Predicting Neurological Deterioration after Moderate Traumatic Brain Injury: Development and Validation of a Prediction Model Based on Data Collected on Admission

Development and validation of a nomogram for predicting early neurological deterioration in patients with moderate traumatic brain injury: a retrospective analysis

Objective

Early neurological deterioration (END) greatly affects prognosis of moderate traumatic brain injury (TBI). This study aimed to develop and validate a nomogram to predict the occurrence of END in patients with moderate TBI.

Methods

A total of 371 patients with moderate TBI were enrolled and divided into the training (n = 260) and validation (n = 111) groups at a ratio of 7:3. Univariate and multivariate logistic regression analyses were used to identify the significant factors for END, which were used to develop a nomogram. The discrimination of the nomogram was evaluated using area under the receiver operating characteristic curves (AUC), the calibration was evaluated using calibration curves and Hosmer-Lemeshow tests. Decision curve analysis (DCA) was used to evaluate the net benefit of the model for patients.

Results

In the training group, multivariate logistic regression demonstrated that GCS score, epidural hematoma, intracerebral hemorrhage, fibrinogen, and D-dimer were independent risk factors for END in patients with moderate TBI. A nomogram was constructed using the logistic regression prediction model. The AUCs of the nomogram in the training and validation groups were 0.901 and 0.927, respectively. The calibration curves showed that the predicted probability was consistent with the actual situation in both the training and validation sets. DCA curves demonstrated significantly better net benefit with the model. Then a web-based calculator was generated to facilitate clinical application.

Conclusion

The present study developed and validated a model to predict END in patients with moderate TBI. The nomogram that had good discrimination, calibration, and clinical utility can provide clinicians with an effective and accurate tool for evaluating the occurrence of END after moderate TBI.

Neuroworsening in Moderate Traumatic Brain Injury

Patients with moderate traumatic brain injury (TBI) are at high risk for developing intracerebral complications and in particular neuroworsening (NW). NW can be unpredictable and may be an important risk factor for poor neurologic outcome and for increased mortality. NW is often a medical and surgical emergency, and it is, therefore, fundamental to identify patients at risk early because they require strict neuromonitoring and repeated neuroimaging. So far, there is no standardized and validated definition of NW. In this review, we aim to discuss the definition, risk factors, and management of patients with moderate TBI at high risk of NW.

A Nomogram to Predict Intracranial Hypertension in Moderate Traumatic Brain Injury Patients

OBJECTIVE

Patients with moderate traumatic brain injury (mTBI) are under the threat of intracranial hypertension (IHT). However, it is unclear which mTBI patient will develop IHT and should receive intracranial pressure (ICP)-lowering treatment or invasive ICP monitoring after admission. The purpose of the present study was to develop and validate a prediction model that estimates the risk of IHT in mTBI patients.

METHODS

Baseline data collected on admission of 296 mTBI patients with Glasgow Coma Scale (GCS) score of 9-11 was collected and analyzed. Multivariable logistic regression modeling with backward stepwise elimination was used to develop a prediction model for IHT. The discrimination efficacy, calibration efficacy, and clinical utility of the prediction model were evaluated. Finally, the prediction model was validated in a separate cohort of 122 patients from 3 hospitals.

RESULTS

Four independent prognostic factors for IHT were identified: GCS score, Marshall head computed tomography score, injury severity score, and location of contusion. The C-statistic of the prediction model in internal validation was 84.30% (95% CI: 0.794-0.892). The area under the curve for the prediction model in external validation was 82.80% (95% CI: 0.747-0.909).

CONCLUSIONS

A prediction model based on baseline parameters was found to be highly sensitive in distinguishing mTBI patients with GCS score of 9-11 who would suffer IHT. The high discriminative ability of the prediction model supports its use in identifying mTBI patients with GCS score of 9-11 who need ICP-lowering therapy or invasive ICP monitoring.

Moderate Traumatic Brain Injury in Adult Population: The Latin American Brain Injury Consortium Consensus for Definition and Categorization

Moderate traumatic brain injury (TBI) is a diagnosis that describes diverse patients with heterogeneity of primary injuries. Defined by a Glasgow Coma Scale between 9 and 12, this category includes patients who may neurologically worsen and require increasing intensive care resources and/or emergency neurosurgery. Despite the unique characteristics of these patients, there have not been specific guidelines published before this effort to support decision-making in these patients. A Delphi consensus group from the Latin American Brain Injury Consortium was established to generate recommendations related to the definition and categorization of moderate TBI. Before an in-person meeting, a systematic review of the literature was performed identifying evidence relevant to planned topics. Blinded voting assessed support for each recommendation. A priori the threshold for consensus was set at 80% agreement. Nine PICOT questions were generated by the panel, including definition, categorization, grouping, and diagnosis of moderate TBI. Here, we report the results of our work including relevant consensus statements and discussion for each question. Moderate TBI is an entity for which there is little published evidence available supporting definition, diagnosis, and management. Recommendations based on experts' opinion were informed by available evidence and aim to refine the definition and categorization of moderate TBI. Further studies evaluating the impact of these recommendations will be required.

Development of predictive model for the neurological deterioration among mild traumatic brain injury patients using machine learning algorithms

BACKGROUND

Mild traumatic brain injury (mTBI) comprises a majority of traumatic brain injury (TBI) cases. While some mTBI would suffer neurological deterioration (ND) and therefore have poorer prognosis. This study was designed to develop the predictive model for the ND among mTBI using machine learning algorithms.

METHODS

mTBI patients recorded in the Medical Information Mart for Intensive Care-III were selected for the study. The ND was defined as a drop of Glasgow Coma Scale ≥ 2 within the first 7 day after admission. Eight machine learning algorithms were trained and validated with 5-fold cross validation including extreme gradient boosting, logistic regression, light gradient boosting machine, random forest, adaptive boosting, decision tree, complement naïve Bayes, and support vector machine. The value of eight machine learning algorithms was compared by the area under the receiver operating characteristic curve (AUC).

RESULTS

361 mTBI patients suffered the ND with the incidence of 30.7%. The ND group had higher 30-day mortality (p = 0.001). In the training cohort of mTBI patients, the random forest performed the best on predicting the ND with the AUC of 1.000. The XGBoost and AdaBoost had an AUC of 0.827 and 0.815, respectively. The logistic regression performed the best on predicting the ND in the validation cohort with the AUC of 0.741. The XGBoost, random forest and AdaBoost had an AUC of 0.729, 0.735, 0.736 in the validation cohort, respectively. After adjusting confounding effects, the multivariate logistic regression found only two independent risk factors for the ND including Sequential Organ Failure Assessment (SOFA) (p < 0.001) and hypertension (p = 0.001). The logistic regression predictive model composed of SOFA and hypertension had an AUC of 0.741.

CONCLUSIONS

SOFA score and complicated hypertension are two independent risk factors for the neurological deterioration among mTBI patients. The logistic regression predictive model incorporating SOFA and hypertension is helpful to identify mTBI patients with the high risk of ND.

Vision Transformer-based Decision Support for Neurosurgical Intervention in Acute Traumatic Brain Injury: Automated Surgical Intervention Support Tool

Purpose To develop an automated triage tool to predict neurosurgical intervention for patients with traumatic brain injury (TBI). Materials and Methods A provincial trauma registry was reviewed to retrospectively identify patients with TBI from 2005 to 2022 treated at a specialized Canadian trauma center. Model training, validation, and testing were performed using head CT scans with binary reference standard patient-level labels corresponding to whether the patient received neurosurgical intervention. Performance and accuracy of the model, the Automated Surgical Intervention Support Tool for TBI (ASIST-TBI), were also assessed using a held-out consecutive test set of all patients with TBI presenting to the center between March 2021 and September 2022. Results Head CT scans from 2806 patients with TBI (mean age, 57 years ± 22 [SD]; 1955 [70%] men) were acquired between 2005 and 2021 and used for training, validation, and testing. Consecutive scans from an additional 612 patients (mean age, 61 years ± 22; 443 [72%] men) were used to assess the performance of ASIST-TBI. There was accurate prediction of neurosurgical intervention with an area under the receiver operating characteristic curve (AUC) of 0.92 (95% CI: 0.88, 0.94), accuracy of 87% (491 of 562), sensitivity of 87% (196 of 225), and specificity of 88% (295 of 337) on the test dataset. Performance on the held-out test dataset remained robust with an AUC of 0.89 (95% CI: 0.85, 0.91), accuracy of 84% (517 of 612), sensitivity of 85% (199 of 235), and specificity of 84% (318 of 377). Conclusion A novel deep learning model was developed that could accurately predict the requirement for neurosurgical intervention using acute TBI CT scans. Keywords: CT, Brain/Brain Stem, Surgery, Trauma, Prognosis, Classification, Application Domain, Traumatic Brain Injury, Triage, Machine Learning, Decision Support Supplemental material is available for this article. © RSNA, 2024 See also commentary by Haller in this issue.

Secondary Neurologic Deterioration After Moderate Traumatic Brain Injury: Development of a Multivariable Prediction Model and Proposition of a Simple Triage Score

BACKGROUND

Identifying patients at risk of secondary neurologic deterioration (SND) after moderate traumatic brain injury (moTBI) is a challenge, as such patients will need specific care. No simple scoring system has been evaluated to date. This study aimed to determine clinical and radiological factors associated with SND after moTBI and to propose a triage score.

METHODS

All adults admitted in our academic trauma center between January 2016 and January 2019 for moTBI (Glasgow Coma Scale [GCS] score, 9-13) were eligible. SND during the first week was defined either by a decrease in GCS score of >2 points from the admission GCS in the absence of pharmacologic sedation or by a deterioration in neurologic status associated with an intervention, such as mechanical ventilation, sedation, osmotherapy, transfer to the intensive care unit (ICU), or neurosurgical intervention (for intracranial mass lesions or depressed skull fracture). Clinical, biological, and radiological independent predictors of SND were identified by logistic regression (LR). An internal validation was performed using a bootstrap technique. A weighted score was defined based on beta (β) coefficients of the LR.

RESULTS

A total of 142 patients were included. Forty-six patients (32%) showed SND, and 14-day mortality rate was 18.4%. Independent variables associated with SND were age above 60 years (odds ratio [OR], 3.45 [95% confidence interval {CI}, 1.45-8.48]; P = .005), brain frontal contusion (OR, 3.22 [95% CI, 1.31-8.49]; P = .01), prehospital or admission arterial hypotension (OR, 4.86 [95% CI, 2.03-12.60]; P = .006), and a Marshall computed tomography (CT) score of 6 (OR, 3.25 [95% CI, 1.31-8.20]; P = .01). The SND score was defined with a range from 0 to 10. The score included the following variables: age >60 years (3 points), prehospital or admission arterial hypotension (3 points), frontal contusion (2 points), and Marshall CT score of 6 (2 points). The score was able to detect patients at risk of SND, with an area under the receiver operating characteristic curve (AUC) of 0.73 (95% CI, 0.65-0.82). A score of 3 had a sensitivity of 85%, a specificity of 50%, a VPN of 87%, and a VPP of 44 % to predict SND.

CONCLUSIONS

In this study, we demonstrate that moTBI patients have a significant risk of SND. A simple weighted score at hospital admission could be able to detect patients at risk of SND. The use of the score may enable optimization of care resources for these patients.

Association Between High-Level D-Dimer at Admission and Early Intubation in Patients With Moderate Traumatic Brain Injury

It is unclear who can benefit from tracheal intubation in the moderate (mTBI) traumatic brain injury (TBI) population. Given that mTBI patients are conscious, intubation can cause intense stress, possibly triggering neurological deterioration. Therefore, identifying potential risk factors for intubation in mTBI patients can serve as a valuable clinical warning. We sought to investigate whether elevated D-dimer is a possible risk factor for intubation in mTBI patients. Using the STROBE statement, adult patients with isolated TBI (Glasgow Coma Scale [GCS] score 9-13) treated at a high-volume neurotrauma center between January 2015 and December 2020 were reviewed. The demographics, clinical presentation, neuroimaging, and laboratory information were collected based on the patients' electronic medical record. D-dimer values were assessed from serum when patients were admitted to the hospital. The primary study end-point was that the mTBI patient was intubated within 72 h upon admission. A total of 557 patients with mTBI were finally included in this study. Of these, 85 (15.3%) patients were intubated. Multi-variate logistic regression analysis showed that high-level D-dimer (≥17.9mg/L) was significantly associated with early tracheal intubation in mTBI patients (odds ratio, 3.10 [1.16-8.25]; p = 0.024) after adjusting for age, sex, GCS scores, Marshall scores, and Injury Severity Scores. Sensitivity analysis showed that high-level D-dimer had a robust correlation with intubation in the different subgroups or after propensity score matching. High-level D-dimer on admission is an independent risk factor for early tracheal intubation in isolated mTBI patients.

Online dynamic nomogram for predicting pain recurrence after microvascular decompression in trigeminal neuralgia

Trigeminal neuralgia (TN) is one of the most common causes of facial pain. Microvascular decompression (MVD) is the first-choice surgical treatment. The present study aimed to develop a novel practical assessment system based on preoperative clinical and imaging factors for clinicians to predict the likelihood of pain recurrence following MVD in TN. A total of 56 patients with primary unilateral TN who underwent MVD were retrospectively analyzed. Patients were followed up to observe pain recurrence 1 year after MVD. An online dynamic nomogram was constructed for predicting the probability of pain recurrence after MVD in patients with TN based on multivariate logistic model. The concordance index (C-index) and receiver operating characteristic (ROC) were used to measure model discrimination. Bootstrap resampling was used for internal validation of the model and calibration curve was constructed. Decision curve analysis (DCA) was used to assess clinical applicability. Factors such as numeric rating scale (to score pain degree of patients with TN), response to neuroanalgesic drugs and neurovascular contact on magnetic resonance imaging were independent risk factors affecting the pain recurrence rate (all P<0.05). C-index was 0.973 (95%CI, 0.938-1.000) and the area under the ROC was 0.973 (95%CI, 0.938-1.000). Calibration curve with a 1,000 bootstrap resampling showed a good fit between dynamic nomogram prediction and actual observations. The DCA showed that at a threshold probability between 0 and 100%, this model can achieve a greater net benefit than if all patients had surgery or none had surgery. In conclusion, this online dynamic nomogram reliably predicted risk of pain recurrence in patients with TN following MVD.

A Long Short-Term Memory Network Using Resting-State Electroencephalogram to Predict Outcomes Following Moderate Traumatic Brain Injury

Although traumatic brain injury (TBI) is a global public health issue, not all injuries necessitate additional hospitalisation. Thinking, memory, attention, personality, and movement can all be negatively impacted by TBI. However, only a small proportion of nonsevere TBIs necessitate prolonged observation. Clinicians would benefit from an electroencephalography (EEG)-based computational intelligence model for outcome prediction by having access to an evidence-based analysis that would allow them to securely discharge patients who are at minimal risk of TBI-related mortality. Despite the increasing popularity of EEG-based deep learning research to create predictive models with breakthrough performance, particularly in epilepsy prediction, its use in clinical decision making for the diagnosis and prognosis of TBI has not been as widely exploited. Therefore, utilising 60s segments of unprocessed resting-state EEG data as input, we suggest a long short-term memory (LSTM) network that can distinguish between improved and unimproved outcomes in moderate TBI patients. Complex feature extraction and selection are avoided in this architecture. The experimental results show that, with a classification accuracy of 87.50 ± 0.05%, the proposed prognostic model outperforms three related works. The results suggest that the proposed methodology is an efficient and reliable strategy to assist clinicians in creating an automated tool for predicting treatment outcomes from EEG signals.

Sociodemographic and clinical characteristics of children with tic disorders and behavioral problems: A real-world study and development of a prediction model

BACKGROUND

Tic disorders (TD) are complex neuropsychiatric disorders frequently associated with a variety of comorbid problems, whose negative effects may exceed those of the tics themselves. In this study, we aimed to explore the sociodemographic and clinical characteristics of children with TD and behavioral problems, and develop a prediction model of behavioral problems based on the predictors under real-world conditions.

METHODS

A hospital-based cross-sectional study was conducted on children with TD. Behavioral problems were surveyed using the Achenbach Child Behavior Checklist (CBCL). Sociodemographic information was collected from face-to-face interviews using an electronic questionnaire administered during the initial ambulatory visit. Clinical data were collected from medical records, and quality control was performed. The sociodemographic and clinical characteristics of patients with and without behavioral problems were statistically compared, and a nomogram prediction model was developed based on multivariate logistic regression analysis. The discriminatory ability and clinical utility of the nomogram were assessed by concordance index (C-index), receiver operating characteristic (ROC) curve, decision curve analysis (DCA) and clinical impact curve (CIC).

RESULTS

A total of 343 TD cases were included in the final analysis, of which 30.32% had behavioral problems. The prediction model showed age 12-16 years, abnormal birth history, parenting pattern of indulgence, parent/close relatives with psychiatric disorders, chronic motor or vocal tic disorder (CTD)/Tourette syndrome (TS) and moderate/severe tic severity were associated with behavioral problems in children with TD. The C-index of the prediction model (nomogram) was 0.763 (95% confidence interval, 0.710 ~ 0.816). The nomogram was feasible for making beneficial clinical decisions, according to the satisfactory results of the DCA and CIC.

CONCLUSIONS

A nomogram prediction model for comorbid behavioral problems in children with TD was established. The prediction model demonstrated a good discriminative ability and predictive performance for beneficial clinical decisions. This model further provides a comprehensive understanding of associated sociodemographic and clinical characteristics by visual graphs and allows clinicians to rapidly identify patients with a higher risk of behavioral problems and tailor necessary interventions to improve clinical outcomes.

Nomogram model predicts the risk of visual impairment in diabetic retinopathy: a retrospective study

BACKGROUND

To develop a model for predicting the risk of visual impairment in diabetic retinopathy (DR) by a nomogram.

METHODS

Patients with DR who underwent both optical coherence tomography angiography (OCTA) and fundus fluorescein angiography (FFA) were retrospectively enrolled. FFA was conducted for DR staging, swept-source optical coherence tomography (SS-OCT) of the macula and 3*3-mm blood flow imaging by OCTA to observe retinal structure and blood flow parameters. We defined a logarithm of the minimum angle of resolution visual acuity (LogMAR VA) ≥0.5 as visual impairment, and the characteristics correlated with VA were screened using binary logistic regression. The selected factors were then entered into a multivariate binary stepwise regression, and a nomogram was developed to predict visual impairment risk. Finally, the model was validated using the area under the receiver operating characteristic (ROC) curve (AUC), calibration plots, decision curve analysis (DCA), and clinical impact curve (CIC).

RESULTS

A total of 29 parameters were included in the analysis, and 13 characteristics were used to develop a nomogram model. Finally, diabetic macular ischaemia (DMI) grading, disorganization of the retinal inner layers (DRIL), outer layer disruption, and the vessel density of choriocapillaris layer inferior (SubVD) were found to be statistically significant (P < 0.05). The model was found to have good accuracy based on the ROC (AUC = 0.931) and calibration curves (C-index = 0.930). The DCA showed that risk threshold probabilities in the (3-91%) interval models can be used to guide clinical practice, and the proportion of people at risk at each threshold probability is illustrated by the CIC.

CONCLUSION

The nomogram model for predicting visual impairment in DR patients demonstrated good accuracy and utility, and it can be used to guide clinical practice.

TRIAL REGISTRATION

Chinese Clinical Trial Registry, ChiCTR2200059835. Registered 12 May 2022, https://www.chictr.org.cn/edit.aspx?pid=169290&htm=4.

Intracranial-Pressure-Monitoring-Assisted Management Associated with Favorable Outcomes in Moderate Traumatic Brain Injury Patients with a GCS of 9-11

Objective: With a mortality rate of 10−30%, a moderate traumatic brain injury (mTBI) is one of the most variable traumas. The indications for intracranial pressure (ICP) monitoring in patients with mTBI and the effects of ICP on patients’ outcomes are uncertain. The purpose of this study was to examine the indications of ICP monitoring (ICPm) and its effects on the long-term functional outcomes of mTBI patients. Methods: Patients with Glasgow Coma Scale (GCS) scores of 9−11 at Tangdu hospital, between January 2015 and December 2021, were enrolled and treated in this retrospective cohort study. We assessed practice variations in ICP interventions using the therapy intensity level (TIL). Six-month mortality and a Glasgow Outcome Scale Extended (GOS-E) score were the main outcomes. The secondary outcome was neurological deterioration (ND) events. The indication and the estimated impact of ICPm on the functional outcome were investigated by using binary regression analyses. Results: Of the 350 patients, 145 underwent ICP monitoring-assisted management, and the other 205 patients received a standard control based on imaging or clinical examinations. A GCS ≤ 10 (OR 1.751 (95% CI 1.216−3.023), p = 0.003), midline shift (mm) ≥ 2.5 (OR 3.916 (95% CI 2.076−7.386) p < 0.001), and SDH (OR 1.772 (95% CI 1.065−2.949) p = 0.028) were predictors of ICP. Patients who had ICPm (14/145 (9.7%)) had a decreased 6-month mortality rate compared to those who were not monitored (40/205 (19.5%), p = 0.011). ICPm was linked to both improved neurological outcomes at 6 months (OR 0.815 (95% CI 0.712−0.933), p = 0.003) and a lower ND rate (2 = 11.375, p = 0.010). A higher mean ICP (17.32 ± 3.52, t = −6.047, p < 0.001) and a more significant number of ICP > 15 mmHg (27 (9−45.5), Z = −5.406, p < 0.001) or ICP > 20 mmHg (5 (0−23), Z = −4.635, p < 0.001) 72 h after injury were associated with unfavorable outcomes. The best unfavorable GOS-E cutoff value of different ICP characteristics showed that the mean ICP was >15.8 mmHg (AUC 0.698; 95% CI, 0.606−0.789, p < 0.001), the number of ICP > 15 mmHg was >25.5 (AUC 0.681; 95% CI, 0.587−0.774, p < 0.001), and the number of ICP > 20 mmHg was >6 (AUC 0.660; 95% CI, 0.561−0.759, p < 0.001). The total TIL score during the first 72 h post-injury in the non-ICP group (9 (8, 11)) was lower than that of the ICP group (13 (9, 17), Z = −8.388, p < 0.001), and was associated with unfavorable outcomes. Conclusion: ICPm-assisted management was associated with better clinical outcomes six months after discharge and lower incidences of ND for seven days post-injury. A mean ICP > 15.8 mmHg, the number of ICP > 15 mmHg > 25.5, or the number of ICP > 20 mmHg > 6 implicate an unfavorable long-term prognosis after 72 h of an mTBI.

Predicting Neuroworsening in Moderate Traumatic Brain Injury Less is more?

No applicable.