Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Open Access

Peer-reviewed

Research Article

Complex patterns of functional change among TBI survivors from discharge to 6-month follow-up: Findings from a registry-based cohort study

  • Hosein Kouchaki ,

    Contributed equally to this work with: Hosein Kouchaki, Sina Bazmi

    Roles Conceptualization, Formal analysis, Project administration, Software, Visualization, Writing – original draft, Writing – review & editing

    Affiliations Health Policy Research Center, Institute of Health, Shiraz University of Medical Sciences, Shiraz, Iran, USERN Office, Fasa University of Medical Sciences, Fasa, Iran

    ⨯
  • Sina Bazmi ,

    Contributed equally to this work with: Hosein Kouchaki, Sina Bazmi

    Roles Conceptualization, Methodology, Supervision, Validation, Writing – original draft, Writing – review & editing

    Affiliation USERN Office, Fasa University of Medical Sciences, Fasa, Iran

    ⨯
  • Saman Rabiei,

    Roles Investigation, Writing – original draft

    Affiliations Student Research Committee, Fasa University of Medical Sciences, Fasa, Iran, Department of Psychiatry, School of Medicine, Fasa University of Medical Sciences, Fasa, Iran

    ⨯
  • Parnia Kamyab,

    Roles Investigation, Writing – original draft

    Affiliations Health Policy Research Center, Institute of Health, Shiraz University of Medical Sciences, Shiraz, Iran, USERN Office, Fasa University of Medical Sciences, Fasa, Iran

    ⨯
  • Mohammadreza Khosravi,

    Roles Investigation, Writing – original draft

    Affiliations Noncommunicable Diseases Research Center, Fasa University of Medical Sciences, Fasa, Iran, Trauma Research Center, Shahid Rajaee (Emtiaz) Trauma Hospital, Shiraz University of Medical Sciences, Shiraz, Iran

    ⨯
  • Reza Taheri,

    Roles Supervision, Validation

    Affiliations Shiraz Neuroscience Research Center, Shiraz University of Medical Sciences, Shiraz, Iran, School of Medicine, Fasa University of Medical Sciences, Fasa, Iran

    ⨯
  • Hosseinali Khalili ,

    Roles Data curation, Resources, Supervision, Validation, Writing – review & editing

    kmsrc89@gmail.com (RT); khalilih@yahoo.com (HK)

    Affiliations Trauma Research Center, Shahid Rajaee (Emtiaz) Trauma Hospital, Shiraz University of Medical Sciences, Shiraz, Iran, Department of Neurosurgery, Shiraz University of Medical Sciences, Shiraz, Iran

    ⨯
  • Reza Tabrizi

    Roles Supervision, Validation, Writing – review & editing

    kmsrc89@gmail.com (RT); khalilih@yahoo.com (HK)

    Affiliations Noncommunicable Diseases Research Center, Fasa University of Medical Sciences, Fasa, Iran, Clinical Research Development Unit, Valiasr Hospital, Fasa University of Medical Sciences, Fasa, Iran

    ⨯

Abstract

Background

Functional recovery after traumatic brain injury (TBI) is heterogeneous, and reliance on single time-point outcomes may obscure meaningful post-discharge change.

Objective

To identify early clinical, laboratory, and radiologic factors associated with functional change between hospital discharge and six-month follow-up using the Glasgow Outcome Scale-Extended (GOSE).

Methods

In this longitudinal study, 2,948 TBI survivors aged 12 years or older admitted to a tertiary trauma center underwent GOSE assessment at discharge and six months post-discharge. Functional trajectories were classified as recovering, stable, or regressing. Multinomial logistic regression evaluated associations between early variables and functional trajectories. Analyses were additionally stratified by discharge GOSE to assess statistical effects.

Results

Substantial bidirectional functional change occurred after discharge. Several variables, including age, Glasgow Coma Scale, subarachnoid or intraventricular hemorrhage, hemoglobin, and coagulation indices, were associated with both recovery and regression. Discharge GOSE strongly predicted subsequent change but increased the odds of both improvement and decline. Subgroup analyses indicated many unexpected associations reflected ceiling and floor effects and regression to the mean. Nevertheless, some predictors showed consistent directional effects: in non-severe TBI, age 40–74 years, low admission blood glucose, and higher platelet counts; and in severe TBI, age > 40 years, low admission blood glucose, pneumocephalus, and bilateral fixed pupils.

Conclusions

Post-TBI functional outcomes follow complex, often unpredictable trajectories. These findings highlight limitations of static early prognostication and support a trajectory-oriented approach to post-discharge TBI care.

Citation: Kouchaki H, Bazmi S, Rabiei S, Kamyab P, Khosravi M, Taheri R, et al. (2026) Complex patterns of functional change among TBI survivors from discharge to 6-month follow-up: Findings from a registry-based cohort study. PLoS One 21(7): e0353189. https://doi.org/10.1371/journal.pone.0353189

Editor: Chinh Quoc Luong, Bach Mai Hospital, VIET NAM

Received: February 17, 2026; Accepted: June 19, 2026; Published: July 9, 2026

Copyright: © 2026 Kouchaki et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Data Availability: In our institutional policy, it is not stated that the data should be made public, and a data and material transfer agreement should not allow further transfer of data without the provider’s prior written consent. However, the data can be made available upon request from the corresponding author, who is a member of this team. Additionally, the dataset generated for this study is available upon request from the Trauma Research Center, Shiraz University of Medical Sciences, through the management team at Rajaei (Emtiaz) Hospital. Requests can be submitted via email at Natrauma2@sums.ac.ir (mailto:ncdrc.fums.ac.ir@gmail.com).

Funding: The author(s) received no specific funding for this work.

Competing interests: The authors have declared that no competing interests exist.

Introduction

Traumatic brain injury (TBI) is a complex condition that exhibits considerable variation in its mechanisms of injury, underlying pathophysiological processes, and clinical manifestations [1]. TBI has traditionally been considered a condition affecting younger individuals during their most productive years, leading to significant long-term consequences for their overall functioning [2]. However, recent epidemiological data indicate a substantial shift toward older adults, as aging populations and fall-related injuries have also increased the burden of TBI in the elderly [3]. The traditional classification system for TBI categorizes its severity using the Glasgow Coma Scale (GCS), which divides TBI into three levels: mild (GCS scores of 13–15), moderate (GCS scores of 9–12), and severe (GCS scores of 3–8) [4]. More recent frameworks have been proposed to provide a multidimensional characterization of TBI [5], although GCS-based classification remains widely used in clinical research.

Annually, over 1.7 million cases of TBI occur in the United States, with the majority classified as mild [4]. In contrast, the severe form of TBI (sTBI) has a mortality rate of approximately 40%, and fewer than 10% of individuals achieve a favorable recovery outcome [6,7]. Although preventable, TBI remains a significant cause of death and disability worldwide [8], and even survivors may face lifelong challenges. Research indicates that due to neurological deficits, reduced quality of life, and mental health issues, only 12% to 33% of TBI survivors are able to return to pre-injury levels of functioning, such as employment [2,8–11]. Therefore, accurately predicting outcomes, including survival, mortality, and complications, during the critical early days of hospitalization is essential for guiding clinical decision-making [12].

While most clinicians agree on the importance of prognostic assessment in TBI care, only a limited number feel confident in their ability to accurately predict patients’ outcomes [13]. Early outcome prediction is a valuable tool for improving communication with patients and their families, helping to manage expectations and decision-making [14]. Additionally, it aids in adjusting case-mix for more accurate benchmarking of care quality, ultimately enhancing the healthcare experience [14]. The prognostic effects of admission characteristics on functional outcomes have been evaluated at various intervals, including 14 days and 6–9 months post-injury, utilizing the Glasgow Outcome Scale-Extended (GOSE) [14–17]. Typically, previous studies have categorized functional outcomes into two groups: unfavorable outcomes (GOSE 1–4) including mortality (GOSE 1), and favorable outcomes (GOSE 5–8) [14,15,18,19]. While traditional classifications of TBI outcomes provide valuable insights, the heterogeneous nature of recovery trajectories in TBI patients remains underexplored in the literature [20]. Simplifying the GOSE into binary categories of “favorable” or “unfavorable” may obscure the identification of specific treatment effects [13,18,21]. This can lead to misguidance in prognostic predictions and influence critical decisions, such as the withdrawal of life-sustaining therapy [18].

This retrospective cohort study aims to investigate the association between early assessment variables and changes in the GOSE from discharge to six months post-discharge in TBI patients admitted to a tertiary trauma hospital. A key innovation in our approach is the classification of the longitudinal changes in GOSE (delta GOSE) into three distinct groups: “regressing,” “stable,” and “recovering.” By examining the “delta GOSE,” we offer a more dynamic and comprehensive approach than studies that rely on a single time-point outcome (e.g., GOSE0 or FGOSE). This also allows for a more precise evaluation of functional trajectories, as opposed to previous studies that have typically considered outcome categories unidirectionally (e.g., recovery or regression) [21–24] and provides a better understanding of functional changes over time.

Materials and methods

Study design and population

This was a retrospective registry-based cohort study of patients with TBI admitted to Emtiaz Hospital, a Level I trauma center located in southern Iran, between January 2016 and December 2022. Emtiaz Hospital serves as a major referral center for Fars Province and its neighboring regions, with a catchment population of approximately 7.5 million. The study included patients aged 12 years and older who were hospitalized with TBI and survived until discharge. Patients who died in the emergency department or during hospitalization, those with penetrating TBI, and patients lacking data on functional outcomes at discharge or the 6-month follow-up were excluded. The patient selection process is detailed in Fig 1. Given the relatively small proportion of excluded cases, a complete-case analysis approach was considered unlikely to substantially bias the results.

thumbnail
Download:
Fig 1. Flowchart of participant selection and study method.

1Exclusion criteria were applied in a stepwise manner; therefore, overlapping exclusions were not counted multiple times. 2Trauma severity was defined based on initial Glasgow Coma Scale (GCS) scores: GCS 9–15 indicated mild to moderate TBI, and GCS 3–8 indicated severe TBI. 3Functional trajectories were classified based on the change in Glasgow Outcome Scale–Extended (GOSE) scores from discharge to six months. Patients were considered “Recovering” with a ≥ 2-point increase, “Stable” with a change between −1 and +1, and “Regressing” with a ≥ 2-point decrease. Patients who declined by only one point but had a GOSE score of 1 at follow-up were reclassified as “Regressing.” 4GOSE at discharge (GOSE0) was dichotomized into unfavorable (GOSE ≤ 4) and favorable (GOSE ≥ 5) categories.

https://doi.org/10.1371/journal.pone.0353189.g001

Independent variables and measures

Data were extracted from a prospectively maintained trauma registry. The following variables were collected upon hospital admission: demographic information (age, sex); medical comorbidities, including hypertension (HTN), diabetes mellitus (DM), and cerebrovascular or cardiovascular disease (CVA/CVD); and physiologic findings, such as systolic blood pressure (SBP), GCS score, and pupil reactivity (categorized as brisk, anisocoria, bilateral fixed, or uncheckable). Initial laboratory parameters included hemoglobin (Hb), international normalized ratio (INR), partial thromboplastin time (PTT), platelet count, and the first recorded random blood sugar (BS) level.

Non-contrast brain computed tomography (CT) scans were performed at admission and evaluated within 24 hours post-injury by two neurosurgery residents, with adjudication by a board-certified neurosurgeon blinded to clinical data. Radiologic findings included presence of intracerebral hemorrhage (ICH), intraventricular hemorrhage (IVH), epidural hematoma (EDH), subdural hematoma (SDH), subarachnoid hemorrhage (SAH), pneumocephalus, basilar skull fracture (BSFx), depressed skull fracture (DSFx), midline shift (>5 mm), and brain cistern status (compressed, absent, or normal). The Rotterdam CT score, a composite scale ranging from 1 to 6 that evaluates cistern status, midline shift, traumatic subarachnoid hemorrhage, and intraventricular hemorrhage, was also recorded but ultimately excluded from multivariable regression models due to multicollinearity with its components. The mechanism of injury was categorized as falling, motor vehicle collision, pedestrian trauma, or other. Hospital course variables included total hospital length of stay (LOS) and intensive care unit (ICU) LOS. ICU LOS was excluded from multivariable regression models due to collinearity with total hospital LOS. Moreover, because hospital LOS reflects downstream effects of clinical management rather than initial patient status, it was excluded from the primary regression models to avoid introducing post-treatment bias. Nonetheless, exploratory models including hospital LOS were conducted to assess its influence on trajectory outcomes and are presented in Supplementary Tables mentioned below.

Outcome assessment

The primary outcome of interest was the patient’s functional trajectory over a six-month period, assessed using the GOSE. The GOSE is an eight-point ordinal scale commonly employed in TBI research to measure global functional outcomes, ranging from death (1) to upper good recovery (8) [25]. Each patient’s GOSE score was recorded at the time of hospital discharge and again at six months post-discharge. Based on the change in GOSE between these two time points, patients were classified into one of three functional trajectory groups. Patients were categorized as “Recovering” if their GOSE improved by two or more points, “Stable” if the change in GOSE was between −1 and +1, and “Regressing” if the GOSE declined by two or more points. Additionally, to more accurately reflect significant deterioration, patients whose GOSE decreased by one point but who died after discharge (i.e., GOSE score of 1 at follow-up) were reclassified into the Regressing group [26]. For regression modeling, GOSE0 was dichotomized into two categories: unfavorable (GOSE ≤ 4) and favorable (GOSE ≥ 5), based on commonly used functional outcome thresholds in TBI literature [14,15,18,19].

Statistical analyses

Continuous variables were summarized as medians with interquartile ranges (IQR, 25th - 75th percentile), while categorical variables were reported as frequencies and percentages. The normality of continuous variables was assessed using the Shapiro-Wilk test. As most continuous variables were not normally distributed, the Mann–Whitney U test was applied to compare differences across groups. Differences between categorical variables were assessed using the Chi-square test and Fisher’s Exact Test. For larger contingency tables, Monte Carlo exact tests were applied when appropriate.

Multinomial logistic regression was used to identify independent predictors of the six-month functional trajectory. First, univariate multinomial regression models were performed for each variable to estimate unadjusted odds ratios (ORs) and 95% confidence intervals (CIs).

Candidate variables for multivariable modeling were initially selected based on univariate analysis using a liberal threshold (p < 0.20). This approach is commonly used in epidemiological research to avoid excluding potentially important variables that may not show strong univariate associations but could become significant in multivariable contexts due to confounding or interaction effects. Indeed, prior methodological studies recommend using higher p-value thresholds (e.g., 0.15–0.25) during initial screening to minimize the risk of omitting relevant predictors [27,28]. However, recognizing the limitations of purely data-driven selection methods, we conducted additional sensitivity analyses in which covariate selection was based on clinical relevance and prior evidence, informed by established TBI prognostic models (e.g., CRASH and IMPACT).

Before conducting multivariable regression, collinearity diagnostics were performed to assess the interdependence of predictor variables. The final multivariable models provided adjusted ORs and 95% CIs for the likelihood of being in either the regressing or recovering group, relative to the stable group.

In addition, subgroup analyses were conducted to further evaluate predictors that showed unexpected associations in the primary models. Patients were stratified by GOSE0 into three categories (poor, intermediate, good) to account for differences in available range for subsequent recovery or regression along the ordinal scale. Multinomial regression models were then applied within each discharge group to examine whether these predictors displayed patterns consistent with ceiling and floor effects or regression to the mean.

All analyses, except the model that predicted six-month functional trajectory based on GOSE0, were stratified by initial TBI severity using arrival GCS scores: mild to moderate or non-severe TBI (nsTBI) (GCS 9–15) and sTBI (GCS 3–8), to account for baseline neurologic injury severity.In addition to the primary analyses, patients were categorized into mild TBI (GCS 13-15) and moderate-severe TBI (msTBI) (GCS 3-12) groups based on admission GCS. This grouping was performed to enhance comparability with existing literature, where moderate and severe TBI are frequently analyzed together. All primary analyses were repeated within these severity strata. Statistical analyses were conducted using IBM SPSS Statistics for Windows, Version 23.0 (IBM Corp., Armonk, NY, USA). A two-sided p-value <0.05 was considered statistically significant.

Ethics approval and consent to participate

The current study received ethical approval from the Ethics Committee of Fasa University of Medical Sciences (approval code: IR.FUMS.REC.1404.160). The research protocol complied with the ethical guidelines of the Declaration of Helsinki. Written informed consent for study participation and data dissemination was obtained from patients or their legal representatives upon hospital admission. All data utilized in this study were retrieved from the trauma registry database, originally established with ethical approval code IR.SUMS.REC.1401.183. Data were retrieved for research purposes on December 20, 2025. The authors lack access to any information that could identify individuals post-data collection.

Results

Patient characteristics and functional trajectories

Out of 2,948 TBI patients, 1,873 (63.5%) presented with mild to moderate injury (GCS score 9–15), while 1,075 (36.5%) had severe injury (GCS score 3–8). In the mild to moderate TBI group, 17.4% of patients demonstrated recovery, 79.0% remained stable, and 3.7% experienced regression. Among patients with sTBI, 40.0% recovered, 49.6% remained stable, and 10.4% regressed.

Group comparisons by TBI severity

Non-severe TBI.

Patients who regressed were significantly older than those who remained stable or recovered (median age: 66 vs. 36 and 35 years, respectively; p < 0.001) and had a higher prevalence of HTN, DM, or CVA/CVD (all p < 0.001). Additionally, patients who regressed presented with lower admission GCS and GCS motor scores (both p < 0.001) and higher SBP (p = 0.019). Regressing patients had significantly higher rates of IVH (p = 0.007), SAH (p < 0.001), and SDH (p = 0.013), but lower rates of EDH (p < 0.001) and BSFx (p = 0.029). Abnormal cistern status and higher Rotterdam scores were significantly associated with both regression and recovery (p < 0.001). Longer hospital and ICU stays were also associated with regression and recovery groups (p < 0.001). Falls were the most frequent injury mechanism among regressing patients (p < 0.001). In addition, unfavorable discharge functional status (GOSE0 ≤ 4) was markedly more frequent among patients who subsequently regressed or recovered than among those who remained stable. For detailed numerical data and additional clinical and radiological features, see Table 1.

thumbnail
Download:
Table 1. Baseline clinical, radiologic, and laboratory characteristics of mild to moderate TBI patients (GCS 9–15) stratified by six-month functional trajectory group.

https://doi.org/10.1371/journal.pone.0353189.t001

Severe TBI.

Older age and higher rates of comorbidities, particularly HTN (p < 0.001) and CVA/CVD (p = 0.008), were significantly associated with regression. Regressing patients had lower GCS and GCS motor (GCSM) scores (both p < 0.001), a higher frequency of bilateral fixed pupils (31.3%, p < 0.001), and more IVH (p < 0.001) and SAH (p = 0.008). Regarding lab profiles, lower BS (p = 0.039), higher Hb (p = 0.034), and shorter PTT (p = 0.018) were observed among recovering patients. Rotterdam scores were also higher in this group (p < 0.001). Median ICU and hospital LOS were longest among regressors (21 and 33 days, respectively; both p < 0.001). Unfavorable GOSE0 was substantially more common in the regression and recovery groups than in the stable group (Table 2).

thumbnail
Download:
Table 2. Baseline clinical, radiologic, and laboratory characteristics of sTBI patients (GCS 3–8) stratified by six-month functional trajectory group.

https://doi.org/10.1371/journal.pone.0353189.t002

Predictors of functional trajectory

Univariate associations.

In both TBI severity groups, older age and lower GCS were strong predictors of regression. Patients aged ≥75 had markedly higher odds of regression (OR = 28.2 in nsTBI; OR = 6.5 in sTBI, both p < 0.01). HTN, DM, and CVA/CVD were also significantly associated with regression across both groups (all p < 0.05). In sTBI, bilateral fixed pupils significantly predicted regression (OR = 2.85, p < 0.001), but this was not significant in mild/moderate cases. IVH and SAH increased the odds of regression in both groups, while EDH was protective (OR = 0.20 in nsTBI, p < 0.001; OR = 0.52 in sTBI, p = 0.031). The same associations, but with smaller effect sizes, were found for recovery. Higher Rotterdam scores were additional radiologic predictors of extreme outcomes in both groups. Hospital and ICU LOS were positively associated with both recovery and regression, likely reflecting more complex clinical courses. Interestingly, in sTBI, recovery was linked to bilateral fixed pupils, SAH, and IVH. Among lab variables, low Hb predicted regression in both groups. Higher BS was only associated with regression in nsTBI, while prolonged PTT was significant in severe cases. Detailed information is brought to Supplementary Tables 1 and 2 in S1 File.

Independent predictors (multivariate associations).

Non-severe TBI.

In the adjusted model for nsTBI, older age remained a strong independent predictor of regression. Patients aged ≥75 had approximately 34 times higher odds of deterioration compared to the youngest group (p < 0.001). Lower GCS scores were independently associated with both recovery (OR = 0.837, p < 0.001) and regression (OR = 0.649, p < 0.001). EDH emerged as a significant protective factor against regression (OR = 0.248, p = 0.006), while SAH was associated with increased odds of both recovery (OR = 1.573, p = 0.002) and regression (OR = 1.842, p = 0.040). Higher platelet count was independently associated with regression (OR = 1.005, p = 0.001), and elevated BS levels predicted recovery (OR = 1.005, p < 0.001). In addition, higher Hb concentration was inversely associated with both non-stable outcome groups. Table 3 presents the full multivariate regression model for the mild to moderate TBI group. Hospital LOS was excluded from the final model to avoid post-treatment bias; however, the results of a parallel model including hospital LOS are provided in Supplementary Table 3 in S1 File. Additional sensitivity analyses using clinically selected covariates based on the CRASH and IMPACT prognostic models yielded broadly consistent findings (Supplementary Table 4 in S1 File). Recovery was independently associated with age ≥ 75 years (OR = 2.586, p < 0.001), SAH (OR = 1.581, p = 0.001), absent cisterns (OR = 2.668, p = 0.001), lower GCS scores (OR = 0.825, p < 0.001), higher admission BS (OR = 1.004, p < 0.001), and lower Hb levels (OR = 0.931, p = 0.020). Regression was independently associated with older age across all age categories, lower GCS scores (OR = 0.651, p < 0.001), lower Hb levels (OR = 0.863, p = 0.015), and absence of EDH, as EDH remained protective against regression (OR = 0.258, p = 0.005). SAH showed a borderline association with regression but did not reach statistical significance.

thumbnail
Download:
Table 3. Multivariate multinomial logistic regression of independent baseline predictors in mild to moderate TBI patients.

https://doi.org/10.1371/journal.pone.0353189.t003

Severe TBI.

In patients with sTBI, older age independently predicted functional decline. Patients aged ≥75 had nearly six-fold higher odds of regression compared to the youngest group (OR = 5.94, p = 0.019). Regression was also associated with lower GCS scores (OR = 0.795, p = 0.001) and the presence of bilateral fixed pupils (OR = 2.251, p = 0.005). IVH significantly predicted both regression (OR = 2.689, p = 0.001) and recovery (OR = 1.733, p = 0.009). Recovery was further associated with lower GCS scores (OR = 0.869, p = 0.002), absence of pneumocephalus (OR = 0.547, p = 0.008), and the presence of BSFx (OR = 1.461, p = 0.008). Elevated PTT levels were associated with both recovery and regression, with a stronger association observed for regression (OR = 1.030, p = 0.026 vs. OR = 1.022, p = 0.021). The presence of EDH demonstrated a protective effect against regression (OR = 0.512, p = 0.044). The full multivariate regression model for sTBI is presented in Table 4. Supplementary Table 5 in S1 File provides the same model with the inclusion of hospital LOS.

thumbnail
Download:
Table 4. Multivariate multinomial logistic regression of independent baseline predictors in sTBI patients.

https://doi.org/10.1371/journal.pone.0353189.t004

Sensitivity analyses using clinically selected CRASH/IMPACT-based covariates also supported the primary findings in sTBI patients (Supplementary Table 6 in S1 File). In this model, recovery was independently associated only with lower GCS scores (OR = 0.871, p = 0.002) and higher admission BS (OR = 1.002, p = 0.047). Regression was independently associated with older age, bilateral fixed pupils (OR = 2.239, p = 0.004), and lower GCS scores (OR = 0.791, p < 0.001). SAH (p = 0.051), EDH (p = 0.067), and absent cisterns (p = 0.092) showed borderline associations but did not reach statistical significance.

Alternative TBI severity grouping analysis.

When patients were reclassified as mild TBI (GCS 13–15) and moderate to severe TBI (GCS 3–12), the findings remained broadly consistent with the primary analyses. In the mild TBI subgroup, recovery was independently associated with age ≥ 75 years (OR = 3.365, p < 0.001), SAH (OR = 2.021, p < 0.001), absent cisterns (OR = 3.526, p = 0.004), and higher admission BS (OR = 1.004, p = 0.002). Regression was independently associated with older age across all age categories, with particularly large effect estimates in patients aged ≥75 years (OR = 108.256, p < 0.001). Pupil-related estimates for regression could not be calculated for anisocoria and bilateral fixed pupils because of sparse data (Supplementary Table 7 in S1 File).

In the moderate to severe TBI subgroup, recovery was independently associated with absent cisterns (OR = 1.591, p = 0.029), lower GCS scores, and higher admission BS. Regression was independently associated with older age, bilateral fixed pupils (OR = 2.102, p = 0.004), SAH (OR = 1.541, p = 0.030), lower GCS scores, and lower Hb levels. EDH remained protective against regression (OR = 0.459, p = 0.006) (see Supplementary Table 8 in S1 File). These findings support the robustness of the primary results while improving comparability with studies that group moderate and severe TBI together.

Predictive role of discharge functional status.

As shown in Table 5 and Supplementary Table 9 in S1 File, GOSE0 was a strong predictor of the six-month functional trajectory. In the adjusted multinomial regression model, patients with unfavorable discharge status (GOSE0 ≤ 4) had over 11-fold higher odds of substantial recovery (OR: 11.421, 95% CI: 9.032–14.441) and more than 17-fold higher odds of substantial regression (OR: 17.341, 95% CI: 11.310–26.587), compared to those who remained functionally stable.

thumbnail
Download:
Table 5. Multinomial logistic regression (unadjusted and adjusted) of six-month functional trajectory based on GOSE0.

https://doi.org/10.1371/journal.pone.0353189.t005

Fig 2 illustrates these transitions in detail: the heatmap (2A) highlights bidirectional movement, and the Sankey diagram (2B) visually demonstrates the considerable flow toward both improvement and decline from different discharge levels during the follow-up period.

thumbnail
Download:
Fig 2. Functional trajectory transitions from hospital discharge to 6-month follow-up among TBI patients (N = 2,948).

Panel A. Heatmap of GOSE transitions from hospital discharge (GOSE0) to six-month follow-up (FGOSE). Cell values represent patient counts; warmer colors indicate higher frequencies. Panel B. Sankey diagram showing transitions in GOSE from discharge (GOSE0) to six-month follow-up (FGOSE). Line thickness reflects patient volume per trajectory. Abbreviations: GOSE0, GOSE at discharge; FGOSE, GOSE at 6-month follow-up.

https://doi.org/10.1371/journal.pone.0353189.g002

Subgroup analysis for unexpected trajectory findings.

We next performed subgroup analyses of predictors identified as unexpected in the primary models (see Supplementary Table 10 in S1 File). Predictors were selected for this analysis if they demonstrated one of the following patterns in the main multinomial models: (i) two-sided associations with both recovery and regression, (ii) counterintuitive positive associations with recovery despite being pathological variables, or (iii) unexpected protective associations against regression. Patients were stratified by GOSE0 into poor, intermediate, and good groups, and univariate multinomial logistic regression was applied. This approach allowed us to evaluate whether the observed associations varied according to discharge status and followed patterns consistent with ceiling/floor effects and regression to the mean.

In mild to moderate TBI patients, age ≥ 75 followed this expected pattern: it was associated with both recovery and regression in the intermediate group, while regression also remained significant in the good group. SAH showed a similar bidirectional tendency in the intermediate group, with recovery significant and regression borderline. GCS aligned closely with expectations, being associated with recovery in the intermediate group and with regression in the good group. Hb was inversely associated with regression in the poor group only. Among predictors with illogical recovery effects, absent cisterns predicted recovery in the intermediate group. EDH, as a protective predictor, consistently reduced the odds of regression, especially in poor and intermediate discharge groups.

In sTBI patients, IVH strongly predicted regression in the good discharge group, aligning with ceiling effects, but did not influence outcomes in the poor or intermediate groups. BSFx was associated with recovery in the poor group, consistent with an illogical recovery predictor pattern. PTT showed a significant association with regression and a nonsignificant trend toward association with recovery (p = 0.066) in the poor group, reflecting regression to the mean effects. EDH again appeared protective against regression, this time only in the poor group. GCS did not demonstrate consistent effects across strata in this severity category.

Overall, these subgroup analyses showed that most unexpected predictors (age ≥ 75, SAH, GCS, IVH, Hb levels, and absent cisterns) followed patterns consistent with ceiling and floor effects, whereby patients closer to the extremes of GOSE0 had differential opportunities for recovery or regression. BSFx and EDH more closely reflected regression to the mean dynamics, with patients at extreme discharge states tending to normalize toward the middle over time.

Discussion

In the current registry-based longitudinal cohort study of TBI survivors, we investigated six-month functional trajectories, categorized as recovery, stability, or regression, based on changes in GOSE scores. We sought to explore the relationship between various early assessment variables and changes in the GOSE from discharge to six months post-discharge. While our analysis identified several clinical and radiologic variables associated with trajectory classification, a key and unanticipated finding was that some of the predictors were significantly associated with both recovery and regression.

Our findings reveal that as the age of TBI patients increases, the likelihood of experiencing regressing functional outcomes within six months after discharge also rises. However, an exception was observed among patients over 75 with nsTBI, where we found that aging was associated with both recovery and regression. In this group, nevertheless, the chance of regression significantly surpassed that of recovery. Previous studies have similarly highlighted that advancing age can negatively impact TBI outcomes, which can be attributed to factors such as reduced brain plasticity, the presence of extracranial comorbidities, compromised physiological compensatory mechanisms, and subpar rehabilitation results [29–32]. On the other hand, the increased volume of cerebrospinal fluid (CSF) in older adults may explain their higher chances of recovery when affected by nsTBI. The average CSF volume is about 265 mL in individuals in their 20s, while it increases to over 400 mL by the 70s [33]. This rise in CSF volume might offer a degree of protection against mild to moderate trauma, potentially contributing to better outcomes in older individuals. Overall, there is a general trend indicating worse outcomes for TBI with advancing age.

In our investigation, the presence of pneumocephalus and bilateral fixed pupils were linked to decline in functional outcomes in patients experiencing sTBI. This observation aligns with existing literature [34–37]. Traumatic pneumocephalus typically emerges in conjunction with open fractures at the base of the skull, along with damage to the sinuses or mastoid bone [38]. This condition can lead to the buildup of gas in various spaces within the cranial cavity, including the extradural, subdural, subarachnoid, intracerebral, or intraventricular areas [38]. As air accumulates, it can exert abnormal pressure on the brain, elevating intracranial pressure (ICP) and potentially causing neurological decline, a situation referred to as tension pneumocephalus [39]. Furthermore, fixed pupils are recognized as indicators of acute brainstem compression and escalating ICP and are classified as a “grave sign” in traumatic cases, particularly when accompanied by additional neurological manifestations [35,40]. Thus, bilateral fixed pupils and the presence of pneumocephalus in patients with sTBI can significantly hinder their recovery process.

The findings also suggest that a higher random BS level upon arrival in TBI patients is positively associated with improvements in the GOSE after discharge. Earlier studies have mostly linked both extremely high and extremely low BS levels to higher rates of surgical procedures, complications, and mortality [41,42]. This suggests a bell-shaped relationship between average BS levels and functional outcomes [43]. Research indicates that fluctuations in glucose levels during hospitalization exert a more significant influence on patients’ functional recovery than the glucose levels recorded at the time of admission [44]. We propose that within our study population, elevated admission BS levels prompted closer monitoring of patients. This careful oversight aimed to minimize fluctuations in their glucose levels, ultimately leading to improved functional outcomes for these individuals [44]. In addition, our study explored the changes in the GOSE following the discharge of patients. In other words, those with severely inadequate BS levels who passed away were not included in the study population. This situation, known as survivorship bias, may have played a role in the conflict of our results compared to other studies. Thus, maintaining tight control of BS levels in patients with TBI during their hospital stay can significantly improve their functional recovery trajectory upon discharge.

We found that in patients with nsTBI, an increased Hb level upon admission was correlated with both recovery and regression in functional outcomes post-discharge. However, previous studies have consistently shown that patients with TBI and concomitant anemia often experience worse functional outcomes [45,46]. High Hb levels significantly heighten the risks of thrombosis and vascular events in patients, thereby increasing the chances of complications in their overall recovery [47]. Additionally, patients who are dehydrated may present with falsely increased Hb levels upon arrival [48]. Therefore, while low Hb levels can hinder a patient’s functional recovery, it is crucial to understand that higher Hb levels on admission do not necessarily reflect better recovery prospects; instead, they might signal dehydration or a greater risk of future thrombotic events.

Our findings indicated that hypocoagulability arising from an increased PTT, is associated with both setbacks and improvements in the functional recovery trajectories of sTBI patients. In contrast, previous studies have consistently highlighted that hypocoagulability is related to poorer clinical outcomes for TBI patients [49–51]. Trauma-induced coagulopathy (TIC) can occur early or late, each with distinct characteristics [51]. In the initial hours following a TBI, patients often enter a hypocoagulable state characterized by platelet dysfunction, depletion of fibrinogen, and reduced thrombin generation [51]. This condition elevates the risk of persistent bleeding, contributing to 20–34% of trauma-related mortality [52]. On the other hand, late TIC, usually emerging after 24 hours, is marked by a hypercoagulable state that can cause excessive macro- and micro-clotting [51]. Our study suggests that elevated PTT levels upon arrival may mitigate the risk of delayed clotting complications in patients. As a result, having reduced clotting risks upon hospital admission may be advantageous for patients who manage to survive the early stages of TIC. At the same time, sTBI survivors who presented with relative hypocoagulability still faced an increased likelihood of unfavorable outcomes. It is noteworthy that the two-sided relationships of Hb and PTT with recovery and regression also might be attributed to statistical reasons, as elaborated below.

To improve comparability with prior studies, we also analyzed patients using a binary classification of mild versus moderate-severe TBI. This approach is commonly used in the literature and reflects differences in injury severity, clinical course, and long-term outcomes. The consistency of findings across both classification strategies supports the robustness of our results.

The bidirectional or unexpected relationship of the remaining significant predictors, however, challenges traditional assumptions about prognostic markers in TBI and suggests greater complexity in post-injury functional dynamics. These overlapping associations may, in part, be explained by statistical phenomena. First, patients discharged with intermediate GOSE scores (typically 3–5) occupy a “zone of clinical volatility,” where their condition may significantly improve or decline. Second, the ordinal nature of the GOSE scale introduces ceiling and floor effects, restricting upward or downward change for patients near score extremes [53,54]. Third, regression to the mean likely contributes to observed directional changes, especially in those with extreme or unstable early scores [55,56]. Collectively, these effects complicate the interpretation of such predictors and limit the ability to use early clinical variables, either at admission or discharge, as reliable indicators of long-term functional outcomes.

Importantly, this ambiguity is not merely a statistical artifact but also reflects real-world clinical experiences. Neurosurgeons and rehabilitation teams frequently face substantial uncertainty in post-discharge planning [57]. Decisions about follow-up intensity, therapy referrals, imaging, or secondary surgical interventions often rely on early indicators, such as neuroimaging, GCS, or GOSE0, that, as our data show, do not consistently point in a single prognostic direction. For example, IVH and SAH were associated with both improved and worsened trajectories in patients with sTBI and nsTBI, respectively, rendering such variables unreliable as standalone predictors.

GOSE0, although a strong predictor of subsequent trajectory, also demonstrated a paradox: patients with unfavorable scores (≤ 4) were at significantly increased odds of both recovery and regression. This likely reflects a wider potential range for change in this group, a ceiling/floor effect inherent to ordinal outcome scales such as the GOSE [53,54]. Additionally, some observed changes in GOSE may be attributable to regression to the mean, particularly among patients discharged with extreme scores. The statistical phenomenon, commonly seen in longitudinal studies of recovery, may lead to overestimation of true clinical change in trajectory classification [55,56]. This finding also supports the view that discharge status represents a dynamic state, not an outcome. Our results suggest that predicting functional recovery or regression based solely on early clinical data is inherently limited.

Accordingly, we categorized patients into three groups: poor, intermediate, and good, based on GOSE0, and the sensitivity analyses were conducted within these subgroups. As expected, most bidirectional or seemingly illogical results were observed exclusively in the intermediate GOSE0 group. For instance, among patients with nsTBI, age over 75 years, which showed a significant bidirectional association with both recovery and regression, was associated with greater recovery only in the intermediate GOSE0 group, whereas in the poor and good groups it was associated solely with regression. Similarly, in the same nsTBI group, the presence of SAH, which had a significant bidirectional association with both recovery and regression, was significantly associated with recovery only in the intermediate GOSE0 group. The same pattern was observed for absent cisterns, which unexpectedly showed an association with recovery; after stratification, this association remained significant only in the intermediate GOSE0 group. Regarding GCS, which also showed a bidirectional association, a higher GCS reduced the probability of recovery only in the intermediate GOSE0 group. After stratification, no further illogical significant results were observed for Hb. In addition, although EDH unexpectedly reduced the odds of regression in the initial analyses, this counterintuitive finding was seen only in the intermediate GOSE0 group. In patients with sTBI, variables such as IVH, GCS, and PTT, which had shown bidirectional or illogical associations, no longer demonstrated unexpected results after stratification. All of the above findings were explained and illustrated by the ceiling–floor effect. However, in patients with sTBI, two challenging findings remained: BSFx and EDH. Among patients with poor GOSE0, the presence of BSFx was associated with recovery, and the presence of EDH reduced the likelihood of regression. Given that these findings were observed only in patients with poor GOSE0 who were in a critical condition at discharge and considering that EDH and BSFx are severe pathologies and that these paradoxical results were seen exclusively in sTBI patients, it can be concluded that these findings can also be explained by regression to the mean.

Rather than viewing this uncertainty as a limitation, it should be recognized as a fundamental characteristic of TBI care [57]. A trajectory-based framework is more appropriate, one that emphasizes longitudinal monitoring, adaptability in care planning, and transparency in communication with families [58]. Structured post-discharge programs, serial GOSE assessments, and early neurorehabilitation strategies should be prioritized, particularly for patients with unstable or intermediate discharge profiles. Therefore, neurosurgeons should not provide definitive judgments about recovery or non-recovery to patients and their caregivers at the time of discharge based on admission, hospitalization, or even discharge characteristics. Specifically, they should not only avoid equating low GOSE0 with poor long-term outcomes but also recognize the high potential for meaningful recovery in this group, especially younger patients or those with reversible impairments. Conversely, favorable discharge scores should not preclude structured follow-up, as functional regression remains possible due to cognitive, psychological, or delayed physiological complications. Instead, neurosurgeons and rehabilitation teams must adopt a trajectory-oriented approach that accommodates uncertainty and focuses on ongoing post-discharge functional evaluation [58]. Likewise, caregivers should not blame neurosurgeons if their predictions about the course of recovery turn out to be incorrect or the opposite of what actually occurs, because there are complex bidirectional dynamics in the recovery process that cannot be reliably predicted using routine clinical features alone.

The primary aim of this study and presenting the results was not so much to provide solutions, but rather to realistically and statistically demonstrate the challenges, to highlight them, and to emphasize the need for more complex and comprehensive studies. Nevertheless, despite these challenges, some variables can still reliably predict recovery or regression, allowing neurosurgeons, when available, to communicate prognostic expectations to patients and caregivers with greater confidence. These include age 40–74 years, low admission BS, and high platelet count in patients with nsTBI; and age over 40 years, low admission BS, and the presence of pneumocephalus and bilateral fixed pupils in patients with sTBI. When the mentioned variables are present, patients should receive closer attention from the outset, with more frequent follow-up.

Our findings are consistent with and extend prior work from large prospective cohorts such as the TRACK-TBI initiative, which has demonstrated that recovery following TBI is highly heterogeneous and best understood through longitudinal trajectory-based approaches rather than single time-point outcomes. For example, studies by Nelson et al. and McCrea et al. have shown that patients exhibit diverse recovery patterns over time, with substantial variability even among individuals with similar baseline injury severity [59,60]. More recently, trajectory-based analyses (e.g., Curpen et al. [61]) have identified multiple distinct functional recovery pathways using GOSE over the first year post-injury, highlighting the dynamic and non-linear nature of recovery. While prior TRACK-TBI studies have primarily focused on longitudinal trajectories from the time of injury, our study offers a complementary perspective by specifically examining functional change from hospital discharge to 6 months, a clinically critical transition point. This focus allows for more targeted insights into post-discharge recovery and may be particularly relevant for discharge planning and early rehabilitation decision-making.

Although the proportion of excluded patients was relatively small, the possibility of residual selection bias cannot be fully excluded [62]. Future studies may benefit from applying multiple imputation techniques to further address this issue. An important consideration when interpreting our findings is the specific population studied. Our cohort included only patients who were admitted and survived to hospital discharge, excluding both individuals with mild injuries discharged directly from the emergency department and those who died during hospitalization. As such, this study reflects a selected subgroup of TBI patients with an intermediate risk profile, representing survivors with sufficient clinical stability to be discharged but still at risk for subsequent functional change. This selection has implications for the generalizability of our findings. The observed predictors and trajectory patterns may not apply to the full spectrum of TBI, particularly patients with very mild injuries or those with early mortality, who exhibit fundamentally different clinical courses. Therefore, our results should be interpreted within the context of this intermediate-risk population and may be most relevant for clinicians involved in discharge planning and post-acute care. At the same time, focusing on this subgroup provides clinically meaningful insights, as this is the population for whom prognostic uncertainty is often greatest at the time of discharge, and where trajectory-based prediction may have the highest practical value.

Conclusions

The present study underscores the inherent complexity and unpredictability of long-term functional outcomes in TBI survivors. Our findings revealed some reliable predictors for anticipating recovery or regression of patients with TBI at a six-month follow-up. However, several clinical variables showed an illogical or bidirectional association, which reinforces the limitations of relying on static, early-phase indicators, whether at admission or discharge, to determine prognosis. Our observation showed that these findings may be attributable to statistical phenomena, especially given the confounding influence of ceiling/floor effects and regression to the mean. From a clinical standpoint, our study highlights the need for caution in using early prognostic variables or even discharge scores (GOSE0) to guide decisions about rehabilitation intensity, follow-up frequency, or surgical planning. For researchers, these results call for advanced predictive models that integrate longitudinal data, repeated measures, and broader domains of recovery, including neurocognitive, emotional, and quality-of-life outcomes. Ultimately, embracing this complexity, rather than oversimplifying it, will be essential for advancing both the science and the practice of TBI care.

Supporting information

S1 File. Supplementary tables for multinomial logistic regression analyses of six-month functional trajectories following traumatic brain injury.

This file contains 10 supplementary tables presenting the results of univariate and multivariate multinomial logistic regression analyses evaluating baseline predictors of six-month functional trajectories across patients with mild, moderate, and severe traumatic brain injury (TBI). The tables include primary analyses, alternative multivariable models incorporating hospital length of stay or selecting predictors based on the CRASH and IMPACT prognostic models, sensitivity analyses in mild and moderate-to-severe TBI subgroups, and additional analyses examining functional trajectory according to Glasgow Outcome Scale-Extended (GOSE) at discharge, including subgroup analyses of unexpected predictors. Odds ratios (ORs), 95% confidence intervals (CIs), and p-values are reported, with statistically significant results (p < 0.05) shown in bold. Abbreviations are defined within the individual tables.

https://doi.org/10.1371/journal.pone.0353189.s001

(PDF)

Acknowledgments

We appreciate the Trauma Research Center of Shiraz University of Medical Sciences for their cooperation in providing registry data.

References

  1. 1. Maas AIR, Menon DK, Adelson PD, Andelic N, Bell MJ, Belli A, et al. Traumatic brain injury: integrated approaches to improve prevention, clinical care, and research. Lancet Neurol. 2017;16(12):987–1048. pmid:29122524
  2. 2. Shames J, Treger I, Ring H, Giaquinto S. Return to work following traumatic brain injury: trends and challenges. Disabil Rehabil. 2007;29(17):1387–95. pmid:17729084
  3. 3. Maas AIR, Menon DK, Manley GT, Abrams M, Åkerlund C, Andelic N, et al. Traumatic brain injury: progress and challenges in prevention, clinical care, and research. Lancet Neurol. 2022;21(11):1004–60. pmid:36183712
  4. 4. Lafta G, Sbahi H. Factors associated with the severity of traumatic brain injury. Med Pharm Rep. 2023;96(1):58–64. pmid:36818327
  5. 5. Manley GT, Dams-O’Connor K, Alosco ML, Awwad HO, Bazarian JJ, Bragge P, et al. A new characterisation of acute traumatic brain injury: the NIH-NINDS TBI Classification and Nomenclature Initiative. Lancet Neurol. 2025;24(6):512–23. pmid:40409315
  6. 6. Maegele M, Lefering R, Sakowitz O, Kopp MA, Schwab JM, Steudel WI, et al. The incidence and management of moderate to severe head injury: a retrospective analysis of data from the trauma register of the German trauma society. Deutsches Ärzteblatt International. 2019;116(10):167.
  7. 7. Ruet A, Bayen E, Jourdan C, Ghout I, Meaude L, Lalanne A, et al. A Detailed Overview of Long-Term Outcomes in Severe Traumatic Brain Injury Eight Years Post-injury. Front Neurol. 2019;10:120. pmid:30846966
  8. 8. Khalili H, Sadraei N, Niakan A, Ghaffarpasand F, Sadraei A. Role of Intracranial Pressure Monitoring in Management of Patients with Severe Traumatic Brain Injury: Results of a Large Level I Trauma Center in Southern Iran. World Neurosurg. 2016;94:120–5. pmid:27392896
  9. 9. Chan A, Ouyang J, Nguyen K, Jones A, Basso S, Karasik R. Traumatic brain injuries: a neuropsychological review. Front Behav Neurosci. 2024;18:1326115. pmid:39444788
  10. 10. Page TA, Gordon S, Balchin R, Tomlinson M. Caregivers’ perspectives of the challenges faced with survivors of traumatic brain injury: A scoping review. NeuroRehabilitation. 2021;49(3):349–62. pmid:34308914
  11. 11. Emery CA, Barlow KM, Brooks BL, Max JE, Villavicencio-Requis A, Gnanakumar V, et al. A Systematic Review of Psychiatric, Psychological, and Behavioural Outcomes following Mild Traumatic Brain Injury in Children and Adolescents. Can J Psychiatry. 2016;61(5):259–69. pmid:27254800
  12. 12. Hafizi-Rastani I, Khalili H, Paydar S, Pourahmad S. Identifying Important Attributes for Prognostic Prediction in Traumatic Brain Injury Patients. Methods Inf Med. 2016;55(05):440–9.
  13. 13. Bark D, Boman M, Depreitere B, Wright DW, Lewén A, Enblad P, et al. Refining outcome prediction after traumatic brain injury with machine learning algorithms. Sci Rep. 2024;14(1):8036. pmid:38580767
  14. 14. Dijkland SA, Helmrich IRAR, Nieboer D, van der Jagt M, Dippel DWJ, Menon DK, et al. Outcome Prediction after Moderate and Severe Traumatic Brain Injury: External Validation of Two Established Prognostic Models in 1742 European Patients. J Neurotrauma. 2021;38(10):1377–88. pmid:33161840
  15. 15. MRC CRASH Trial Collaborators, Perel P, Arango M, Clayton T, Edwards P, Komolafe E, et al. Predicting outcome after traumatic brain injury: practical prognostic models based on large cohort of international patients. BMJ. 2008;336(7641):425–9. pmid:18270239
  16. 16. Nissen SE, Tuzcu EM, Schoenhagen P, Crowe T, Sasiela WJ, Tsai J, et al. Statin therapy, LDL cholesterol, C-reactive protein, and coronary artery disease. N Engl J Med. 2005;352(1):29–38. pmid:15635110
  17. 17. Bruschetta R, Tartarisco G, Lucca LF, Leto E, Ursino M, Tonin P, et al. Predicting Outcome of Traumatic Brain Injury: Is Machine Learning the Best Way?. Biomedicines. 2022;10(3):686. pmid:35327488
  18. 18. Zuckerman DA, Giacino JT, Bodien YG. Traumatic Brain Injury: What Is a Favorable Outcome?. J Neurotrauma. 2022;39(13–14):1010–2. pmid:34861770
  19. 19. Bazmi S, Zoghi S, Mohammadi Z, Ansari A, Shahriari S, Niakan A, et al. Abdominal and pelvic surgery in traumatic brain injury patients; a registry-based retrospective analysis. BMC Surg. 2025;25(1):250. pmid:40484927
  20. 20. Curpen PP, To XV, Lu M, Winter C, Bellapart J, Newcombe VF, et al. Trajectories of Glasgow Outcome Scale-Extended after traumatic brain injury: an analysis of the TRACK-TBI cohort. J Neurol Neurosurg Psychiatry. 2026;97(3):239–45. pmid:41161723
  21. 21. Zoghi S, Ansari A, Niakan A, Taheri R, Khalili H. Post-discharge 6-Month Functional Recovery of Traumatic Brain Injury Survivors with Unfavorable Functional Status at Discharge: A Registry-Based Cohort Study. World Neurosurg. 2024;189:e580–90. pmid:38936616
  22. 22. Booker J, Sinha S, Choudhari K, Dawson J, Singh R. Predicting functional recovery after mild traumatic brain injury: the SHEFBIT cohort. Brain Inj. 2019;33(9):1158–64. pmid:31215812
  23. 23. Sigurdardottir S, Andelic N, Roe C, Schanke A-K. Cognitive recovery and predictors of functional outcome 1 year after traumatic brain injury. J Int Neuropsychol Soc. 2009;15(5):740–50. pmid:19602303
  24. 24. Slovis JC, Gupta N, Li NY, Kernie SG, Miles DK. Assessment of Recovery Following Pediatric Traumatic Brain Injury. Pediatr Crit Care Med. 2018;19(4):353–60. pmid:29419604
  25. 25. Wilson JT, Pettigrew LE, Teasdale GM. Structured interviews for the Glasgow Outcome Scale and the extended Glasgow Outcome Scale: guidelines for their use. J Neurotrauma. 1998;15(8):573–85. pmid:9726257
  26. 26. Zoghi S, Ansari A, Niakan A, Taheri R, Khalili H. Post-discharge 6-Month Functional Recovery of Traumatic Brain Injury Survivors with Unfavorable Functional Status at Discharge: A Registry-Based Cohort Study. World Neurosurg. 2024;189:e580–90. pmid:38936616
  27. 27. Chowdhury MZI, Turin TC. Variable selection strategies and its importance in clinical prediction modelling. Fam Med Community Health. 2020;8(1):e000262. pmid:32148735
  28. 28. Bursac Z, Gauss CH, Williams DK, Hosmer DW. Purposeful selection of variables in logistic regression. Source Code Biol Med. 2008;3:17. pmid:19087314
  29. 29. Borazjani R, Ajdari MR, Niakan A, Yousefi O, Amoozandeh A, Sayadi M, et al. Current Status and Outcomes of Critical Traumatic Brain Injury (GCS = 3-5) in a Developing Country: A Retrospective, Registry-Based Study. World J Surg. 2022;46(10):2335–43. pmid:35789431
  30. 30. Pauwels L, Chalavi S, Swinnen SP. Aging and brain plasticity. Aging (Albany NY). 2018;10(8):1789–90. pmid:30067225
  31. 31. Kulesza B, Nogalski A, Kulesza T, Prystupa A. Prognostic factors in traumatic brain injury and their association with outcome. Journal of Pre-Clinical and Clinical Research. 2015;9(2).
  32. 32. Winter L, Mensinger JL, Moriarty HJ, Robinson KM, McKay M, Leiby BE. Age Moderates the Effect of Injury Severity on Functional Trajectories in Traumatic Brain Injury: A Study Using the NIDILRR Traumatic Brain Injury Model Systems National Dataset. J Clin Med. 2022;11(9):2477. pmid:35566607
  33. 33. Yamada S, Otani T, Ii S, Kawano H, Nozaki K, Wada S, et al. Aging-related volume changes in the brain and cerebrospinal fluid using artificial intelligence-automated segmentation. Eur Radiol. 2023;33(10):7099–112. pmid:37060450
  34. 34. Hoffmann M, Lefering R, Rueger JM, Kolb JP, Izbicki JR, Ruecker AH, et al. Pupil evaluation in addition to Glasgow Coma Scale components in prediction of traumatic brain injury and mortality. Br J Surg. 2012;99 Suppl 1:122–30. pmid:22441866
  35. 35. Luz Teixeira T, Peluso L, Banco P, Njimi H, Abi-Khalil L, Chanchay Pillajo M, et al. Early Pupillometry Assessment in Traumatic Brain Injury Patients: A Retrospective Study. Brain Sci. 2021;11(12):1657. pmid:34942959
  36. 36. Cunqueiro A, Scheinfeld MH. Causes of pneumocephalus and when to be concerned about it. Emerg Radiol. 2018;25(4):331–40. pmid:29546674
  37. 37. Schirmer CM, Heilman CB, Bhardwaj A. Pneumocephalus: case illustrations and review. Neurocritical Care. 2010;13(1):152–8.
  38. 38. Fan ZK, Zhang ZR, Shen YY, Yi RQ, Wang LC. Traumatic tension pneumocephalus: A case report. World Journal of Emergency Medicine. 2022;13(3):251.
  39. 39. Prichard RG. Tension Pneumocephalus. R I Med J (2013). 2021;104(5):47–8. pmid:34044439
  40. 40. Scotter J, Hendrickson S, Marcus HJ, Wilson MH. Prognosis of patients with bilateral fixed dilated pupils secondary to traumatic extradural or subdural haematoma who undergo surgery: a systematic review and meta-analysis. Emerg Med J. 2015;32(8):654–9. pmid:25385844
  41. 41. Wahl WL, Taddonio M, Maggio PM, Arbabi S, Hemmila MR. Mean glucose values predict trauma patient mortality. J Trauma. 2008;65(1):42–7; discussion 47-8. pmid:18580507
  42. 42. Alvis-Miranda HR, Navas-Marrugo SZ, Velasquez-Loperena RA, Adie-Villafañe RJ, Velasquez-Loperena D, Castellar-Leones SM, et al. Effects of Glycemic Level on Outcome of Patients with Traumatic Brain Injury: A Retrospective Cohort Study. Bull Emerg Trauma. 2014;2(2):65–71. pmid:27162868
  43. 43. Yuan T, He H, Liu Y, Wang J, Kang X, Fu G, et al. Association between blood glucose levels and Glasgow Outcome Score in patients with traumatic brain injury: secondary analysis of a randomized trial. Trials. 2022;23(1):38. pmid:35033158
  44. 44. Matsushima K, Peng M, Velasco C, Schaefer E, Diaz-Arrastia R, Frankel H. Glucose variability negatively impacts long-term functional outcome in patients with traumatic brain injury. J Crit Care. 2012;27(2):125–31. pmid:22033047
  45. 45. Guglielmi A, Graziano F, Bogossian EG, Turgeon AF, Taccone FS, Citerio G, et al. Haemoglobin values, transfusion practices, and long-term outcomes in critically ill patients with traumatic brain injury: a secondary analysis of CENTER-TBI. Crit Care. 2024;28(1):199. pmid:38877571
  46. 46. Carr KR, Rodriguez M, Ottesen A, Michalek J, Son C, Patel V, et al. Association Between Relative Anemia and Early Functional Recovery After Severe Traumatic Brain Injury (TBI). Neurocrit Care. 2016;25(2):185–92. pmid:27256295
  47. 47. Hultcrantz M, Modlitba A, Vasan SK, Sjölander A, Rostgaard K, Landgren O, et al. Hemoglobin concentration and risk of arterial and venous thrombosis in 1.5 million Swedish and Danish blood donors. Thromb Res. 2020;186:86–92. pmid:31927271
  48. 48. Ali MA, Abu Damir H, Adem MA, Ali OM, Amir N, Shah AAM, et al. Effects of long-term dehydration on stress markers, blood parameters, and tissue morphology in the dromedary camel (Camelus dromedarius). Front Vet Sci. 2023;10:1236425. pmid:38116506
  49. 49. Xu D-X, Du W-T, Li X, Wu Z-X, Yu G-F. D-dimer/fibrinogen ratio for the prediction of progressive hemorrhagic injury after traumatic brain injury. Clin Chim Acta. 2020;507:143–8. pmid:32333859
  50. 50. Riojas CM, Ekaney ML, Ross SW, Cunningham KW, Furay EJ, Brown CVR, et al. Platelet Dysfunction after Traumatic Brain Injury: A Review. J Neurotrauma. 2021;38(7):819–29. pmid:33143502
  51. 51. Moore EE, Moore HB, Kornblith LZ, Neal MD, Hoffman M, Mutch NJ, et al. Trauma-induced coagulopathy. Nat Rev Dis Primers. 2021;7(1):30. pmid:33927200
  52. 52. Callcut RA, Kornblith LZ, Conroy AS, Robles AJ, Meizoso JP, Namias N, et al. The why and how our trauma patients die: A prospective Multicenter Western Trauma Association study. J Trauma Acute Care Surg. 2019;86(5):864–70. pmid:30633095
  53. 53. Hudak AM, Caesar RR, Frol AB, Krueger K, Harper CR, Temkin NR, et al. Functional outcome scales in traumatic brain injury: a comparison of the Glasgow Outcome Scale (Extended) and the Functional Status Examination. J Neurotrauma. 2005;22(11):1319–26. pmid:16305320
  54. 54. Wang L, Zhang Z, McArdle JJ, Salthouse TA. Investigating Ceiling Effects in Longitudinal Data Analysis. Multivariate Behav Res. 2009;43(3):476–96. pmid:19924269
  55. 55. Cochrane KM, Williams BA, Fischer JAJ, Samson KLI, Pei LX, Karakochuk CD. Regression to the Mean: A Statistical Phenomenon of Worthy Consideration in Anemia Research. Curr Dev Nutr. 2020;4(10):nzaa152. pmid:33154991
  56. 56. Pretz CR, Dams-O’Connor K. Longitudinal description of the glasgow outcome scale-extended for individuals in the traumatic brain injury model systems national database: a National Institute on Disability and Rehabilitation Research traumatic brain injury model systems study. Arch Phys Med Rehabil. 2013;94(12):2486–93. pmid:23838239
  57. 57. Jones K, Quinn T, Mazor KM, Muehlschlegel S. Prognostic Uncertainty in Critically Ill Patients with Traumatic Brain Injury: A Multicenter Qualitative Study. Neurocrit Care. 2021;35(2):311–21. pmid:34080083
  58. 58. Dijkland SA, Foks KA, Polinder S, Dippel DWJ, Maas AIR, Lingsma HF, et al. Prognosis in Moderate and Severe Traumatic Brain Injury: A Systematic Review of Contemporary Models and Validation Studies. J Neurotrauma. 2020;37(1):1–13. pmid:31099301
  59. 59. Nelson LD, Temkin NR, Barber J, Brett BL, Okonkwo DO, McCrea MA. Functional recovery, symptoms, and quality of life 1 to 5 years after traumatic brain injury. JAMA Netw Open. 2023;6(3):e233660.
  60. 60. McCrea MA, Giacino JT, Barber J, Temkin NR, Nelson LD, Levin HS, et al. Functional Outcomes Over the First Year After Moderate to Severe Traumatic Brain Injury in the Prospective, Longitudinal TRACK-TBI Study. JAMA Neurol. 2021;78(8):982–92. pmid:34228047
  61. 61. Curpen PP, To XV, Lu M, Winter C, Bellapart J, Newcombe VF, et al. Trajectories of Glasgow Outcome Scale-Extended after traumatic brain injury: an analysis of the TRACK-TBI cohort. J Neurol Neurosurg Psychiatry. 2026;97(3):239–45. pmid:41161723
  62. 62. Richter S, Stevenson S, Newman T, Wilson L, Menon DK, Maas AIR, et al. Handling of Missing Outcome Data in Traumatic Brain Injury Research: A Systematic Review. J Neurotrauma. 2019;36(19):2743–52. pmid:31062649