THE VALIDATION OF NEW INJURY SEVERITY SCORE FOR SEVERE AND CRITICAL TRAUMA PATIENTS Текст научной статьи по специальности «Клиническая медицина»

DOI: 10.528 l/zenodo.3865986

Open 0Access

UDC: 616-001-07-037

The validation of New Injury Severity Score for severe and critical trauma patients

*1,2Oleg Arnaut

1Departmanet of Human Physiology and Biophysics 2Valeriu Ghereg Department of Anesthesiology and Intensive Care Nicolae Testemitanu State University of Medicine and Pharmacy, Chisinau, the Republic of Moldova

Author's ORCID iD, academic degrees and contributions are available at the end of the article

^Corresponding author: oleg.arnaut@usmf.md Manuscript received April 15, 2020; revised manuscript May 27, published online June 10, 2020

Abstract

Background: Considerable part of traumatized patients has criteria for severe/critical trauma. There is no international consensus concerning the most accurate traumatic scores. Their utilization in Moldova requires optimization. This article's goal is to validate the predictive ability of New Injury Severity Score (NISS) in sever/critical trauma patients admitted in Moldovan trauma center.

Material and methods: The retrospective cohort study, trauma patients (n=476) were admitted to Intensive Care Unit (ICU). The cohort was divided into severe (NISS > 15) and critical (NISS > 24) traumas. To achieve the aim, the multivariate logistic regression was used. The results were adjusted to gender, age and the mechanical ventilation use.

Results: Severe trauma model had an acceptable determination coefficient (Nagelkerke R Square=0.541). The calibration was poor (Hosmer-Lemeshow test, x2=17.430, df=8, p=0.026). The discrimination parameters, sensibility and specificity, were 85.9% and 85.1%. The determination coefficient for critical trauma model was 0.568, the calibration ability being normal ^=7.249, df=8, p=0.510). The sensibility and specificity were 70.9% and 94.7%, respectively. Conclusions: In this study, were proposed two mathematical models that validated NISS as an instrument to predict the outcomes in sever/critical trauma patients admitted in Moldovan trauma center. In general, the model's characteristics (determination, calibration and discrimination) could be appreciated as good ones with some limitations. Taking into account the advantages and disadvantages, both models could be recommended for daily practice usage in condition of ICU from Emergency Medicine Institute.

Key words: new injury severity score, severe trauma, critical trauma, score validation.

Cite this article

Arnaut O. The validation of New Injury Severity Score for severe and critical trauma patients. Mold Med J. 2020;63(2): 18-24. doi: 10.5281/zenodo.3865986.

Introduction

Trauma remains medical, social and economic problem being the main worldwide cause of death for the patients younger than 40 years [1]. In addition, while the death rate from oncological and cardiac diseases has favorable trend, the death rate from trauma is rising faster than the parallel population increase. This is a sign that, in the long perspectives, probably, the trauma will prevail in general lethality structure [2].

The trauma structure analysis has shown that considerable part of traumatized patients, almost 40%, has criteria for severe or critical trauma. These two categories of trauma require treatment in the intensive care unit (ICU) department considering the increased risk of complications and death. The trauma completes condition for severe if NISS (New Injury Severity Score) > 15, and when NISS > 24, the trauma is classified as critical [3, 4]. Taking into account the risks, it is crucial to estimate the trauma's lesions severity. Unfortunately, today, there is no international consensus on what kind of scores or algorithms of trauma assessment are optimal. As a consequence, different medical systems use different scores [5, 6], considering particularities of these

systems, of demographic structure or sometimes geographical factors [7].

The rational (optimal) utilization of the traumatic scores in the Republic of Moldova remains a serious problem. Because of that, at the patient's evaluation there are some disagreements on the prediction, different scores often estimating the outcomes completely different. According to the open resources at least, RTS (Revised Trauma Score), AIS (Abbreviated Injury Score) and ISS (Injury Severity Score) algorithms have been used, all these models having no any statistical validation [8]. The same statement is avai-lable for Acute Physiology, Age, Chronic Health Evaluation II score (APACHE II), used by the anesthesiologists to predict mortality rate for critical patients in ICU condition [9] and A Severity Characterization of Trauma (ASCOT) used for patients with associated trauma and Traumatic Brain Injury (TBI) lesions [10]. There are only two validated trauma scores. Both, the MPMoIII (Mortality Probability Admission Model) and NISS were tested for survival prediction in severe trauma patients (criteria being NISS > 15) transferred from regional hospitals to Emergency Medicine Institute (EMI). These models were used in a pilot research

with relatively reduced number of respondents and, of course, the accuracy of coefficient needs improvement, results being extrapolated on the specific category of trauma patients and as a result cannot be used for patients admitted directly in EMI [11].

The described problems need solution. This can be done using the strategy, described above [12]. It is based on three main elements - revision of potential models, their adaptation or alternative model/models ela-boration and, finally, comparative evaluation. The first element was performed and the main existing trauma scoring systems were revised to highlight the potential scoring systems that in perspective can be validated in the Moldovan medical system. Totally, we have analyzed 33 potential mo-dels divided into three main groups: anatomical, physiological and mixed.

In this list of scores for trauma patient evaluation, the ISS and NISS, mentioned above, derivate from AIS, are the most popular anatomical scores being widely used in different medical systems. To estimate ISS, we have to use the following formula: ISS = AIS12 + AIS/ + AIS32, where AIS1, AIS2, AIS3 are the highest AIS values present in six topographic regions [5]. NISS in comparison with ISS, estimates trauma severity taking into account three maximal values of AIS, without taking into account the lesion's localization [6]. For example, in case of trauma when AISabdomen = 2,

AISh,

c = 3, AISh,

= 3 and AISllmb5 = 5, the NISS

value will be higher (NISS=52 + 32 + 32 = 43) in comparison with ISS (ISS=52 + 32 + 22 = 38). Both, the ISS and NISS can vary from 0 up to 75. In condition if there is a topographical region with AIS = 6, ISS or NISS automatically is considered equal to 75. A meta-analysis shows similar sensibility and specificity of NISS and ISS mortality prediction in trauma patients [13]. We suppose that this result can be explained by insufficient determination coefficient in equations that use NISS or ISS. At the same time, there are data that the NISS is more precise, especially for patients with multiple injuries [14].

This article's goal is to validate the NISS as an instrument to predict the outcomes in sever/critical trauma patients admitted in Moldovan trauma center in order to use it for survival prediction. The validation has two elements - the confirmation of predictive power and adaptation of NISS regression coefficients for sever and critical trauma patients admitted in ICU, EMI of Chisinau, the Republic of Moldova.

Material and methods

The retrospective cohort study was performed. The injured patients (n = 467) were admitted in acute trauma period (the first 72 hours after impact) to ICU of EMI, Chisinau, the Republic of Moldova. The severity of lesions was appreciated using the NISS (AIS 2015), criteria for severe trauma being the value over the 15 points (n = 467 patients), the cut-off for critical trauma was considered the NISS value over the 24 points (n = 225 patients). Taking into account particularities related to the patients with mental disorders (psychoses) as the reason for admission to ICU, these ones were not considered in this research.

Statistical analysis

Descriptive statistics had the following components. The continuous data were represented using the central tendency indicators (mean, median) and dispersion parameters (standard deviation, interquartile range). To describe the dichotomous data have been used the frequencies and proportions. To analyze the precision of obtained data the 95% confidence intervals were estimated. To achieve the aim of this research, taking into account that the outcome (death/ survival) depends on a variety of factors and it represents a dichotomous variable, the multivariate logistic regression procedure was used. The results were adjusted to gender, age and the mechanical ventilation (MV) use. In order to match the regression analysis conditions the data was tested for multicollinearity. The potential model was characterized using determination (Nagelkerke R Square), calibration (Hosmer-Lemeshow test) and discrimination (sensibility, specificity, mean validation, ROC curve and classification graph, cut-off modification) parameters. In addition, the proposed models stability analysis was performed (resampling using bootstrapping). According to literature data, we expected better discrimination and calibration in critical vs severe trauma patients [4].

Results

According to the table 1 data, becomes evident the prevalence of males in both cohorts, the proportion being similar (70.2% for severe and 72% for critical trauma, 95% CI having approximately the same range, 65.9, 74.2 and 65.8, 77.5, respectively). Similar affirmation is valid for age variable (median 43 vs 44, interquartile range being 26 and 24, respectively for severe and critical trauma).

Intrahospital lethality was estimated at level 13.7% (CI 95% 10.9, 17.1) and 24.4% (CI 95% 19.3, 30.5), evidently predominating in critical trauma patients. It is important to mention that CI 95% are not large and there is no any intersection between them. This is a sign for significant differences in outcomes for these two cohorts and the lethality in critical trauma is twice higher (0R=2.04, CI 95% 1.36, 3.05).

The NISS variable, without any doubts, is significantly higher in critical trauma patients (mean=43, 95% CI 40.8, 45.2 for critical and mean=26.6, 95% CI 25.6, 27.6 for severe trauma patients). Regarding MV, it can be concluded that, in condition of critical trauma, the proportion of alternative approach is similar with traditional one that is not characteristic for severe trauma patients (39% vs 61%), (tab. 1).

The predictive model for primary outcome (death/ survival) in severe trauma

First, for the patients that meet criteria for severe trauma, were formulated the hypotheses as follows. The null hypothesis - the covariates included in the model (gender, age, NISS, MV) are not able to predict the death's probability in severe trauma patients better than a model that includes only the constant. The alternative hypothesis - at least one variable is able to predict the death's probability in severe trauma patients better than a model that includes only the constant.

Table 1

Descriptive statistics for severe (a) and critical (b) trauma patients

a. Severe trauma patients (NISS > 15)

Mean (Standard Deviation)/ 95% Confident interval Median

Count (%) Lower limit, Upper limit (Interquartile Range)

Gender Female 139 (29.8%) 25.8, 34

Male 328 (70.2%) 65.9, 74.2

Age, years 42 (17) 40.5, 43.5 42 (26)

NISS, points 26.6 (11.5) 25.6, 27.6 23 (15)

Mechanical Traditional 285 (61%) 56.5, 65.3

ventilation Prophylactical 182 (39%) 34.7, 43.5

Outcome Survival 403 (86.3%) 82.9, 89.1

Death 64 (13.7%) 10.9, 17.1

b. Critical trauma patients (NISS > 24)

Gender Female 63 (28%) 22.5, 34.2

Male 162 (72%) 65.8, 77.5

Age, years 43 (17) 40.8, 45.2 44 (24)

NISS, points 36 (10) 34.7, 37.3 34 (14)

Mechanical Traditional 117 (52%) 45.5, 58.5

ventilation Prophylactical 108 (48%) 41.6, 54.5

Outcome Survival 170 (75.6%) 69.5, 80.7

Death 55 (24.4%) 19.3, 30.5

The null hypothesis was rejected (Omnibus Test of Model Coefficients (^2 = 165.044, df=3, p<0.001, significance being 0.05/2 = 0.025) that means that at least one variable is able to predict the outcome, the elaborated model having the following characteristics. Determination coefficient, Nagelkerke R Square = 0.541 (54.1%). This result can be considered as acceptable one, but anyway it can be improved. The calibration ability was poor, Hosmer-Lemeshow test

being significant ^ = 17.430, df=8, p=0.026). The discrimination parameters, sensibility and specificity, according to the classification table, were 85.9% (55 dead patients out of 64) and 85.1% (343 survived patients out of 403), respectively, overall percentage being estimated at 85.2% (cut-off=0.13, fig. 1). Analysis of classification graph can reveal the overfitting problem.

Area under the Receiver Operating Characteristics

Observed Groups and Predicted Probabilities

160 + +

I I

I I

F I I

R 120 + +

E IS I

Q IS I

U IS I

E 80 +S +

N IS I

C ISS I

Y ISSS I

40 +SSS +

ISSSS I

ISSSSSS I

ISSSSSSSSSSSS D D I

Prob: 0 ,1 ,2 ,3 ,4 ,5 ,6 ,7 ,8 ,9 1

Group: SSSSSSSSSSSSSDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDD

Predicted Probability in the quantity of dead The Cut-off Value is ,13 Symbols: S - Survival

D - Dead Each Symbol Represents 10 Cases.

Fig. 1. Classificatioo graph forsevere trauma prediction model.

Table 2

Variables in equation of severe trauma. Initial model (a), final model (b) and after resampling (c)

a. Model before the Gender elimination

B S.E. Wald df Sig. Exp(B) 95% CI.for EXР(B)

Lоwer Uрper

Gender .062 .406 .023 1 .879 1.064 .480 2.357

Age, years .050 .012 18.441 1 .000 1.052 1.028 1.076

NISS, points .156 .018 76.982 1 .000 1.169 1.129 1.210

MV -1.335 .432 9.533 1 .002 .263 .113 .614

Constant -8.807 1.021 74.443 1 .000 .000

Vаriаble(s) entered оn steр 1: NISS, Age, MV, Gender b. Model after the Gender elimination (final model)

Age, years .050 .012 18.677 1 .000 1.051 1.028 1.075

NISS, points .156 .018 77.132 1 .000 1.169 1.129 1.210

MV -1.328 .430 9.549 1 .002 .265 .114 .615

Constant -8.755 .960 83.249 1 .000 .000

Vаriаble(s) entered оn steр 2: NISS, Age, MV

c. Bootstrap for Variables in the Equation

B Bias S.E. Sig. 95% Confidence Interval for B

Lоwer Uрper

Age, years .050 .002 .012 .001 .029 .078

NISS, points .156 .005 .021 .001 .124 .206

MV -1.328 -.056 .488 .004 -2.440 -.489

Constant -8.755 -.278 1.038 .001 -11.364 -7.226

Unless otherwise noted, bootstrap results are based on 1000 bootstrap samples

Taking into account the data from table 2, the proposed (final) predictive model has the following formula:

p = 1/(1+e8775 - NISS * a156 - Age * a05 + MV * L335) (formula 1), where

p - probability of death in severe trauma e (exponent) - constant equal to 2.71828

The final model includes the constant (B = -8.755), NISS (B = 0.156), Age (B =0.050) and MV (B = -.1.335) as efficient variables. Gender did not achieve the significance (p=.879). Moreover, after the elimination of this variable (tab. 2a and tab. 2b), the regression's coefficient, the Odds Ratio (0R) and the confidence intervals were without significant changes.

According to the obtained data, the most important co-variate was NISS (determination coefficient .424), followed by Age (approximately 9%) and MV (3%). The quantitative expression of positive relationships of NISS and Age on the one hand and the primary outcome on the other hand (positive regression coefficients and OR more than 1, confident intervals being relatively narrow) allows to modulate different situations as follows. For example, if in ICU, we meet two similar patients according to NISS and MV criteria, having only one year age difference, the death probability is 5% higher (95% CI 2.8. 7.5) for the older patient. In the same way, the only one NISS point difference can increase

21)) i

(ROC) curve for severe trauma model was estimated at 0.907 (95%CI 0.862, 0.952) being significant compared with 0.5 value (p<0.001), (fig. 2).

ROC Curve

1,0-1-p-7

Fig. 2. RОС curve for severe trauma patients mortality predictive model.

Observed Groups and Predicted Probabilities

32 + +

I I

I I

F I I

R 24 + +

E I S I

Q I SS I

U IDSS I

E 16 +SSS +

N ISSS D I

C ISSSSS I

Y ISSSSS I

8 +SSSSSSSSD +

ISSSSSSSSSS D I

ISSSSSSSSSSSD SDS S DD S D D D D I ISSSSSSSSSSSSSSSSSSSSSSSS SS SSSSS S S SSS SD SSDD S DS S SDD S D D D S D DSDD D DDDDDDS D DI

Prob: 0 ,1 ,2 ,3 ,4 ,5 ,6 ,7 ,8 ,9 1

Group: SSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDD

Predicted Probability in the quantity of dead The Cut-off Value is ,47 Symbols: S - Survival

D - Dead Each Symbol Represents 10 Cases.

Fig. 3. Classification graph for critical trauma prediction model.

the probability of death by 16.9% (95% CI 12.9, 21). were significant, regression coefficients having narrow 95%

The resampling (bootstrapping, 1000 samples) was used confidence intervals, these two conditions serving as argu-to prove the elaborated model's stability. All parameters ments to consider this model as a stable one (tab. 2c).

Table 3

Variables in equation for critical trauma. Initial model (a), final model (b) and after resampling (c) a. Model before the Gender elimination

B S.E. Wald df Sig. Exp(B) 95% C.I.for EXP(B)

Lower Upper

Gender -.065 .483 .018 1 .893 .937 .364 2.414

Age, years .045 .014 10.526 1 .001 1.046 1.018 1.075

NISS, points .180 .026 48.722 1 .000 1.197 1.138 1.259

MV -1.448 .487 8.828 1 .003 .235 .090 .611

Constant -9.517 1.420 44.940 1 .000 .000

Variable(s) entered on step 1: NISS, Age, MV, Gender

b. Model after the Gender elimination (final model)

Age, years .046 .014 10.700 1 .001 1.047 1.018 1.075

NISS, points .180 .026 48.667 1 .000 1.197 1.138 1.260

MV -1.457 .484 9.068 1 .003 .233 .090 .601

Constant -9.572 1.363 49.321 1 .000 .000

Variable(s) entered on step 2: NISS, Age, MV

d. Bootstrap for Variables in the Equation

B Bias S.E. Sig. 95% Confidence Interval for B

Lower Upper

Age, years .046 .001 .015 .002 .019 .082

NISS, points .180 .009 .032 .001 .134 .258

MV -1.457 -.094 .565 .003 -2.710 -.553

Constant -9.572 -.403 1.552 .001 -13.605 -7.578

Unless otherwise noted, bootstrap results are based on 1000 bootstrap samples

The predictive model for primary outcome (death/ survival) in critical trauma

As in condition of severe trauma, for critical trauma patients, were formulated the following hypotheses. The null hypothesis - the covariates included in the model (gender, age, NISS, MV) are not able to predict the death's probability in critical trauma patients better than a model that includes only the constant. The alternative hypothesis - at least one variable is able to predict the death's probability in critical trauma patients better than a model that includes only the constant.

The null hypothesis was rejected (Omnibus Test of Model Coefficients (x2 = 107.889, df = 3, p<0.001, significance being 0.05/2 = 0.025) that means that at least one variable is able to predict the probability of death in critical trauma patients. Determination coefficient, Nagelkerke R Square, was estimated at 0.568 (56.8%) vs 0.541 (54.1%) in severe trauma. This result can be considered as acceptable but needs improvement. The calibration ability (fidelity results evaluation) was normal, Hosmer-Lemeshow test being nonsignificant (x2 = 7.249, df = 8, p = 0.510). The discrimination parameters, sensibility and specificity, according to the classification table were 70.9% (39 dead patients out of 55) and 94.7% (161 survived patients out of 170), respectively, overall percentage being estimated at 88.9% (cut-off = 0.47, fig. 3).

Area under the ROC curve for critical trauma model was estimated at 0.905 (95% CI 0.856, 0.954) being significant vs 0.5 value (p<0.001), (fig. 4).

Taking into account the data from table 3, the proposed (final) predictive model for death prediction in critical trauma patient can be converted into the following formula:

p = 1/(1+e9572 - NISS * a180 - Age * a46 + MV * L457) (formula 2), where

p - probability of death in sever trauma

e (exponenta) - constant equal to 2.71828

ROC Curve

1 - Specificity

Fig. 4. RОС curve for critical trauma patients mortality predictive model.

The final model includes the constant (B = -9.572), NISS (B = 0.180), Age (B = 0.046) and MV (B = -.1.448) as sig-

i (¡23

nificant variables, Gender being far from significance range (p = 0.893) without any effects on the regression coefficients, the OR or the confidence intervals (tab. 3a and tab. 3b) as it was mentioned in severe trauma model.

The most powerful covariate (NISS) was able to determine 43.6% of variable dispersion, a result similar with the determination coefficient values of NISS in severe trauma. The effect of Age was estimated at level of 9%, MV having 4.3% instead of 3%.

The resampling (bootstrapping) has shown stability; coefficients having significance and narrow 95% confidence intervals (tab. 3c).

Discussion

In this study, were proposed two mathematical models that validated NISS as an instrument to predict the outcomes in sever/critical trauma patients admitted in Moldovan trauma center in order to use it for survival prediction. The validation of trauma scoring scales for different medical systems prior to their practical use represents an efficient instrument with a worldwide use, net result of this process being the regression coefficients correction and optimization [3, 14, 15, 16].

The predictive potential of NISS was proven for both, critical and severe trauma patients. In addition, NISS regression coefficient was adapted to particularities of Moldovan medical system and adjusted to Age and MV strategy. The Gender variable was not significant in both trauma groups, being independent in relation to other variables. The NISS and Age covariates have shown negative effects, having relatively narrow confidence intervals. The MV has had the opposite effect with relatively large confidence intervals.

Taking into account the model's characteristics, logistic regression classified them as significantly better models than aleatory ones. In general, the determination, calibration and discrimination model's parameters could be appreciated as good ones with some following limitations.

Firstly, the research is retrospective that is why there could not be considered all risk factors and potential bio-markers that reduce the evidence level. The perspective group of potential predictors consists of components of proteases/antiproteases system. Among them are cathep-sin D and a2-macroglobulin that have shown their potential to predict the survival rate and Acute Respiratory Distress Syndrome in polytrauma patients [17, 18]. Secondly, evidently, it is not possible to extrapolate the results for all Moldovan medical system hospitals. Thirdly, both models explained less than 60% of dependent variable dispersion, optimal value being over 80%. This fact makes us think that they should be added some other efficient variables that could improve the accuracy of prediction. Fourthly, the poor calibration and potential overfitting for severe trauma model could demand precaution in practical use, other parameters being close to the optimal value [4].

In addition, it is important to mention the reduced NISS determination coefficient values for severe trauma (.424) and critical trauma (.436) vs severe trauma patients trans-

ported from regional hospitals (.641) [11]. These differences, about .2 (20%) from dispersion of dependent variable, can be considered as a benefit for patients admitted directly in trauma center because of conversion of the nonchange-able factor (trauma injuries) into potentially changeable. Increasing the ratio of the last one gives us more opportunities to influence the outcomes in order to improve them.

Conclusions

Taking into account the advantages and listed above disadvantages, both models could be recommended for daily practice usage in ICU. The arguments against these models are that they were elaborated for this particular trauma group and do not have alternative validated scores. Also, being incorporated in hospital's informational system the models could be improved in real time by adding the potential efficient variables. In addition, the model's accuracy could be raised by including new severe and critical trauma cases data.

The obtained results determine us to continue the researches in this field. Taking into consideration the experience of this study, the perspective for completely elaborated models is opening, to validate other potential models and to elaborate new scores able to predict the trauma patient's outcomes.

References

1. Spahn DR, Bouillon B, Cerny V, et al. The European guideline on management of major bleeding and coagulopathy following trauma: fifth edition. Crit Care. 2019;23(1):98. doi:10.1186/s13054-019-2347-3.

2. Kunitake RC, Kornblith LZ, Cohen MJ, Callcut RA. Trauma Early Mortality Prediction Tool (TEMPT) for assessing 28-day mortality. Trauma Surg Acute Care Open. 2018;3(1):e000131. doi: 10.1136/ tsaco-2017-000131.

3. Skaga NO, Eken T, S0vik S. Validating performance ofTRISS, TARN and NORMIT survival prediction models in a Norwegian trauma population. Acta Anaesthesiol Scand. 2018;62(2):253-266. doi: 10.1111/aas.13029.

4. Shi J1, Shen J, Caupp S, et al. A new weighted injury severity scoring system. J Trauma Acute Care Surg. 2018;85(2):334-340. doi:10.1186/ s13054-016-1208-6.

5. Lefering R. Trauma scoring systems. Curr Opin Crit Care. 2012;18(6):637-640. doi: 10.1097/MCC.0b013e3283585356.

6. Association for the Advancement of Automotive Medicine (AAAM). ISS, NISS and MAIS Mapping with AAAM's ICD ISS Map [Internet]. Chicago: AAAM; c2016-2020 [cited 12 Dec 2019]. Available from: https://www.aaam.org/abbreviated-injury-scale-ais/ais-icd-iss-map/

7. Rutledge R. The Injury Severity Score is unable to differentiate between poor care and severe injury. J Trauma. 1996;40(6):944-50. doi: 10.1097/00005373-199606000-00013.

8. Pascari V. Evaluarea intraspitaliceasca si unele actiuni terapeutice urgente la pacientii cu fracturi multiple de locomotor. [Interhospital evaluation of urgent therapeutic actions in patients with multiple fractures]. Curierul Medical (Chisinau). 2009;(4/310):23-29. Romanian.

9. Tazlavan T. Delirul in terapia intensiva. [Delirium in intensive care]. Curierul Medical (Chisinau). 2009;(6/324):72-74. Romanian.

10. Cojocari V, Ciobanu G, Scurtov N. [The impact of pre-hospital arterial hypo- and hypertension on clinical severity and prognosis of patients with traumatic brain injury]. Curierul Medical (Chisinau). 2013;56(3):29-33. Romanian.

11. Arnaut O. Survival predictive models in severe trauma patients transportation within Moldovan medical system. Mold Med J. 2019;62(4):39-44. doi: 10.5281/zenodo.3556496.

12. Arnaut O, Croitoru D, Grabovschi I, Sandru S. Trauma scoring systems. Mold Med J. 2020;63(1):64-74. doi: 10.5281/zenodo.3685673.

13. Deng Q, Tang B, Xue C, et al. Comparison of the Ability to Predict Mortality between the Injury Severity Score and the New Injury Severity Score: a meta-analysis. Int J Environ Res Public Health. 2016;13(8). pii: E825. doi:10.3390/ijerph13080825.

14. Whitaker IY, Gennari TD, Whitaker AL. The difference between ISS and NISS in a series of trauma patients in Brazil. Annu Proc Assoc Adv Automot Med. 2003;47:301-9. PMID: 12941232; PMCID: PMC3217565.

15. Wong TH, Krishnaswamy G, Nadkarni NV, et al. Combining the new injury severity score with an anatomical polytrauma injury variable predicts mortality better than the new injury severity score and the injury severity score: a retrospective cohort study. Scand J Trauma Resusc Emerg Med. 2016;24:25. doi:10.1186/s13049-016-0215-6.

16. Tan JH, Tan HCL, Noh NAM, Mohamad Y, Alwi RI. Validation of the trauma mortality prediction scores from a Malaysian population. Burns Trauma. 2017;5:37. doi:10.1186/s41038-017-0102-z.

17. Arnaut O, Baltaga R, Rojnoveanu G, Sandru S. Is a2-macroglobulin a new biomarker for ARDS occurrence in polytrauma patients? Eur J Trauma Emerg Surg. 2018;44(Suppl 2):S321. ISSN 1863-9933. doi: 10.1007/s00068-018-0934-7

18. Arnaut O, Sandru S, Saulea A, Grabovschi I, Rojnoveanu G. The Ca-thepsin D as a potential biomarker for survival rate in polytrauma. Pilot research. In: Tiginyanu I, Sontea V, Railean S, editors. Proceedings of the 4th International Conference on Nanotechnologies and Biomedical Engineering; 2019 Sep 18-21; Chisinau, Moldova. Cham: Springer; 2020. p. 703-705. (IFMBE Proceedings; vol. 77, Series Online ISSN 144339277). https://doi.org/10.1007/978-3-030-31866-6.

Author's ORCID iD and academic degree

Oleg Arnaut, MD, PhD, Associate Professor of Physiology - https://orcid.org/0000-0002-5483-8672. Author's contributions

OA conceptualized the idea, conducted literature review, wrote the manuscript, revised and approved the final text. Funding

This study was supported by Nicolae Testemitanu State University of Medicine and Pharmacy. The trial was the author's initiative. The author is independent and takes responsibility for the integrity of the data and accuracy of the data analysis.

Ethics approval and consent to participate

The research project was approved by Ethics Committee of Nicolae Testemitanu State University of Medicine and Pharmacy (Protocol No 46, 16.12.2016).

Conflict of Interests

No competing interests were disclosed.

i (SS)- i