{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,4,23]],"date-time":"2026-04-23T04:47:10Z","timestamp":1776919630181,"version":"3.51.2"},"reference-count":40,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2026,4,23]],"date-time":"2026-04-23T00:00:00Z","timestamp":1776902400000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["frontiersin.org"],"crossmark-restriction":true},"short-container-title":["Front. Neurol."],"abstract":"<jats:sec>\n                    <jats:title>Background<\/jats:title>\n                    <jats:p>TBI is associated with high ICU mortality, yet traditional prognostic scores often lack accuracy due to linear assumptions. This study aimed to develop an interpretable machine learning model to predict in-hospital mortality in TBI patients, combining high predictive performance with clinical transparency.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Methods<\/jats:title>\n                    <jats:p>\n                      This retrospective analysis utilized TBI clinical records (2008\u20132019) retrieved from the MIMIC-IV database. We collected comprehensive baseline data including demographics, comorbidities, vital signs, laboratory parameters, disease severity scores, and therapeutic interventions. To identify the most robust predictors, we employed a rigorous intersectional feature selection strategy combining Univariate Logistic Regression (\n                      <jats:italic>p<\/jats:italic>\n                      \u202f&amp;lt;\u202f0.05), LASSO regression, and the Boruta algorithm. Seven supervised ML algorithms (Logistic Regression, Decision Tree, Random Forest, XGBoost, LightGBM, SVM, and ANN) were developed and compared. Predictive performance was benchmarked using AUROC, Brier score, and DCA. Additionally, SHAP were implemented to enhance the interpretability of the final model\u2019s decision-making process.\n                    <\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Results<\/jats:title>\n                    <jats:p>Among the 2,691 included TBI patients, 1,716 (63.8%) were male, and the median age was 66 (IQR 48\u201381) years. The primary outcome, in-hospital mortality, occurred in 316 patients (11.7%). A final predictive set of 12 variables was identified: demographic and clinical metrics (age, GCS, temperature), lab results (anion gap, glucose, PT, RDW, WBC, urea nitrogen), and key clinical interventions or conditions (acidosis, mannitol, and sedatives). Among the seven algorithms, the XGBoost model achieved the best performance with the highest AUROC of 0.873 (95% CI: 0.842\u20130.906) and superior calibration (Brier score\u202f=\u202f0.078). SHAP analysis identified GCS, age, anion gap, and glucose as the top mortality drivers, and revealed critical non-linear relationships, such as a U-shaped association between body temperature and mortality risk.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Conclusion<\/jats:title>\n                    <jats:p>We successfully developed and validated an interpretable XGBoost-based model that demonstrates robust discriminative capacity, excellent calibration, and high sensitivity for predicting in-hospital mortality in ICU patients with TBI. By integrating SHAP analysis, the model provides intuitive explanations for both population-level risk drivers and individual predictions, effectively bridging the gap between predictive accuracy and clinical interpretability to facilitate personalized decision-making in the ICU.<\/jats:p>\n                  <\/jats:sec>","DOI":"10.3389\/fneur.2026.1815307","type":"journal-article","created":{"date-parts":[[2026,4,23]],"date-time":"2026-04-23T04:21:19Z","timestamp":1776918079000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":0,"title":["Interpretable machine learning for in-hospital mortality prediction in ICU patients with traumatic brain injury"],"prefix":"10.3389","volume":"17","author":[{"given":"Ning","family":"Liu","sequence":"first","affiliation":[{"name":"Affiliated Taihe Hospital of Wannan Medical College","place":["Fuyang, China"]}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Congcong","family":"Li","sequence":"additional","affiliation":[{"name":"Affiliated Taihe Hospital of Wannan Medical College","place":["Fuyang, China"]}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"1965","published-online":{"date-parts":[[2026,4,23]]},"reference":[{"key":"ref1","doi-asserted-by":"publisher","first-page":"1080","DOI":"10.3171\/2017.10.JNS17352","article-title":"Estimating the global incidence of traumatic brain injury","volume":"130","author":"Dewan","year":"2018","journal-title":"J Neurosurg"},{"key":"ref2","doi-asserted-by":"publisher","first-page":"987","DOI":"10.1016\/s1474-4422(17)30371-x","article-title":"Traumatic brain injury: integrated approaches to improve prevention, clinical care, and research","volume":"16","author":"Maas","year":"2017","journal-title":"Lancet Neurol"},{"key":"ref3","doi-asserted-by":"publisher","first-page":"1088","DOI":"10.1016\/S0140-6736(12)60864-2","article-title":"Early management of severe traumatic brain injury","volume":"380","author":"Rosenfeld","year":"2012","journal-title":"Lancet"},{"key":"ref4","doi-asserted-by":"publisher","first-page":"986","DOI":"10.1016\/j.apmr.2013.10.032","article-title":"Incidence, prevalence, costs, and impact on disability of common conditions requiring rehabilitation in the United States: stroke, spinal cord injury, traumatic brain injury, multiple sclerosis, osteoarthritis, rheumatoid arthritis, limb loss, and back pain","volume":"95","author":"Ma","year":"2014","journal-title":"Arch Phys Med Rehabil"},{"key":"ref5","doi-asserted-by":"publisher","first-page":"213","DOI":"10.1016\/j.mcna.2019.11.001","article-title":"Traumatic brain injury: an overview of epidemiology, pathophysiology, and medical management","volume":"104","author":"Capizzi","year":"2020","journal-title":"Med Clin North Am"},{"key":"ref6","doi-asserted-by":"publisher","first-page":"e165","DOI":"10.1371\/journal.pmed.0050165","article-title":"Predicting outcome after traumatic brain injury: development and international validation of prognostic scores based on admission characteristics","volume":"5","author":"Steyerberg","year":"2008","journal-title":"PLoS Med"},{"key":"ref7","doi-asserted-by":"publisher","first-page":"1581","DOI":"10.1503\/cmaj.101786","article-title":"Mortality associated with withdrawal of life-sustaining therapy for patients with severe traumatic brain injury: a Canadian multicentre cohort study","volume":"183","author":"Turgeon","year":"2011","journal-title":"CMAJ"},{"key":"ref8","doi-asserted-by":"publisher","first-page":"417","DOI":"10.3171\/2023.11.JNS231425","article-title":"Performance of the IMPACT and CRASH prognostic models for traumatic brain injury in a contemporary multicenter cohort: a TRACK-TBI study","volume":"141","author":"Yue","year":"2024","journal-title":"J Neurosurg"},{"key":"ref9","doi-asserted-by":"publisher","first-page":"425","DOI":"10.1136\/bmj.39461.643438.25","article-title":"Predicting outcome after traumatic brain injury: practical prognostic models based on large cohort of international patients","volume":"336","year":"2008","journal-title":"BMJ"},{"key":"ref10","doi-asserted-by":"publisher","first-page":"749","DOI":"10.1038\/s41551-018-0304-0","article-title":"Explainable machine-learning predictions for the prevention of hypoxaemia during surgery","volume":"2","author":"Lundberg","year":"2018","journal-title":"Nat Biomed Engin"},{"key":"ref11","first-page":"359","article-title":"What clinicians want: contextualizing explainable machine learning for clinical end use","volume-title":"Machine Learning for Healthcare","author":"Tonekaboni","year":"2019"},{"key":"ref12","doi-asserted-by":"publisher","first-page":"81","DOI":"10.1016\/S0140-6736(74)91639-0","article-title":"Assessment of coma and impaired consciousness","volume":"2","author":"Teasdale","year":"1974","journal-title":"Pract Scale Lancet"},{"key":"ref13","doi-asserted-by":"publisher","first-page":"267","DOI":"10.1111\/j.2517-6161.1996.tb02080.x","article-title":"Regression shrinkage and selection via the lasso","volume":"58","author":"Tibshirani","year":"1996","journal-title":"J R Stat Soc Series B Stat Methodol"},{"key":"ref14","doi-asserted-by":"publisher","first-page":"1","DOI":"10.18637\/jss.v036.i11","article-title":"Feature selection with the Boruta package","volume":"36","author":"Kursa","year":"2010","journal-title":"J Stat Softw"},{"key":"ref15","doi-asserted-by":"publisher","first-page":"56","DOI":"10.1038\/s42256-019-0138-9","article-title":"From local explanations to global understanding with explainable AI for trees","volume":"2","author":"Lundberg","year":"2020","journal-title":"Nat Machine Intel"},{"key":"ref16","doi-asserted-by":"publisher","first-page":"206","DOI":"10.1038\/s42256-019-0048-x","article-title":"Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead","volume":"1","author":"Rudin","year":"2019","journal-title":"Nat Machine Intel"},{"key":"ref17","first-page":"4765","article-title":"Unified approach to interpreting model predictions","volume":"30","author":"Lundberg","year":"2017","journal-title":"Adv Neural Inf Process Syst"},{"key":"ref18","doi-asserted-by":"publisher","first-page":"270","DOI":"10.1089\/neu.2006.0029","article-title":"Prognostic value of the Glasgow coma scale and pupil reactivity in traumatic brain injury assessed pre-hospital and on enrollment: an IMPACT analysis","volume":"24","author":"Marmarou","year":"2007","journal-title":"J Neurotrauma"},{"key":"ref19","doi-asserted-by":"publisher","first-page":"666","DOI":"10.3171\/jns.2003.99.4.0666","article-title":"Patient age and outcome following severe traumatic brain injury: an analysis of 5600 patients","volume":"99","author":"Hukkelhoven","year":"2003","journal-title":"J Neurosurg"},{"key":"ref20","doi-asserted-by":"publisher","first-page":"1590","DOI":"10.1111\/j.1532-5415.2006.00894.x","article-title":"Traumatic brain injury in older adults: epidemiology, outcomes, and future implications","volume":"54","author":"Thompson","year":"2006","journal-title":"J Am Geriatr Soc"},{"key":"ref21","first-page":"131","article-title":"Early predictive factors on mortality in head injured patients: a retrospective analysis of 112 traumatic brain injured patients","volume":"56","author":"Prisco","year":"2012","journal-title":"J Neurosurg Sci"},{"key":"ref22","doi-asserted-by":"publisher","first-page":"25","DOI":"10.1177\/000313480907500105","article-title":"Persistent hyperglycemia in severe traumatic brain injury: an independent predictor of outcome","volume":"75","author":"Salim","year":"2009","journal-title":"Am Surg"},{"key":"ref23","doi-asserted-by":"publisher","first-page":"91","DOI":"10.3389\/fneur.2014.00091","article-title":"Glucose and the injured brain-monitored in the neurointensive care unit","volume":"5","author":"Rostami","year":"2014","journal-title":"Front Neurol"},{"key":"ref24","doi-asserted-by":"publisher","first-page":"999","DOI":"10.1097\/TA.0b013e3182858a3e","article-title":"Lactate clearance as a predictor of mortality in trauma patients","volume":"74","author":"Odom","year":"2013","journal-title":"J Trauma Acute Care Surg"},{"key":"ref25","doi-asserted-by":"publisher","first-page":"1037","DOI":"10.1227\/NEU.0b013e3182287ca7","article-title":"Brain hypoxia is associated with short-term outcome after severe traumatic brain injury independently of intracranial hypertension and low cerebral perfusion pressure","volume":"69","author":"Oddo","year":"2011","journal-title":"Neurosurgery"},{"key":"ref26","doi-asserted-by":"publisher","first-page":"1913","DOI":"10.1097\/CCM.0b013e31821b85c6","article-title":"Red cell distribution width and all-cause mortality in critically ill patients","volume":"39","author":"Bazick","year":"2011","journal-title":"Crit Care Med"},{"key":"ref27","doi-asserted-by":"publisher","first-page":"161","DOI":"10.1007\/s10067-017-3871-7","article-title":"Red blood cell distribution width: a potential laboratory parameter for monitoring inflammation in rheumatoid arthritis","volume":"37","author":"He","year":"2018","journal-title":"Clin Rheumatol"},{"key":"ref28","doi-asserted-by":"publisher","first-page":"110","DOI":"10.1007\/s12028-017-0410-y","article-title":"Systemic inflammatory response syndrome (SIRS) score independently predicts poor outcome in isolated traumatic brain injury","volume":"28","author":"Jacome","year":"2018","journal-title":"Neurocrit Care"},{"key":"ref29","doi-asserted-by":"publisher","first-page":"22","DOI":"10.1016\/j.nurt.2009.10.016","article-title":"Involvement of pro-and anti-inflammatory cytokines and chemokines in the pathophysiology of traumatic brain injury","volume":"7","author":"Ziebell","year":"2010","journal-title":"Neurotherapeutics"},{"key":"ref30","doi-asserted-by":"publisher","first-page":"630","DOI":"10.1016\/S1474-4422(17)30197-7","article-title":"Coagulopathy and haemorrhagic progression in traumatic brain injury: advances in mechanisms, diagnosis, and management","volume":"16","author":"Maegele","year":"2017","journal-title":"Lancet Neurol"},{"key":"ref31","doi-asserted-by":"publisher","first-page":"1254","DOI":"10.1097\/TA.0b013e318156ee4c","article-title":"Early coagulopathy after traumatic brain injury: the role of hypoperfusion and the protein C pathway","volume":"63","author":"Cohen","year":"2007","journal-title":"J Trauma Acute Care Surg"},{"key":"ref32","doi-asserted-by":"publisher","first-page":"654","DOI":"10.1097\/01.CCM.0000155911.01844.54","article-title":"Non-neurologic organ dysfunction in severe traumatic brain injury","volume":"33","author":"Zygun","year":"2005","journal-title":"Crit Care Med"},{"key":"ref33","doi-asserted-by":"publisher","first-page":"108","DOI":"10.1080\/02699052.2024.2304885","article-title":"Severe traumatic brain injury and acute kidney injury patients: factors associated with in-hospital mortality and unfavorable outcomes","volume":"38","author":"de C\u00e1ssia Almeida Vieira","year":"2024","journal-title":"Brain Inj"},{"key":"ref34","doi-asserted-by":"publisher","first-page":"e90956","DOI":"10.1371\/journal.pone.0090956","article-title":"Fever burden is an independent predictor for prognosis of traumatic brain injury","volume":"9","author":"Bao","year":"2014","journal-title":"PLoS One"},{"key":"ref35","doi-asserted-by":"publisher","first-page":"2211","DOI":"10.1001\/jama.2018.17075","article-title":"Effect of early sustained prophylactic hypothermia on neurologic outcomes among patients with severe traumatic brain injury: the POLAR randomized clinical trial","volume":"320","author":"Cooper","year":"2018","journal-title":"JAMA"},{"key":"ref36","doi-asserted-by":"publisher","first-page":"224","DOI":"10.1016\/j.jss.2019.09.002","article-title":"Older patients with severe traumatic brain injury: National Variability in palliative care","volume":"246","author":"Hwang","year":"2020","journal-title":"J Surg Res"},{"key":"ref37","doi-asserted-by":"publisher","first-page":"1700383","DOI":"10.3389\/fmed.2025.1700383","article-title":"Cultural and socioeconomic determinants of family satisfaction with ICU care across the globe: a scoping review","volume":"12","author":"Bhatt","year":"2025","journal-title":"Front Med (Lausanne)"},{"key":"ref38","first-page":"36","article-title":"Death after discharge: prognostic model of 1-year mortality in traumatic brain injury patients undergoing decompressive craniectomy","volume":"8","author":"Cui","year":"2022","journal-title":"Chin Neurosurg J"},{"key":"ref39","doi-asserted-by":"publisher","first-page":"2044","DOI":"10.1111\/cns.13935","article-title":"Prediction of early prognosis after traumatic brain injury by multifactor model","volume":"28","author":"Yang","year":"2022","journal-title":"CNS Neurosci Ther"},{"key":"ref40","doi-asserted-by":"publisher","first-page":"1122","DOI":"10.3390\/ijms22031122","article-title":"Mismatch between tissue partial oxygen pressure and near-infrared spectroscopy Neuromonitoring of tissue respiration in acute brain trauma: the rationale for implementing a multimodal monitoring strategy","volume":"22","author":"Forcione","year":"2021","journal-title":"Int J Mol Sci"}],"container-title":["Frontiers in Neurology"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fneur.2026.1815307\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,4,23]],"date-time":"2026-04-23T04:21:20Z","timestamp":1776918080000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fneur.2026.1815307\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,4,23]]},"references-count":40,"alternative-id":["10.3389\/fneur.2026.1815307"],"URL":"https:\/\/doi.org\/10.3389\/fneur.2026.1815307","relation":{},"ISSN":["1664-2295"],"issn-type":[{"value":"1664-2295","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,4,23]]},"article-number":"1815307"}}