Glasgow Coma Scale

IntroductionOur aim in this study was to assess whether the new Glasgow Coma Scale, Age, and Systolic Blood Pressure (GAP) scoring system, which is a modification of the Mechanism, Glasgow Coma Scale, Age, and Arterial Pressure (MGAP) scoring system, better predicts in-hospital mortality and can be applied more easily than previous trauma scores among trauma patients in the emergency department (ED).MethodsThis multicenter, prospective, observational study was conducted to analyze readily available variables in the ED, which are associated with mortality rates among trauma patients. The data used in this study were derived from the Japan Trauma Data Bank (JTDB), which consists of 114 major emergency hospitals in Japan. A total of 35,732 trauma patients in the JTDB from 2004 to 2009 who were 15 years of age or older were eligible for inclusion in the study. Of these patients, 27,154 (76%) with complete sets of important data (patient age, Glasgow Coma Scale (GCS) score, systolic blood pressure (SBP), respiratory rate and Injury Severity Score (ISS)) were included in our analysis. We calculated weight for the predictors of the GAP scores on the basis of the records of 13,463 trauma patients in a derivation data set determined by using logistic regression. Scores derived from four existing scoring systems (Revised Trauma Score, Triage Revised Trauma Score, Trauma and Injury Severity Score and MGAP score) were calibrated using logistic regression models that fit in the derivation set. The GAP scoring system was compared to the calibrated scoring systems with data from a total of 13,691 patients in a validation data set using c-statistics and reclassification tables with three defined risk groups based on a previous publication: low risk (mortality 50%).ResultsCalculated GAP scores involved GCS score (from three to fifteen points), patient age 120 mmHg, six points; 60 to 120 mmHg, four points). The c-statistics for the GAP scores (0.933 for long-term mortality and 0.965 for short-term mortality) were better than or comparable to the trauma scores calculated using other scales. Compared with existing instruments, our reclassification tables show that the GAP scoring system reclassified all patients except one in the correct direction. In most cases, the observed incidence of death in patients who were reclassified matched what would have been predicted by the GAP scoring system.ConclusionsThe GAP scoring system can predict in-hospital mortality more accurately than the previously developed trauma scoring systems.

IntroductionThe Glasgow Coma Scale (GCS) is the most widely used scoring system for comatose patients in intensive care. Limitations of the GCS include the impossibility to assess the verbal score in intubated or aphasic patients, and an inconsistent inter-rater reliability. The FOUR (Full Outline of UnResponsiveness) score, a new coma scale not reliant on verbal response, was recently proposed. The aim of the present study was to compare the inter-rater reliability of the GCS and the FOUR score among unselected patients in general critical care. A further aim was to compare the inter-rater reliability of neurologists with that of intensive care unit (ICU) staff.MethodsIn this prospective observational study, scoring of GCS and FOUR score was performed by neurologists and ICU staff on 267 consecutive patients admitted to intensive care.ResultsIn a total of 437 pair wise ratings the exact inter-rater agreement for the GCS was 71%, and for the FOUR score 82% (P = 0.0016); the inter-rater agreement within a range of +/- 1 score point for the GCS was 90%, and for the FOUR score 92% (P = ns.). The exact inter-rater agreement among neurologists was superior to that among ICU staff for the FOUR score (87% vs. 79%, P = 0.04) but not for the GCS (73% vs. 73%). Neurologists and ICU staff did not significantly differ in the inter-rater agreement within a range of +/- 1 score point for both GCS (88% vs. 93%) and the FOUR score (91% vs. 88%).ConclusionsThe FOUR score performed better than the GCS for exact inter-rater agreement, but not for the clinically more relevant agreement within the range of +/- 1 score point. Though neurologists outperformed ICU staff with regard to exact inter-rater agreement, the inter-rater agreement of ICU staff within the clinically more relevant range of +/- 1 score point equalled that of the neurologists. The small advantage in inter-rater reliability of the FOUR score is most likely insufficient to replace the GCS, a score with a long tradition in intensive care.

BackgroundThe shock index (SI), defined as heart rate (HR) divided by systolic blood pressure (SBP), is reported to be a more sensitive marker of shock than traditional vital signs alone. In previous literature, use of the reverse shock index (rSI), taken as SBP divided by HR, is recommended instead of SI for hospital triage. Among traumatized patients aged > 55 years, SI multiplied by age (SIA) might provide better prediction of early post-injury mortality. Separately, the Glasgow Coma Scale (GCS) score has been shown to be a very strong predictor. When considering these points together, rSI multiplied by GCS score (rSIG) or rSIG divided by age (rSIG/A) could provide even better prediction of in-hospital mortality.MethodsThis retrospective, multicenter study used data from 168,517 patients registered in the Japan Trauma Data Bank for the period 2006-2015. We calculated areas under receiver operating characteristic curves (AUROCs) to measure the discriminant ability by comparing those of SI (or rSI), SIA, rSIG, and rSIG/A for in-hospital mortality and for 24-h blood transfusion.ResultsThe highest ROC AUC (AUROC), 0.901(0.894-0.908) for in-hospital mortality in younger patients (aged ConclusionsrSIG ((SBP/HR) × GCS score) is easy to calculate without the need for additional information, charts or equipment, and can be a more reliable triage tool for identifying risk levels in trauma patients.

BackgroundMulticenter electronic health records (EHRs) can support quality improvement and comparative effectiveness research in critical care. However, limitations of EHR-based research include challenges in abstracting key clinical variables, including a patient's level of consciousness.ObjectiveThis study aimed to develop a natural language processing model to predict Glasgow Coma Scale (GCS) scores from daily EHR notes.MethodsThe study included adult patients (aged ≥18 years) admitted to Mass General Brigham (MGB) hospitals (2017-2024) and patients from the Medical Information Mart for Intensive Care-III (MIMIC-III version 1.4; 2001-2012) database. A dataset of all patients from both institutions was split into training (70%) or hold-out test (30%) sets. Variables consisted of daily notes, age, sex, and admission type. A pooled ordinal regression model (ordinalNet) with an elastic net penalty was trained to predict the lowest daily level of consciousness across 3 classes of impairment: severe (GCS score 3-8), moderate (GCS score 9-12), and mild (GCS score 13-15), and a pooled linear model was trained to predict continuous GCS scores (3-15). Gold standard GCS was obtained from structured flowsheet data. External generalizability was assessed using a single-institution ordinal model trained on MGB and tested on MIMIC. Following post hoc calibration, the performance of the ordinal and linear models was evaluated on the hold-out test sets using the area under the receiver operating characteristic curve (AUROC) and area under the precision-recall curve (AUPRC) for the ordinal models and root mean square error and Pearson correlation coefficient for the linear models.ResultsThe modeling cohort included 145,897 patients (MGB: n=123,257, MIMIC: n=22,640), with 1,446,965 days of hospitalization between training and testing sets; the average age was 62 (SD 18) years, and the sex distribution was balanced. The pooled ordinalNet achieved an AUROC of 0.96 (95% CI 0.96-0.96) and an AUPRC of 0.77 (95% CI 0.76-0.77). The single-institution ordinal model achieved an AUROC of 0.90 (95% CI 0.89-0.90) and an AUPRC of 0.80 (95% CI 0.79-0.80). The pooled linear model achieved a root mean square error of 2.30 (95% CI 2.30-2.30) and a correlation of 0.76 (95% CI 0.76-0.76). Predictions for severe GCS were driven by terms indicating unresponsiveness and critical interventions, moderate GCS by intermediate alertness descriptors, and mild GCS by mentions of normal or awake behavior.ConclusionsPooled ordinal and linear models can accurately predict GCS from unstructured data and can support large-scale phenotyping of neurological assessments for future critical care research.