Comparison of glucose/potassium ratio and revised trauma score in predicting mortality in patients with isolated blunt head trauma

Abstract

Isolated blunt head trauma (IBHT) is a common condition that can result in high mortality, requiring effective prognostic evaluations. Methods such as glucose-to-potassium ratio (GPR) and revised trauma score (RTS) have been developed to determine the mortality risk and optimize the treatment strategies. High GPR levels and low RTS levels are associated with increased mortality. This study aimed to compare the effectiveness of GPR and RTS in predicting the prognosis in IBHT patients. A retrospective evaluation was conducted on demographic data, vital signs, laboratory values, and CT images of IBHT patients presenting to the emergency department (ED). GPR and RTS were calculated, and ED and in-hospital outcomes were assessed. 764 patients were included, with 70.9% male and a median age of 33. Median glucose and potassium levels were measured as 104 mg/dL and 4.2 mmol/L, respectively. CT imaging showed trauma-related pathology in 11.8% of patients, with isolated subarachnoid hemorrhage most common (30%). Ninety-one patients were hospitalized, 24 of whom underwent surgery (epidural hematoma was the most common pathology). 10 of the 91 hospitalized patients died. Both low RTS and high GPR were significantly associated with hospitalization (p < 0.001). RTS was significantly lower in patients with mortality (p < 0.001). No significant difference was found between patients’ GPR with mortality or not (p = 0.261). RTS > 11 provided high sensitivity for survival (99.2%). GPR was not significantly associated with mortality in this cohort in IBHT patients, but RTS is effective in predicting mortality in these patients.

Introduction

Isolated blunt head trauma (IBHT) is a common condition encountered worldwide due to reasons such as traffic accidents, falls, and violent incidents. IBHT can lead to brain damage, neurological complications, and high mortality rates. Clinical management of patients with head trauma requires rapid and effective prognostic assessments. In this context, determining patients’ mortality risk is critical to optimize treatment strategies.

Different methods have been developed to predict outcomes. Among these, the glucose-to-potassium ratio (GPR)^1,2 and revised trauma score (RTS)^3,4 are two prominent parameters. Glucose and potassium are electrolytes vital for body metabolism, and changes in these values after trauma may affect the patient’s clinical condition. High glucose levels and irregular potassium values may indicate impaired metabolic activity in the brain, which may increase mortality risk. Studies in the literature show that mortality increases as GPR increases^1,2.

In addition, trauma score systems, especially RTS, are widely used as clinical decision support systems. Although RTS is a score that emerged primarily for triage assessment, it is a system that evaluates the clinical status of patients. This score reflects the general condition of the patient by combining factors such as systolic blood pressure (SBP), Glasgow Coma Score (GCS), and respiratory rate (RR)^3,4. RTS is widely used in determining the prognosis of trauma patients, and its effectiveness has been shown in many studies^3,4. In the original research in which RTS was defined, it was stated that the threshold value of RTS ≤ 11 had the highest sensitivity and specificity and detected 97.2% of patients with head trauma-related mortality³.

This study aims to compare the effectiveness of GPR and RTS in predicting prognosis in isolated head trauma patients. It also seeks to calculate the highest sensitivity and specificity threshold value in cases where GPR and RTS are significant in predicting any clinical outcome. In this context, the results of studies on the prognostic value of the two parameters were examined, and the advantages and limitations of both methods were evaluated.

Materials and methods

This study was conducted retrospectively and multi-centered in tertiary Emergency Departments (ED). The multi-centered approach was chosen to ensure a diverse and representative patient population, as well as to increase the generalizability of the study’s findings. The hospitals where the study was performed have an average of 50,000–60,000 patients admitted to the ED monthly. Ethical approval was received from Health Sciences University İzmir Tepecik Education and Research Hospital’s ethics committee (Decision number: 2024/12 − 08 Date: 09.01.2025). The study was performed using the tenets of the Declaration of Helsinki. Due to the retrospective nature of the study, Health Sciences University İzmir Tepecik Education and Research Hospital’s ethics committee waived the need of obtaining informed consent.

Study population

Patient information was retrieved by both authors from the Hospital Information Management Systems (HIMS) of their respective institutions. The study was conducted between January 1, 2023, and December 31, 2023. All patients who presented to the ED during the specified time frame and underwent brain CT imaging, blood glucose testing, and serum potassium measurement concurrently were included in the study. Each case was reviewed individually, and patients who met the exclusion criteria were excluded.

Patient selection

Inclusion criteria

Among the patients over the age of 18, patients who had brain Computed Tomography (CT) imaging and requested glucose and potassium as laboratory tests were included in the study.

Exclusion criteria

Patients who met any of the following criteria were excluded from the study:

• Applied to the ED with cardiac arrest,

• Underwent brain CT for reasons other than trauma,

• Findings of chronic pathology requiring surgery in brain CT (chronic subdural hematoma, etc.)

• Detected pathological findings due to trauma in CT images other than brain CT,

• Detected fractures in X-ray imaging,

• Those who had hemolysis, etc., that could change laboratory values (In the hospitals where the study is conducted, if there is hemolysis, etc., in the blood sample, it is stated in the laboratory results),

• Patients with diabetes mellitus,

• Patients with malignancy,

• Patients with chronic renal failure,

• Left the ED with a treatment refusal signature,

• Referred from the ED to another hospital,

• Applied to the ED for head trauma but was hospitalized for reasons other than trauma.

• Left from the department where he was admitted with a treatment refusal signature,

• Referred from the department where he was admitted to another hospital.

Data collection and standards

The patients’ sociodemographic and clinical features, CT images and reports, ED outcomes, in-hospital outcomes, and other relevant data were meticulously collected from the HIMS files, ensuring a comprehensive understanding of each case.

Glucose and potassium results from laboratory tests taken during ED admission were considered. The same glucose and potassium kits were used in both hospitals. Therefore, the unit and threshold values of both values were the same in both hospitals. Glucose value levels were measured via “Roche” laboratory kit, and the levels between 70 and 110 mg/dL were accepted as normal. Potassium value levels were measured via “Roche” laboratory kit, and the levels between 3.5 and 5.1 mmol/L were accepted as normal. The GPR of the patients was calculated by dividing these data. GCS, RR (/min), and SBP (mmHg) of the patients were recorded. The patients’ RTS was calculated by summing the scores corresponding to each parameter (Table 1).

Table 1.

Revised trauma score (RTS).

Point	GCS	SBP (mmHg)	RR (/min)
4	13–15	≥ 90	10–29
3	9–12	76–89	≥ 30
2	6–8	50–75	6–9
1	4–5	1–49	1–5
0	3	0	0

Data availability

The datasets generated and/or analyzed during the current study are not publicly available because it is impossible to integrate data in OMV format into the manuscript. However, they are available from the corresponding author at a reasonable request.

Statistical analysis

Statistical analyses were performed using Jamovi Ver. 2.3.28 software package (The Jamovi Project, Sydney, Australia). The Kolmogorov-Smirnov test was used to check the normality of data distribution. When the amount of data was less than 50, the Shapiro-Wilk test was used to check the normality of data distribution. Categorical data were presented as frequency and percentages, numerical data as mean and standard deviation if normally distributed, and median and interquartile range (25–75 percentiles; IQR) values otherwise. An independent sample t-test was used to compare two groups in terms of normally distributed data. Mann-Whitney U test was used to compare two groups in terms of non-normally distributed data. One-way ANOVA was used to compare more than two independent normally distributed data. Kruskal-Wallis analysis of variance was used to analyze non-normally distributed data in more than two independent groups. Statistical significance was taken as p < 0.05. Violin plots were used to show the distributions of groups with statistically significant differences but no difference in median values and were presented together with density curves. Youden’s index was used to determine the optimal sensitivity and specificity values.

Results

In this study, 1,942 patients matched the inclusion criteria. After excluding 1,178 patients who met the exclusion criteria, the study was conducted with the eligible 764 patients (Fig. 1).

Fig. 1.

Study Flow Chart.

In this study, 764 patients were included. 70.9% (n = 542) of the patients were male, and the median age was 33 (24–47; IQR:23) years. The patients’ median glucose level was 104 mg/dL (92–120; IQR:28.3), and the potassium level was 4.2 mmol/L (3.90–4.40; IQR:0.502). There was a pathology due to trauma on CT images of 90 patients (11.8%). In patients with any pathology on CT images, the most common pathology was isolated subarachnoid hemorrhage (n = 27, 30%). Ninety-one patients from the ED were admitted to the hospital (11.9%). Twenty-four of these patients underwent surgery during hospitalization (26.4%). The most common pathology in operated patients was epidural hematoma (n = 8, 33.3%). Of the 91 patients admitted to the hospital, 10 resulted in mortality. No patient died in the ED (Table 2).

Table 2.

Characteristics of the patients.

	Number (n)	Percent (%)
Gender
Men	542	70.9
Women	222	29.1
Pathological findings in CT
No pathology	674	88.2
Any pathology present	90	11.8
Pathological findings in CT
Subarachnoid hemorrhage (isolated)	27	30
Acute subdural hematoma (isolated)	17	18.9
Acute subdural hematoma and Subarachnoid hemorrhage	14	15.6
Epidural hematoma (isolated)	11	12.2
Linear skull fracture (isolated)	11	12.2
Cerebral contusion	5	5.6
Epidural hematoma and Subarachnoid hemorrhage	3	3.3
Depressed skull fracture (isolated)	2	2.2
ED outcome
Discharge from ED	673	88.1
Hospitalization	91	11.9
Operation during hospitalization
Yes	24	26.4
Epidural hematoma (isolated)	8	33.3
Acute subdural hematoma (isolated)	6	25
Acute subdural hematoma and Subarachnoid hemorrhage	3	12.5
Subarachnoid hemorrhage (isolated)	2	8.3
Depressed skull fracture (isolated)	2	8.3
Other	3	12.5
No	67	73.6
In-hospital outcome
Discharge from hospital	81	89
Exitus	10	11

There was no statistically significant relationship between the pathology present on CT and the patient’s in-hospital outcome and mortality (p = 0.393). Still, there was a statistically significant relationship between the pathology present on CT and the patient’s operation. The operation rate was significantly higher in patients with epidural hematoma than in patients with subarachnoid hemorrhage (p < 0.001). There was no significant difference in whether or not to have an operation among other pathologies seen on CT.

When patients were grouped according to the presence or absence of traumatic pathology on their CTs, there was a statistically significant difference between the RTS of patients with traumatic pathology on their CTs and the RTS of patients without traumatic pathology on their CTs (p < 0.001). Similarly, there was a statistically significant difference between the GPR of patients with traumatic pathology on their CTs and the GPR of patients without traumatic pathology on their CTs (p < 0.001). While the median GPR value of patients without traumatic pathology on their CT was 24.5 (21.6–28.5; IQR:6.95), The median GPR value of patients with traumatic pathology was found to be 29.4 (25.8–36; IQR:10.1). Figure 2 was created using a violin plot to show the RTS distribution of patients with and without traumatic pathology on CT. The distributions of both groups are shown with a density curve (Fig. 2).

Fig. 2.

RTS distribution of patients with and without any pathology on CT.

The most appropriate GPR cut-off value for predicting the presence of traumatic pathology in patients’ CTs was calculated as 25.8139534883721 (Sensitivity: 75.56%, Specificity: 60.83%, PPV: 20.48%, NPV: 4.91%, Youden’s index: 0.364, AUC: 0.726, Metric Score: 1.36).

There was a statistically significant difference between both RTS and GPR of patients discharged from the emergency department and hospitalized patients (p < 0.001). The median GPR value of patients discharged from the emergency department was 24.5 (21.6–28.5; IQR:6.91), while the median GPR value of patients hospitalized was 29.2 (21.6–35.9; IQR:10.3). Figure 3 was created using a violin plot to show the RTS distribution of patients discharged from the emergency department and hospitalized patients. The distributions of both groups are shown together with the density curve. (Fig. 3).

Fig. 3.

RTS distribution of patients discharged from the ED and admitted to hospital.

The most appropriate GPR cut-off value for predicting whether patients will need hospitalization was calculated as 25.8139534883721 (Sensitivity: 73.63%, Specificity: 60.62%, PPV: 20.18%, NPV: 94.44%, Youden’s index: 0.343, AUC: 0.714, Metric Score: 1.34).

There was no statistically significant difference between the RTS of patients who were hospitalized for trauma and operated on and those who were not (p = 0.49). Similarly, there was no statistically significant difference between the GPR of these patients (p = 0.14).

In hospitalized patients, there was no significant difference between the GPR of patients discharged and those who died in the hospital (p = 0.261). The median GPR value of patients discharged from the hospital was 28.9 (25.6–35.2; IQR:9.65), while the median GPR value of patients who died in the hospital was 34.5 (26.8–42.1; IQR:15.4). When the RTS was examined to compare the in-hospital outcomes of hospitalized patients; there was a statistically significant difference between the RTS of patients discharged from the hospital and those who died in the hospital (p < 0.001). The median RTS value of patients discharged from the hospital was 12 (12–12; IQR:0), while the median RTS value of patients who died in the hospital was 8 (8-11.5; IQR:3.5).

In hospitalized patients, the most appropriate RTS cut-off value that could be used to predict in-hospital survival was calculated as 11 (Sensitivity: 92.59%, Specificity: 70%, PPV: 96.15%, NPV: 53.855%, Youden’s index: 0.626, AUC: 0.814, Metric Score: 1.63).

In all patients, there was no significant difference between the GPR of patients with and without mortality (p = 0.008). While the median GPR value of patients without mortality was 24.9 (21.8–29.0; IQR:7.30), the median GPR value of patients with mortality was 34.5 (26.8–42.1; IQR:15.4). There was a significant difference between RTS of patients with and without mortality (p < 0.001). While the median RTS value of patients without mortality was 12 (12–12; IQR:0), the median RTS value of patients with mortality was 8 (8-11.5; IQR:3.5).

In all patients, the most appropriate RTS cut-off value that could be used to predict survival was calculated as 11 (Sensitivity: 99.2%, Specificity: 70%, PPV: 99.6%, NPV: 53.85%, Youden’s index: 0.692, AUC: 0.846, Metric Score: 1.69).

Discussion

This study investigated the relationships between clinical characteristics, CT findings, RTS and GPR, and in-hospital outcomes of trauma patients. Our results emphasize the importance of RTS and GPR for the clinical management of trauma patients. Also, the impact of pathologies seen on CT on patients’ treatment needs and hospitalization status was discussed. Many of the findings are consistent with those reported data in the current literature.

We observed that men were more likely to visit the emergency department due to head trauma than women. Studies by Rao V. et al. (2000)⁵ and Ki Seong, Eom, et al. (2021)⁶ also showed that emergency department visits due to traumatic brain injury were more than twice as common in men as in women. Our current findings are consistent with the literature.

The frequency of CT findings resulting from trauma has also been reported in the literature. One study reported that approximately 35–40% of patients with moderate head trauma had subarachnoid hemorrhage⁷. Another study found that subdural hematoma was seen at a rate of 12–29% after head trauma⁸. A study evaluating the frequency of epidural hematoma reported that it was seen at a rate of 0.2–12%. Our study found that the proportional distribution of pathologies present in CT scans of patients with any pathology on their CT scans was consistent with the rates in the literature.

A study conducted in the United States found that there were approximately 2.8 million emergency room visits due to head trauma; approximately 282,000 (10.1%) of these patients were hospitalized, and 56,000 (2%) died⁹. In our study, we found that the hospitalization and mortality rates in patients admitted due to head trauma were like the literature.

There are studies in the literature that both GPR and RTS may indicate poor outcomes in patients with head trauma^1–4. However, we could not find a study in which these scores can predict whether there will be any pathology on CT images in patients presenting with head trauma. Our study provided statistically significant data that both GPR and RTS can predict the presence of pathological findings on CT. According to our study, high GPR and low RTS can predict the presence of pathological findings on CT. Patients with GPR greater than 25.814 are more likely to have any pathology on their CTs (Sensitivity: 75.56%, Specificity: 60.83%). Although statistically significant, no significant cut-off value could be determined in RTS to predict whether there will be any pathology on CT images in our study.

Similarly, we did not find any studies on whether these two parameters can be used to predict the need for hospitalization. Our study provided statistically significant data indicating that GPR and RTS can predict the need for hospitalization. According to our study, high GPR and low RTS can predict the need for hospitalization. Patients with GPR greater than 25.814 are likelier to need hospitalization (Sensitivity: 73.63%, Specificity: 60.62%). Although statistically significant, no significant cut-off value could be determined in RTS to predict the need for hospitalization in our study.

In both groups, GPR cut-off values did not have very high sensitivity and specificity, and Youden’s indexes were lower than 0.5.

Studies indicate that GPR can predict poor outcomes and mortality in different patient groups. GPR is significant in predicting mortality in patients with blunt abdominal trauma¹⁰, blunt thoracic trauma¹¹, aneurysmatic subarachnoid hemorrhage¹, or intracerebral hemorrhage¹². There are also studies in the literature investigating the effectiveness of GPR in patients with head trauma. There is a study indicating that GPR can be used to predict traumatic brain injury and that the risk of adverse outcomes increases as GPR increases. However, this study examined the modified ranking scale (mRS) values of the patients as “poor outcome” and included patients with mRS ≥ 4 in this group². In another study indicating that GPR can predict poor outcomes in traumatic brain injury, patients who died or were in a vegetative state (Glasgow Outcome Scale [GOS] = 1, 2) were accepted as poor outcomes. This study found that patients with GPR ≥ 50 had a high rate of poor outcomes¹³. These two studies’ outcome criteria differ from ours in that they did not only look at mortality.

In our study, only mortality was considered a poor outcome and evaluated according to this criterion. GPR was not found to be effective in predicting mortality in patients admitted with head trauma. This may be the reason for the difference in the existing literature.

In the study where RTS was defined, it was found that RTS analyzed the outcome of head trauma well and predicted mortality at the highest rate at RTS ≤ 11 values³. In another study, it was found that it could be used to predict poor outcomes in patients with head trauma and that it was highly sensitive and specific at the threshold value of 9.5⁴. In different studies where RTS was evaluated in predicting mortality in isolated head trauma, low RTS was found to be significant in predicting mortality^14,15. Like the literature, our study found that low RTS significantly predicted mortality. In addition, like the study conducted by Champion, H R et al. (1989)³, we found that the threshold value of RTS > 11 was significant in predicting survival. Our current finding was consistent with the literature.

Study limitations

This study was conducted retrospectively, cross-sectionally, and observationally. The patients’ reasons for admission, vital signs, and comorbid diseases were obtained from HIMS. In our retrospective study, the possibility that these data were missing or incorrect cannot be ruled out. CT images were evaluated based on CT reports written in the hospital. A limitation is that evaluations may differ among different specialists and that reports may be subjective. In addition, mortality was considered the only poor outcome in our study. For example, post-traumatic morbidity or quality of life of the patients was not evaluated. Only the mortality of the patients in the emergency department and the hospital was followed, and mortality that may develop in longer processes was not evaluated.

Conclusion

This study shows that GPR and RTS scores are important in the clinical management of trauma patients. Both low RTS and high GPR show a significant relationship between the presence of traumatic pathology on CT images and the need for hospitalization. However, optimal cut-off values that demonstrate both high sensitivity and specificity could not be determined for RTS. In predicting mortality, low RTS was found to be associated with the presence of mortality, and the ideal cut-off value was determined in accordance with the current literature. It was determined that GPR was not effective in predicting mortality due to blunt head trauma. Future studies should combine these scores with a larger population and predict the outcomes of patients more accurately by prospectively observing them.

Author contributions

A.U. and A.D. collected and scanned data at the hospital where they worked. A.U. wrote the manuscript’s main text, prepared the tables and figures, and performed the statistics. All authors reviewed the manuscript.

Data availability

The datasets generated and/or analyzed during the current study are not publicly available because it is impossible to integrate data in OMV format into the manuscript. However, they are available from the corresponding author at a reasonable request.

Declarations

Competing interests

The authors declare no competing interests.