with the defense systems result in a systemic inflammatory response
and severe complications such as sepsis, multi-organ dysfunction, and multi-organ
failure [11, 13]. The comprehensive knowledge of the underlying pathophysiological
mechanisms and the corresponding principles of clinical management are indispensa-
ble for the successful treatment of multiple injured patients [13].
Characterization of the severity of injury is crucial for the scientific study of trau-
ma, triage, classification of patients, quality management and the assessment of prog-
nosis (prediction of mortality of an individual patient) [14]. Various scoring systems
have been created for prognostic value in patients with thoracic trauma (Thoracic
trauma score, Injury severity score, Abbreviated injury score thoracic, and pulmonary
contusion score). However, existing data stay dubious about the final risk in terms of
morbidity and mortality [15]. The search for the best marker or set of markers for the
diagnosis, prognosis and treatment quality of ―at risk‖ trauma patients is ongoing.
2 Aim
The aim of this study is to create a prognostic tool for outcome prediction for patients
with combined thoracic trauma based on the determination of main homeostasis pa-
rameters on the 1st and 2nd day after injury.
3
3 Materials and methods
3.1 Patients.
73 Male patients with combined thoracic trauma were included in the study. They
were treated in anesthesiology and intensive care department for patients with com-
bined trauma of Kharkiv Municipal Clinical Emergency Hospital named by
prof. O.I. Meshchaninov. As inclusion criteria were chosen ISS > 16, two or more
injured body regions, presence of severe blunt thoracic injuries (AIS 3 and more).
Excluding criteria was a concomitant chronic disease in subcompensation or decom-
pensation phase. Survival / nonsurvival ratio was 42 / 31. The examination was per-
formed on the 1st and 2nd day after trauma (10.75 – 33.5 hours). Table 1 gives some of
the main characteristics of the patients’ groups. It can be seen that there were no sig-
nificant differences according to age, admission time and the anatomical classification
between patients groups.
Table 1. Patient Characteristics (Median, 95% confidence interval).
Survivors Nonsurvivors P
Number of patients 42 31
Age, years 41 (38.21-44.89) 42 (36.7-46.46) 1
ISS score 24.5 (22.73-28.22) 34 (30.38-38.53) 0.0006
RTS score 7.84 (7.051-7.684) 6.17 (5.356-6.464) <0.0001
TRISS probability 0.964 (0.871-0.961) 0.717 (0.556-0.766) <0.0001
Admission time, hours 1 (0.854-1.97) 1 (0.435-3.297) 0.8434
Craniothoracic 6 3
Thoracoabdominal 3 1
Thoracoscelethal 7 1
Craniothoracoabdominal 5 5 0.0901
Craniothoracoscelethal 7 7
Thoracoabdominoscelethal 5 2
Craniothoracoabdominoscelethal 9 12
3.2 Laboratory methods.
In general, 89 clinical and laboratory parameters, as well as their relationships were
used for the regression analysis as explanatory variables. The patients’ plasma was
assayed for biochemical markers using spectrophotometric methods in the Biochemis-
try department of Kharkiv National Medical University. White blood cell count was
performed according to the standard method in the clinical laboratory of Kharkiv
Municipal Clinical Emergency Hospital. The level of eosinophils was expressed as
4
percentages to the total white blood cells. Albumin concentration was estimated
through the turbidimetric method and was expressed in g/L. Plasmatic Iron concentra-
tion and total iron-binding capacity (TIBC) were estimated spectrophotometrically
and both were expressed in mol/L. These biochemical markers were estimated with
the help of ―Filisit-Diagnostica‖ diagnostic kits. Transferrin saturation was calculated
as Iron/TIPS ratio and was expressed in percentages. ELISA kit ―Vector-Best‖ was
used for the determination of Tumor necrosis factor-α (TNF-α) concentration and was
expressed in pg/mL.
3.3 Data analysis.
Data were collected in a Microsoft Excel spreadsheet. STATGRAPHICS Plus 5.0 was
used for multivariate logistic regression analyses with forward elimination of the vari-
ables [1, 3]. Analysis of deviance was used to measure the usefulness of the model
and the goodness of fit of the model was assessed according to Chi-square goodness
of fit test. The presence of serious multicollinearity was checked by the correlation
matrix for coefficient estimates of the regression equation. Mann-Whitney test was
used to assess differences between groups. Chi-square test for trends was performed
to consider differences in nominal data. The significance level was specified as
p <0.05.
4 Results.
4.1 Logistic regression.
Table 2 shows the results of the estimated regression model. The results of fitting a
logistic regression model show the relationship between mortality and six independ-
ent variables.
Table 2. Estimated Regression Model (Maximum Likelihood).
Parameter Estimate Standard Error Estimated Odds Ratio
Constant -4.38337 6.0877
Transferrin saturation 0.268732 0.0907964 1.3083
Eosinophils 6.88929 2.39064 981.706
TNF-α 0.153333 0.0655899 1.16571
TIBC -0.102003 0.0635617 0.903027
FiO2 32.7435 9.29733 1.66089×1014
Albumin -0.825942 0.209276 0.437822
The equation of the fitted model is:
5
(1)
Where k is calculated from the formula:
-
(2)
Table 3 shows the results of the analysis of deviance. Because the p-value for the
model in the Analysis of Deviance table is less than 0.01, there is a statistically signif-
icant relationship between the variables at the 99% confidence level. Furthermore, the
p-value for the residuals is greater than 0.10, indicating that the model is not signifi-
cantly worse than the best possible model for this data at the higher than 90% confi-
dence level.
Table 3. Analysis of deviance.
Source Deviance Df p-Value
Model 86.4843 6 0.0000
Residual 5.6549 61 1.0000
Total (corr.) 92.1392 67
The analysis also shows that the percentage of deviance in lethal outcome explained
by the model equals 93.8627%. This statistic is similar to the usual R-Squared statis-
tic. The adjusted percentage, which is more suitable for comparing models with dif-
ferent numbers of independent variables, is 78.6683%.
The results of Likelihood ratio tests are described in Table 4. In determining
whether the model can be simplified, notice that the highest p-value for the likelihood
ratio tests is 0.0246, belonging to Transferrin saturation.
Table 4. Likelihood Ratio Tests.
Factor Chi-square Df p-Value
Transferrin saturation 5.05213 1 0.0246
Eosinophils 10.5325 1 0.0012
TNF-α 10.2318 1 0.0014
TIBC 15.9856 1 0.0001
FiO2 15.4042 1 0.0001
Albumin 22.3353 1 0.0000
Because the p-value is less than 0.05, that term is statistically significant at the 95%
confidence level. Consequently, there is no need to remove any variables from the
model.
6
The results of Chi-square goodness of fit test are illustrated in Table 5. This test de-
termines whether the logistic function adequately fits the observed data.
Table 5. Chi-Square Goodness of Fit Test
Class Logit interval N True False
Observed Expected Observed Expected
1 less than -10.2567 17 0.0 17.0 17.0
2 -10.2567 to 0.734184 24 1.0 1.69911 23.0 22.3009
3 0.734184 to 16.7516 26 26.0 25.2774 0.0 0.72258
4 16.7516 or greater 68 28.0 27.9766 40.0 40.0234
Total 135 55.0 80.0
Chi-squared was calculated as 1.05286 with 2 d.f. and p-value = 0.590709. Because
the p-value is greater than 0.10, there is no reason to reject the adequacy of the fitted
model at the higher than 90% confidence level.
Table 6 shows the estimated correlations between the coefficients in the fitted
model.
Table 6. Correlation matrix for coefficient estimates
Constant SatTrans Eosinoph TNF-α TIBC FiO2 Albumin
Constant 1.0 -0.536 -0.078 -0.302 -0.484 -0.634 -0.298
SatTrans -0.536 1.0 0.283 0.251 0.261 0.564 -0.447
Eosinoph -0.078 0.283 1.0 0.362 -0.467 0.482 -0.438
TNF-α -0.302 0.251 0.362 1.0 -0.319 0.422 -0.338
TIBC -0.484 0.261 -0.467 -0.319 1.0 -0.1 0.348
FiO2 -0.634 0.564 0.482 0.422 -0.1 1.0 -0.357
Albumin -0.298 -0.447 -0.438 -0.338 0.348 -0.357 1.0
SatTrans – Transferrin saturation; Eosinoph – percentage of eosinophils.
These correlations can be used to detect the presence of serious multicollinearity. In
our case, there is 1 correlation with the absolute value greater than 0.5 – in the case of
FiO2 with Transferrin saturation.
5 Discussion
In the present study, we proposed a useful tool for mortality risk stratification on the
1st and 2nd day after trauma for patients with combined thoracic injuries based on six
available variables. Once the probability of mortality after emergency surgery due to
trauma is established, the provided supportive care of vital function management can
be revised according to the risk degree. Frequently reoperations can be necessary for
definitive correction of the injured organs’ impaired functions. Adequate risk assess-
7
ment before the such second-stage surgery can be useful for planning the amount and
duration of operative treatment. The use of this score offers an effective clinical tool
for decision making in case of massive patients’ admission when intensive care must
be provided according to injury severity, patients’ survival ability and hospital facili-
ties.
Forward elimination of the variables into the logistic regression equation allows
increasing the quality of the regression model. On the other hand, it removes variables
that duplicate information as a part of redundant features [3]. So, we can hypothesize
that pathophysiological mechanisms, responsible for wound dystrophy progression,
which activity markers were defined most significant for outcome prediction, are
independent of each other. Besides, this independence is relative to how much it can
be in biological systems. These results can be interpreted as that there are several
relatively independent pathophysiological mechanisms involved in the progression of
the wound dystrophy in case of the severe combined thoracic trauma on the 1st and 2nd
day after injury. These findings are supported by data from the correlation matrix for
coefficient estimates (Table 6).
Logistic regression parameters with positive sides mean that pathophysiological
mechanisms that they reflect make harmful effects and are involved in the progression
of wound dystrophy and parameters with negative sides reflect the capacity of com-
pensatory mechanisms directed on stabilization of vital function disturbances and
maintenance of homeostasis in case of the severe combined thoracic trauma.
The likelihood ratio tests may help to understand the contribution degree of each
pathogenesis routs involved in the progression of wound dystrophy in case of the
severe combined thoracic trauma. It can be done according to chi-square values of the
logistic regression factors (Table 4). Mechanisms associated with blood loss, and
most of all, with a decrease of plasma albumin concentration are the most responsible
for the progression of the wound dystrophy on the 1 st and 2nd day of the early post-
traumatic period. The multifunctionality of this class of protein molecules (mainte-
nance of oncotic pressure, transport of hormones, drugs and inorganic ions, etc.),
obviously, determines the need for stable albumin concentration for aerobic metabo-
lism providing in case of traumatic shock.
Total iron binding capacity has the second importance degree according to chi-
square value. Sufficient transferrin concentration is necessary for binding and subse-
quent disposal of toxic iron released into the bloodstream after significant tissue dam-
age and parenchymal hemorrhages [16, 17]. The presence of the transferrin saturation
in the logistic regression equation underlines the importance of the damaging and
compensatory mechanisms counterbalance, involved in the pathogenesis of the shock
period in case of the severe combined thoracic trauma. On the one hand, hemolysis of
erythrocytes in hematomas occurs after closed traumatic injuries and increasing of
free toxic iron concentration in the blood produces a disturbance of antioxidant ho-
meostasis [18, 19]. At the same time, on the other hand, the compensatory mecha-
nisms involved in the processes of transport, deactivation and absorption of the iron
from the blood are depleted due to decrease of transferrin concentration in the blood
plasma as the result of blood loss, which always accompanies severe trauma [17, 20].
The current data highlight the importance of the critical loss of such valuable plasma
8
proteins as albumin and transferrin with bleeding that determines the body's future
ability of the damaging factors compensation associated with traumatic shock due to
combined thoracic trauma.
The respiratory failure is the third most important pathophysiological mechanism
involved in the progression of the severe combined thoracic injury wound dystrophy
on the 1st and 2nd day after trauma, determined by logistic regression. This finding can
be postulated according to the presence of FiO2 parameter in the logistic regression
equation with chi-square value 15.4 in Likelihood ratio test (Table 4). This parameter
reflects patients’ oxygen dependence occurring due to both ventilatory disorders
(damage to the rib cage, mechanical lung ventilation, presence of intrathoracic vol-
umes) and direct damage to the alveolocapillary membrane as the result of direct lung
contusion [21].
Furthermore, early activation of the immune system with the development of sys-
temic inflammatory response syndrome occurs [11, 22]. This pathophysiology mech-
anism was defined as the fourth degree importance of contributing to the progression
of wound dystrophy of this type of polytrauma. The presence in the equation of lo-
gistic regression the percentage of eosinophils, as well as the plasma concentration of
TNF-α suggests that disturbations of host defense processes are the next most im-
portant pathophysiological mechanisms of the shock period of the severe combined
thoracic trauma. In accordance with previous studies, the present results have demon-
strates that the release of proinflammatory cytokines, including TNF-α, occurs pro-
portionally to the volume of damaged tissues and cells of the body which leads to the
activation of nonspecific mechanisms of immune defense [11, 22], including the in-
crease in the number of blood eosinophils involved in the pathophysiological mecha-
nisms of systemic inflammatory response. Prior studies have noted that maximum
activation of hyperinflammation occurs within 3-4 days after trauma insult [11, 23–
25]. These accords with our observations, which showed less influence of immune
parameters in the regression equation on outcome prediction on the 1 st and 2nd day
after trauma.
Conclusions
The main goal of the current study was to create a prognostic tool for outcome predic-
tion for patients with combined thoracic trauma based on determination of main ho-
meostasis parameters on the 1st and 2nd day after injury. Multiple regression analysis
revealed that the outcome prediction can be fulfilled according to transferrin satura-
tion, percentage of eosinophils, TNF-a concentration, total iron binding capacity,
inspiratory fraction of oxygen and albumin concentration on the 1 st and 2nd of the
posttraumatic period in case of the severe combined thoracic trauma. The study con-
tributes to our understanding of the interaction between pathophysiological mecha-
nisms that make harmful effects and are involved in the progression of wound dystro-
phy and compensatory reactions directed on stabilization of vital function disturb-
ances and maintenance of homeostasis during this type of wound dystrophy. Multi-
9
variate logistic regression analysis with forward elimination of the variables allows
finding and analyzing detailed features of pathophysiological mechanisms.
References
1. Bewick, V., Cheek, L., Ball, J.: Statistics review 14: Logistic regression. Crit Care
9(1), 112–8 (2005).
2. Tepla, T., et al.: Alloys selection based on the supervised learning technique for design
of biocompatible medical materials. Arch Mater Sci Eng 1(93), 32–40 (2018).
3. Bowers, D.: Medical statistics from scratch: an introduction for health professionals.
Second edition. John Wiley & Sons LTD, Chichester (2008).
4. Krishtafor, А.А.: The structure of modern civilian multiple trauma at the stage of
providing tertiary care in the intensive care unit of regional hospital. Emergency
medicine 1, 96–100 (2019) Ukrainian.
5. Carlino, W.: Damage control resuscitation from major haemorrhage in polytrauma.
Eur J Orthop Surg Traumatol 24(2), 137–41 (2014).
6. Elbaih, A.H.: Patterns and management of chest injuries patients and its outcome in
Emergency Department in Suez Canal University Hospital, Egypt. Med Sci Int J. 6(2),
328–37 (2017).
7. Bayer, J., Lefering, R., Reinhardt, S., Kühle, J., Südkamp, N.P., Hammer, T., et al.:
Severity-dependent differences in early management of thoracic trauma in severely
injured patients - Analysis based on the TraumaRegister DGU ®. Scand J Trauma
Resusc Emerg Med 25(10), 1–10 (2017).
8. Señor, B.V., Puertas, A.CN., Polo, C.S., Civera, A.B., Pinilla M,AS., Virgós Señor,
B., et al.: Predictors of outcome in blunt chest trauma. Arch Bronconeumol 40(11),
489–94 (2004).
9. Dennis, B.M., Bellister, S.A., Guillamondegui, O.D.: Thoracic trauma. Surg Clin N
Am 97, 1047–64 (2017).
10. Bouzat, P., Raux, M., David, J.S., Tazarourte, K., Galinski, M., Desmettre, T., et al.:
Formalized Expert Recommendations Chest Trauma: First 48 hours management.
Anaesth Crit Care Pain Med 36(2), 135–45 (2017).
11. Keel, M., Trentz, O.: Pathophysiology of polytrauma. Injury 36(6), 691–709 (2005).
12. Dewar, D., Moore, F.A., Moore, E.E., Balogh, Z.: Postinjury multiple organ failure.
Injury 40, 912–8 (2009).
13. Gebhard, F., Huber-Lang, M.S.: Polytrauma — pathophysiology and management
principles. Langenbecks Arch Surg 393, 825–31 (2008).
14. Huber-Wagner, S., Stegmaier, J., Mathonia, P., Paffrath, T., Euler, E., Mutschler, W.,
et al.: The sequential trauma score – a new instrument for the sequential mortality
prediction in major trauma. Eur J Med Res 15(5), 185–95 (2010).
15. Abdel, M., Branscheid, D., Mertzlufft, F., Beshay, M.: Long term management of
thoracic trauma in a high frequency trauma center; what have we learned? J Egypt Soc
Cardio-Thoracic Surg 26(1), 73–81 (2018).
16. Wang, B., Yu, X., Wang, D., Qi, X., Wang, H., Yang, T., et al:. Alterations of trace
elements (Zn, Se, Cu, Fe) and related metalloenzymes in rabbit blood after severe
10
trauma. J trace Elem Med Biol organ Soc Miner Trace Elem 21(2), 102–7 (2007).
17. Yildirim, C., Kekec, Z., Sozuer, E., Ikizcel, I., Avsarogullar, L.: Blood levels of acute
phase reactants with traffic accidents. Soud Lek 49(2), 25–9 (2004).
18. Gorbunov, N.V., Asher, L.V., Ayyagari, V., Atkins, J.L.: Inflammatory leukocytes and
iron turnover in experimental hemorrhagic lung trauma. Exp Mol Pathol 80, 11–25
(2006).
19. Rathore, K.I., Kerr, B.J., Redensek, A., López-Vales, R., Jeong, S.Y., Ponka, P., et al.
Ceruloplasmin protects injured spinal cord from iron-mediated oxidative damage. J
Neurosci 28(48), 12736–47 (2008).
20. Dubick, M.A., Barr, J.L., Keen, C.L., Atkins, J.L.: Ceruloplasmin and hypoferremia:
studies in burn and non-burn trauma patients. Antioxidants 4, 153–69 (2015).
21. Perl, M., Hohmann, C., Denk, S., Kellermann, P., Lu, D., Braumüller, S., et al.: Role
of activated neutrophils in chest trauma-induced septic acute lung injury. Shock 38(1),
98–106 (2012).
22. Lenz, A., Franklin, G.A., Cheadle, W.G.: Systemic inflammation after trauma. Injury
38(12), 1336–45 (2007).
23. Neher, M.D., Weckbach, S., Flierl, M.A., Huber-lang, M.S., Stahel, P.F.: Molecular
mechanisms of inflammation and tissue injury after major trauma – is complement the
―bad guy‖? J Biomed Sci 18(1), 90 (2011).
24. Tsukamoto, T., Chanthaphavong, R.S., Pape, H-C.: Current theories on the
pathophysiology of multiple organ failure after trauma. Injury 41(1), 21–6 (2010).
25. Aller, M., Arias, J., Alonso-Poza, A., Arias, J.: A Review of metabolic staging in
severely injured patients. Scand J Trauma Resusc Emerg Med 18, 27 (2010).