Diagnostic Accuracy of the Madras Head Injury Prognostication Scale (MHIPS) in Predicting Mortality among Traumatic Brain Injury Patients

Abstract

Background

Accurate prediction of outcomes in traumatic brain injury (TBI) is crucial for optimizing therapeutic interventions and improving patient survival rates.

Objectives

This article determines the diagnostic accuracy of Madras Head Injury Prognostication Scale (MHIPS) in predicting mortality among patients with TBI, and compares the performance of MHIPS scores with that of Corticosteroid Randomisation after Significant Head Injury (CRASH) and International Mission for Prognosis and Analysis of Clinical Trials (IMPACT) scores.

Materials and Methods

This was a prospective observational study conducted among patients (n = 100) with clinical evidence of TBI presenting to the Department of Emergency Medicine, R. L. Jalappa Hospital and Research Centre, Tamaka, Karnataka, India, between August 2023 and July 2024.

Results

Of the 100 patients, 92 patients (92.0%) were survivors of which 4 patients (4.0%) had disability and 8 patients died/nonsurvivors (8.0%). Age more than 40 years, higher heart rate, lower Glasgow Coma Scale scores, lower MHIPS scores, higher CRASH scores, and higher IMPACT scores were significantly (p < 0.05) associated with mortality among patients with TBI. However, gender, mode of injury, diagnosis, time to presentation, systolic blood pressure (BP), diastolic BP, and respiratory rate did not vary significantly between nonsurvivors and survivors in the present study (p > 0.05). The mean (standard deviation) duration of ventilation among nonsurvivors was 3.3 (2.2), and that among survivors was 0.5 (1.1)—the difference was statistically significant (p < 0.05). The area under the curve of MHIPS scores was 0.912, in comparison with 0.893 for CRASH scores and 0.927 for IMPACT scores (p < 0.05). The MHIPS scores, with a cutoff of 13.5, showed a sensitivity of 87.5%, specificity of 81.5%, positive predictive value (PPV) of 29.2%, and negative predictive value (NPV) of 98.7%. The CRASH scores, with a cutoff of 5.5, demonstrated a sensitivity of 87.5%, specificity of 53.3%, PPV of 14.0%, and NPV of 98.0%. The IMPACT scores, with a cutoff of 8.5, had a sensitivity of 87.5%, specificity of 91.3%, PPV of 46.7%, and NPV of 98.8%. All three scoring systems showed statistically significant predictive accuracy.

Conclusion

MHIPS, CRASH, and IMPACT are effective tools for prognosticating mortality in TBI patients. MHIPS score offers simplicity and ease of use, making it valuable in resource-limited environments.

Keywords

Corticosteroid Randomisation after Significant Head Injury (CRASH) - diagnostic accuracy - International Mission for Prognosis and Analysis of Clinical Trials (IMPACT) - Madras Head Injury Prognostication Scale (MHIPS) - traumatic brain injury (TBI)

Introduction

Traumatic brain injury (TBI) remains a significant global health concern, contributing to high rates of mortality and long-term disability among affected individuals.[1] It encompasses a spectrum of injuries ranging from mild concussions to severe, life-threatening trauma, necessitating precise prognostication tools to guide clinical management and resource allocation.[2] [3] [4] Accurate prediction of outcomes in TBI is crucial for optimizing therapeutic interventions and improving patient survival rates.[5] [6] Several prognostic scoring systems have been developed to facilitate this process, each incorporating a combination of clinical, radiological, and demographic variables to stratify patient risk.[7] Among these, the Madras Head Injury Prognostication Scale (MHIPS), Corticosteroid Randomisation after Significant Head Injury (CRASH) scores, and International Mission for Prognosis and Analysis of Clinical Trials (IMPACT) scores have emerged as prominent tools validated across diverse clinical settings.[7] [8] [9]

MHIPS is designed to provide a practical and user-friendly assessment of TBI severity, integrating parameters such as Glasgow Coma Scale (GCS) motor response, pupillary reactivity, and computed tomography (CT) findings.[10] CRASH scores, derived from a large multicenter trial, incorporate age, GCS, radiological features, and laboratory findings to predict mortality and functional outcomes in TBI patients. IMPACT scores utilize data from multiple clinical trials to refine prognostic models, offering comprehensive risk stratification based on patient demographics, clinical presentation, and neuroimaging findings.[11] [12]

Physicians utilize imaging techniques such as CT and magnetic resonance imaging to diagnose intracranial injuries resulting from head trauma. However, there are several challenges in imaging trauma patients suspected of TBI. While scanning has become integral to trauma diagnostics, not all hospitals have access to CT machines or trained technicians.[13] Furthermore, hemodynamically unstable patients may deteriorate during transport to radiology, and there is concern over the high levels of ionizing radiation exposure. Although GCS is widely used to predict TBI severity and associated outcomes, recent TBI-specific prognostic models have shown promising performance.[14] These models consider imaging and laboratory results alongside variables like GCS and pupillary reactivity. Prognostic models, which are statistical tools using multiple variables to predict specific outcomes, play a crucial role in efficiently allocating resources and communicating with families of affected individuals.[15] In the context of MHIPS, the oculocephalic response has been integrated as a prognostic indicator due to its correlation with brainstem function and outcomes in head injury cases.

Despite their widespread use, comparative studies evaluating the diagnostic accuracy of MHIPS, CRASH, and IMPACT scores in predicting mortality among TBI patients remain limited. Against this background, the objective of the present study was to determine the diagnostic accuracy of MHIPS in predicting mortality among patients with TBI, and to compare the performance of MHIPS scores with that of CRASH and IMPACT scores.

Materials and Methods

This was a prospective observational study conducted among patients with clinical evidence of TBI presenting to the Department of Emergency Medicine, R. L. Jalappa Hospital and Research Centre, Tamaka, Karnataka, India, between August 2023 and July 2024. The study was approved by the Institutional Human Ethics Committee. The present study enrolled patients of all age groups, both gender, presenting with clinical evidence of TBI and willing to provide informed written consent (either by patients or their primary caregivers). However, pregnant women with TBI, patients discharged against medical advice, and those not willing to participate in the study were excluded.

Ramesh et al[8] conducted a prospective observational study and noted the diagnostic accuracy of MHIPS to be 87.5%. Using this information, considering the type I error to be 5%, type II error to be 20% or power to be 80%, and nonresponse rate to be 10%, the minimum estimated sample size was 100 patients with 95% confidence. The patients were enrolled using nonprobability sampling—convenience/purposive sampling—complete enumeration of all patients presenting to the study settings in line with prespecified inclusion and exclusion criteria. We administered a predesigned, semistructured questionnaire among patients enrolled in the study/admitted to the hospital to capture sociodemographic characteristics (including age, gender), clinical history (including time to presentation, mode of injury), findings of general physical examination, systemic examination, radiological findings (including CT of the brain), and results of laboratory investigations (including blood glucose and hemoglobin). At the time of presentation, CRASH, IMPACT, and MHIPS scores were computed. The patients were followed up until discharge with/without disability or deceased. Disability was assessed based on patients having any focal neurological deficit that affects daily activities, at the time of discharge.

MHIPS is a clinical tool developed to predict outcomes in patients with TBI. It incorporates a range of clinical and radiological parameters to provide a prognostic score.[8] Key factors include the patient's age, best motor response in GCS score, pupillary reactivity, oculocephalic response, CT findings, and systemic injuries. MHIPS is designed to be practical and user-friendly, facilitating its application in various clinical settings, particularly in resource-limited environments. Its predictive accuracy has been validated in several studies, showing good correlation with patient outcomes such as mortality and functional status at follow-up. The scale's psychometric properties are robust, demonstrating high reliability and validity in predicting outcomes. The simplicity and ease of use of MHIPS make it a valuable tool for clinicians in making informed decisions about the prognosis and management of TBI patients.

The CRASH trial was a large, multicenter randomized controlled trial designed to assess the effects of corticosteroids on outcomes in patients with TBI. It also provided a prognostic model, derived from the large data set of enrolled patients.[9] The CRASH score consists of age, GCS, pupillary light reaction, presence of major extracranial injury, and CT scan findings (petechial hemorrhages, status of the third ventricle and basal cisterns, presence of traumatic subarachnoid hemorrhage [SAH], midline shift, and mass lesion). The psychometric properties of this model are strong, with high predictive accuracy validated across diverse populations.

The IMPACT initiative focuses on improving prognostic models for TBI by analyzing data from multiple clinical trials and observational studies. The IMPACT models consider a range of variables, including demographic information, clinical presentation, and radiological findings. These models have been validated and are widely used in both clinical practice and research settings.[14] They help clinicians estimate the likelihood of outcomes such as survival, functional recovery, and quality of life postinjury.

Statistical analysis: The data obtained in the present study was manually entered into Microsoft Excel, coded, recoded, and analyzed using Stata v17. All the categorical variables were summarized using frequencies and percentages. Continuous variables were summarized using mean (standard deviation [SD]) and/or median (interquartile range). Based on data normality tested using the Kolmogorov–Smirnov test and the Shapiro–Wilk test, the chi-square test or Fisher's exact test (for categorical variables) and independent “t” test or Mann–Whitney U test (for continuous variables) were used to test for association between independent and dependent study variables. Statistical significance was considered at p-value of less than 0.05. Receiver operating characteristic (ROC) analysis was done to compute the area under the curve (AUC). Diagnostic accuracy of CRASH, IMPACT, and MHIPS scores were assessed by estimating sensitivity, specificity, positive (PPV), and negative predictive values (NPV).

Results

The present study included a total of 100 patients with clinical evidence of TBI. Of the 100 patients, 92 patients (92.0%) were survivors (4.0% had disability) and 8 patients died/nonsurvivors (8.0%).

As shown in [Table 1], the mean (SD) MHIPS score was 10.6 (2.9). The distribution of factors under MHIPS score showed that 36% were older than 45 years, while 64% were between 15 and 45 years. When assessing the best motor response using the GCS, 11% scored between 1 and 2, 6% scored between 3 and 4, and 83% scored between 5 and 6. Pupillary reaction to light revealed that 6% had no reaction, 18% had an abnormal reaction, and 76% had a normal reaction. The oculocephalic response showed that 6% had no reaction, 15% had an abnormal reaction, and 79% had a normal reaction. CT findings indicated that 10% had severe injuries characterized by basal cisterns not seen, a midline shift greater than 5 mm, or an injury volume greater than 3 mm; 24% had moderate injuries with basal cisterns partially seen, a midline shift less than 5 mm, or an injury volume less than 3 mm; and 66% had normal CT findings. Regarding systemic injuries, 5% had severe injuries such as thoracic or abdominal injuries or fractures of more than two long bones, 6% had fractures of one or two long bones, and 89% had no systemic injuries.

Table 1 Descriptive analysis of patients, by Madras Head Injury Prognostication Scale (MHIPS)

		Number (N)	Percent (%)
Age (in years)	> 45	36	36.0
	15–45	64	64.0
	< 15	0	0.0
Best motor response in GCS	1–2	11	11.0
	3–4	6	6.0
	5–6	83	83.0
Pupils reaction to light	No reaction	6	6.0
	Abnormal	18	18.0
	Normal	76	76.0
Oculocephalic response	No reaction	6	6.0
	Abnormal	15	15.0
	Normal	79	79.0
CT findings	Basal cisterns not seen; midline shift > 5 mm; injury volume > 3 mm	10	10.0
	Basal cisterns partially seen; midline shift < 5 mm; injury volume < 3 mm	24	24.0
	Normal	66	66.0
Systemic injuries	Thoracic injuries; abdominal injuries; fracture of more than 2 long bones	5	5.0
	Fracture of 1 or 2 long bones	6	6.0
	No systemic injury	89	89.0

Abbreviations: CT, computed tomography; GCS, Glasgow Coma Scale.

As shown in [Table 2], the mean (SD) CRASH scores were 4.8 (2.0). The average age of the participants was 41.3 years, with a SD of 14.4 years. In terms of motor response, 8% had no response, 3% showed extension, 4% had abnormal flexion, 2% had normal flexion, 21% localized pain, and 62% obeyed commands. Pupillary reaction revealed that 76% had both pupils reacting, 12% had one pupil reacting, and 12% had no reaction. Hypoxia was present in 23% of the cases, while 77% had no hypoxia. Hypotension was observed in 26% of the participants, with 74% not experiencing it. CT findings showed that 22% had diffuse injury I, 56% had diffuse injury II, 10% had diffuse injury III, 9% had diffuse injury IV, 3% had an evacuated mass lesion, and none had a nonevacuated mass lesion. Traumatic SAH was present in 25% of the cases, and 19% had an epidural mass on CT. Blood glucose levels were below 6 mmol/L in 80% of the participants, between 6 and 8.9 mmol/L in 14%, between 9 and 11.9 mmol/L in 3%, and between 12 and 14.9 mmol/L in another 3%. Hemoglobin levels were below 9 g/dL in 8% of the participants, between 9 and 11.9 g/dL in 36%, between 12 and 14.9 g/dL in 49%, and above 15 g/dL in 7%.

Table 2 Descriptive analysis of patients, by Corticosteroid Randomisation after Significant Head Injury (CRASH)

		Number (N)	Percent (%)
Age (in years) Mean (SD)		41.3 (14.4)
Motor score	None	8	8.0
	Extension	3	3.0
	Abnormal flexion	4	4.0
	Normal flexion	2	2.0
	Localizes	21	21.0
	Obeys	62	62.0
Pupils	Both reacting	76	76.0
	One reacting	12	12.0
	None	12	12.0
Hypoxia	No	77	77.0
	Yes	23	23.0
Hypotension	No	74	74.0
	Yes	26	26.0
CT findings	Diffuse injury I	22	22.0
	Diffuse injury II	56	56.0
	Diffuse injury III	10	10.0
	Diffuse injury IV	9	9.0
	Evacuated mass lesion	3	3.0
	Non evacuated mass lesion	0	0.0
Traumatic subarachnoid hemorrhage	No	75	75.0
	Yes	25	25.0
Epidural mass on CT	No	81	81.0
	Yes	19	19.0
Glucose (mmol/L)	< 6	80	80.0
	6–8.9	14	14.0
	9–11.9	3	3.0
	12–14.9	3	3.0
Hemoglobin (g/dL)	< 9	8	8.0
	9–11.9	36	36.0
	12–14.9	49	49.0
	≥ 15	7	7.0

Abbreviations: CT, computed tomography; SD, standard deviation.

As shown in [Table 3], the mean (SD) IMPACT scores were 5.9 (4.4). The distribution of factors under IMPACT showed that 45% were younger than 40 years and 55% were older than 40 years. The mean GCS score was 12.2 with a SD of 3.9. Regarding pupillary response to light, 75% had both pupils reacting, 12% had one pupil reacting, and 13% had no reaction. Major extracranial injuries were present in 89% of the participants, while 11% had none. CT findings revealed that 23% had normal scans, whereas 77% had the presence of a bleed.

Table 3 Descriptive analysis of patients, by International Mission for Prognosis and Analysis of Clinical Trials (IMPACT)

		Number (N)	Percent (%)
Age (in years)	≤ 40	45	45.0
	> 40	55	55.0
GCS, mean (SD)		12.2 (3.9)
Pupils react to light	Both	75	75.0
	One	12	12.0
	None	13	13.0
Major extracranial injury	No	11	11.0
	Yes	89	89.0
CT findings	Normal	23	23.0
	Presence of bleed	77	77.0

Abbreviations: CT, computed tomography; GCS, Glasgow Coma Scale; SD, standard deviation.

Factors associated with mortality among patients with TBI: As shown in [Table 4], the test of association showed that age more than 40 years, higher heart rate (106.5 [21.6] among nonsurvivors and 86.3 [14.1] among survivors), lower GCS scores (3.8 [1.4] among nonsurvivors and 12.9 [3.2] among survivors), lower MHIPS scores (10.3 [1.0] among nonsurvivors and 14.8 [2.7] among survivors), higher CRASH scores (7.3 [1.2] among nonsurvivors and 4.6 [2.0] among survivors), and higher IMPACT scores (15.1 [5.1] among nonsurvivors and 5.1 [3.3] among survivors) were significantly (p < 0.05) associated with mortality among patients with TBI. However, the results showed that gender, mode of injury, diagnosis, time to presentation, systolic blood pressure (BP), diastolic BP, and respiratory rate did not vary significantly between nonsurvivors and survivors in the present study (p > 0.05).

Table 4 Factors associated with mortality among patients with traumatic brain injury

		Nonsurvivors N = 8	Survivors N = 92	Total N = 100	p-Value
		n (%)	n (%)	n (%)
Age (in years)	≤ 40	1 (12.5)	44 (47.8)	45 (45.0)	0.032[a]
	> 40	7 (87.5)	48 (52.2)	55 (55.0)
Gender	Female	1 (12.5)	12 (13.0)	13 (13.0)	0.965
	Male	7 (87.5)	80 (87.0)	87 (87.0)
Mode of injury	Assault	0 (0.0)	2 (2.2)	2 (2.0)	0.755
	Fall from height	1 (12.5)	6 (6.5)	7 (7.0)
	RTA	7 (87.5)	84 (91.3)	91 (91.0)
Diagnosis (the numbers are not mutually exclusive)	Subarachnoid	3 (37.5)	28 (30.4)	31 (31.0)	0.572
	Subdural	7 (87.5)	28 (30.4)	35 (35.0)
	Intraparenchymal	0 (0.0)	6 (6.5)	6 (6.0)
	Extradural	0 (0.0)	13 (14.1)	13 (13.0)
	Concussion	0 (0.0)	13 (14.1)	13 (13.0)
	Contusion	0 (0.0)	15 (16.3)	15 (15.0)
	DAI	1 (12.5)	2 (2.2)	3 (3.0)
Time to presentation Median (IQR)		2 hours 9 minutes (1:34–2:42)	1 hour 43 minutes (1:11–2:07)	1 hour 59 minutes (1:30–2:29)	0.377
Heart rate Mean (SD)		106.5 (21.6)	86.3 (14.1)	87.9 (15.7)	< 0.001[a]
Systolic blood pressure Mean (SD)		108.8 (45.5)	119.3 (15.8)	118.5 (19.6)	0.143
Diastolic blood pressure Mean (SD)		71.3 (26.4)	75.3 (10.3)	75.0 (12.2)	0.367
Respiratory rate Mean (SD)		16.3 (2.7)	18.5 (7.9)	18.3 (7.6)	0.433
Glasgow Coma Scale Mean (SD)		3.8 (1.4)	12.9 (3.2)	12.2 (3.9)	< 0.001[a]
MHIPS scores Mean (SD)		10.3 (1.0)	14.8 (2.7)	14.5 (2.9)	0.003[a]
CRASH scores Mean (SD)		7.3 (1.2)	4.6 (2.0)	4.8 (2.0)	< 0.001[a]
IMPACT scores Mean (SD)		15.1 (5.1)	5.1 (3.3)	5.9 (4.4)	< 0.001[a]
Duration of ICU stay (in days) Mean (SD)		3.6 (2.3)	3.3 (2.9)	3.3 (2.9)	0.732
Duration of ventilation (in days) Mean (SD)		3.3 (2.2)	0.5 (1.1)	0.7 (1.4)	< 0.001[a]
Total duration of hospital stay (in days) Mean (SD)		5.5 (4.7)	9.2 (5.5)	8.9 (5.5)	0.064

Abbreviations: CRASH, Corticosteroid Randomisation after Significant Head Injury; DAI, diffuse axonal injury; GCS, Glasgow Coma Scale; ICU, intensive care unit; IMPACT, International Mission for Prognosis and Analysis of Clinical Trials; IQR, interquartile range; MHIPS, Madras Head Injury Prognostication Scale; RTA, road traffic accident; SD, standard deviation.

^a Statistically significant at p < 0.05.

It was found that the mean (SD) duration of ventilation among nonsurvivors was 3.3 (2.2), and that among survivors was 0.5 (1.1)—the difference was found to be statistically significant (p < 0.05). However, the duration of intensive care unit (ICU) stay and total duration of hospital stay did not vary significantly between nonsurvivors and survivors (p > 0.05).

ROC analysis: As shown in [Fig. 1], the study compared three prognostic scoring systems for head injuries: the MHIPS, the CRASH scores, and the IMPACT scores. The MHIPS scores had an AUC of 0.912 (95% confidence interval [CI] 0.842–0.983). The CRASH scores had an AUC of 0.893 (95% CI 0.767–1.000). The IMPACT scores demonstrated the highest AUC at 0.965 (95% CI 0.927–1.000). As per [Table 5], the MHIPS, CRASH, and IMPACT all showed high predictive accuracy with statistically significant results.

Table 5 Receiver operating characteristic (ROC) analysis—area under the curve to predict mortality

	AUC	Upper CI	Lower CI	p-Value
MHIPS scores	0.912	0.842	0.983	< 0.001[a]
CRASH scores	0.893	0.767	1.000	< 0.001[a]
IMPACT scores	0.965	0.927	1.000	< 0.001[a]

Abbreviations: AUC, area under the curve; CI, confidence interval; CRASH, Corticosteroid Randomisation after Significant Head Injury; IMPACT, International Mission for Prognosis and Analysis of Clinical Trials; MHIPS, Madras Head Injury Prognostication Scale.

^a Statistically significant at p < 0.05.

Fig. 1 Receiver operating characteristic (ROC) analysis—area under the curve to predict mortality.

Diagnostic accuracy of the scoring systems: As shown in [Table 6], the study evaluated the predictive accuracy of the three prognostic scoring systems. The MHIPS scores, with a cutoff of 13.5, showed a sensitivity of 87.5%, specificity of 81.5%, PPV of 29.2%, and NPV of 98.7%. The CRASH scores, with a cutoff of 5.5, demonstrated a sensitivity of 87.5%, specificity of 53.3%, PPV of 14.0%, and NPV of 98.0%. The IMPACT scores, with a cutoff of 8.5, had a sensitivity of 87.5%, specificity of 91.3%, PPV of 46.7%, and NPV of 98.8%. All three scoring systems showed statistically significant predictive accuracy.

Table 6 Diagnostic accuracy of MHIPS, CRASH, and IMPACT scores to predict mortality

	Cutoff	Sensitivity (%)	Specificity (%)	PPV (%)	NPV (%)	p-Value
MHIPS scores	13.5	87.5	81.5	29.2	98.7	< 0.001[a]
CRASH scores	5.5	87.5	53.3	14.0	98.0	< 0.001[a]
IMPACT scores	8.5	87.5	91.3	46.7	98.8	< 0.001[a]

Abbreviations: CRASH, Corticosteroid Randomisation after Significant Head Injury; IMPACT, International Mission for Prognosis and Analysis of Clinical Trials; MHIPS, Madras Head Injury Prognostication Scale; NPV, negative predictive value; PPV, positive predictive value.

^a Statistically significant at p < 0.05.

Discussion

The present study aimed to evaluate the diagnostic accuracy of the MHIPS in predicting mortality among patients with TBI and compare its performance with the CRASH and IMPACT scores. The results demonstrated significant associations between higher age, heart rate, lower GCS scores, and higher prognostication scores (MHIPS, CRASH, and IMPACT) with mortality in TBI patients. Age has been consistently identified as a critical factor influencing outcomes in TBI. In the current study, patients older than 40 years were more likely to experience mortality, aligning with findings from previous research indicating that older age is associated with increased mortality and poorer functional outcomes.[16] [17] This correlation can be attributed to the diminished physiological reserve and higher prevalence of comorbidities in older individuals, which exacerbate the severity of TBI. The study observed a significantly higher mean heart rate among nonsurvivors compared with survivors. Elevated heart rate can be indicative of systemic stress, autonomic dysfunction, or underlying conditions such as hemorrhage and hypovolemia, all of which can worsen TBI outcomes.[18] This finding underscores the importance of monitoring and managing cardiovascular parameters in TBI patients to improve prognostic accuracy and outcomes. Lower GCS scores were significantly associated with mortality, reinforcing the scale's established utility in assessing TBI severity and predicting outcomes.[19] GCS remains a cornerstone in TBI evaluation due to its simplicity and strong correlation with patient prognosis, as lower scores often reflect severe brain injury and impaired consciousness levels. Interestingly, the study found no significant associations between gender, mode of injury, time to presentation, systolic and diastolic BP, respiratory rate, and mortality. This contrasts with some studies suggesting that factors like hypotension and hypoxia significantly impact TBI outcomes.[20] The lack of significant associations in the present study might be due to the sample size or specific population characteristics.

The mean duration of ventilation was significantly longer among nonsurvivors (3.3 days) compared with survivors (0.5 days), highlighting the severity of injury and the critical care needs of nonsurvivors. However, the duration of ICU and total hospital stay did not vary significantly between the groups, suggesting that while ventilation duration is a critical indicator of outcome, ICU and hospital stay durations alone may not sufficiently capture mortality risk in TBI patients.[21]

This study highlighted that higher MHIPS, CRASH, and IMPACT scores were significantly associated with increased mortality, indicating the robust prognostic value of these scales. The ROC analysis ([Fig. 1]) revealed that the IMPACT scores had the highest AUC at 0.965 (95% CI 0.927–1.000), indicating excellent discriminative ability. The MHIPS scores also showed a high AUC of 0.912 (95% CI 0.842–0.983), followed by the CRASH scores with an AUC of 0.893 (95% CI 0.767–1.000) ([Table 5]). These findings suggest that all three scoring systems are highly effective in predicting mortality among TBI patients, with the IMPACT score having a slight edge in predictive accuracy. The diagnostic accuracy metrics for the scoring systems further support their reliability.[22] The MHIPS scores, with a cutoff of 13.5, exhibited a sensitivity of 87.5%, specificity of 81.5%, PPV of 29.2%, and NPV of 98.7%. The CRASH scores, with a cutoff of 5.5, demonstrated a sensitivity of 87.5%, specificity of 53.3%, PPV of 14.0%, and NPV of 98.0%. The IMPACT scores, with a cutoff of 8.5, had a sensitivity of 87.5%, specificity of 91.3%, PPV of 46.7%, and NPV of 98.8% ([Table 6]). The high sensitivity and specificity of the IMPACT scores indicate that it is particularly effective in correctly identifying both survivors and nonsurvivors, making it a robust tool for clinical decision-making. The higher specificity of the IMPACT scores compared with MHIPS and CRASH suggests that it has a lower rate of false positives, which is crucial in avoiding unnecessary interventions in patients predicted to have poor outcomes.[23] The MHIPS scores, despite being slightly lower in AUC compared with IMPACT, showed comparable sensitivity and a very high NPV, making it a valuable tool, especially in resource-limited settings where simplicity and ease of use are critical.[8] The CRASH scores, while having a lower specificity and PPV, still provide valuable prognostic information and are supported by robust data from a large multicenter trial. The findings from this study have significant clinical implications. The high predictive accuracy of all three scoring systems underscores their utility in guiding clinical decisions and resource allocation in TBI management. The IMPACT score, with its highest AUC and specificity, could be particularly useful in tertiary care centers where comprehensive data collection and analysis are feasible.[15] The MHIPS score, with its simplicity and ease of use, can be a valuable tool in settings with limited resources, providing reliable prognostic information without requiring extensive data input.

Conclusion

The present study evaluated the diagnostic accuracy of the MHIPS in predicting mortality among patients with TBI and compared its performance with the CRASH and IMPACT scores. Our findings demonstrate that all three prognostic scoring systems—MHIPS, CRASH, and IMPACT—exhibit high predictive accuracy for mortality in TBI patients. Each scoring system has distinct advantages: the IMPACT score with its highest specificity and AUC is particularly suitable for comprehensive clinical settings, while the MHIPS score offers simplicity and ease of use, making it valuable in resource-limited environments. The significant associations between age, heart rate, GCS scores, and higher prognostic scores with mortality highlight the importance of these parameters in outcome prediction. Future research should focus on larger and more diverse populations to validate these findings further and explore the potential for these scoring systems to guide individualized patient management strategies in diverse clinical settings.

Conflict of Interest

None declared.

• References

• 1 Maas AIR, Menon DK, Adelson PD. et al; InTBIR Participants and Investigators. Traumatic brain injury: integrated approaches to improve prevention, clinical care, and research. Lancet Neurol 2017; 16 (12) 987-1048
• 2 Jennett B. The Glasgow Coma Scale: history and current practice. Trauma 2002; 4: 91-103
• 3 Koome G, Thuita F, Egondi T, Atela M. Association between traumatic brain injury (TBI) patterns and mortality: a retrospective case-control study. F1000 Res 2021; 10: 795
• 4 Fu TS, Jing R, McFaull SR, Cusimano MD. Recent trends in hospitalization and in-hospital mortality associated with traumatic brain injury in Canada: a nationwide, population-based study. J Trauma Acute Care Surg 2015; 79 (03) 449-454
• 5 Wang J, Yin MJ, Wen HC. Prediction performance of the machine learning model in predicting mortality risk in patients with traumatic brain injuries: a systematic review and meta-analysis. BMC Med Inform Decis Mak 2023; 23 (01) 142
• 6 Zuniga RDDR, Vieira RCA, Solla DJF. et al. Long-term outcome of traumatic brain injury patients with initial GCS of 3-5. World Neurosurg X 2024; 23: 100361
• 7 Steyerberg EW, Mushkudiani N, Perel P. et al. Predicting outcome after traumatic brain injury: development and international validation of prognostic scores based on admission characteristics. PLoS Med 2008; 5 (08) e165 , discussion e165
• 8 Ramesh VG, Thirumaran KP, Raja MC. A new scale for prognostication in head injury. J Clin Neurosci 2008; 15 (10) 1110-1113 , discussion 1113–1114
• 9 Perel P, Arango M, Clayton T. et al; MRC CRASH Trial Collaborators. Predicting outcome after traumatic brain injury: practical prognostic models based on large cohort of international patients. BMJ 2008; 336 (7641) 425-429
• 10 Tsuchiya R, Ooigawa H, Kimura T. et al. Study of certain easily available biochemical markers in prognostication in severe traumatic brain injury requiring surgery. Surg Neurol Int 2023; 14: 410
• 11 Sivashankar S A, Swamiyappan SS, Visweswaran V. et al. Biochemical and radiological factors for prognostication of traumatic brain injury: an institutional experience. Cureus 2023; 15 (06) e40999
• 12 Srinivasan US. A mathematical model for predicting the outcome in moderate head injury. Neurol India 2006; 54 (01) 28-32
• 13 Hukkelhoven CW, Rampen AJ, Maas AI. et al. Some prognostic models for traumatic brain injury were not valid. J Clin Epidemiol 2006; 59 (02) 132-143
• 14 Dijkland SA, Foks KA, Polinder S. et al. Prognosis in moderate and severe traumatic brain injury: a systematic review of contemporary models and validation studies. J Neurotrauma 2020; 37 (01) 1-13
• 15 Perel P, Edwards P, Wentz R, Roberts I. Systematic review of prognostic models in traumatic brain injury. BMC Med Inform Decis Mak 2006; 6: 38
• 16 McIntyre A, Mehta S, Aubut J, Dijkers M, Teasell RW. Mortality among older adults after a traumatic brain injury: a meta-analysis. Brain Inj 2013; 27 (01) 31-40
• 17 Reith FC, Van den Brande R, Synnot A, Gruen R, Maas AI. The reliability of the Glasgow Coma Scale: a systematic review. Intensive Care Med 2016; 42 (01) 3-15
• 18 Moppett IK. Traumatic brain injury: assessment, resuscitation and early management. Br J Anaesth 2007; 99 (01) 18-31
• 19 Teasdale G, Maas A, Lecky F, Manley G, Stocchetti N, Murray G. The Glasgow Coma Scale at 40 years: standing the test of time. Lancet Neurol 2014; 13 (08) 844-854
• 20 Manley GT, Maas AI. Traumatic brain injury: an international knowledge-based approach. JAMA 2013; 310 (05) 473-474
• 21 Dijkland SA, Helmrich IRAR, Nieboer D. et al; CENTER-TBI Participants and Investigators. Outcome prediction after moderate and severe traumatic brain injury: external validation of two established prognostic models in 1742 European patients. J Neurotrauma 2021; 38 (10) 1377-1388
• 22 Retel Helmrich IRA, Lingsma HF, Turgeon AF, Yamal JM, Steyerberg EW. Prognostic research in traumatic brain injury: markers, modeling, and methodological principles. J Neurotrauma 2021; 38 (18) 2502-2513
• 23 Steyerberg EW, Moons KG, van der Windt DA. et al; PROGRESS Group. Prognosis Research Strategy (PROGRESS) 3: prognostic model research. PLoS Med 2013; 10 (02) e1001381

Publikationsverlauf

Artikel online veröffentlicht:
02. Juni 2025