Inpatient Metabolic Panel Trajectories Improve 6-Month Mortality Prognostication after Severe Traumatic Brain Injury

Abstract

The International Mission for Prognosis and Analysis of Clinical Trials (IMPACT) in traumatic brain injury (TBI) only includes data from hospital admission as predictors. Including updated physiological data after hospital admission would likely improve prognostic ability above the IMPACT model alone. We sought to evaluate differences in the daily trajectory of clinical metabolic panels (e.g., glucose, sodium, platelets, hemoglobin) for the first 7 days post-severe TBI (sTBI). This is a prospective cohort study of patients with sTBI with Glasgow Coma Scale (GCS) ≤8. Intake GCS was conducted by a neurosurgeon or neurosurgical resident to ensure GCS at presentation was related to the brain injury and not other factors. We compared daily metabolic panel trajectories between survivors to 6 months and nonsurvivors to 6 months using ordinal mixed-effects models. Dichotomizing trajectories as “high” or “low” was set to 4+ of the first 7 days postinjury to represent a majority of the first postinjury week. We then examined the added prognostic value of these trajectories compared with the IMPACT-extended model for 6-month mortality. Included patients (n = 572) were 40.4 ± 17.1 years of age, 79.4% male, and 38.4% had died by 6 months postinjury. The likelihood ratio tests (LRT) comparing the ordinal trajectories of sodium and platelets were statistically significant after false discovery correction (sodium: likelihood ratio = 51.0; adj. p < 0.001; platelets: likelihood ratio = 16.0; adj. p = 0.003), indicating that the overall trajectory over 7 days differs between groups. The LRT comparing the trajectories of glucose and hemoglobin were not statistically significant (p = 0.21–0.98). The models identified a divergence in platelet values on days 6 and 7, where nonsurvivors to 6 months had lower odds (OR = 0.27–0.41) of being in higher platelet categories than survivors to 6 months on those days. The IMPACT-only model had an area under the curve (AUC) for 6-month mortality of 0.85 and a Hosmer–Lemeshow p value = 0.68. The IMPACT-biomarker trajectory model had an AUC for 6-month mortality of 0.87 (DeLong’s test p value of 0.005) and a Hosmer–Lemeshow p value of 0.16. Trajectory of metabolic panel labs in the first week postinjury yields meaningful improvements in prognostic ability for the individual patient.

Keywords

biomarkers hemoglobin platelets prognostication severe traumatic brain injury

Introduction

Recovery from severe traumatic brain injury (sTBI) is a complex and dynamic process. However, the most well-known and studied prognostic model for sTBI is the International Mission for Prognosis and Analysis of Clinical Trials in TBI (IMPACT-TBI), which only includes data from hospital admission as predictors.¹ The IMPACT authors reported that the model was for research use only and best suited to discriminate unfavorable outcomes and/or mortality at 6 months postinjury in large cohorts,² but it has reportedly been used by clinicians to inform withdrawal of care decision-making for individual patients.^3,4 Given that recent studies of sTBI patients who survived until hospital discharge indicate that the average time to follow commands is a median of 5 days,⁵ updating the IMPACT model with more granular data reflecting real-time systemic health can improve prognostication and potentially improve performance for the individual patient.

sTBI inpatients receive repeated evaluations of objective biomarkers to provide the clinician with information regarding the patient’s status, including comprehensive metabolic panels. Certain components of metabolic panels, such as glucose, hemoglobin, sodium, and platelets, are recognized predictors of recovery status from trauma of all types, including sTBI.^6,7 In fact, glucose and hemoglobin at hospital intake are parts of the original IMPACT model.¹ Glucose variability, defined by daily standard deviation of glucose levels, has been significantly associated with functional outcome after sTBI.⁸ Initial hemoglobin and lowest hemoglobin level in the intensive care unit (ICU) have been associated with poor outcome after TBI.⁹ Harrois et al.¹⁰ evaluated dysnatremia in sTBI patients by evaluating the average daily standard deviation of sodium within the first 7 days postinjury in the ICU. The authors found that variability in sodium was independently associated with 28-day mortality. Early thrombocytopenia is another strong risk factor for mortality, both in-hospital and 12 months postinjury.¹¹ These studies have provided valuable information about which biomarkers are critical for monitoring following sTBI and how the variability of measurement may be important for prognostication. However, these data do not provide information regarding the direction or slope of change over time for these variables. Understanding how these biomarkers evolve across time at standardized daily intervals could yield critical information for individual prognostication. We sought to evaluate differences in daily trajectory of glucose, hemoglobin, platelets, and sodium for the first 7 days post-sTBI and examine the added prognostic value of these trajectories compared with the IMPACT model for 6-month mortality. We hypothesized that adding the high/low trajectories of these variables would significantly improve IMPACT model performance.

Methods

Design and participants

This is a secondary analysis of a prospectively collected database of patients from a single level 1 trauma center between November 2002 and December 2018 (n = 598). All patients had sTBI at presentation, defined as Glasgow Coma Scale (GCS) ≤8. Patients were excluded from inclusion in the database if presenting with age <18 years, currently pregnant, penetrating head trauma, or GCS = 3 with bilaterally nonreactive, enlarged pupils. Intake GCS was conducted by a neurosurgeon or neurosurgical resident to ensure GCS at presentation was related to the brain injury and not other factors. All patients remained GCS ≤8 for 24 h postpresentation. Mortality (yes/no) at 6 months was determined by search of the patient’s electronic health record and/or public databases, such as the social security index. Researchers extracted daily blood labs (e.g., glucose in mg/dL, sodium in mmol/L, platelets in 10³/μL, hemoglobin in g/dL) for the first 7 days postinjury from the electronic medical record, which were typically drawn at 4 am and analyzed at the central laboratory. Values were labeled as low, normal, or high based upon reference ranges reported in the EMR. The reference ranges were as follows: hemoglobin = 13.7–17.5 g/dL, sodium = 136–146 mmol/L, glucose = 70–90 mg/dL, platelets = 150,000–450,000/μL. If values were below the lowest value of the reference range, they were labeled as “low.” If values appeared within the reference range, they were labeled as “normal.” If values appeared above the reference range, they were labeled as “high.”

Standard protocol approvals, registrations, and patient consents

Participants or their legally authorized representatives provided written informed consent to participate after being approached by a member of the research team in the hospital. Human subjects research approvals were obtained by the institutional review board or ethics committee of each site. Article preparation adhered to the Strengthening the Reporting of Observational Studies in Epidemiology guidelines.

IMPACT model

The IMPACT extended model includes 10 predictors to discriminate mortality at 6 months. Predictors include age in years, GCS motor score, pupil reactivity (bilateral, unilateral, none), hypoxia, hypotension, Marshall CT Classification, evidence of traumatic subarachnoid hemorrhage on CT, epidural hemorrhage on CT, glucose (in mmol/L), and hemoglobin (in g/dL).¹ The resulting score is converted into a percentage, where 0 indicates no probability of the outcome and 100 indicates definite probability of the outcome.

Statistical analysis

We fit ordinal mixed-effects models to compare the trajectories of daily biomarkers within the first week of injury between sTBI patients who died compared with those who survived by 6 months. Daily biomarker measurements over the first 7 days were categorized into ordinal levels (low, normal, high). We modeled the ordinal outcome using a logit link function, with “Day” (treated as a categorical factor) and “Survival Group” included as fixed effects, and a random intercept for each subject to account for within-patient correlation.

The primary hypothesis test evaluated the interaction between Time (Day) and Group, determining if the daily pattern of biomarker levels significantly diverged between the two groups. Statistical significance was assessed using likelihood ratio tests (LRT) comparing the full interaction model against a reduced model (main effects only). p Values were adjusted for multiple comparisons using the Holm method.

To evaluate the potential additive value of daily biomarker levels to established prognostic models, we compared the IMPACT-only model to the IMPACT + biomarker trajectory model. The latter model included variables for high and low values of each biomarker, operationally defined as four or more of the first 7 postinjury days having a value above or below the reference range for that marker. DeLong’s test was used to compare differences in area under the curve (AUC) between the IMPACT-only model and the IMPACT + biomarker trajectory model (p < 0.05). To assess calibration, the Hosmer–Lemeshow test was calculated, where a nonsignificant test indicates better model calibration and lower scores indicate better model calibration.¹² All analyses were run in R (Posit Software, PBC; version 2025.05.0 + 496) using the lme4, lmerTest, epiR, caret, ggplot2, and pROC packages.^13–16

Results

Descriptive statistics for the overall cohort can be viewed in Table 1. Estimates for each marker by day can be found in Table 2. Descriptive statistics for each marker by day can be found in Supplementary Table S1. The LRT comparing the ordinal trajectories of sodium and platelets were statistically significant after false discovery correction (sodium: likelihood ratio = 51.0; adj. p < 0.001; platelets: likelihood ratio = 16.0; adj. p = 0.003), indicating that the overall trajectory over 7 days differs between groups. The LRT comparing the trajectories of glucose and hemoglobin were not statistically significant (p = 0.21–0.98).

Table 1.

Descriptive Statistics for the Cohort

	Mean ± SD or count (%)
Age	40.4 ± 17.1
Male biological sex	469 (79.4)
Death by 6 months	228 (38.4)
Total presenting GCS	5.6 ± 1.7
GCS motor score	3.2 ± 1.7
Hypoxia	93 (16.3)
Hypotension	141 (25.1)
Glucose	161.4 ± 62.0
Hemoglobin	13.0 ± 2.2
Marshall CT classification
1	24 (4.2)
2	283 (49.5)
3	66 (11.5)
4	42 (7.3)
5	126 (22.0)
6	31 (5.4)
Traumatic subarachnoid hemorrhage	468 (81.7)
Epidural hemorrhage	62 (10.8)

GCS, Glasgow Coma Scale; SD, standard deviation.

Table 2.

Ordinal Mixed-Effects Model Results for Biomarker Trajectories Across the First 7 Days Postinjury with an Interaction for Group Status (i.e., Alive or Not at 6 Months)

Biomarker	Model term	Estimate (log-odds)	SE	z	p Value
Hgb	Low\|Normal	−6.899	0.459	−15.028	0.000
Hgb	Normal\|High	−4.285	0.374	−11.462	0.000
Hgb	Day 2	−0.656	0.420	−1.563	0.118
Hgb	Day 3	−0.584	0.423	−1.379	0.168
Hgb	Day 4	−1.124	0.399	−2.819	0.005
Hgb	Day 5	−0.741	0.416	−1.780	0.075
Hgb	Day 6	−0.867	0.410	−2.114	0.034
Hgb	Day 7	−0.442	0.439	−1.007	0.314
Hgb	Nonsurvivor	0.284	0.596	0.476	0.634
Hgb	Day 2: Nonsurvivor	1.168	0.869	1.345	0.179
Hgb	Day 3: Nonsurvivor	−0.125	0.741	−0.168	0.866
Hgb	Day 4: Nonsurvivor	0.682	0.758	0.899	0.368
Hgb	Day 5: Nonsurvivor	−0.535	0.716	−0.747	0.455
Hgb	Day 6: Nonsurvivor	−0.308	0.727	−0.424	0.671
Hgb	Day 7: Nonsurvivor	0.922	0.981	0.941	0.347
Plt	Low\|Normal	−4.058	0.288	−14.103	0.000
Plt	Normal\|High	7.565	0.512	14.776	0.000
Plt	Day 2	−2.886	0.279	−10.354	0.000
Plt	Day 3	−4.384	0.298	−14.723	0.000
Plt	Day 4	−4.219	0.295	−14.312	0.000
Plt	Day 5	−2.992	0.280	−10.703	0.000
Plt	Day 6	−1.710	0.274	−6.247	0.000
Plt	Day 7	−0.664	0.280	−2.375	0.018
Plt	Nonsurvivor	−0.835	0.414	−2.017	0.044
Plt	Day 2: Nonsurvivor	0.116	0.427	0.271	0.786
Plt	Day 3: Nonsurvivor	−0.125	0.462	−0.270	0.787
Plt	Day 4: Nonsurvivor	−0.304	0.467	−0.652	0.514
Plt	Day 5: Nonsurvivor	−0.537	0.457	−1.175	0.240
Plt	Day 6: Nonsurvivor	−0.900	0.459	−1.961	0.050*
Plt	Day 7: Nonsurvivor	−1.323	0.468	−2.826	0.005*
Na	Low\|Normal	−2.664	0.199	−13.394	0.000
Na	Normal\|High	4.005	0.216	18.559	0.000
Na	Day 2	0.879	0.209	4.195	0.000
Na	Day 3	1.677	0.214	7.825	0.000
Na	Day 4	2.740	0.220	12.466	0.000
Na	Day 5	3.106	0.223	13.949	0.000
Na	Day 6	2.697	0.220	12.264	0.000
Na	Day 7	2.168	0.218	9.927	0.000
Na	Nonsurvivor	−0.654	0.299	−2.184	0.029
Na	Day 2: Nonsurvivor	1.192	0.342	3.485	0.000*
Na	Day 3: Nonsurvivor	1.914	0.351	5.449	0.000*
Na	Day 4: Nonsurvivor	1.508	0.353	4.276	0.000*
Na	Day 5: Nonsurvivor	1.626	0.355	4.581	0.000*
Na	Day 6: Nonsurvivor	2.019	0.364	5.552	0.000*
Na	Day 7: Nonsurvivor	2.243	0.375	5.986	0.000*
Glu	Low\|Normal	0.523	0.215	2.426	0.015
Glu	Normal\|High	7.902	0.731	10.816	0.000
Glu	Day 2	−3.158	0.330	−9.576	0.000
Glu	Day 3	−5.985	0.588	−10.174	0.000
Glu	Day 4	−6.404	0.650	−9.860	0.000
Glu	Day 5	−7.014	0.751	−9.341	0.000
Glu	Day 6	−7.432	0.836	−8.889	0.000
Glu	Day 7	−8.967	1.303	−6.880	0.000
Glu	Nonsurvivor	−0.792	0.351	−2.261	0.024
Glu	Day 2: Nonsurvivor	−0.121	0.499	−0.243	0.808
Glu	Day 3: Nonsurvivor	0.424	0.888	0.478	0.633
Glu	Day 4: Nonsurvivor	−0.107	1.181	−0.091	0.928
Glu	Day 5: Nonsurvivor	−0.419	1.637	−0.256	0.798
Glu	Day 6: Nonsurvivor	0.099	1.705	0.058	0.954
Glu	Day 7: Nonsurvivor	1.711	1.985	0.862	0.389

Statistically significant at p < 0.05.

SE, standard error.

The models identified a divergence in sodium values between survivors and nonsurvivors by day 2 postinjury, where nonsurvivors had 3.3 times higher odds of being in a higher sodium category compared with survivors (p < 0.001; Fig. 1). The difference between groups increased over time, peaking at days 6–7 where nonsurvivors had 7.5–9.4 times higher odds of being in a higher sodium category compared with survivors (p < 0.001; Fig. 1).

FIG. 1.

Predicted daily platelet reference range category (low, normal, high) for platelet level by 6-month outcome.

The models identified a divergence in platelet values on days 6 and 7, where nonsurvivors had lower odds (odds ratio [OR] = 0.27–0.41; Fig. 2) of being in higher platelet categories than survivors on those days.

FIG. 2.

Predicted daily sodium reference range category (low, normal, high) for platelet level by 6-month outcome.

The IMPACT-only model had an AUC for 6-month mortality of 0.85 and a Hosmer–Lemeshow p value = 0.68. The IMPACT-biomarker trajectory model had an AUC for 6-month mortality of 0.87 (DeLong’s test p value of 0.005) and a Hosmer–Lemeshow p value of 0.16.

Discussion

In this study of nearly 600 patients with sTBI, including high/low trajectories of platelets, hemoglobin, glucose, and sodium to the IMPACT model improved group-level discrimination by 2%. Our results highlight the feasibility of updating prognostic models which include only predictors from hospital admission, as these dynamics can make meaningful improvements in the calibration and accuracy of predictions (i.e., improve performance of the models for individual cases). The advantage of including predictors from a single timepoint is the reduction in burden to the medical and/or research team. In this study, we show that routine assessments typically conducted daily for all patients in the intensive care unit can add significant clinical value to these baseline predictors at minimal additional time cost.

Clear differences in daily metabolic panel variables from a standardized time across patients were observed for the first week postinjury between patients who died by 6 months compared with survivors. On days 6 and 7 postinjury, nonsurvivors had 59–73% lower odds of being above the upper limit of the reference range for platelets than survivors. This finding needs to be explored more deeply in future research, as there could be many reasons for thrombocytopenia (further bleeding, systemic inflammation, coagulopathy, etc.) that could influence these trajectories. Many additional clinical variables, such as biological injury severity, transfusion, or preinjury antiplatelet use, could impact platelet values and need to be considered in future work. Observational studies have shown mixed results for the use of platelet transfusions to improve sTBI outcomes. Using thromboelastogram platelet mapping (i.e., a guided method to detect platelet inhibition and trigger platelet transfusion), several studies have shown improved platelet dysfunction and lower mortality within 12 months postinjury.^17,18 Anglin et al.¹⁹ found that platelet transfusions were not associated with long-term functional outcome, but some individual transfusion strategies were associated with poorer outcomes. Other observational studies have shown that early thrombocytopenia in sTBI patients is associated with substantially higher hospital and 12-month mortality¹¹; evidence which is used for the recommended target of >100 × 10⁹/L targets for sTBI management.²⁰

Our results highlight the growing appreciation for heterogeneity amongst TBI patients, both in clinical presentation and recovery potential.^21,22 In 2025, the National Institutes of Neurological Disorders and Stroke TBI Classification and Nomenclature Initiative introduced a new model for TBI characterization called CBI-M (Clinical, Biomarkers, Imaging, Modifiers).²³ This model has not yet been applied to prognostication, but our results show how important the “biomarkers” component can be for prognostication in this population. The original model included several variables classified as “clinical” (e.g., GCS motor score, hypoxia, hypotension) and “imaging” (e.g., Marshall CT classification, presence of traumatic subarachnoid hemorrhage or epidural hematoma), but neither the original IMPACT model nor the model presented here included “modifiers” (i.e., risk factors of better or worse outcomes).²⁴ The importance of modifiers for prognostication after “mild” TBI is well-described,²⁵ but evidence of the importance of modifiers after “severe” TBI is less understood. This is a critical area of future research in this population.

Limitations and future directions

We also observed survivors had substantially higher odds of being above the reference range for sodium from days 2–7 postinjury than nonsurvivors. Trajectories of sodium, glucose, and hemoglobin could be substantially influenced by intensive care unit management (e.g., hyperosmolar therapy, insulin protocols, blood transfusions).²⁶ Hyperosmolar therapy can shift water concentrations with direct impacts to sodium levels, which require frequent monitoring to ensure appropriate balance.²⁷ These treatments also transiently influence hemoglobin by concentrating or diluting blood hemoconcentration.²⁸ Insulin administration rapidly corrects glucose levels but can also indirectly result in a rise of sodium and diluted hemoglobin due to shifts in the cellular location of water.^29,30 Blood transfusions are indicated when hemoglobin levels are below certain cutoffs, leading to a direct increase in hemoglobin.³¹ The use of these treatments was not recorded in a systematic manner and, as such, was not included in these analyses. This is an acknowledged limitation, as the initiation of these protocols could have influenced trajectories of the biomarkers for those who survived compared with those who died, for example. Future work needs to understand differences in trajectories based upon administration of standardized intensive care unit protocols. While this is an important area for future research, the current work’s goal was to illustrate if trajectories of clinical labs over time would improve statistical modeling of clinical outcomes, not necessarily the physiological explanation for the trajectories themselves.

It is also worth noting that these data span a relatively long period of time, which observed changes in consensus for clinical management for sTBI.^6,7,32 These changes in protocols could have influenced outcomes for earlier patients compared with more recent patients. Future work should consider evaluating any potential changes in outcomes after formal adoption of new TBI management guidelines, which was not possible in the current analysis. The evolution of classical clinical assessments, such as the GCS, during the ICU stay may yield important information for future studies, as well. For example, patients whose GCS score recovered to 13–15 by 1 week postinjury would likely be at lower risk of 6-month mortality than others whose GCS score remained ∼8 or lower. Indeed, “neuroworsening” (i.e., a worsening of GCS motor score after initial presentation) is a risk factor for worse outcomes after TBI.³³ This is a single-center study that could impact external validity; future work should attempt to replicate these findings in a multisite cohort with broader classification of polytrauma or comorbidities that may have influenced lab values.

We observed significant heterogeneity in trajectories, where patients move from “low” to “normal” to “high” or vice versa. This is possibly due to the inclusion criteria of GCS 3–8 being heterogenous regarding ranges of clinical pathology and severity, metabolic treatments, and other secondary injuries, which could not be quantified or controlled for in the current analysis. Treatments to bring metabolic values back within the normal range were not able to be accounted for in the current study and should be a focus of future work. The intent of this work is not to present a definitive methodology for use, but rather to show that this type of modeling is possible and useful for prognostication. Future work should quantify clinically meaningful thresholds of these 4 laboratory markers related to 6-month outcomes but could benefit from analyses that add markers such as those specific to brain injury (e.g., glial fibrillary acidic protein, ubiquitin c-terminal hydrolase L1). Using this method to predict other outcomes (such as in-hospital mortality or the need for surgery) would be important additions to the literature. Dichotomizing trajectories as “high” or “low” was arbitrarily set to 4+ of the first 7 days postinjury to represent a majority of the first postinjury week. Validation studies will be required to refine thresholds with clinical meaning for the change in laboratory values.

Conclusion

Inclusion of daily clinical lab values for platelets, sodium, glucose, and hemoglobin for the first week post-sTBI may improve 6-month prognostic model performance compared with the IMPACT model. Results of this work represent a preliminary step toward iterative improvement of prognostication for sTBI.

Transparency, Rigor, and Reproducibility Statement

This is a secondary analysis of a prospective cohort study, so the study was not preregistered. The analysis plan for the current study was not preregistered online but was conceived by the primary authors and executed by a biostatistician. The primary author certifies that the analyses were prespecified. The sample size was a convenience sample of participants enrolled at a single site. Data collection and analyses were performed by investigators who were aware of relevant characteristics. Data and analytic code are available at reasonable request. R was used to complete analyses.

Authors’ Contributions

S.R.E.: Writing original draft (co-lead) and review and editing (equal); J.S.: Methodology (co-lead) and review and editing (equal); M.A.M.: Methodology (co-lead) and review and editing (equal); M.W.P.: Methodology (co-lead) and review and editing (equal); H.D.: Methodology (co-lead) and review and editing (equal); R.S.: Methodology (equal) and review and editing (equal); A.S.: Methodology (co-lead) and review and editing (equal); S.B.: Methodology (co-lead) and review and editing (equal); M.R.K.: Methodology (co-lead) and review and editing (equal); T.A.: Methodology (co-lead) and review and editing (equal); A.P.: Methodology (co-lead) and review and editing (equal); E.L.N.: Review and editing (equal); D.O.O.: Conceptualization (co-lead), methodology (co-lead), and review and editing (co-lead).

Footnotes

Data Availability Statement

Deidentified data are available by reasonable request to the corresponding author.

Author Disclosure Statement

The authors have no conflicts of interest to report.

Funding Information

No funding was received for this work.

Supplemental Material

References

1. Steyerberg

, Mushkudiani

, Perel

, et al. Predicting outcome after traumatic brain injury: Development and international validation of prognostic scores based on admission characteristics. PLoS Med 2008;5(8):e165; discussion e165; doi: 10.1371/journal.pmed.0050165

2. Eagle

, Pease

, Nwachuku

, et al. Prognostic models for traumatic brain injury have good discrimination but poor overall model performance for predicting mortality and unfavorable outcomes. Neurosurgery 2023;92(1):137–143; doi: 10.1227/neu.0000000000002150

3. Moskowitz

, Quinn

, Khan

, et al. Should we use the IMPACT-Model for the outcome prognostication of TBI patients? A qualitative study assessing physicians’ perceptions. MDM Policy Pract 2018;3(1):2381468318757987; doi: 10.1177/2381468318757987

4. Letsinger

, Rommel

, Hirschi

, et al. The aggressiveness of neurotrauma practitioners and the influence of the IMPACT prognostic calculator. PLoS One 2017;12(8):e0183552; doi: 10.1371/journal.pone.0183552

5. Snider

, Deng

, Hammond

, et al. Time to command-following and outcomes after traumatic brain injury. JAMA Netw Open 2024;7(12):e2449928; doi: 10.1001/jamanetworkopen.2024.49928

6. Carney

, Totten

, O’Reilly

, et al. Guidelines for the management of severe traumatic brain injury. Neurosurgery 2017;80(1):6–15.

7. Hawryluk

. Prognostication in traumatic brain injury: now what? J Neurosurg 2024;141(2):412–416.

8. Matsushima

, Peng

, Velasco

, et al. Glucose variability negatively impacts long-term functional outcome in patients with traumatic brain injury. J Crit Care 2012;27(2):125–131; doi: 10.1016/j.jcrc.2011.08.012

9. Litofsky

, Martin

, Diaz

, et al. The negative impact of anemia in outcome from traumatic brain injury. World Neurosurg 2016;90:82–90; doi: 10.1016/j.wneu.2016.02.076

10.

10. Harrois

, Anstey

, van der Jagt

, et al.; TBI Collaborative. Variability in serum sodium concentration and prognostic significance in severe traumatic brain injury: A multicenter observational study. Neurocrit Care 2021;34(3):899–907; doi: 10.1007/s12028-020-01118-8

11.

11. Lillemäe

, Luostarinen

, Reinikainen

, et al. Early thrombocytopenia is associated with an increased risk of mortality in patients with traumatic brain injury treated in the intensive care unit: A Finnish Intensive Care Consortium study. Acta Neurochir (Wien) 2022;164(10):2731–2740; doi: 10.1007/s00701-022-05277-9

12.

12. Van Calster

, McLernon

, van Smeden

, et al.; Topic Group ‘Evaluating diagnostic tests and prediction models’ of the STRATOS initiative. Calibration: The Achilles heel of predictive analytics. BMC Med 2019;17(1):230; doi: 10.1186/s12916-019-1466-7

13.

13. Robin

, Turck

, Hainard

, et al. pROC: An open-source package for R and S+ to analyze and compare ROC curves. BMC Bioinformatics 2011;12:77; doi: 10.1186/1471-2105-12-77

14.

14. Kuhn

. Building predictive models in R using the caret package. J Stat Soft 2008;28(5):1–26.

15.

15. Bates

, Mächler

, Bolker

, et al. Fitting linear mixed-effects models using lme4. J Stat Soft 2015;67(1):1–48.

16.

16. Stevenson

, Stevenson

, BiasedUrn

. Package ‘epiR’. Tools for the analysis of epidemiological data R package version 09–62. 2015.

17.

17. Furay

, Daley

, Teixeira

, et al. Goal-directed platelet transfusions correct platelet dysfunction and may improve survival in patients with severe traumatic brain injury. J Trauma Acute Care Surg 2018;85(5):881–887; doi: 10.1097/ta.0000000000002047

18.

18. Miles

MVP

, Hicks

, Parmer

, et al. Traumatic brain injury patients with platelet inhibition receiving platelet transfusion demonstrate decreased need for neurosurgical intervention and decreased mortality. J Trauma Acute Care Surg 2022;92(4):701–707; doi: 10.1097/ta.0000000000003516

19.

19. Anglin

, Spence

, Warner

, et al. Effects of platelet and plasma transfusion on outcome in traumatic brain injury patients with moderate bleeding diatheses. J Neurosurg 2013;118(3):676–686; doi: 10.3171/2012.11.Jns12622

20.

20. Tqip

. Best practices in the management of traumatic brain injury. ACS Committee on Trauma 2015:3–23.

21.

21. Åkerlund

CAI

, Holst

, Stocchetti

, et al.; CENTER-TBI Participants and Investigators. Clustering identifies endotypes of traumatic brain injury in an intensive care cohort: A CENTER-TBI study. Crit Care 2022;26(1):228; doi: 10.1186/s13054-022-04079-w

22.

22. Folweiler

, Sandsmark

, Diaz-Arrastia

, et al. Unsupervised machine learning reveals novel traumatic brain injury patient phenotypes with distinct acute injury profiles and long-term outcomes. J Neurotrauma 2020;37(12):1431–1444; doi: 10.1089/neu.2019.6705

23.

23. Manley

, Dams-O’Connor

, Alosco

, et al.; NIH-NINDS TBI Classification and Nomenclature Initiative. A new characterisation of acute traumatic brain injury: The NIH-NINDS TBI Classification and Nomenclature Initiative. Lancet Neurol 2025;24(6):512–523; doi: 10.1016/s1474-4422(25)00154-1

24.

24. Nelson

, Wilson

, Albrecht

, et al. Toward more holistic early traumatic brain injury evaluation and care: Recommendations from the 2024 national institute of neurological disorders and stroke traumatic brain injury classification and nomenclature initiative psychosocial and environmental modifiers working group. J Neurotrauma 2025;42(13–14):1023–1037; doi: 10.1089/neu.2024.0569

25.

25. Mikolić

, Polinder

, Steyerberg

, et al.; Collaborative European NeuroTrauma Effectiveness Research in Traumatic Brain Injury (CENTER-TBI) Study Participants and Investigators. Prediction of global functional outcome and post-concussive symptoms after mild traumatic brain injury: External validation of prognostic models in the Collaborative European NeuroTrauma Effectiveness Research in Traumatic Brain Injury (CENTER-TBI) study. J Neurotrauma 2021;38(2):196–209.

26.

26. Hawryluk

, Ghajar

. Evolution and impact of the Brain Trauma Foundation guidelines. Neurosurgery 2021;89(6):1148–1156.

27.

27. Dabrowski

, Siwicka-Gieroba

, Robba

, et al. Potentially detrimental effects of hyperosmolality in patients treated for traumatic brain injury. J Clin Med 2021;10(18):4141; doi: 10.3390/jcm10184141

28.

28. Wellard

, Kuwabara

, Adelson

, et al. Physiologic characteristics of hyperosmolar therapy after pediatric traumatic brain injury. Front Neurol 2021;12:662089; doi: 10.3389/fneur.2021.662089

29.

29. Ing

, Ganta

, Bhave

, et al. The corrected serum sodium concentration in hyperglycemic crises: Computation and clinical applications. Front Med (Lausanne) 2020;7:477; doi: 10.3389/fmed.2020.00477

30.

30. Quispe-Cornejo

, Alves da Cunha

, Njimi

, et al. Effects of rapid fluid infusion on hemoglobin concentration: A systematic review and meta-analysis. Crit Care 2022;26(1):324; doi: 10.1186/s13054-022-04191-x

31.

31. Striker

, Woldorf

, Holt

. Modification of sodium, glucose, potassium, and osmolarity in packed red blood cells and fresh frozen plasma using a desktop hemoconcentrator setup. J Extra Corpor Technol 2012;44(2):60–65.

32.

32. Hawryluk

, Aguilera

, Buki

, et al. A management algorithm for patients with intracranial pressure monitoring: The Seattle International Severe Traumatic Brain Injury Consensus Conference (SIBICC). Intensive Care Med 2019;45(12):1783–1794.

33.

33. Yue

, Krishnan

, Kanter

, et al.; TRACK-TBI Investigators. Neuroworsening in the emergency department is a predictor of traumatic brain injury intervention and outcome: A TRACK-TBI pilot study. J Clin Med 2023;12(5):2024; doi: 10.3390/jcm12052024

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.02 MB

0.00 MB