Evidence synthesis on sex differences in mortality after traumatic brain injury: Systematic review integrating meta-analysis and explainable machine learning
- 1University Health Network
Abstract
Sex differences in mortality after traumatic brain injury (TBI) have been inconsistently reported, even after accounting for age, injury severity, and other injury related factors.
Understanding the sources of this heterogeneity is important for prognostic research and the development of data-driven clinical decision-support tools. We conducted a systematic review and meta-analysis of sex-specific mortality after adult TBI and used explainable machine learning (ML) to investigate potential modifiers of between-study effect estimates. We searched eight databases from inception to August 25, 2025, for studies reporting mortality estimates adjusted for at least age and injury severity. Risk of bias was assessed using the Quality in Prognosis Studies tool. Male-to-female mortality estimates were pooled using random-effects meta-analysis. Three explainable ML algorithms were used to identify important modifiers of study-level effect estimates. Among 9,418 records identified, 21 studies comprising 899,813 participants and 98,286 deaths met the inclusion criteria. Pooled analyses showed lower mortality among females than males (OR 0.88, 95% CI 0.80-0.98; HR 0.76, 95% CI 0.59-0.98). Lower mortality among females was also observed for in-hospital mortality (OR 0.86, 95% CI 0.76-0.97) and overall or unspecified mortality (OR 0.83, 95% CI 0.78-0.88). Substantial residual heterogeneity remained across analyses. Explainable ML identified mean age, study sex composition, and study quality as important modifiers of reported effect estimates. These findings were consistent across sensitivity analyses. Female sex was associated with lower mortality after adult TBI, although substantial between-study heterogeneity remained.
Explainable ML identified demographic and methodological factors that may contribute to variation in reported associations. These findings highlight the importance of population characteristics and study context when developing, validating, and interpreting explainable digital prognostic tools for TBI. PROSPERO: CRD420251252136.
Author SummaryExisting evidence syntheses on sex differences in mortality after traumatic brain injury (TBI) remains inconclusive, with some systematic reviews reporting female sex being protective to mortality, while others showing no differences. This raised a question on whether social or study methodology factors may be responsible for this heterogeneity.
To answer the question, we conducted a systematic review and integrated meta-analysis with explainable machine learning (ML). We extracted and organized data concerning social and cohort-related factors using the PROGRESS-Plus framework from 21 studies. We conducted a meta-analysis of the pooled effects of sex differences on mortality outcomes and explored between-study heterogeneity, using three ML algorithms to identify important modifiers.
Our meta-analysis confirmed substantial heterogeneity in the association between sex and mortality, with important variation in both the magnitude and direction of reported effects.
Explainable ML identified study-level age, sex composition, and methodological quality as the major drivers of heterogeneity, with consistent findings across several sensitivity analyses.
Our findings have implications for future digital health research. Explainable ML may be a useful tool for interpreting variability in existing evidence and informing the design and interpretation of future studies examining sex differences in TBI outcomes.
Related
- A machine learning-driven and explainable prediction model for in-hospital mortality in patients with diabetes mellitus and congestive heart failure: integrating XGBoost and SHAP approaches.
- Artificial intelligence-enhanced electrocardiography for predicting mortality and major adverse cardiovascular events: A systematic review and narrative synthesis
- Negotiating Mortality: A Qualitative Systematic Review and Thematic Synthesis of the Psychosocial Experiences of Pancreatic Cancer Across the Illness Trajectory.
- Explainable machine learning and deep learning models for predicting in-hospital mortality in burn patients
- An explainable artificial intelligence framework using electrocardiograms to predict mortality and major adverse cardiovascular events
BibTeX
@misc{phan2026evidence,
title = {{Evidence synthesis on sex differences in mortality after traumatic brain injury: Systematic review integrating meta-analysis and explainable machine learning}},
author = {D. Phan and Y. Yao and U. Shaikh and N. Ahmad and T. Mollayeva},
howpublished = {medRxiv},
year = {2026},
month = oct,
doi = {10.64898/2026.10.02.26364454},
url = {https://www.medrxiv.org/content/10.64898/2026.10.02.26364454v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - Evidence synthesis on sex differences in mortality after traumatic brain injury: Systematic review integrating meta-analysis and explainable machine learning
AU - D. Phan
AU - Y. Yao
AU - U. Shaikh
AU - N. Ahmad
AU - T. Mollayeva
PY - 2026
DA - 2026/10/02
PB - medRxiv
DO - 10.64898/2026.10.02.26364454
UR - https://www.medrxiv.org/content/10.64898/2026.10.02.26364454v1
AB - Sex differences in mortality after traumatic brain injury (TBI) have been inconsistently reported, even after accounting for age, injury severity, and other injury related factors. Understanding the sources of this heterogeneity is important for prognostic research and the development of data-driven clinical decision-support tools. We conducted a systematic review and meta-analysis of sex-specific mortality after adult TBI and used explainable machine learning (ML) to investigate potential modifiers of between-study effect estimates. We searched eight databases from inception to August 25, 2025, for studies reporting mortality estimates adjusted for at least age and injury severity. Risk of bias was assessed using the Quality in Prognosis Studies tool. Male-to-female mortality estimates were pooled using random-effects meta-analysis. Three explainable ML algorithms were used to identify important modifiers of study-level effect estimates. Among 9,418 records identified, 21 studies comprising 899,813 participants and 98,286 deaths met the inclusion criteria. Pooled analyses showed lower mortality among females than males (OR 0.88, 95% CI 0.80-0.98; HR 0.76, 95% CI 0.59-0.98). Lower mortality among females was also observed for in-hospital mortality (OR 0.86, 95% CI 0.76-0.97) and overall or unspecified mortality (OR 0.83, 95% CI 0.78-0.88). Substantial residual heterogeneity remained across analyses. Explainable ML identified mean age, study sex composition, and study quality as important modifiers of reported effect estimates. These findings were consistent across sensitivity analyses. Female sex was associated with lower mortality after adult TBI, although substantial between-study heterogeneity remained. Explainable ML identified demographic and methodological factors that may contribute to variation in reported associations. These findings highlight the importance of population characteristics and study context when developing, validating, and interpreting explainable digital prognostic tools for TBI. PROSPERO: CRD420251252136. Author SummaryExisting evidence syntheses on sex differences in mortality after traumatic brain injury (TBI) remains inconclusive, with some systematic reviews reporting female sex being protective to mortality, while others showing no differences. This raised a question on whether social or study methodology factors may be responsible for this heterogeneity. To answer the question, we conducted a systematic review and integrated meta-analysis with explainable machine learning (ML). We extracted and organized data concerning social and cohort-related factors using the PROGRESS-Plus framework from 21 studies. We conducted a meta-analysis of the pooled effects of sex differences on mortality outcomes and explored between-study heterogeneity, using three ML algorithms to identify important modifiers. Our meta-analysis confirmed substantial heterogeneity in the association between sex and mortality, with important variation in both the magnitude and direction of reported effects. Explainable ML identified study-level age, sex composition, and methodological quality as the major drivers of heterogeneity, with consistent findings across several sensitivity analyses. Our findings have implications for future digital health research. Explainable ML may be a useful tool for interpreting variability in existing evidence and informing the design and interpretation of future studies examining sex differences in TBI outcomes.
ER -