Review ArticleA systematic review finds methodological improvements necessary for prognostic models in determining traumatic brain injury outcomes
Introduction
Traumatic brain injury (TBI) remains the main cause of death and disability in young people. Worldwide, injury is the cause of the largest number of disability-adjusted life years lost, which includes years lost to death and to varying degrees of disability [1]. TBI is by definition a heterogeneous disease, in terms of cause, pathology, severity, and prognosis. The wide variation in prognosis is already captured in the Hippocratic aphorism: “no head injury is too severe to despair of, nor too trivial to ignore.” Even today considerable uncertainty may exist on the expected outcome in individual patients. Yet, confident predictions are essential toward provision of accurate information on expectations to relatives and care givers. Physicians frequently—and often unwittingly—utilize prognostic estimates toward therapeutic decision making and allocation of resources. For accurate outcome prediction, multiple risk factors need to be considered jointly in a prognostic model because single factors have insufficient predictive value to distinguish patients who will do well from those who will do poorly. Prognostic models are generally created by multivariable analysis including predictors such as age and Glasgow Coma Scale (GCS) [2].
Following the pioneering work by Jennett et al. [3], many studies have been published on prognosis in TBI, and the Glasgow Outcome Scale (GOS) has become the generally accepted standard for assessing patient outcome [4]. Although this work moved the field toward a more realistic statistical base and away from clinical guesswork, prognostic models in TBI have not had a widespread impact on clinical practice. Partly, this may be because clinicians are not familiar with prognostic models. However, probably more important is that the validity and usefulness of prognostic models in TBI has not been demonstrated with sufficient clarity and certainty to convince clinicians of their added value. We aimed to identify how previous studies have dealt with four important modeling issues: the study population, choice of predictors and outcome, model development, and model validation. We propose recommendations for future prognostic modeling studies.
Section snippets
Methods
We searched the National Library of Medicine's PubMed database from 1970 till 2005 using the following search terms: brain or head injury, progno or pred, and GCS or GOS. No language restrictions were applied. From the initial search we found over 100 publications. From these articles we chose those that had presented a prognostic model based on admission data for patients with moderate to severe TBI. Next we excluded studies that did not specify the prognostic model, or were conducted on
Results
In total we considered 31 studies (Table 1) [4], [11], [12], [13], [14], [15], [16], [17], [18], [19], [20], [21], [22], [23], [24], [25], [26], [27], [28], [29], [30], [31], [32], [33], [34], [35], [36], [37], [38], [39], [40]. The objective of all studies was to identify clinically important predictors and to create a prognostic model at an early stage of the injury. Several studies addressed additional items, such as the comparison of different modeling techniques [11], [17], or the
Discussion
We systematically reviewed 31 prognostic models for use in severe and moderate TBI that considered admission data. Several models had been developed on patient series collected over 20 years ago. The validity of these models for current practice may be questioned as standards of care have improved. Substantial limitations were identified in the development of many models that confines their generalizability: first, most models (25 of 31) were developed on relatively small patient series (N <
Conclusions
In conclusion, sample sizes of many prognostic modeling studies in TBI were too small to permit construction of valid prognostic models. A limited set of predictors (five to seven) may capture most of the currently available prognostic information. The performance of prognostic models in TBI is more determined by the selection of predictors than by the modeling strategies [43], [56]. Restricting the analysis to a complete case analysis may however lead to bias and suboptimal results. The
Acknowledgments
The authors wish to thank Marja van Gemerden and Frans Slieker for administrative assistance and support. Grant support was provided by NIH NS 42691.
References (63)
Traumatic brain injury
Lancet
(2000)- et al.
Assessment of coma and impaired consciousness. A practical scale
Lancet
(1974) - et al.
Penalized maximum likelihood estimation to directly adjust diagnostic and prognostic prediction models for overoptimism: a clinical example
J Clin Epidemiol
(2004) - et al.
Internal validation of predictive models: efficiency of some procedures for logistic regression analysis
J Clin Epidemiol
(2001) - et al.
Internal and external validation of predictive models: a simulation study of bias and precision in small samples
J Clin Epidemiol
(2003) - et al.
Predicting outcome in individual patients after severe head injury
Lancet
(1976) - et al.
Evaluation of the Leeds prognostic score for severe head injury
Lancet
(1991) - et al.
Some prognostic models for traumatic brain injury were not valid
J Clin Epidemiol
(2006) Advantages and disadvantages of using artificial neural networks versus logistic regression for predicting medical outcomes
J Clin Epidemiol
(1996)Prognosis after severe head injury
Clin Neurosurg
(1972)
Is early prediction of outcome in severe head injury possible?
Arch Surg
Regression modeling strategies: With applications to linear models, logistic regression and survival analysis
Assessing the generalizability of prognostic information
Ann Intern Med
Applied logistic regression
Predicting recovery in patients suffering from traumatic brain injury by using admission variables and physiological data: a comparison between decision tree analysis and logistic regression
J Neurosurg
Relative prognostic value of best motor response and brain stem reflexes in patients with severe head injury
Neurosurgery
Systematic selection of prognostic features in patients with severe head injury
Neurosurgery
Prognosis and prediction of outcome in comatose head injured patients
Acta Neurochir Suppl (Wien)
Chart for outcome prediction in severe head injury
J Neurosurg
Enhanced specificity of prognosis in severe head injury
J Neurosurg
Prediction tree for severely head-injured patients
J Neurosurg
Severe head injuries: an outcome prediction and survival analysis
Intensive Care Med
Outcome and prognostic factors in head injuries with an admission Glasgow Coma Scale score of 3
Arch Surg
The Westmead Head Injury Project outcome in severe head injury. A comparative analysis of pre-hospital, clinical and CT variables
Br J Neurosurg
Aggressive management of severe closed head trauma: time for reappraisal
Lancet
Predicting outcome after traumatic brain injury: development and validation of a prognostic score based on admission characteristics
J Neurotrauma
Outcome after severe head injury: an analysis of prediction based upon comparison of neural network versus logistic regression analysis
Neurol Res
Early predictors of mortality and morbidity after severe closed head injury
J Neurotrauma
Assessing the influence of non-treatment variables in a study of outcome from severe head injuries
J Neurosurg
Evaluation of designs for clinical trials of neuroprotective agents in head injury. European Brain Injury Consortium
J Neurotrauma
Predicting survival from head trauma 24 hours after injury: a practical method with therapeutic implications
J Trauma
Cited by (140)
Traumatic brain injury and in-hospital mortality- CT perfusion and beyond
2022, Diagnosis and Treatment of Traumatic Brain Injury: The Neuroscience of Traumatic Brain InjuryPredictors of outcome in moderate and severe traumatic brain injury
2022, Diagnosis and Treatment of Traumatic Brain Injury: The Neuroscience of Traumatic Brain InjuryDesign, methods, and reporting of impact studies of cardiovascular clinical prediction rules are suboptimal: a systematic review
2021, Journal of Clinical EpidemiologyPredicting in-hospital mortality after traumatic brain injury: External validation of CRASH-basic and IMPACT-core in the national trauma data bank
2021, InjuryCitation Excerpt :Still, there are a myriad of ways to categorize patients with respect to mechanism of insult, clinical severity and pathophysiology, each of which may influence management and prognosis. The complex and heterogeneous nature of TBI often leads to uncertainty in expected patient outcomes and has significantly limited research design and clinical trial results [8–10]. As a result, a lack of robust randomized controlled trials (RCTs) has weakened evidence that supports the benefit of many treatment concepts [11–13].
Pre-injury health status and excess mortality in persons with traumatic brain injury: A decade-long historical cohort study
2020, Preventive MedicineCitation Excerpt :Our study has limitations. Despite being internally validated in a large cohort, the 43 factors preceding TBI were not medical diagnoses with well-defined criteria (Mushkudiani et al., 2008). Therefore, excess mortality rates may incorporate some imprecision, similar to that of other parameters, e.g., differences between physiological and chronological age (Cole et al., 2015; Wang et al., 2017) or frailty being a relative vs. absolute state.