Nomogram-Based Prediction of Post-Traumatic Occult Shock: Development and Evaluation of Novel Predictive Scores

doi:10.21203/rs.3.rs-29654/v1

Download PDF

Research article

Nomogram-Based Prediction of Post-Traumatic Occult Shock: Development and Evaluation of Novel Predictive Scores

https://doi.org/10.21203/rs.3.rs-29654/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Background

Shock is a common severe complication of trauma and can result in multiple organ dysfunction and even death. To improve prognosis, it is crucial that clinicians intervene and treat occult shock, which shows no obvious changes in vital signs. The purpose of this study was to develop a new, nomogram-based shock prediction score model that is independent of the trauma injury severity score, abbreviated injury scale, and Glasgow coma scale. Our shock prediction score should help clinicians diagnose post-traumatic occult shock (PTOS) early and intervene promptly to improve patient outcomes.

Patients and methods

This study recruited 1005 injured patients (431 with shock and 574 without shock) from the emergency department of the Affiliated Hospital of Guizhou Medical University. Based on least absolute shrinkage and selection operator technology, 31 commonly used indicators were screened for influence on traumatic shock. We used multivariate logistic regression analysis to screen each independent influencing factor (p < 0.05), construct a nomogram-based prediction model, and formulate a PTOS score. We also estimated the predictive power of the model using C-index and the area under the receiver operating characteristic curve (AUC) of PTOS. The stability and net benefit range of this model were evaluated using a calibration diagram, internal grouping data verification, and decision curve analysis.

Results

Six individual influencing factors were selected using multivariate logistic regression to construct the nomogram-based model (systolic pressure, diastolic pressure, white blood cell count, absolute basophil count, red blood cell count, and hematocrit). The incidence of traumatic shock increased with an increase in the PTOS score. The C-index of the PTOS score in the training dataset was 0.92 (95% confidence interval [CI]: 0.90– 0.94), and the AUC was 0.92. The C-index of the validation dataset was 0.90 (95% CI: 0.86–0.94), and the AUC value was 0.90. The calibration chart showed that our model had excellent prediction. Decision curve analysis showed that the shock nomogram predictive score was clinically useful when the intervention was decided at a shock probability threshold more significant than 2%.

Conclusion

Using clinical electronic medical record system data, we developed and validated a new PTOS score with high accuracy, stability, and calibration to predict the risk of shock in patients with trauma.

Surgery

Trauma

Post-traumatic occult shock

Nomogram-based prediction

Shock often occurs as a severe complication of trauma,1 with Anderson et al. reporting that traumatic shock is the fifth most common shock type after low volume shock, carcinogenicity shock, obstructive shock, and distributed shock.2,3 Violence or accidental trauma causes 12% of global diseases, and post-traumatic occult shock (PTOS) is the leading cause of early death.4,5 Furthermore, those who survive the initial ordeal often suffer severe sequelae due to the first treatment, and the long-term mortality rate also increases significantly, which brings a substantial economic burden on the society.6 The prognosis of patients with shock is improved by early fluid replacement and blood transfusion. Earlier treatment is more effective, even in patients with high risk of collapse. Therefore, intervention measures should be adopted to prevent high-risk patients with trauma from developing shock.9-11 At present, the shock index is the most common tool used in the early clinical judgment of shock. It can determine the presence of shock and the severity of trauma, providing a simple device for screening patients with potential shock. However, the shock index can only be used for qualitative judgment of shock. It can make a quantitative judgment when marked changes in vital signs occur but cannot judge patients with occult shock whose vital signs are not evident.7,8 Therefore, researchers must devise a method to predict the probability of collapse in patients with trauma so that clinicians can provide targeted prevention and treatment to high-risk patients. This would reduce the delayed diagnosis of traumatic shock and improve the efficacy of clinical trauma treatment.12

Study populations

This study was conducted at the Affiliated Hospital of Guizhou Medical University (national third-grade class-A hospital), and patients with trauma were registered between January 2014 and December 2019. All enrolled patients with a history of trauma met two of the following diagnostic criteria for traumatic shock: (1) systolic pressure < 13.3 kPa (100 mmHg), (2) pulse pressure < 4 kPa (30 mmHg), (3) “5P” symptoms (pallor, perspiration, prostration, pulselessness, and shortness of breath), (4) urine volume < 25 mL/h, (5) blood lactate level > 2 mmol/L, (6) heart index < 2.5 L/min/m². In addition, the following inclusion criteria were implemented: (1) age ≥ 18 years, (2) < 24 hours between injury and admission, (3) status of emergency or referral patients, (4) consent to participate from the patient themselves or their legally authorized representatives. The exclusion criteria were as follows: (1) admission to the hospital for emergency blood tests, anti-shock treatments such as blood transfusion, or massive fluid replacement; (2) underlying diseases such as hematological diseases, malignant tumors, or bacterial or viral infection; (3) pregnancy; (4) history of oral hormonal drug use.

Variables adopted for analysis and imputation

The demographic data of the enrolled patients included gender and age. Vital signs included temperature, pulse rate, breathing rate, systolic blood pressure, and diastolic blood pressure. Laboratory variables included routine blood examination: white blood cell count, neutrophil percentage, lymphocyte percentage, monocyte percentage, eosinophil percentage, basophil percentage, absolute neutrophil value, absolute lymphocyte count, absolute monocyte count, absolute eosinophil count, absolute basophil count, red blood cell count, hemoglobin level, hematocrit, average red blood cell volume, average red blood cell hemoglobin content, average red blood cell hemoglobin concentration, red blood cell volume distribution width, platelet count, average platelet volume, platelet compaction, and platelet distribution width. Other variables included type of injury and injury duration. All variables were collected within 24 hours of patient admission. The injury time of the included patients was 0.5–20 h, with a median time of 4 h. None of the inclusion criteria were missing in any case. The institutional ethics review committee of Guizhou Medical University approved the study.

Statistical analysis

The independent-sample t-test was used to compare continuous variables and the chi-square test was used to compare categorical variables between patients with and without traumatic shock. Complete data sets were randomly separated: 705 patients (70%) in the training dataset and 300 (30%) in the validation dataset. To select between independent predictors and predictive models, all predictors were first analyzed based on the least absolute shrinkage and selection operator (LASSO) technology. Factors that may influence traumatic shock were identified, and multi-factor logistic regression models were used to analyze the effect of these suspected influencing factors on the shock risk score. To ascertain the odds ratio (OR) and p-value of each potential influencing factor, each variable (p < 0.05) was treated as an independent predictor. The predictors were then combined to construct a shock model. To construct a nomogram-based prediction chart, the predictive factor was converted into the predicted shock score, and the corresponding risk probability of shock occurrence was obtained from the nomogram chart according to the sum of the ratings. After receiving the score, the C-index and area under the curve (AUC) were calculated to estimate the predictive ability of the discriminant model (< 0.7: low predictive power, 0.7–0.9: medium predictive power, and > 0.9: high predictive power). The model was calibrated using the calibration chart. Decision curve analysis of the nomogram was used to ascertain the relationship between the net benefit of the model and the threshold probability. To confirm that the multi-factor model had superior predictive ability, we calculated the predictive ability of each independent predictive factor using a single-factor logistic regression model. All statistical analyses were performed using R software (version 3.6.2), and p-values < 0.05 were considered statistically significant.

Model development

To better promote clinical practice, we chose to use in the clinical familiar, comfortable to get 31 indicators as a model of the development of potential predictors (Table 1) and use the 31 predictors for subsequent research. Using the LASSO technology, we carried out preliminary screening of factors that may influence shock (Fig. 1). Thereafter, we subjected these factors to multivariate logistic regression analysis to derive the regression coefficient, OR, and p-value. Ultimately, six factors (p < 0.05) were selected as independent predictors: systolic pressure, diastolic pressure, white blood cell count, absolute basophil count, red blood cell count, and hematocrit (Table 2). These were combined to construct the nomogram-based prediction model (Fig. 2). Moreover, to allow convenient clinical application, we used the quartile method to grade the shock score, as follows: low-risk group (score < 98), intermediate-risk group (score: 98–129), high-risk group (score: 129-159), and very high-risk group (score > 159). The shock incidence in the low-risk group was < 30%, whereas those in the intermediate-risk, high-risk, and very high-risk groups were 30%– 50%, 50%–70%, and > 70%, respectively. As the shock score increased, so gradually did the frequency of shock. The impact of the shock prediction model C-index was 0.92 (95% confidence interval [CI]: 0.90–0.94), with an AUC of 0.92. The model had a highly accurate predictive power (Fig. 3A). As demonstrated in the calibration diagram analysis, the predictive model for traumatic shock performed reasonably well (Fig. 4), and decision curve analysis showed that shock nomogram-based predictive scores were clinically useful when interventions were decided at a shock probability threshold more significant than 2% (Fig. 5).13 In terms of C-index and AUC values, the predictive ability of our shock score was significantly better than that of any single independent predictor (Table 3)

Table 1

Baseline characteristics of trauma patients
Variable factor	n(%)
Variable factor	Shock（431）	Non-shock（574）	Total（1005）
Sex
female	92(21.3)	131(22.8)	223(22.2)
male	339(78.7)	443(77.2)	782(77.8)
Type of injury
Traffic accident injury	146(33.9)	190(33.1)	336(33.4)
Sharp injury	123(28.5)	107(18.7)	230(22.9)
High fall injury	70(16.2)	77(13.4)	147(14.6)
Fall injury	18(4.2)	85(14.8)	103(10.3)
Other injuries	74(17.2)	115(20.0)	189(18.8)
Age
<41	247(57.3)	281(49.0)	528(52.5)
>=41	184(42.7)	293(51.0)	477(47.5)
Pulse
<60	2(0.5)	1(0.2)	3(0.3)
60-100	157(36.4)	512(89.2)	669(66.6)
>=100	272(63.1)	61(10.6)	333(33.1)
Systolic pressure
<=90	191(44.3)	15(2.6)	206(20.5)
90-120	189(43.9)	262(45.6)	451(44.9)
>=120	51(11.8)	297(51.8)	348(34.6)
Diastolic pressure
<=60	229(53.1)	30(5.2)	259(25.8)
60-90	193(44.8)	519(90.4)	712(70.8)
>90	9(2.1)	25(4.4)	34(3.4)
Wbc count
<13.81	117(27.1)	386(67.2)	503(50.0)
13.81-20	147(34.1)	154(26.9)	301(30.0)
>=20	167(38.8)	34(5.9)	201(20.0)
Lymphocyte percentage
<8	225(52.2)	272(47.4)	497(49.5)
>=8	206(47.8)	302(52.6)	508(50.5)
Absolute value of neutrophils
<11.84	120(27.8)	381(66.4)	501(49.9)
>=11.84	311(72.2)	193(33.6)	504(50.1)
Absolute value of basophils
<0.06	361(83.8)	556(96.9)	917(91.2)
>=0.06	70(16.2)	18(3.1)	88(8.8)
RBC count
<4.28	339(78.7)	160(27.9)	499(44.7)
>=4.28	92(21.3)	414(72.1)	506(55.3)
Hematocrit
<40	343(79.6)	161(28.1)	504(50.1)
>=40	88(20.4)	413(71.9)	501(49.9)
Platelet distribution width
<16.3	210(48.7)	276(48.1)	486(48.4)
>=16.3	221(51.3)	298(51.9)	519(51.6)
Time of trauma
<4	199(46.2)	245(42.7)	444(44.2)
>=4	232(53.8)	329(57.3)	561(55.8)
Temperature
<37	375(87.0)	503(87.6)	878(87.4)
>=37	56(13.0)	71(12.4)	127(12.6)
Breathing
<=20	210(48.7)	441(76.8)	651(64.8)
>20	221(51.3)	133(23.2)	354(35.2)
Percentage of neutrophils
<86.4	204(47.3)	298(51.9)	502(50.0)
>=86.4	227(52.7)	276(48.1)	503(50.0)
Monocyte percentage
<4.9	215(49.9)	285(49.7)	500(49.8)
>=4.9	216(50.1)	289(50.3)	505(50.2)
Percentage of eosinophils
<0.1	128(29.7)	177(30.8)	305(30.3)
>=0.1	303(70.3)	397(69.2)	700(69.7)
Percentage of basophils
<0.2	209(48.5)	211(36.8)	420(41.8)
>=0.2	222(51.5)	363(63.2)	585(58.2)
Lymphocytes absolute value
<1.15	169(39.2)	322(56.1)	491(48.9)
>=1.15	262(60.8)	252(43.9)	514(51.1)
Monocytes absolute value
<0.65	138(32.0)	355(61.8)	493(49.1)
>=0.65	293(68.0)	219(38.2)	512(50.9)
Absolute value of eosinophils
<0.02	182(42.2)	294(51.2)	476(47.4)
>=0.02	249(57.8)	280(48.8)	529(52.6)
Hb
<131	333(77.3)	158(27.5)	491(48.9)
>=131	98(22.7)	416(72.5)	514(51.1)
Mean corpuscular volume
<92.9	186(43.2)	313(54.5)	499(44.7)
>=92.9	245(56.8)	261(45.5)	506(55.3)
Mean corpuscular hemoglobin content
<30.8	193(44.8)	305(53.1)	498(49.6)
>=30.8	238(55.2)	269(46.9)	507(50.4)
Mean corpuscular hemoglobin concentration
<330	235(54.5)	258(44.9)	493(49.1)
>=330	196(45.5)	316(55.1)	512(50.9)
Red blood cell volume division width
<12.8	228(52.9)	249(43.4)	477(47.5)
>=12.8	203(47.1)	325(56.6)	528(52.5)
Platelet count
<195	233(54.1)	268(46.7)	501(49.9)
>=195	198(45.9)	306(53.3)	504(50.1)
Mean platelet volume
<10.4	209(48.5)	285(49.7)	494(49.2)
>=10.4	222(51.5)	289(50.3)	511(50.8)
*Abbreviations: Wbc,white blood cell;Rbc,red blood cell;Hb,hemoglobin.

Table 2

Prediction factors for shock in trauma
Intercept and variable	Prediction model
	Odds ratio (95% CI)	p-value
Systolic pressure >90	0.23(0.08 to 0.63)	0.005	**
Diastolic pressure 60-90	0.24(0.09 to 0.55)	0.001	**
Wbc>=20	8.15(1.74 to 41.3)	0.009	**
Absolute value of basophils>=0.06	5.62(2.05 to 16.0)	<0.001	***
RBC count>=4.28	0.37(0.17 to 0.79)	0.010	*
Hematocrit>=40	0.17(0.08 to 0.36)	<0.001	***

Signif.codes: 0‘* * ’0.001‘ ’0.01‘’

Table 3

Independent influencing factors and PTOS score predictive power comparison chart
Independent influencing factorsand PTOS score	C-index(95%CI)	The area under curve(AUC)

Systolic pressure	0.70(0.67 to 0.73)	0.70
Diastolic pressure	0.74(0.71 to 0.77)	0.74
Wbc	0.75(0.72 to 0.78)	0.75
Absolute value of basophils	0.57(0.55 to 0.59)	0.57
RBC count	0.76(0.73 to 0.79)	0.76
Hematocrit	0.76(0.73 to 0.79)	0.76
PTOS score	0.92(0.90 to 0.94)	0.92

Internal validation of the shock score

To verify the accuracy of the PTOS score, we selected 300 cases (30%) of the validation dataset for internal grouping verification of the scoring model. We found that the incidence of shock increased with an increase in the PTOS score. Moreover, the C-index of the validation dataset was 0.90 (95% CI: 0.86–0.94) and that the AUC was 0.90, indicating excellent predictive performance (Fig. 3B). Calibration diagram analysis also indicated that our PTOS score had considerable calibration capability in the validation dataset.

Shock is a common complication of severe trauma that increases the mortality rate and incidence of multiple organ dysfunction.1 As such, shock severely affects the prognosis of patients with trauma and it should be identified and treated early.12 However, because it lacks evidential external clinical manifestations, PTOS is easy to overlook or misdiagnose. This leads to delayed diagnosis and treatment, with serious consequences. Therefore, it is crucial that researchers develop a useful tool for scoring PTOS.14 The scoring tool constructed in the present study provided an accurate and reliable scoring method for clinically predicting susceptibility to shock. As such, the tool improved diagnostic accuracy in patients with trauma and markedly reduced diagnostic failures or delays in patients with PTOS. This study did not rebut the usefulness of the shock index in clinical work. Instead, it further improved and supplemented the early diagnosis of traumatic shock. All the variables contributing to PTOS score were screened from commonly used clinical indicators, so the score could be useful in clinical practice. Among the independent variables included in the present study, laboratory indices were only selected from a routine blood test, which is the most basic item of clinical laboratory examination, entailing a simple and rapid examination process. The results can be obtained within 15 minutes of blood collection in the emergency department of most hospitals. As such, these values allow a quick, accurate, and simple scoring prediction tool to be constructed. Although shock markers and end organ injury markers (i.e., lactate, base excess, international normalized ratio, and creatinine) are important in the diagnosis of shock, they were excluded from the present research because they require complex test processes and have slow result turnover times.

It is not difficult to diagnose patients with prominent shock, but many patients with PTOS show no noticeable changes in symptoms or signs at first. Therefore, the condition can be overlooked in the rush of emergency work. However, as PTOS develops, it worsens. If it is not identified as early as possible, it can develop into more severe shock, which is extremely adverse to prognosis and increases the burden of medical treatment.15,16 Our shock scoring tool is a prediction method that considers the first physical signs in patients and laboratory test indicators. The preliminary prediction of shock must be timely, objective, comprehensive, and accurate to compensate for the lack of subjective shock predictors.17 To ensure that the clinical data analysis was accurate and reduced the impact of different time nodes on the results, all indicators in the present study were collected within 24 hours of injury. Our current research had some limitations. First, the data analyzed were all sourced from the electronic medical records of the Affiliated Hospital of Guizhou Medical University. Therefore, although our nomogram-based prediction model had excellent predictive power and stability, it lacked external data validation from multiple users and regions, so the evidence for the generalizability of the model was insufficient. Second, the study involved various causes and types of trauma. Although it did include the most common trauma types, a few rare trauma types were enrolled that have not been widely accepted. Further research is needed to improve the classification of trauma types and make the model more objective. Third, other than red blood cell count, hematocrit, and white blood cell count, which are closely related to trauma and disease,21 we identified absolute basophil count as an independent risk factor of shock. In this regard, there is still a lack of domestic and foreign research on the relationship between basophils and trauma or traumatic shock. The specific reasons and mechanisms of the correlation are not yet apparent, so they should be explored in future research.

This study developed a predictive shock score tool using a combination of six commonly used clinical indicators with high predictive accuracy and excellent predictive stability. This tool may help clinicians predict the risk of PTOS at an early stage. By grading individual risk, clinicians could reduce trauma mortality and improve outcomes through early intervention and management. The nomogram-based prediction tool must be validated in more extensive, multi-user clinical applications to further clarify its general practicality, feasibility, and clinical predictive power.

Wbc: white blood cell; Rbc: red blood cell; Ab: hemoglobin; PTOS: post-traumatic occult shock; ISS: injury severity score; AIS: abbreviated injury scale; GCS: Glasgow coma scale; AUC: area under the curve; LASSO: Least absolute shrinkage and selection operator; ROC: Receiver operating characteristic curve.

ACKNOWLEDGMENTS

We thank Hongxiang Lu and Anqiang Zhang for critically analyzing and modifying the contents of this manuscript.

FUNDING

The author(s) disclosed receipt of the following financial support for the research, authorship, and publication of this article: This work was supported by the National Natural Science Funds of China (81660317), the Project of Trauma, the Burns and Combined Injury State Key Laboratory (SKLKF201802, no.SKLKF201502), the Guizhou Provincial Critical and Seriously Diseased Green Channel Construction of Scientific and Technological Innovation Talent Team ([2017]5654), the clinical research project funds of the Military Medical University (2018XLC3057), the training plan of the innovation ability of military medical frontier research (2019CXJSB014).

AVAILABILITY OF DATA AND MATERIALS

Data sharing not applicable to this article as no datasets were generated or analyzed during the current study.

AUTHOR’ CONTRIBUTIONS

Guoxuan Peng, Lebin Gan, and Zhihong Deng contributed to the data curation, Jiali Sun, Jun Wang, and Guofen Chen contributed the investigation, Ling Zeng and Jin Deng contributed the project administration, Qiang Wang contributed the patient data resources and the software. Qiang Wang and Ling Zeng contributed writing the original draft. All authors read and approved the final manuscript.

ETHICAL APPROVAL AND CONSENT TO PARTICIPATE

The Ethics and Program Review Committee of Guizhou Medical University approved this project, and informed consent from the patients or their legally authorized representatives was obtained.

COMPETING INTERESTS

The authors declare that they have no competing interests.

Consent for publication

Not applicable

ORCID iD

Jin Deng https://orcid.org/0000-0001-9968-1552.

Ling Zeng https://orcid.org/0000-0002-9640-9083.

Wacker DA, Winters ME. Shock. Emerg Med Clin North Am 2014;32 (4):747

Astir ME. Histopathology and classification cation of shock states. Textbook of Critical Care 2005:897–904.

Brecht BG, Billiard TR. Shock. The Trauma Manual: Trauma and Acute Care Surgery 2008: 28–33.

Kauvar DS, Lefering R, Wade CE. Impact of hemorrhage on trauma outcome: an overview of epidemiology, clinical presentations, and therapeutic considerations. J Trauma 2006;60(6 Suppl): S3–S11.
Assert, S. M. et al. Guidelines for field triage of injured Recommendations of the National Expert Panel on Field Triage. MMWR Re comm. Rep. 58,1–35 (2009).
Cannon Hemorrhagic shock. N Encl J Med 2018;378(4):370–379.
Allgöwer M, Burri C. Shock index. Dtsch Med Wochenschr 1967; 92:1947–50.

Brady MY, Smith line HA, Blake H, Noway R, Rivers E. A comparison of the shock index and conventional vital signs to identify acute, critical illness in the emergency department. Ann Em erg Med 1994;24:685Y690.
Gruen RL, Brohi K, Schreiber M, et a1. Haemorrhage control In severely injured patients [J]. Lancet 2012,380(9847):1099.1108.
Husain FA, Martin MJ, Mullenix loS, et a1. Serum lactate and base deficit as predictors of mortality and morbidity[J]. Am J Surg 2003,185(5):485–491.
Kashuk JL, Moore EE, Johnson JL.et a1. Postinjury life threatening Coagulopathy: is l:l fresh frozen plasma: packed red blood cells the answer? [J]. J trauma 2008, 65(2):261– 271.

Kauvar DS, Wade CE. The epidemiology and modem management of traumatic hemorrhage: US and international perspectives[J]. Crit Care 2005, 9 Suppl 5: S1–S9.

Lu HX, Du J, Wen DL, et al. Development and validation of a novel predictive score for sepsis risk among trauma patients. World J Emerg Surg 2019;14:11. Published 2019 Mar 12.
Brohi K, Cohen MJ, Ganter MT, et a1. Acute coagulopathy of trauma: hypoperfusion induces systemic anticoagulation and hyperfibrinolysis[J]. J Trauma 2008,64(5): 121l– 1217.

Foe¨x BA. Systemic responses to trauma. Br Med Bull 1999;55:726Y743.

Kannel WB. Blood pressure as a cardiovascular risk factor: prevention and treatment. JAMA 1996;275:1571Y1576.
Shively SB, Perl DP. Traumatic brain injury, shell shock, and posttraumatic stress disorder in the military-past, present, and future. J Head Trauma Rehabil 2012;27(3): 234–239.

Pieracci FM, Biffl WL, Moore EE. Current concepts in resuscitation. J Intensive Care Med 2012;27:79–96.
Turner J, Nicholl J, Webber L, et al. A randomized controlled trial of prehospital intravenous fluid replacement therapy in serious trauma. Health Technol Assess. 2000; 4:1–57.

Dutton RP, Mackenzie CF, Scalea TM. Hypotensive resuscitation during active hemorrhage: impact on in-hospital mortality. J Trauma 2002;52:1141–6.
O'Keeffe T, （efasi M, Tchorz K, et al. A massive transfusion protocol to decrease blood component use and costs [J]. Arch Surg 2008,143(7):686-690.

Download PDF

Version 1

posted

You are reading this latest preprint version

Nomogram-Based Prediction of Post-Traumatic Occult Shock: Development and Evaluation of Novel Predictive Scores

Status:

Version 1

Abstract

Figures

Background

Methods

Study populations

Variables adopted for analysis and imputation

Statistical analysis

Results

Model development

Internal validation of the shock score

Discussion

Conclusion

Abbreviations

Declaration

References

Status:

Version 1