Risk factors related to traumatic spinal cord injury outcomes: An analysis of the US National Trauma Data Bank

Abstract

Introduction

Traumatic spinal cord injuries (TSCIs) are associated with high morbidity rates that can affect motor, sensory, and autonomic function. SCIs commonly result from sports, motor vehicle accidents, and other blunt traumas.

Objective

To determine risk factors for in-hospital mortality, in-hospital complications, and increased length of stay (LOS) following TSCI.

Methods

Data from the United States (US) National Trauma Data Bank (NTDB) from 2017 to 2021 was collected and used to assess risk factors for increased in-hospital mortality, complications, and hospital length of stay. Univariable analysis was completed on the 2017–2018 data. Variable importance was calculated, and a logistic regression was performed using the most important variables. Internal and external validation was completed using a testing cohort and 2020–2021 data.

Results

Spinal fusion was associated with decreased mortality but increased complications and LOS. Increased age and injury severity score was associated with increased mortality, complications, and LOS, while the emergency department Glasgow Coma Score was inversely related to all outcomes. Each model had moderate to high prediction ability based on internal area under the curves (AUCs) ranging from 0.757 to 0.899 and external AUCs between 0.744 and 0.884.

Conclusion

Understanding the factors that contribute to poorer outcomes following TSCI can help inform physicians of best treatment practices and patient prognosis. We highlighted five risk factors that correlate with short-term patient outcomes. Future research will further explore how treatment can impact both short – and long-term outcomes.

Introduction

Traumatic spinal cord injuries (TSCIs) are a growing health concern, particularly due to the high mortality rate, morbidity rate, and associated costs to patients, their families, and the healthcare system. In 2019, the global prevalence of spinal cord injuries (SCI) was 20.6 million with 0.9 million new cases (1). The reported in-hospital mortality for patients with SCI ranges from 3% to 13%, and life expectancy after the incident remains low compared to the general population (2, 3). Aside from mortality, there is also systemic morbidity associated with SCI, as patients may suffer from decreased neurological function, bladder incontinence, neuropathic pain, and other complications (4–6). Injuries higher in the spinal cord often lead to more severe long-term complications and autonomic dysfunction (4). Globally, the cost of care for SCI ranges from $290,000 to $612,590 for inpatient rehabilitation, and this cost increases with longer hospital stays, especially in older patients (7–10).

These concerns associated with SCI have grown over the last decade, as the average age for all SCI patients has increased globally (11). The mortality of all SCI patients increases substantially in older individuals, and rehabilitation becomes challenging due to increased fall risk, bone fragility, and medical comorbidities (12).

Although previous studies have explored factors that correlate with patient outcomes, this study benefits from the large number of participants from trauma centers nationwide. This study uses data from the United States (US) National Trauma Data Bank (NTDB) to explore the impacts of patient demographics, comorbidities, injury details, and spinal fusion on in-hospital mortality, complications, and hospital length of stay. The objective of this study is to develop a logistic regression model to predict outcomes following TSCI and explore the most important factors that contribute to patients’ prognosis following an injury.

Materials and methods

Study design and database description

The NTDB collects nationwide information from emergency rooms and trauma center regarding traumatic events or injuries. Centers that volunteer to participate in submitting their medical records to the NTDB receive the database in return. The information submitted includes patient demographics, injury details, hospital events, and outcomes for each participant. The NTDB collects data from over 900 trauma centers across the United States.

Data from the NTDB from the years 2017–2021 was used in this retrospective cohort study. Patients were included in the study if they had an abbreviated injury scale (AIS) code starting with 630, 640, or 650. Missing data for comorbidities and procedures was assumed to indicate a lack of the comorbidity or procedure. Missing completely at random (MCAR) analysis was completed and showed a nonrandom pattern of missing data (supplementary Figure 1) (13). Missing data for age, body mass index (BMI), injury severity score (ISS), or emergency department (ED) systolic blood pressure (SBP), ED Glasgow coma score (GCS), and time to ED arrival underwent multiple imputation by chained equations with 20 imputations before the univariable analysis and multivariable analyses (14). The remaining variables are shown in Table 2 and the outcomes were used to calculate the imputations utilizing predictive mean matching. Each model underwent logistical regression, and the results were pooled by averaging the estimates without transformation. Confidence intervals were calculated using Rubin’s Rule (15). In addition to the multiple imputed analysis, the results of the complete-case analysis, where participants with incomplete data were removed, is summarized in supplementary Tables 2 and 3.

Table 2.

Univariable analysis of in-hospital mortality. Unless otherwise stated, data is listed as mean$\pm$SD or N (%).

	Deceased (n = 2,270)	Discharged (n = 17,307)	Missing	OR [95% CI]
Age	5$4.0\pm$23.0	48.1$\pm$20.4	373 (1.9)	0.999 [0.998, 0.999]
Sex			3 (0.0)	1.01 [0.999, 1.019]
Female	548 (24.1)	44,71 (25.8)
Male	1,720 (75.8)	12,835 (74.2)
BMI	28.5$\pm 7.6$	27.7$\pm 6.7$	2326 (11.9)	0.998 [0.998, 0.999]
ISS	41.6$\pm 21.5$	2$1.9\pm 11.9$	0 (0.0)	0.991 [0.991, 0.991]
ED SBP			426 (2.2)	1.101 [1.091, 1.111]
Above median	589 (25.9)	8,706 (50.3)
Below median	1,572 (69.3)	8,284 (47.9)
ED GCS	$7.2\pm$5.4	$13.7\pm$3.3	573 (2.9)	1.039 [1.038, 1.040]
Time to ED (Hours)			4,620 (23.6)	1.053 [1.042, 1.064]
Greater than median	629 (27.7)	5,906 (34.1)
Less than median	1,244 (54.8)	7,178 (41.5)
Hospital Transfer				1.030 [1.020, 1.040]
Yes	537 (23.7)	5,110 (29.5)
No	1,733 (76.3)	12,197 (70.5)
Vertebral fracture				0.984 [0.975, 0.993]
Yes	1,266 (55.8)	8,982 (51.9)
No	1,004 (44.2)	8,325 (48.1)
Spinal Fusion or Fixation				1.138 [1.128, 1.148]
Yes	463 (20.4)	8,977 (51.9)
No	1,807 (79.6)	8,330 (48.1)
Anticoagulant Use				0.935 [0.917, 0.954]
Yes	195 (8.6)	894 (5.2)
No	2,075 (91.4)	16,413 (94.8)
Bleeding Disorder				0.976 [0.935, 1.019]
Yes	30 (1.3)	185 (1.1)
No	2,240 (98.7)	17,122 (98.9)
Chronic Obstructive Pulmonary Disease (COPD)				0.977 [0.957, 0.998]
Yes	126 (5.6)	787 (4.5)
No	2,144 (94.4)	16,520 (95.5)
Cerebrovascular Accident				0.943 [0.910, 0.978]
Yes	52 (2.3)	248 (1.4)
No	2,218 (97.7)	17,059 (98.6)
Diabetes				0.991 [0.978, 1.004]
Yes	333 (14.7)	2,361 (13.6)
No	1,937 (85.3)	14,946 (86.4)
Chronic Heart Failure (CHF)				0.897 [0.872, 0.922]
Yes	114 (5.0)	400 (2.3)
No	2,156 (95.0)	16,907 (97.7)
Hypertension				0.995 [0.985, 1.005]
Yes	704 (31.0)	5193 (30.0)
No	1,566 (69.0)	12,114 (70.0)
Myocardial Infarction				0.974 [0.923, 1.029]
Yes	19 (0.8)	115 (0.7)
No	2,251 (99.2)	17,192 (99.3)
Peripheral artery disease				0.887 [0.837, 0.941]
Yes	27 (1.2)	88 (0.5)
No	2,243 (98.8)	17,219 (99.5)
Smoking				1.082 [1.071, 1.094]
Yes	247 (10.9)	4,229 (24.4)
No	2,023 (89.1)	13,078 (75.6)
Level of Injury
Cervical	1,128 (49.7)	7,302 (42.2)		0.969 [0.961-0.977]
Thoracic	41 (1.8)	188 (1.1)		0.996 [0.990-1.002]
Lumbar	48 (2.1)	1,295 (7.5)		1.093 [1.074-1.113]
Unspecified	690 (30.4)	6,007 (34.7)		0.991 [0.821-1.195]
Severity of Injury
Complete	1,189 (52.4)	3,990 (23.1)		0.857 [0.848, 0.865]
Incomplete	626 (27.6)	7,552 (43.6)		1.070 [1.060, 1.080]
Transient	536 (23.6)	6,198 (35.8)		1.057 [1.047, 1.067]
Unspecified	0 (0.0)	0 (0.0)		NA

Participation

Participants for this study were selected from the NTDB and the single center’s trauma registry. Participants were included if they had an AIS indicating TSCI. All participants were initially included regardless of age or in-hospital time of death. The group was then further analyzed based on hospital discharge disposition. Patients who were discharged to any location were included in the discharged group, and other patients made up the deceased group. For hospital length of stay and complications, only the discharged group was analyzed.

Variables

Variables including patient demographics, injury details and severity, and hospital outcomes were collected from the NTDB database. BMI was calculated from the reported height and weight of the participants. Location of injury (cervical, thoracic, or lumbar) and severity (complete, incomplete, or transient) were collected from specific diagnosis ICD10 codes. In this case, transient referred to an injury that is temporary and passes with time. Not all the included ICD10 codes specified a region or severity of injury. In these cases, the data was left as unspecified without undergoing removal or imputations for these variables.

Outcomes

The outcomes of interest in the study include in-hospital mortality, total hospital length of stay, and in-hospital complications. An increased hospital length of stay was defined as greater than or equal to the median length, and an in-hospital complication included pneumonia, urinary tract infections (UTIs), central line infections, sepsis, surgical site infection, osteomyelitis, pulmonary embolism, deep vein thrombosis, a cardiac or cerebrovascular event, or respiratory failure.

Statistical analysis

Data from 2017 and 2018 was split into a training and testing model at an 80:20 split. The significance value was set at a p-value of 0.05. A univariable logistical regression analysis was completed for each variable. Variables that were found to be significant were included in a multivariable logistic regression. Variable linearity was evaluated visually, and age, ED SBP, and time to arrival to ED were found to be curvilinear (supplementary Figure 1). In response, age was used to the power of fourth in the logistic regression model while ED SBP and time to arrival were categorized into two groups: greater than or less than the median for each. Multicollinearity was evaluated using the variable inflation factor, which indicated no multicollinearity since all values were less than 2.5 (16, 17). From the resulting model, variable importance for each feature was calculated using interpretable machine learning, which quantifies the contribution that each variable has to the final logistical model (18, 19). To avoid overfitting of the model, especially with the large sample size, the included features in a final logistic regression model had a cutoff 20 in the variable importance value. Validation of this model was performed internally using the 20% testing cohort. External validation was completed using the NTDB data from 2020 and 2021 to separate the participants by time. The accuracy and AUC were calculated and reported for each validation. This testing is demonstrated as a schematic in Fig. 1. All analysis for this project was completed using R version 4.3.1. Linearity and multicollinearity were assessed using the GGally and car package. MCAR analysis was completed using the finalfit package. Multiple imputation was completed using the mice and miceafter package.

Figure 1.

Schematic for cohorts of participants.

Results

Participants

Between 2017 and 2021, there were 67,340 incidents of recorded TSCIs from the NTDB database (Table 1). Over those five years, the incidence recorded in NTDB increased by 27.0%. Overall, the reported procedural rate of spinal fusion or fixation increased from 48.4% to 52.1%. The reported in-hospital mortality rate remained around 10.9% for all years. The average reported hospital length of stay was 14 days, 25.8% of the patients were female, and the average age was 49.3 years old. The breakdown of reported location of injury (LOI) was 43.5% cervical, 1.1% thoracic, and 7.1% lumbar. In terms of grade, 9.6% were complete injuries, 15.2% were incomplete, and 12.5% were transient injuries. The level of injury was not indicated in the ICD10 code in 48.3% of cases, and the grade was unspecified based on the ICD10 code in 62.7% of cases.

Table 1.

Demographics of participants. Unless otherwise stated, data is listed as mean$\pm$SD or N (%).

	Total	2017	2018	2019	2020	2021
N	67,340	11,909	12,563	13,212	14,534	15,122
Age	49.3$\pm$20.7	$48.7\pm$20.8	$49.1\pm 20.6$	50.2$\pm$20.7	48.9$\pm$20.8	49.7$\pm$20.8
Sex
Female	17,376 (25.8)	3,053 (25.6)	3,225 (25.7)	3,458 (26.2)	3,644 (25.1)	3,996 (26.4)
Male	49,797 (73.9)	8856 (74.4)	9,335 (74.3)	9,750 (73.8)	10,888 (74.9)	10,968 (72.5)
Spinal procedure
Yes	33,824 (50.2)	5,763 (48.4)	6,072 (48.3)	6,595 (49.9)	7,511 (51.7)	7,883 (52.1)
No	33,516 (49,8)	6,146 (51.6)	6,491 (51.7)	6,617 (50.1)	7,023 (48.3)	7,239 (47.9)
In-hospital mortality
Yes	7,322 (10.9)	1,398 (11.7)	1,418 (11.3)	1,409 (10.7)	1,458 (10.0)	1,639 (10.8)
No	60,018 (89.1)	10,511 (88.3)	11,145 (88.7)	11,803 (89.3)	13,076 (90.0)	13,483 (89.2)
Complication
Yes	8,613 (12.8)	1,639 (13.8)	1,635 (13.0)	1,620 (12.3)	1,774 (12.2)	1,945 (12.9)
No	58,727 (87.2)	10,270 (86.2)	10,928 (87.0)	11,592 (87.7)	12,760 (87.8)	13,177 (87.1)
Hospital length of stay (days)	14.0$\pm$17.0	13.9$\pm$17.0	13.5$\pm$17.1	13.8$\pm$16.9	14.$0\pm$16.4	1$4.9\pm$17.6

In-hospital mortality

Of the 67,340 participants with TSCI, 7,322 (10.9%) died during their hospitalization. Univariable analysis of data from 2017 and 2018 demonstrated that patient age, sex, BMI, ISS, transfer status, location of injury, the GCS and SBP recorded in the ED, time to ED, concurrent vertebral fracture, grade of injury, and whether a spinal fusion or fixation was completed significantly differed between the in-hospital mortality and discharged cohort (Table 3). Multivariable analysis was completed via a logistical regression on these variables and the importance was calculated (Table 4 and Fig. 2). The final logistic regression model included injury ISS, ED GCS, age, and a spinal fusion or fixation completed as variables. Where the risk of in-hospital mortality increased with age and ISS, it decreased with the other variables (Table 5).

Table 3.

Univariable analysis of complication and length of stay data. Unless otherwise stated, data is listed as mean$\pm$SD or N (%).

	Complications				Length of Stay
	Complication (n = 2,042)	No Complication (n = 15,282)	Missing	OR [95% CI]	LOS <10 days (n = 8,508)	LOS ≥10 days (n = 8,631)	Missing	OR [95% CI]
Age	46.2$\pm$19.2	48.$6\pm$20.4	249 (1.4)	0.999 [0.999,1.000]	49.$8\pm$20.4	$47.0\pm 20.1$	247 (1.4)	0.998 [0.998, 0.999]
Sex			1 (0.1)	0.971 [0.961, 0.983]			1 (0.0)	0.925 [0910, 0.941]
Female	429 (21.0)	4,010 (26.2)			2,436 (28.6)	1,961 (22.7)
Male	1,613 (79.0)	11,271 (73.8)			6,072 (71.4)	6,669 (77.3)
BMI	28.8$\pm$7.5	27.$6\pm$6.6	1747 (10.1)	1.003 [1.002. 1.004]	27.6$\pm$6.5	27.9$\pm$7.0	1703 (9.9)	1.002 [1.001, 1.003]
ISS	31.$2\pm$14.7	20.$7\pm$11.0	1 (0.1)	1.008 [1.007, 1.008]	17.4$\pm$8.7	26.4$\pm$13.0	1 (0.0)	1.016 [1.015, 1.016]
ED SBP			308 (1.8)	0.920 [0.911, 0.929]			304 (1.8)	0.821 [0.809, 0.8333]
Above median	674 (33.0)	8,093 (53.0)			5,144 (60.5)	3,546 (41.1)
Below median	1,323 (64.8)	6,926 (45.3)			3,221 (37.9)	4,924 (57.1)
ED GCS	11.5$\pm$4.8	14.0$\pm$2.9	499 (2.9)	0.976 [0.974, 0.977]	14.6$\pm 1.8$	12.9$\pm 4.0$	492 (2.9)	0.961 [0.959, 0.963]
Time to ED (Hours)			4289 (24.8)	0.988 [0.977, 1.000]			4228 (24.7)	0.987 [0.970, 1.004]
Greater than median	688 (33.7)	5,141 (33.6)			2,817 (33.1)	2,973 (34.4)
Less than median	936 (45.8)	6,270 (41.0)			3,373 (39.6)	3,748 (43.4)
Hospital Transfer				0.973 [0.962, 0.983]				0.937 [0.921, 0.952]
Yes	496 (24.3)	4,558 (29.8)			2,720 (32.0)	2,286 (26.5)
No	1,546 (75.7)	10,724 (70.2)			5,788 (68.0)	6,345 (73.5)
Vertebral fracture				1.038 [1.028, 1.048]				1.114 [1.097, 1.130]
Yes	1,216 (59.5)	7,741 (50.7)			3,933 (46.2)	4,915 (56.9)
No	826 (40.5)	7,541 (49.3)			4,575 (53.8)	3,716 (43.1)
Spinal Fusion or Fixation				1.065 [1.055, 1.075]				1.281 [1.262, 1.299]
Yes	1,330 (65.1)	7,651 (50.7)			3,370 (39.6)	5,556 (64.4)
No	712 (34.9)	7,631 (49.3)			5,138 (61.4)	3,075 (35.6)
Anticoagulant Use				0.984 [0.962, 1.006]				0.966 [0.934, 1.000]
Yes	91 (4.5)	796 (5.2)			466 (5.5)	415 (4.8)
No	1,951 (95.5)	11,486 (94.8)			8,042 (94.5)	8,216 (95.2)
Bleeding Disorder				0.955 [0.911,1.002]				0.957 [0.889, 1.031]
Yes	13 (0.6)	166 (1.1)			95 (1.1)	82 (1.0)
No	2,029 (99.4)	15,116 (98.9)			8,413 (98.9)	8,549 (99.0)
COPD				1.004 [0.982, 1.028]				1.049 [1.012, 1.088]
Yes	96 (4.7)	690 (4.5)			351 (4.1)	429 (5.0)
No	1,946 (95.3)	14,592 (95.5)			8,157 (95.9)	8,202 (95.0)
Cerebrovascular Accident				1.008 [0.969, 1.049]				1.033 [0.972, 1.098]
Yes	33 (1.6)	229 (1.5)			121 (1.4)	140 (1.6)
No	2,009 (98.4)	15,053 (98.5)			8,387 (98.6)	8,491 (98.4)
Diabetes				1.004 [0.990,1.018]				1.009 [0.988, 1.032]
Yes	290 (14.2)	2,104 (13.8)			1,161 (13.6)	1,218 (14.1)
No	1,752 (85.8)	13,178 (86.2)			7,347 (86.4)	7,413 (85.9)
CHF				0.978 [0.948, 1.010]				1.044 [0.994, 1.097]
Yes	39 (1.9)	366 (2.4)			184 (2.2)	220 (2.5)
No	2,003 (98.1)	14,916 (97.6)			8,324 (97.8)	8,411 (97.5)
Hypertension				0.979 [0.969,0.989]				0.951 [0.936, 0.967]
Yes	542 (26.5)	4,727 (30.9)			2,773 (32.6)	2,448 (28.4)
No	1,500 (73.5)	10,555 (69.1)			5,735 (67.4)	6,183 (71.6)
Myocardial Infarction				1.002 [0.945, 1.062]				0.968 [0.884, 1.061]
Yes	14 (0.7)	103 (0.7)			61 (0.7)	55 (0.6)
No	2,028 (99.3)	15,179 (99.3)			8,447 (99.3)	8,576 (99.4)
Peripheral artery disease				1.024 [0.956, 1.097]				1.082 [0.972, 1.204]
Yes	12 (0.6)	73 (0.5)			34 (0.4)	49 (0.6)
No	2,030 (99.4)	15,209 (99.5)			8,474 (99.6)	8,582 (99.4)
Smoking				1.006 [0.995, 1.018]				1.022 [1.004, 1.040]
Yes	519 (25.4)	3,716 (24.3)			2,009 (23.6)	2,174 (25.2)
No	1,523 (74.6)	11,566 (75.7)			6,499 (76.4)	6,457 (74.8)
Level of Injury
Cervical	924 (45.2)	6,369 (41.7)		1.012 [1.003, 1.021]	3,472 (40.8)	4,663 (54.0)		1.031 [1.017, 1.046]
Thoracic	29 (1.4)	140 (0.9)		1.029 [1.022, 1.036]	65 (0.8)	104 (1.2)		1.100 [1.088, 1.111]
Lumbar	136 (6.7)	1,177 (7.7)		0.985 [0.967, 1.003]	649 (7.6)	650 (7.5)		0.995 [0.968, 1.024]
Unspecified	546 (26.7)	5,448 (35.6)		0.960 [0.950-0.970]	3,451 (40.6)	2,493 (28.9)		0.879 [0.865-0.893]
Severity of Injury
Complete	1,022 (50.0)	2,960 (19.4)		1.197 [1.184, 1.211]	876 (10.3)	3,051 (35.3)		1.425 [1.402, 1.450]
Incomplete	705 (34.5)	6,850 (44.8)		0.957 [0.948, 0.967]	3,662 (43.0)	3,815 (44.2)		1.011 [0.996, 1.027]
Transient	419 (20.5)	5,805 (38.0)		0.924 [0.915, 0.933]	4,075 (47.9)	2,085 (24.2)		0.773 [0.761, 0.784]
Unspecified	0 (0.0)	0 (0.0)		NA	0 (0.0)	0 (0.0)		NA

Table 4.

Coefficients for multivariable logistic regression.

In-hospital mortality		Complications		Length of stay
Variable	OR [95% CI]	Variable	OR [95% CI]	Variable	OR [95% CI]
ISS	0.995 [0.995, 0.995]	ISS	1.004 [1.004, 1.004]	ISS	1.299 [1.288, 1.310]
ED GCS	1.029 [1.029, 1.030]	ED GCS	0.990 [0.990, 0.991]	Surgery	649.971 [507.746, 832.035]
Age	0.997 [0.997, 0.997]	Surgery	1.072 [1.067, 1.078]
Surgery	1.087 [1.083, 1.091]

Figure 2.

Variable importance plots.

Table 5.

Validation measures of models.

	In-hospital Mortality		Complication		Length of Stay
	Internal	External	Internal	External	Internal	External
Accuracy	92.3%	92.3%	88.4%	89.3%	51.1%	53.2%
AUC	0.899	0.884	0.764	0.762	0.757	0.744

This final model was internally validated against the testing cohort from the 2017 and 2018 data as well as an external validation completed on the data from 2020 and 2021. The internal validation demonstrated an accuracy of 92.3% and an AUC of 0.889. The external test showed an accuracy of 92.3% and an AUC of 0.884 (Table 5 and Fig. 3).

Figure 3.

AUC curves for internal and external validation.

Complications

The incidence of UTI, surgical site infection (SSI), central line infection, osteomyelitis, sepsis, pneumonia, a cardiac or cerebrovascular event, deep vein thrombosis (DVT), pulmonary embolism, or respiratory failure occurred in 12.8% of the training cohort. Analysis showed a significant difference in patient age, sex, BMI, injury ISS, location of injury, concurrent vertebrae fracture, the time to arrival at the ED, the GCS and SBP recorded at the ED, injury grade, comorbidity of hypertension, and whether a spinal fusion or fixation was performed (Table 3). A logistic regression model of these variables was completed, and variable importance was measured (Fig. 2). The odds ratios for ISS, ED GCS, and spinal fusion or fixation are shown in Table 4. The internal and external validation had AUCs of 0.764 and 0.762, respectively (Table 5 and Fig. 3).

Length of stay

The median hospital length of stay was ten days. The participants were split into a cohort with a shorter hospital length of stay (<10 days) and a longer length of stay ($\ge$10 days). Univariable analysis between these groups showed a significant difference in patient age, BMI, sex, if they were transferred, injury ISS, location of injury, injury grade, vertebrae fracture, several comorbidities, the emergency room GCS and SBP, and a spinal fusion or fixation procedure (Table 3). The most important variables in the multivariable analysis were found to be ISS and spinal procedure (Table 3 and Fig. 2). The internal validation showed an AUC of 0.757, and the external validation had an AUC of 0.744 (Table 5 and Fig. 3).

Discussion

The incidence of TSCI is increasing yearly, and this analysis showed a 29.6% increase in the number of TSCI in 2021 compared to 2017. This may be due to increased reporting to the NTDB or an overall increase in incident. In either case, it demonstrates a growing need for being able to predict prognosis following TSCI. Especially as the rate of complications and in-hospital mortality has remained around 13.1%, and 8.9%, respectively. This is despite improvements in long-term outcomes and mortality rates seen previously (20). This indicates that there may be improvements in rehabilitation regimen post TSCI; however, current treatment options and surgeries have plateaued in their ability to improve short-term health outcomes. Despite the average age of TSCI patients increasing nationally, there was not a substantial difference in age for the NTDB in 2017–2021; however, the age was higher than reported ages in other studies (21, 22). This could be the effect of localized regions of changing TSCI demographics or limits to the scope of the timeline in this study.

As noted in previous studies (23, 24), the ISS and patient age remained key contributing factors in in-hospital mortality with older individuals and more severe injury correlating with worse outcomes. This study adds to the current understanding of risk factors by highlighting patient ED SBP and GCS as key variables in predicting risk for poor outcomes. Notably, arrival time to the ED post-injury was not found to be one of the most important variables for any of the short-term outcomes and was not substantially different between the cohorts. This may indicate that a patient's vital signs early on may be more predictive than the speed of response. In addition, the LOS was predicted mostly by a surgery being completed apart from patient demographics or injury details.

Throughout the three outcomes studied, each of them was impacted by the patient undergoing a spinal fusion or fixation. These procedures correlated with a decreased mortality rate, but an increase in complications and length of stay. This highlights the importance of medical interventions in TSCIs. There is a chance that patients undergoing surgery were perceived to have a better chance of survival, were more stable and could tolerate it, or had a SCI that require surgery for decompression or stabilization. Due to limited data on the patient status throughout their visit, it was not possible to account for these differences. Neal et al. explored how surgical delay in cervical SCI patients was impacted by arrival status, specifically SBP, but this did not impact hospital LOS or mortality (25). Further analysis should be completed specifically in patients undergoing surgery to assess risk factors and impacts for this group. In addition, it is possible that less aggressive treatments, such as pharmaceuticals, and alternative surgical approaches could improve mortality without the impacts on other outcomes (26–28). This was not investigated in this study but may be a valuable variable to include in future work.

The limitations of this study stem from the collection of the NTDB data. Since the records are recorded from a single injury event, it is not possible to study long-term outcomes, and previous medical history is limited to reported comorbidities. This also restricted the number of details available for each variable and relies on accurate reports across multiple health systems. This data also requires accurate tracking and reporting by each medical center. This may not occur, especially for tracking complications. Due to the limitations of this data, complications can only include events that occur during the initial hospital visit. Specific variables could not reliably be extracted, such as the types of approaches and medications that were used in each case. The imputation and complete-case analyses were similar for mortality and complication, but there was some variation for the length of stay model. This may be due to the reduced accuracy of that model to begin with. Definition of the injuries may also not be consistent among trauma centers. In addition, the external validation of this model is separated by time only. Further research should study this model’s predictive ability on participants from various communities and countries.

Conclusion

Based on these results and previous research into TSCI outcomes, injury severity and GCS reported in the emergency department are two important factors correlated with short-term outcomes following TSCI. Age is also an important factor for in-hospital mortality. It is important to continue exploring how this contributes to treatment options and rehabilitation post-SCI. We also found that spinal fusions are associated with better mortality but increased complications and hospital LOS. Future studies should further explore how treatment options, including nonsurgical and surgical procedures, can influence long-term outcomes in TSCIs.

Funding Statement

There was no funding received for this study.

Disclaimer statements

Declaration of interest statements The authors have no conflicts of interest to disclose.

Supplemental data

Supplemental data for this article can be accessed online at https://doi.org/10.1080/10790268.2024.2417118.