Elderly Traumatic Brain Injury (eTBI) Scoring System: A Much More Effective New Classification

Abstract

Background and Objectives: The mechanisms of traumatic brain injury (TBI), patient characteristics, and long-term outcomes in elderly patients differ from those in other age groups. This study aims to evaluate the effectiveness of the Elderly Traumatic Brain Injury (eTBI) Scoring System, recently described in the literature, in predicting mortality, prognosis, and surgical indication. Materials and Methods: Patients diagnosed with TBI over the age of 65 between January 2017 and December 2024 were retrospectively analyzed, and their eTBI scores were calculated. Statistical analyses were conducted to assess mortality, prognosis, and surgical indication or benefit from surgery across low-, medium-, and high-risk groups. Results: In this cohort of 236 patients, the mortality rate was higher in the high-risk group according to the eTBI scoring system, compared to the medium- and low-risk groups. However, the scoring system does not appear to be effective in determining surgical indications. While the medium-risk group was most predictive of mortality, the low-risk group demonstrated better accuracy in predicting prognosis. Conclusions: The eTBI scoring system appears to be an effective tool for assessing mortality risk and predicting prognosis in specific subgroups of elderly TBI patients.

1. Introduction

Traumatic brain injury (TBI) is a leading cause of death and disability worldwide. The mechanisms of TBI, patient characteristics, and biological sequelae in elderly patients differ from those in younger individuals. As a result, elderly TBI requires a distinct approach in both treatment management and clinical research [1].

Among older patients, falls are the most common cause of TBI, and the condition is more frequently observed in women. In contrast, TBI in younger individuals primarily results from motor vehicle accidents and is more common in men [2]. With aging, the white matter and vascular system become more vulnerable to injury, while injury response mechanisms, such as autophagy, weaken. Additionally, the prevalence of pre-existing neurological and systemic comorbidities increases [3,4,5].

Elderly TBI patients generally experience higher morbidity and mortality rates, along with slower recovery. Compared to younger patients, they exhibit poorer functional, cognitive, and psychosocial outcomes months or even years post-injury [6]. However, a subset of elderly patients, including those with severe TBI, can recover well, indicating that age and injury severity alone are insufficient as prognostic markers [7]. Despite this, few studies have examined the impact of pre-existing diseases on post-TBI outcomes in this age group, and even fewer have considered premorbid functional status. Moreover, older adults, particularly those with functional impairments or multiple comorbidities, are frequently excluded from TBI research. Consequently, there is a lack of geriatric-specific TBI guidelines for acute management, long-term prognosis, or treatment strategies [8].

To address the need for an age-specific prognostic tool, Bobeff [9] introduced the first scoring system designed for elderly TBI patients. The total eTBI score is calculated using GCS-motor scores, platelet counts, RDW-CV ratios, and the presence of comorbidities such as cardiac, pulmonary, or renal dysfunction or malignancy. Prognostic value is assessed using the Glasgow Outcome Scale (GOS). Subsequent studies categorized patients into low-, medium-, and high-risk groups, evaluating outcomes based on these classifications [10].

This study aims to apply the eTBI scoring system to assess its effectiveness in predicting prognosis and mortality, as well as determining surgical outcomes in elderly TBI patients.

2. Materials and Methods

Following approval from the Bakırçay University Clinical Research Ethics Committee (approval date: 8 January 2025, approval number: 1963), patients aged 65 and older who were diagnosed with TBI and received inpatient treatment at Muğla Training and Research Hospital (affiliated with Muğla Sıtkı Koçman University) between 2017 and 2024 were retrospectively analyzed using the hospital’s health information database.

Data collected included patients’ age, gender, surgical status, Glasgow Coma Scale (GCS) score at admission, Glasgow Outcome Scale (GOS) score at discharge, GCS-motor score at hospitalization, platelet and RDW-CV values, and the presence of comorbidities or malignancy. The total eTBI score was then calculated (

Table 1. Components of the eTBI score.

Component	eTBI Score Points
GCS-M score	Range 1–6
Platelets
≤100 × 109 cells/L	−1
>100 × 109 cells/L	0
RDW-CV
≥14.5%	−1
<14.5%	0
Comorbid cardiac, pulmonary or renal dysfunction/malignancy
Yes	−1
No	0
Total eTBI score	From −2 to 6

). Patients were classified into risk groups based on their scores: those with scores between −2 and 0 were categorized as high-risk, those with scores between +1 and +3 as medium-risk, and those with scores between +4 and +6 as low-risk.

Patients under 65 years of age, those with a GCS score below 15 prior to TBI (home care patients with comorbidities, consciousness disorders secondary to previous severe illnesses, dementia, etc.), and those with incomplete or inaccessible data were excluded from the study.

This study was conducted in accordance with the ethical principles outlined in the Declaration of Helsinki.

All patient records and data were analyzed using the SPSS 23.00 statistical software package. Descriptive statistics (number, percentage distribution, mean, and standard deviation), t-tests, and ANOVA were employed for data analysis. Parameters that did not follow a normal distribution were compared using the Mann–Whitney U test. Findings were evaluated with a 95% confidence interval and a 5% significance level. Prognosis and mortality related group comparisons (e.g., survivors vs. non-survivors, favorable vs. unfavorable GOS outcomes, surgery vs. no surgery, and risk groups) were performed using χ² or Fisher Exact Test for categorical variables and t-tests or Mann–Whitney U tests for continuous variables. For categorical variables with small expected cell counts (<5), Fisher’s exact test was used instead of the χ² test. The predictive performance of the eTBI score in determining prognosis was assessed using the receiver operating characteristic (ROC) curve analysis. The comparison of the ROC curves was performed using the “paired-sample area difference under the ROC curves”, which is based on the nonparametric method described by Hanley and McNeil for comparing correlated AUCs. In cases where significant threshold values were identified, the sensitivity, specificity, positive predictive value, and negative predictive value were calculated. Cut-off points for the scoring systems were derived from ROC curve analysis by selecting thresholds where sensitivity and specificity values were closest to each other. To adjust for baseline differences between surgically and non-surgically treated patients, propensity score matching (PSM) was performed. Propensity scores were estimated using logistic regression, including age, gender, GCS, laboratory parameters, comorbidities, and anticoagulant/antiaggregant use. A 1:1 nearest-neighbor matching algorithm with a caliper of 0.2 standard deviation was applied. Covariate balance after matching was assessed using standardized mean differences (SMD).

Post hoc power analysis was performed based on the ROC curve results. The AUC value of GCS was 0.981 ± 0.015, and the AUC value of eTBI was 0.941 ± 0.026, while the standard AUC value to be tested was 0.5. For the study population (n = 236; 17 deaths, 219 survivals), the calculated post hoc power levels were >99.9% for both GCS and eTBI. These high power values reflect the large observed effect size (AUCs well above 0.5) and are consistent with the robustness of the ROC analyses.

Neurosurgical decision algorithm on all patients were structured as follows: Patients with acute epidural hematomas with a volume greater than 30 cc regardless of GCS; acute subdural hematomas with a thickness > 10 mm, midline shift > 5 mm regardless of GCS, or if there is a decrease of 2 or more points in GCS with thinner hemorrhages and pupillary function abnormality; parenchymal lesions with cisternal compression, midline shift > 5 mm or any lesion volume > 50 cc; posterior fossa mass lesions with distortion of the 4th ventricle, effacement of the basilar cisterns or obstructive hydrocephalus; depressed cranial fractures with depression greater than the thickness of the skull were urgently taken to surgery after rapid evaluation. Intraoperative observation revealed cerebral hemorrhage, which was cleaned, pressure effects were relieved, and the presence of cerebral pulsation was detected. Postoperative radiological imaging also showed a decrease and/or disappearance of preoperative pathologies.

3. Results

A total of 236 patients were included in the study (Figure 1). Among them, 63.6% (n = 150) were male, and 36.4% (n = 86) were female, with a mean age of 76.24 ± 7.68 years. The mean Glasgow Coma Scale (GCS) score for all patients was 12.42 ± 4.0, while the mean eTBI score was 4.18 ± 1.91. In terms of trauma type, the most common was subdural hematoma (37.5%, n = 88), while the least common was parenchymal hemorrhage (7.2%, n = 17). Surgical intervention was performed in 22.5% of all patients (n = 53). The number of patients discharged with a Glasgow Outcome Scale (GOS) score of 3 or higher was 189 (80.1% of all patients). The overall mortality rate was 7.2% (n = 17).

When patients were divided into two groups based on whether they underwent surgery, statistically significant differences were observed in the unit of hospitalization, platelet count, RDW-CV, GCS at admission, eTBI score, GOS at discharge, and discharge status (p-values: <0.001, <0.045, <0.001, <0.001, <0.001, <0.001, and 0.04, respectively). However, no statistically significant association was found between surgery and age, gender, or antiplatelet use (p-values: 0.419, 0.586, and 0.873, respectively) (

Table 2. Descriptive data of patients grouped according to the presence of surgery.

		Surgical Treatment
	Overall	Yes	No	p-Value
Gender, n (%)
Female	86 (36.4%)	22 (58.5%)	64 (35%)	0.419
Male	150 (63.6%)	31 (41.5%)	119 (65%)
Age, Mean (SD)	76.24 (7.6)	76.6 (7.3)	76.12 (7.7)	0.586
GCS
Mean (SD)	12.42 (4.0)	9.2 (4.7)	12.9 (3.6)	<0.001
Median	15	12	15
Minimum	3	3	3
Maximum	15	15	15
eTBI score
Mean (SD)	4.18 (1.92)	3.35 (2.3)	4.42 (1.71)	<0.001
Median	5	4.0	5
Minimum	−1	−1.0	−1.0
Maximum	6	6.0	6.0
Platelet Count, Mean (SD)	220.0 (84.81)	219.56 (76.06)	220.2 (87.37)	0.045
RDW-CV, Mean (SD)	14.0 (2.39)	14.53 (2.93)	13.87 (2.19)	<0.001
GOS score
GOS 1–2	47 (19.9%)	20 (37.7%)	27(14.8%)	<0.001
GOS 3-4-5	189 (80.1%)	33(62.3%)	156 (85.2%)
Hospitalized Unit, n (%)
ICU	88 (37.3%)	43 (81.1%)	45(24.6%)	<0.001
Clinic	148 (62.7%)	10 (18.9%)	138 (75.4%)
Discharge status n (%)
Exitus	17 (7.2%)	9 (17%)	8 (4.4%)	0.04
Survival	219 (92.8%)	44(83%)	175 (95.4%)
Antiaggregant used
Yes	93 (39.4%)	20 (%37.7%)	73 (39.9%)	0.873
No	143 (60.6%)	33(62.3%)	110 (60.1%)

GCS: Glasgow Coma Score; eTBI: Elderly Traumatic Brain Injury Score; GOS: Glasgow Outcome Score; RDW-CV: red blood cell distribution width-CV.

).

In the propensity score-matched cohort of 33 patients per group, surgically treated patients had significantly higher rates of poor neurological outcome (GOS 1–2: 84.8% vs. 39.4%, p < 0.001) and mortality (33.3% vs. 3.0%, p = 0.003).

The predictive performance of the GCS and eTBI scores in determining prognosis and mortality was evaluated using ROC analysis. GOS at discharge was used as an indicator of prognosis. Patients were categorized into two groups: those with a GOS of 1 or 2 (low GOS) and those with a GOS of 3 or higher (high GOS). Mortality was assessed based on discharge status (discharge vs. exitus). A statistically significant difference was found between GCS and eTBI scores in predicting prognosis (p < 0.045). The area under the curve (AUC) was 0.981 with a standard error of 0.015 for GCS and 0.941 with a standard error of 0.026 for eTBI (Figure 2). The GCS score demonstrated significantly higher AUROC values compared to the eTBI score.

For the GCS score, the cut-off point was 8.5, with a sensitivity of 97.9%, specificity of 92%, positive predictive value of 91.5%, and negative predictive value of 78.4%. For the eTBI score, the cut-off point was 3.5, with a sensitivity of 95.2%, specificity of 85.2%, positive predictive value of 81.6%, and negative predictive value of 76.3% (

Table 3. ROC curves for GCS and eTBI compared to prognosis.

Significance of GCS ROC curve		p-value = 0.045
Area under the ROC curve (AUROC)	0.981
Standard deviation	0.015
95% confidence interval	0.954 to 1.008
Significance of eTBI ROC curve
Area under the ROC curve (AUROC)	0.941
Standard deviation	0.026
95% confidence interval	0.890 to 0.993

GCS: Glasgow Coma Score; eTBI: Elderly Traumatic Brain Injury Score; ROC: the receiver operating characteristic curve.

and

Table 4. Diagnostic scanning tests for GCS and eTBI score for prognosis.

Diagnostic scanning tests for GCS
Cut-off	8.5
Sensitivity	97.9%
Specificity	92%
PPV	91.5%
NPV	78.4%
Diagnostic scanning tests for eTBI
Cut-off	3.5
Sensitivity	95.2%
Specificity	85.2%
PPV	81.6%
NPV	76.3%

GCS: Glasgow Coma Score; eTBI: Elderly Traumatic Brain Injury Score; PPV: positive predictive value, NPV: negative predictive value.

).

A statistically significant difference was also observed between GCS and eTBI scores in predicting mortality (p < 0.042). The AUC was 0.902 with a standard error of 0.040 for GCS and 0.804 with a standard error of 0.068 for eTBI (Figure 3). The GCS score again showed significantly higher AUROC values compared to the eTBI score.

For mortality prediction, the GCS score had a cut-off point of 8.5, with a sensitivity of 85.4%, specificity of 88.2%, positive predictive value of 87.2%, and negative predictive value of 68.1%. The eTBI score had a cut-off point of 2.5, with a sensitivity of 86.8%, specificity of 64.7%, positive predictive value of 61.7%, and negative predictive value of 66.8% (

Table 5. ROC curves for GCS and eTBI compared to mortality.

Significance of GCS ROC curve		p-value = 0.042
Area under the ROC curve (AUROC)	0.902
Standard deviation	0.040
95% confidence interval	0.824 to 0.980
Significance of eTBI ROC curve
Area under the ROC curve (AUROC)	0.804
Standard deviation	0.068
95% confidence interval	0.671 to 0.937

GCS: Glasgow Coma Score; eTBI: Elderly Traumatic Brain Injury Score; ROC: the receiver operating characteristic curve.

and

Table 6. Diagnostic scanning tests for GCS and eTBI score for mortality.

Diagnostic scanning tests for GCS
Cut-off	8.5
Sensitivity	85.4%
Specificity	88.2%
PPV	87.2%
NPV	68.1%
Diagnostic scanning tests for eTBI
Cut-off	2.5
Sensitivity	86.8%
Specificity	64.7%
PPV	61.7%
NPV	66.8%

GCS: Glasgow Coma Score; eTBI: Elderly Traumatic Brain Injury Score; PV: positive predictive value, NPV: negative predictive value.

).

Both scoring systems yielded p-values > 0.05 in the Hosmer–Lemeshow goodness-of-fit test, with GCS (H-L statistic = 3.505, p =0.681) demonstrating a better fit than the eTBI score (H-L statistic = 1.00, p = 0.072).

Collinearity diagnostics demonstrated no significant multicollinearity between GCS and eTBI (VIF: 4.062, Tolerance: 0.246 for both variables.

Logistic regression analysis confirmed that both GCS and eTBI scores were statistically significant in predicting mortality, with GCS showing a stronger predictive performance (p < 0.001 and p = 0.02, respectively). An increase in the GCS score reduced mortality risk by 0.744 times, while an increase in the eTBI score reduced mortality risk by 0.657 times (

Table 7. Regression analysis of significant individuals associated with mortality.

	OR (95% CI)	p Values
GCS	0.744 (0.297–0.906)	<0.001
eTBI	0.657 (0.468–0.810)	0.002

GCS: Glasgow Coma Score; eTBI: Elderly Traumatic Brain Injury Score; OR: odds ratio, CI: confidence interval.

).

When the discharge status of eTBI risk groups was analyzed in relation to surgical intervention, statistically significant results were observed in the medium-risk group (p = 0.022). However, no significant results were obtained when the GOS outcomes of eTBI risk groups were analyzed in relation to surgery (

Table 8. Discharge and GOS statistics according to surgery in eTBI risk groups.

	eTBI
	High Risk Group			Medium Risk Group			Low Risk Group
	Surgery + (n:11)	Surgery – (n:11)	p-Value	Surgery + (n:8)	Surgery – (n:19)	p-Value	Surgery + (n:34)	Surgery – (n:153)	p-Value
Discharge status
Exitus, n (%)	4 (36.4%)	5 (45.5%)		2 (25%)	-		3 (8.8%)	3 (%2)
Survival, n (%)	7 (63.6%)	6 (54.5%)	0.664	6 (75%)	19 (%100)	0.022	31 (91.2%)	150 (98%)	0.071
GOS
GOS 1–2, n (%)	11 (100%)	11 (100%)	-	6 (75%)	12 (63.2%)	0.084	3 (8.8%)	4 (2.6%)	0.122
GOS 3–4–5, n (%)	-	-		2 (25%)	7 (36.8%)		31 (91.2%)	149 (97.4%)

eTBI: Elderly Traumatic Brain Injury Score; GOS: Glasgow Outcome Score; p-values are based on Fisher’s exact test.

).

An analysis of the eTBI scoring system’s risk groups in terms of mortality showed that the intermediate-risk group yielded statistically significant results (p < 0.001), with 96.4% specificity and a 96.7% positive predictive value (

Table 9. Diagnostic scanning tests for the eTBI group score for mortality.

	eTBI
	High Risk Group	Medium Risk Group	Low Risk Group
Positive	13	25	181
Negative	9	2	6
AUROC	0.342	0.840	0.639
Standard Deviation	0.120	0.092	0.147
Sensitivity	46.2%	68%	77.3%
Specificity	23.2%	96.4%	76.3%
PPV	77.8%	96.7%	67.8%
NPV	46.2%	66.7%	47.9%
p-value	0.186	<0.001	0.344

eTBI: Elderly Traumatic Brain Injury Score; AUROC: area under the ROC curve; PPV: positive predictive value, NPV: negative predictive value; p-values are based on Fisher’s exact test.

). When the same risk groups were evaluated in terms of prognosis, the low-risk group demonstrated statistically significant results (p < 0.001), with 99% sensitivity and a 96.3% positive predictive value (

Table 10. Diagnostic scanning tests for the eTBI group score for prognosis.

	eTBI
	High Risk Group	Medium Risk Group	Low Risk Group
Positive	13	16	186
Negative	9	11	1
AUROC	0.342	0.429	0.118
Standard Deviation	0.120	0.113	0.072
Sensitivity	46.2%	56.3%	99%
Specificity	23.2%	66.3%	71%
PPV	100%	66.7%	96.3%
NPV	-	33.3%	71.4%
p-value	0.186	0.529	<0.001

eTBI: Elderly Traumatic Brain Injury Score; AUROC: area under the ROC curve; PPV: positive predictive value, NPV: negative predictive value; p-values are based on Fisher’s exact test.

).

4. Discussion

This study reached results supporting that certain parameters of the eTBI scoring system, which is recommended as a new scoring system for elderly patients, may be effective and reliable in predicting mortality and determining prognosis, but may be insufficient in evaluating the contribution of surgical indication or the presence of surgery to survival.

The performance of GCS is often based on determining the risk of death following traumatic brain injury in the geriatric population. Some studies report that patient survival is poor in cases of decerebrate posture or in patients who do not respond to painful stimuli [11,12]. In our study, eight patients presented with a GCS score of 4, and 14 patients with a GCS score of 3; all of them either died or required long-term inpatient treatment. It was also observed that these patients had eTBI scores of −1 or 0 and were classified as high-risk. Brazinova et al. reported improved outcomes in 11% of elderly patients with a GCS score of 3 or 4 [13]. Bobeff et al. found survival in only one of 18 patients with a GCS score of 3 or 4 who were classified as high-risk according to the eTBI scoring system. Thus, it was stated that the high-risk category in the eTBI system was more specific for mortality [9]. In our study, 9 of the 22 patients with a GCS score of 4 or lower died (40.9%), while 13 were discharged with a GOS score of 1 or 2, requiring home care. Although no statistically significant difference was found in terms of mortality among patients in the high-risk group (p = 0.664), it was observed that the mortality rate in this group was higher than in the low-risk group (3.2%) and the medium-risk group (7.4%).

Due to the severity and chronicity of comorbidities and the decreased treatment response in elderly patients, selecting the appropriate treatment and informing patients and their relatives about it is crucial. In the study by Herou et al. [14], 100% mortality was observed in patients with a GCS score of 5 or lower who received conservative treatment. Similarly, Bobeff et al. [10] reported no survival among 34 patients with an eTBI score of 1 or lower who were treated conservatively. In our study, survival was observed in 6 out of 11 patients (54.5%) with a GCS score of 5 or lower, an eTBI score of 0, −1, or −2, who were classified as high-risk and received conservative treatment. When these patients were compared with the 11 patients who underwent surgery, no statistically significant difference was found (p = 0.664). This finding does not suggest that high-risk patients should be considered for surgical intervention. The discrepancy between our results and those of other studies may be attributed to factors such as the increasing number of neurointensive care units, the widespread use and standardization of invasive cerebral pressure monitoring techniques, enhanced safety measures against trauma (especially in traffic accidents) to mitigate brain injury severity, and reduced time to hospital admission and treatment. However, among 27 patients classified as medium-risk, 19 were treated conservatively, and no mortality was observed. Twelve (63.2%) of these patients were discharged with a GOS score of 1 or 2. Conservative treatment was found to be statistically significant in the medium-risk group (p = 0.022), whereas no significant statistical findings were obtained for other risk groups. These findings suggest that the effectiveness of the eTBI risk classification system in determining surgical indication is low when evaluated across all subgroups.

Shafiei et al. also examined the eTBI scoring system, reporting that while it demonstrated strong predictive power for negative outcomes in medium- and high-risk patients, its reliability in predicting good outcomes in low-risk patients was weaker, with 57% specificity and 87.7% positive predictive value, despite 98% sensitivity and 90% negative predictive value [15]. Our study revealed different results for this group. In terms of prognosis prediction, the low-risk group showed 99% sensitivity, 71.4% negative predictive value, 71% specificity, and 96.3% positive predictive value. Furthermore, 180 of 187 patients (96.2%) in the low-risk group were discharged with a GOS score of 3 or higher, making this the only statistically significant group (p < 0.001). These results indicate that the low-risk category is more effective in predicting prognosis compared to the medium- and high-risk groups. We believe that the differences in findings reported in the literature may be due to factors such as the diversity in the types, mechanisms, and localizations of traumatic brain injuries, the mild presentation of symptoms in the low-risk group leading to delayed hospital admission, and the late diagnosis of traumatic brain injury due to its symptoms being attributed to comorbid conditions in elderly patients.

This study has certain limitations. The retrospective collection and analysis of data, along with the reduction in sample size due to incomplete information, led to a lower number of patients. Variations in pathologies such as parenchymal hemorrhages, contusions, and herniation, which contribute to traumatic brain injury, as well as the variability in clinical findings depending on the localization and volume of the affected cerebral tissue, are considered significant limitations. This study also needs to be replicated and strengthened with a prospective and multicenter validation cohort. Future studies or scoring systems that evaluate each pathology separately and perhaps develop pathology-specific scoring criteria may yield more meaningful results.

5. Conclusions

The eTBI scoring system suggests that it is an effective tool for assessing mortality risk and predicting prognosis in certain subgroups over the age of 65. The mortality rate among patients in the high-risk group is higher than in other groups. However, determining surgical indications based on patient risk groups does not appear to be feasible. Additionally, a novel and distinct finding is that this scoring system is more effective in predicting mortality in medium-risk group patients and in predicting prognosis in low-risk group patients.