Ernest J. Bobeff1, Konrad Stawiski2, Dariusz J. Jaskólski1
1. Department of Neurosurgery and Neuro-oncology, Medical University of Lodz, Barlicki University Hospital, Kopcinskiego 22 Street, 90-153 Lodz, Poland
2. Department of Biostatistics and Translational Medicine, Medical University of Lodz, Mazowiecka 15 Street, 92-215 Lodz, Poland
Traumatic brain injury (TBI) is a growing public health concern, one that places a great burden on affected patients, their families, and society. Its incidence peaks in adolescence and older adulthood, though for the most part, the mechanism of injury differs between these two groups. The incidence continues to rise as both the number of traffic accidents and the aging population increase.
Reporting the state of a patient after TBI oftentimes poses a major challenge - but may be urgently required to triage, provide effective treatment, or arrange transfer to another hospital. A simple method for describing injury and assessing an emergency situation is a scoring system, which have been an object of research for decades. Scoring systems have become ubiquitous across various clinical specialties - in particular, those included in trauma care. Nowadays they enable healthcare workers to correctly diagnose, treat, and predict outcome, as well as facilitate post-hoc evaluation of prescribed therapy.
There are a few basic principles behind the development of a new scoring system. The first one is the right choice of an outcome measure relevant to specific clinical scenario. For example, in older adults who have suffered TBI, the primary goal of treatment is the avoidance of chronic institutionalization. Therefore, assessing return to work rate seems somewhat pointless when many patients are already retired. Another key point is the availability of the scoring system, which becomes absolutely crucial at the accident and emergency department (A&E). Ideally, such a tool should be based on routinely used medical tests, yet contain enough information about patient’s status. It could help to make full use of the “golden hour” during which effective management of TBI is of utmost importance.
In view of the foregoing, the optimal scoring system should be also well-calibrated and provide a high level of discrimination. These issues need to be balanced with the simplicity that paves the way for its practical application. It is therefore mandatory to perform a thorough statistical analysis of comprehensive data with inclusion of possible confounders reported in the literature. Finally, it is important that such a tool should be applicable worldwide. In the long run, such a scoring system may offer far-reaching benefits in terms of global neurosurgical outcomes and better understanding of TBI.
Herein we present some examples of scoring systems designed for clinical use in TBI (Table 1). We highlight their importance as a common language for physicians involved in both clinical management and research. We also introduce the Elderly Traumatic Brain Injury Score, eTBI Score (2019).
Comparisons and considerations on trauma scoring systems
In 1974 Baker et al. proposed a simple method to determine the overall severity of injury, the Injury Severity Score (ISS). At present, despite being slightly dated, it remains the most frequently cited trauma scale. Interestingly enough, the authors suggested further validation and improvement of the proposed classification, which was later accomplished with the introduction of the New Injury Severity Score (Osler et al., 1997). A point often overlooked is that all participants of the study had been involved in traffic accidents; therefore, the resulting mortality rates may be less reliable in cases of different mechanisms of injury.
Glasgow Coma Scale (GCS) was introduced in the mid-seventies (Teasdale et al., 1974) and became remarkably widespread among diverse medical specialists who deal with TBI patients. Numerous validation studies have provided evidence for the prognostic value of the GCS. Thus, it was included in trauma triage protocols worldwide. A number of studies have also postulated a convergence between the three components of the scale though awareness of this is not recent. In 1979 Teasdale et al. reported that assessing Best Motor Response Score alone made fine discrimination between patients with survival better than vegetative and the rest.
There are two major and externally validated models for outcome prediction after TBI: CRASH (Perel et al., 2008) and IMPACT (Steyerberg et al., 2008). However, with younger age groups (mean age 37 and 30 years, respectively), caution must be applied as these findings might not be relevant in the elderly. Neither of them accounts for comorbid illness, prescription medication use, and other measures of frailty that are known to be associated with adverse outcomes in older patients.
Future directions of data analysis
The concept of risk stratification is important in the current era of precision medicine. Until recently, most of scoring systems were derived based on experts’ opinion or tools that are mathematically simple. As the amount of individual patient data becomes greater and medicine is integrating with the big data movement, more sophisticated methods of multivariate modeling and data mining are being applied. The eTBI Score is based on simplified results from logistic regression performed with the special consideration of hyperparameter tuning and feature selection (counteracting overfitting). It is one of the simplest modeling methods, yet with great potential. The stack of logistic regression models could be understood as artificial neural networks, which are currently revolutionizing the field of artificial intelligence in the areas of speech and image recognition.
The concept of application of data mining methods in risk stratification is not novel and has already been proven at the bedside. For example, in 1993, recursive partitioning analysis allowed for the development of still-being-used malignant glioma risk stratification classes (Curran et al.,1993). Yet the amount of openly accessible clinical data remains too limited to effectively utilize novel techniques like deep learning or Q-learning.
Elderly Traumatic Brain Injury Score
The eTBI Score is a scoring system for risk stratification in elderly patients after head injury. To assess the mortality rate, one needs information that is usually available at the A&E, namely: neurological state, complete blood count, and patient’s comorbidities. It is noteworthy that although radiological findings were included in the modeling as well, they were discarded after the backward stepwise logistic regression procedure. The final score ranges from -2 to 6 points as a derivative of GCS Motor Score (Table 2). The 30-day mortality or vegetative state probability increases as the score decreases, hence all patients with an eTBI Score ≤1 died in a 30-day period, whereas 83% of the study group presenting with an eTBI Score equal to 6 were discharged home. The positive and negative predictive values amounted to 83% and 90%, respectively.
This new tool may help understand potential threats of head trauma in the elderly while demonstrating the relative importance of different aspects of the patient’s condition. Platelets are generally seen as a factor strongly related to blood coagulation. On the other hand, it is supposed that they may have the capacity for neuroprotection by means of neuroinflammation and stimulation of neuronal functions and synaptic plasticity after injury. RDW-CV is a measure of anisocytosis used in the diagnosis of anemia. It is also an extensively studied mortality risk factor for various medical conditions, including head trauma. We consider it might be an indirect measure of biological aging, becoming an even more significant biomarker than age itself.
A note of caution is due here since mortality risk scoring systems might be misused. Poor outcome in critical care patients can be an effect of self-fulfilling prophecy after the withdrawal of treatment in a preselected group. A reasonable approach to tackle this issue could be to avoid extrapolation of mortality prediction model results to individual patients. We assume that the eTBI Score could well serve as an initial metric of outcome; nonetheless, it would totally fail to (1) indicate a treatment-limiting order , or (2) replace experience-based decision making.
Final remarks
Medical scoring systems support clinical decision-making, thus clearing the way for more efficient and accurate management of patients. The eTBI Score is designed specifically for the elderly after TBI. It was developed in the best way to assure its resilience to overfitting. Further simplification allows for broad translation to the bedside; therefore, the new score establishes a quantitative framework for risk stratification at the level of the A&E. It could immediately serve as an outcome predictor both for the patient and for her/his family. This work contributes to existing knowledge of head trauma by providing the relative significance of clinical, radiological, and laboratory findings in the geriatric population.
References
1. Baker SP, O'Neill B, Haddon W Jr, Long WB. The injury severity score: a method for describing patients with multiple injuries and evaluating emergency care. J Trauma. 1974 Mar;14(3):187-96.
2. Bobeff EJ, Fortuniak J, Bryszewski B, Wiśniewski K, Bryl M, Kwiecień K, Stawiski K, Jaskólski DJ. Mortality after traumatic brain injury in elderly patients: a new scoring system. World Neurosurg. 2019 Apr 11
3. Curran WJ Jr, Scott CB, Horton J, Nelson JS, Weinstein AS, Fischbach AJ, Chang CH, Rotman M, Asbell SO, Krisch RE, et al. Recursive partitioning analysis of prognostic factors in three Radiation Therapy Oncology Group malignant glioma trials. J Natl Cancer Inst. 1993 May 5;85(9):704-10.
4. MRC CRASH Trial Collaborators, Perel P, Arango M, Clayton T, Edwards P, Komolafe E, Poccock S, Roberts I, Shakur H, Steyerberg E, Yutthakasemsunt S. Predicting outcome after traumatic brain injury: practical prognostic models based on large cohort of international patients. BMJ. 2008 Feb 23;336(7641):425-9
5. Osler T, Baker SP, Long W. A modification of the injury severity score that both improves accuracy and simplifies scoring. J Trauma. 1997 Dec;43(6):922-5 6. Steyerberg EW, Mushkudiani N, Perel P, Butcher I, Lu J, McHugh GS, Murray GD, Marmarou A, Roberts I, Habbema JD, Maas AI. Predicting outcome after traumatic brain injury: development and international validation of prognostic scores based on admission characteristics. PLoS Med. 2008 Aug 5;5(8):e165
7. Teasdale G, Jennett B. Assessment of coma and impaired consciousness. A practical scale. Lancet. 1974 Jul 13;2(7872):81-4
8. Teasdale G, Murray G, Parker L, Jennett B. Adding up the Glasgow Coma Score. Acta Neurochir Suppl (Wien). 1979;28(1):13-6
Scoring system
Author, Year
Number of patients
Age
Method
Outcome predictors
Outcome measure
Injury Severity Score (ISS)
Baker, 1974
2128
Not stated
Observational study, multicenter
Three highest Abbreviated
Injury Scale severity codes
3-month mortality
Best Motor Response
Score
Teasdale, 1979
1000
Mean 34
Observational study, single center
GCS Motor Score
6-month survival better than vegetative
Marshall CT Score
Marshall, 1991
746
Midline shift, aspect of basal cisterns, high or mixed density lesion >25 cm3
Outcome at discharge
Head Injury Severity Scale
Stein, 1995
24841
GCS
Complication rate
Full Outline of UnResponsiveness (FOUR) Score
Wijdicks, 2005
120
Median 60 (45-70)
Eye response, motor response, brainstem reflexes, respiration
In-hospital mortality,
3-month unfavorable outcome
Rotterdam CT Score
Maas, 2005
2249
RCT, multicenter, LR
IVH, tSAH, midline shift,
EDH, aspect of basal cisterns
6-month mortality
IMPACT models
Steyerberg, 2008
8509
Median 30 (21–45)
Age, GCS motor score, pupillary reactivity, CT classification, hypoxia, hypotension, tSAH, EDH, glucose and hemoglobin
6-month mortality,
6-month unfavorable outcome
CRASH models
Perel, 2008
10008
Mean 37
±17.1
Age, GCS, pupil reactivity, presence of major extra- cranial injury
14-day mortality,
Nijmegen models
Jacobs, 2013
700
Mean 43.8
±20.4
Observational study, single center, LR
Age, GCS, pupil reactivity, hypotensive episode, aspect of ambient cistern and fourth ventricle, volume of
dominant lesion, contusions, EDH, SDH, tSAH, IVH
Geriatric Trauma
Outcome (GTO) Score
Zhao, 2015
3841
Mean 76.5
±8.1
Age, ISS, amount of packed red blood cells transfused in
24 hours
Inpatient survival
Surviving Penetrating Injury to the Brain (SPIN) Score
Muehlschlegel, 2016
413
Mean 33
±16
Observational study, two centers, LR
Age, GCS motor score, ISS, pupil reactivity, self-inflicted injury, IVH, aspect of basal cisterns, sex, INR, transfer from other hospital
6-month survival
Subdural Hematoma in the Elderly (SHE) Score
Alford, 2019
469
Age, GCS, SDH volume
30-day mortality
Elderly Traumatic Brain
Injury (eTBI) Score
Bobeff, 2019
214
Mean 77,8
±8,8
GCS Motor Score, PLT, RDW- CV, comorbidities
30-day mortality or vegetative state
Table 1. Scoring systems developed for patients after traumatic brain injury
CT – computed tomography; EDH – epidural hematoma; GCS – Glasgow Coma Scale; IVH – intraventricular hemorrhage; LR – logistic regression; PLT – platelets; RDW-CV – red blood cell distribution width, coefficient of variation; SDH – subdural hematoma; tSAH – traumatic subarachnoid hemorrhage
Table 2. Determination of the Elderly Traumatic Brain Injury Score, eTBI Score
[All News]