"Decision tree analysis for assessing the risk of post-traumatic haemorrhage after mild traumatic brain injury in patients on oral anticoagulant therapy"

An interpretable machine learning model based on optimal feature selection for identifying CT abnormalities in patients with mild traumatic brain injury

Background

Minor head trauma is a frequent cause of emergency department visits, early identification and prediction of mild traumatic brain injury (mTBI) patients with abnormal brain lesions are vital for minimizing unnecessary computed tomography (CT) scans, reducing radiation exposure, and ensuring timely effective treatment and care. This study aims to develop and validate an interpretable machine learning (ML) prediction model using routine laboratory data for guiding clinical decisions on CT scan use in mTBI patients.

Methods

We conducted a multicentre study in China including data from January 2019 to July 2024. Our study included three patient cohorts: a retrospective training cohort (654 patients for training and 163 for internal testing) and two prospective validation cohorts (86 internal and 290 external patients). Fifty-one routine clinical laboratory characteristics, readily available from the electronic medical record (EMR) system within the first 24 h of admission, were collected. Seven ML algorithms were trained to develop predictive models, with the random forest (RF) algorithm used to optimize key feature combinations. Model predictive performance was evaluated using metrics such as the area under the receiver operating characteristic curve (AUC), positive predictive value (PPV), and F1 scores. The SHapley Additive exPlanation (SHAP) was applied to interpret the final model, while decision curve analysis (DCA) was used to assess the clinical net benefit.

Findings

In the derivation cohort, 599 (73.3%) patients had normal CT scans and 218 (26.7%) had abnormal CT scans. The Gradient boosting classifier (GBC) model performed best among the seven ML models, with an AUC of 0.932 (95% CI: 0.900-0.963). After reducing features to 21 (8 biochemical test indicators, 3 coagulation markers, and 10 complete blood cell count indicators) according to feature importance rank, an explainable GBC-final model was established. The final model accurately predicted mTBI patients with abnormal CT in both internal (AUC 0.926, 95% CI: 0.893-0.958) and external (AUC 0.904, 95% CI: 0.835-0.973) validation cohorts. In the prospective cohort, final GBC model achieved AUC of 0.885 (95% CI: 0.753-1.000) and was significantly superior to traditional TBI biomarkers GFAP (AUC: 0.745) and PGP9.5 (AUC: 0.794). DCA revealed that the final model offered greater net benefits than "full intervention" or "no intervention" strategies within a probability threshold range of 0.16-0.93. SHAP analysis identified D-dimer levels, absolute lymphocyte and neutrophil counts, and hematocrit as key high-risk features.

Interpretation

Our optimal feature selection-based ML model accurately and reliably predicts CT abnormalities in mTBI patients using routine test data. By addressing clinicians' concerns regarding transparency and decision-making through SHAP and DCA analyses, we strengthen the potential clinical applicability of our ML model.

Funding

The Natural Science Foundation of Hubei Province, high-level Talent Research Startup Funding of Hubei University of Chinese Medicine, Wuhan Health and Family Planning Scientific Research Fund Project of Hubei Province, and Machine Learning-based Intelligent Diagnosis System for AFP-negative Liver Cancer Project.

Development of clinical decision support for patients older than 65 years with fall-related TBI using artificial intelligence modeling

BACKGROUND

Older persons comprise most traumatic brain injury (TBI)-related hospitalizations and deaths and are particularly susceptible to fall-induced TBIs. The combination of increased frailty and susceptibility to clinical decline creates a significant ongoing challenge in the management of geriatric TBI. As the population ages and co-existing medical conditions complexify, so does the need to improve the quality of care for this population. Utilizing early hospital admission variables, this study will create and validate a multinomial decision tree that predicts the discharge disposition of older patients with fall-related TBI.

METHODS

From the National Trauma Data Bank, we retrospectively analyzed 11,977 older patients with a fall-related TBI (2017-2021). Clinical variables included Glasgow Coma Scale (GCS) score, intracranial pressure monitor use, venous thromboembolism (VTE) prophylaxis, and initial vital signs. Outcomes included hospital discharge disposition re-categorized into home, care facility, or deceased. Data were split into two sets, where 80% developed a decision tree, and 20% tested predictive performance. We employed a conditional inference tree algorithm with bootstrap (B = 100) and grid search options to grow the decision tree and measure discrimination ability using the area under the curve (AUC) and calibration plots.

RESULTS

Our decision tree used seven admission variables to predict the discharge disposition of older TBI patients. Significant non-modifiable variables included total GCS and injury severity scores, while VTE prophylaxis type was the most important interventional variable. Patients who did not receive VTE prophylaxis treatment had a higher probability of death. The predictive performance of the tree in terms of AUC value (95% confidence intervals) in the training cohort for death, care, and home were 0.66 (0.65-0.67), 0.75 (0.73-0.76), and 0.77 (0.76-0.79), respectively. In the test cohort, the values were 0.64 (0.62-0.67), 0.75 (0.72-0.77), and 0.77 (0.73-0.79).

CONCLUSIONS

We have developed and internally validated a multinomial decision tree to predict the discharge destination of older patients with TBI. This tree could serve as a decision support tool for caregivers to manage older patients better and inform decision-making. However, the tree must be externally validated using prospective data to ascertain its predictive and clinical importance.

Economic Impact of Surveillance of Head Trauma Patients with Coagulopathy and Normal Initial Computed Tomography Scan (ECO-NCT)

INTRODUCTION

According to the Portuguese clinical guidelines published in 1999, patients with traumatic brain injury and coagulopathies should remain in-hospital for 24 hours for clinical and image surveillance, despite having an admission computed tomography (CT) scan showing no intracranial lesions. Growing evidence suggests this practice is not only void of clinical relevance, but that it can also be potentially harmful for the patient. Nevertheless, up until now there is no published data concerning the economic impact of this clinical practice.

METHODS

A cost analysis compared retrospective data from patients admitted to our emergency department during 2022 with a hypothetical scenario in which a patient with an admission CT scan without traumatic lesions was discharged. Clinical data was also retrieved concerning the rate of a delayed intracranial bleeding on 24-hour CT scan and mortality at a six-month-period after discharge. Direct costs for the national health service were determined in terms of funding and time invested by medical teams.

RESULTS

From a sample of 440 patients, 436 remained in-hospital for a 24-hour clinical and image surveillance, of which only two (0.5%) showed a new intracranial lesion on the second CT-scan. Neither of these two patients required therapeutic measures to control bleeding and were discharged 36 hours after admission. Out of 440 patients, one patient (0.2%) died of cardiac arrest during the 24-hour surveillance period, despite having an initial normal CT scan showing no brain lesions. Our current surveillance practice directly amounted to €163 157.00, whereas the cost of our hypothetical scenario amounted to €29 480.00: a difference of €133 677.00. The application of our surveillance guideline also meant that nine emergency shifts were devoted to this task, compared to 4.6 hypothetical shifts if patients were discharged after an initial CT scan without traumatic intracranial lesions.

CONCLUSION

In spite of apparently not adding any clinical value to our practice, our in-hospital surveillance may represent a significant financial and time-consuming burden, costing five times as much and demanding our medical teams twice as much work when compared to a scenario without clinical surveillance and 24-hour CT scans.

Risk factors for traumatic intracranial hemorrhage in mild traumatic brain injury patients at the emergency department: a systematic review and meta-analysis

BACKGROUND

Mild traumatic brain injury (mTBI), i.e. a TBI with an admission Glasgow Coma Scale (GCS) of 13-15, is a common cause of emergency department visits. Only a small fraction of these patients will develop a traumatic intracranial hemorrhage (tICH) with an even smaller subgroup suffering from severe outcomes. Limitations in existing management guidelines lead to overuse of computed tomography (CT) for emergency department (ED) diagnosis of tICH which may result in patient harm and higher healthcare costs.

OBJECTIVE

To perform a systematic review and meta-analysis to characterize known and potential novel risk factors that impact the risk of tICH in patients with mTBI to provide a foundation for improving existing ED guidelines.

METHODS

The literature was searched using MEDLINE, EMBASE and Web of Science databases. Reference lists of major literature was cross-checked. The outcome variable was tICH on CT. Odds ratios (OR) were pooled for independent risk factors.

RESULTS

After completion of screening, 17 papers were selected for inclusion, with a pooled patient population of 26,040 where 2,054 cases of tICH were verified through CT (7.9%). Signs of a skull base fracture (OR 11.71, 95% CI 5.51-24.86), GCS < 15 (OR 4.69, 95% CI 2.76-7.98), loss of consciousness (OR 2.57, 95% CI 1.83-3.61), post-traumatic amnesia (OR 2.13, 95% CI 1.27-3.57), post-traumatic vomiting (OR 2.04, 95% CI 1.11-3.76), antiplatelet therapy (OR 1.54, 95% CI 1.10-2.15) and male sex (OR 1.28, 95% CI 1.11-1.49) were determined in the data synthesis to be statistically significant predictors of tICH.

CONCLUSION

Our meta-analysis provides additional context to predictors associated with high and low risk for tICH in mTBI. In contrast to signs of a skull base fracture and reduction in GCS, some elements used in ED guidelines such as anticoagulant use, headache and intoxication were not predictive of tICH. Even though there were multiple sources of heterogeneity across studies, these findings suggest that there is potential for improvement over existing guidelines as well as a the need for better prospective trials with consideration for common data elements in this area. PROSPERO registration number CRD42023392495.

Incidence of intracranial bleeding in mild traumatic brain injury patients taking oral anticoagulants: a systematic review and meta-analysis

BACKGROUND

Traumatic brain injury (TBI) is one of the leading causes of disability and death worldwide. Most TBI cases occur in older people, because they are at a higher risk of accidental falling. As the population ages, the use of anticoagulants is increasing. Some serious complications of TBI, such as intracranial hemorrhage (ICH), may occur even in mild cases. According to the current guidelines regarding managing mild TBI patients, a CT head scan is recommended for all patients receiving anticoagulation. We aim to assess the incidence of ICH in patients with mild TBI taking oral anticoagulants.

METHODS

Our systematic review and meta-analysis were performed using the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) checklist. The protocol was registered in PROSPERO (CRD42024503086). Twenty-eight studies evaluating patients with a mild TBI from ten countries with a total sample size of 11,172, 5671 on DOACs, and 5501 on VKAs were included in our meta-analysis.

RESULTS

The random-effects overall incidence of ICH among oral anticoagulated patients with mild TBI was calculated to be 9.4% [95% CI 7.2-12.1%, I2 = 89%]. The rates of immediate ICH for patients taking DOACs and VKAs were 6.4% and 10.5%, respectively. The overall rate of immediate ICH in anticoagulated mild TBI patients was 8.5% [95% CI 6.6-10.9%], with a high heterogeneity between studies (I2 = 88%). Furthermore, the rates of delayed ICH in patients with mild TBI taking DOACs and VKAs were 1.6% and 1.9%, respectively. The overall incidence of delayed ICH among oral anticoagulated mild TBI patients was 1.7% [95% CI 1-2.8%, I2 = 79%]. The overall rate of ICH among mild TBI patients taking DOAC was calculated to be 7.3% [95% CI 5.2-10.3%], with significant heterogeneity between studies (I2 = 79%). However, the overall ICH rate is higher in patients who take only VKAs 11.3% [95% CI 8.6-14.7%, I2 = 83%]. Patients on DOACs were at lower risk of ICH after mild TBI compared to patients on VKAs (OR = 0.64, 95% CI 0.48-0.86, p < 0.01, I2 = 28%).

CONCLUSION

Our meta-analysis confirms the need for performing brain CT scan in patients with mild TBI patients who receive oral anticoagulants before injury. Due to limited data, further multi-center, prospective studies are warranted to confirm the true incidence of traumatic ICH in patients on anticoagulants.

Traumatic Brain Injury in Patients under Anticoagulant Therapy: Review of Management in Emergency Department

The best management of patients who suffer from traumatic brain injury (TBI) while on oral anticoagulants is one of the most disputed problems of emergency services. Indeed, guidelines, clinical decision rules, and observational studies addressing this topic are scarce and conflicting. Moreover, relevant issues such as the specific treatment (and even definition) of mild TBI, rate of delayed intracranial injury, indications for neurosurgery, and anticoagulant modulation are largely empiric. We reviewed the most recent evidence on these topics and explored other clinically relevant aspects, such as the promising role of dosing brain biomarkers, the strategies to assess the extent of anticoagulation, and the indications of reversals and tranexamic acid administration, in cases of mild TBI or as a bridge to neurosurgery. The appropriate timing of anticoagulant resumption was also discussed. Finally, we obtained an insight into the economic burden of TBI in patients on oral anticoagulants, and future directions on the management of this subpopulation of TBI patients were proposed. In this article, at the end of each section, a "take home message" is stated.

Minor head injury in anticoagulated patients: Outcomes and analysis of clinical predictors. A prospective study

BACKGROUND

The optimal management of patients taking oral anticoagulants who experience minor head injury (MHI) is unclear. The availability of validated protocols and reliable predictors of prognosis would be of great benefit. We investigated clinical factors as predictors of clinical outcomes and intracranial injury (ICI).

METHODS

We conducted a single-cohort, prospective, observational study in an ED. Our structured clinical pathway included a first head CT scan, 24 h observation and a second CT scan. The primary outcome was the occurrence of MHI-related death or re-admission to ED at day +30. The secondary outcome was the rate of delayed ICI (dICI), defined as second positive CT scan after a first negative CT scan. We assessed some clinical predictors derived from guidelines and clinical prediction rules as potential risk factors for the outcomes.

RESULTS

450 patients with a negative first CT scan who underwent a second CT scan composed our 'study population'. The rate of the primary outcome was 4%. The rate of the secondary outcome was 4.7%. Upon univariate and multivariate analysis no statistically significant predictors for the outcomes were found.

CONCLUSIONS

Previous retrospective studies showed a lot of negative predictive factors for anticoagulated patients suffering a minor head injury. In our prospective study no clinical factors emerged as predictors of poor clinical outcomes and dICI. So, even if we confirmed a low rate of adverse outcomes, the best management of these patients in ED remains not so clear and future trials are needed.