RESUMEN
Clinical prediction models serve as valuable instruments for assessing the risk of crucial outcomes and facilitating decision-making in clinical settings. Constructing these models requires nuanced analytical decisions and expertise informed by the current statistical literature. Access and thorough understanding of such literature may be limited for neurocritical care physicians, which may hinder the interpretation of existing predictive models. The present emphasis is on narrowing this knowledge gap by providing neurocritical care specialists with methodological guidance for interpreting predictive models in neurocritical care. Presented are the statistical learning principles integral to constructing a model predicting hospital mortality (nonsurvival during hospitalization) in patients with moderate and severe blunt traumatic brain injury using components of the IMPACT-Core model. Discussion encompasses critical elements such as model flexibility, hyperparameter selection, data imbalance, cross-validation, model assessment (discrimination and calibration), prediction instability, and probability thresholds. The intricate interplay among these components, the data set, and the clincal context of neurocritical care is elaborated. Leveraging this comprehensive exploration of statistical learning can enhance comprehension of articles encompassing model generation, tailored clinical care, and, ultimately, better interpretation and clinical applicability of predictive models.
RESUMEN
Traumatic brain injury (TBI) is a global health challenge; however, penetrating brain injury (PBI) remains under-represented in evidence-based knowledge and research efforts. This study utilized data from the Trauma Quality Improvement Program (TQIP) of the National Trauma Data Bank (NTDB) to investigate outcomes of PBI as compared with clinical-severity-matched non-penetrating or blunt TBI. A total of 1765 patients with PBI were 1:1 propensity score-matched for clinical severity with blunt TBI patients. The intent of PBI was self-inflicted in 34.1% of the cases, and the mechanism was firearm-inflicted in 89.1%. Mortality was found to be significantly more common in PBI than in the severity- matched TBI cohort (33.9% vs. 14.3 %, p < 0.001) as was unfavorable outcome. Mortality was mediated by withdrawal of life-sustaining therapies (WOLST) 30% of the time, and WOLST occurred earlier (median 3 days vs. 6 days, p < 0.001) in PBI. Increased rate of mortality was observed with a Glasgow Coma Scale (GCS) of <11 in PBI as compared with <7 in blunt TBI. In conclusion, PBI patients exhibited higher mortality rates and unfavorable outcomes; one third of excess mortality was mediated by WOLST. The study also brings into question the applicability of the conventional TBI classification, based on GCS, in PBI. We emphasize the need to address the observed disparities and better understand the distinctive characteristics and mechanisms underlying PBI outcomes to improve patient care and reduce mortality.