Modified Early Warning Score (MEWS) combined with biomarkers in predicting 7-day mortality in traumatic brain injury patients in the emergency department: a retrospective cohort study

Study design, participants, and procedure

A retrospective study was conducted among patients with TBI admitted to the emergency department of the First People’s Hospital of Changde, China, from January 1, 2023, to June 30, 2023. The inclusion criteria of the participants were as follows: (1) with a confirmed diagnosis of TBI, (2) age ≥18, and (3) with a head Abbreviated Injury Scale (AIS) score ≥3 (Mellick, Gerhart & Whiteneck, 2003). The exclusion criteria were as follows: (1) patients who were admitted after 24 h of injury, (2) patients who had a cardiac arrest following trauma before admission, and (3) patients with incomplete information on the predicting and outcome variables. A total of 54,129 patients were admitted to the emergency department during the study period, among whom 15,360 had a confirmed diagnosis of TBI. After excluding 5,687 patients who were aged <18 or had AIS scores <3, a total of 9,673 patients were selected as eligible participants for the study. We further excluded 2,832 patients who were admitted after 24 h, 1,565 patients who had a cardiac arrest before admission, and 3,575 patients with incomplete data on MEWS or mortality, leading to a final sample of 1,701 patients included in the analysis. The detailed patient selection process is illustrated in Fig. 1. When the TBI patient was admitted to the emergency room, the first step was pre-triage, and then the TBI patient was admitted to the monitoring room of the trauma center via green passage as the second step. The third step involved the initial control of asphyxia, shock, and massive bleeding before further evaluation. The fourth step included routine tests according to the guidelines, such as white blood cells, red blood cells, and creatine. These tests typically took less than 1 h from sample collection to result reporting. Finally, the definitive treatment was applied to the patient, such as surgery or other medical treatment (American College of Surgeons, 2015; Picetti et al., 2019; Silverberg et al., 2020). This study was approved by the Ethics Committee of the First People’s Hospital of Changde City (YX-2024-010-01). The requirement for written informed consent was waived due to the retrospective nature of the study.

Data collection

Data on the following variables were obtained from the electronic medical records: age, gender, leading causes of TBI, type of TBI lesion, hours in the emergency department (HER, hours), respiratory rate (RR, cycles per minute), heart rate (HR, beats per minute), systolic blood pressure (SBP, mmHg), diastolic blood pressure (DBP, mmHg), mean arterial pressure (MAP, mmHg), body temperature (T, °C), and oxygen saturation (SpO₂,%) on admission. We also collected the first routine test results when the TBI patient was admitted to the emergency department: white blood cell, red blood cell, RDW, platelet, potassium, creatine, and hemoglobin. GCS and MEWS scores were assessed by triage nurses who have received in-hospital education and training in the triage room at the emergency department. The GCS is calculated based on eye-opening, verbal response, and motor, and the total score ranges from 3 to 15 (Anand, Shahid & Shameel, 2024). The MEWS is calculated based on the following vital signs: heart rate, blood pressure, respiratory rate, body temperature, and AVPU (Alert, Voice, Pain, Unresponsive). Details of each variable are shown in Table 1. The total MEWS score ranges from 0 to 3, with a higher score indicating a worse state of TBI patients. AVPU is estimated from the GCS reported previously as follows: A = 14–15, V = 9–13, P = 4–8, U = 3 (Kim et al., 2021). The primary outcome was 7-day in-hospital mortality.

Statistical analyses

Missing data were filled in using the multiple imputation method. Continuous data were described using means ± standard deviations (SD) for normally distributed variables and medians with interquartile ranges (IQR) for non-normally distributed variables. Categorical variables were presented using frequencies and percentages (%). A Student’s t-test (normal distribution) or Mann-Whitney U test (non-normal distribution) was used to compare continuous variables by 7-day in-hospital mortality. The chi-square test was used to compare categorical variables by 7-day in-hospital mortality. Influencing factors of the 7-day mortality were determined by the logistic regression model, which models the log odds of an outcome as a linear combination of one or more independent variables. The algorithm for the logistic regression model is as follows: y = β0 + β1X1 + β2X2 +…+ βmXm, where the coefficient (weight) of each X variable represents the change in the log odds of the Y variable (mortality) for a unit change in the X variable. Receiver operating characteristic (ROC) curve analysis was performed to compare the prognostic performances of the individual and combined effects of MEWS, RDW, and creatine in predicting 7-day in-hospital mortality, and their predicting performance was calculated by the area under the curve (AUC). The Delong tests were conducted to compare the ROC CIs of various indicators. Bootstrap resampling was used to validate the IT model based on the TRIPOD guideline. In particular, bootstrap resampling (500 times) was employed to obtain the area under the curve (AUC) and the best cut-off value of individual and combined indicators of MEWS, RDW, and creatine. Statistical analyses were performed using R (http://www.R-project.org), Empower (R) (http://www.empowerstats.com, X&Y Solutions, Inc., Boston, MA, USA), and SPSS V 23.0. P < 0.05 (two-sided) was considered statistically significant.

Sample characteristics

Table 2 shows the sample characteristics and their comparisons between the survival and death groups. A total of 1,701 patients with TBI were included in the analysis, including 456 women and 1,245 men. Patients were stratified by 7-day mortality, among whom 1,476 TBI patients survived, and 225 TBI patients died, leading to a mortality rate of 13.23%. Vehicle collision (36.51%) accounted for the largest proportion of all TBI causes. The survival group had higher MEWS scores and spent longer hours (HER) in the emergency room than the death group. The two groups also showed statistically significant differences in the following indicators: the type of TBI lesion, HR, RR, SBP, DBP, MAP, SpO₂, temperature, RDW, and potassium.

Univariate and multivariate analysis

Table 3 shows the univariate and multivariate analysis results of the influencing factors of 7-day mortality. The univariate analysis indicated that MEWS score, the type of TBI lesion, HER, HR, RR, SBP, DBP, MAP, SpO₂, temperature, RDW, potassium, and creatine were correlated with 7-day mortality. The multivariate analysis identified the following independent influencing factors associated with 7-day mortality: MEWS score (OR: 1.92, 95% CI [1.70–2.18]), the TBI type of Subdural hematoma (OR: 4.23, 95% CI [1.73–10.30]), Intracerebral hematoma (OR: 3.84, 95% CI [1.55–9.52]), Cerebral contusion (OR: 5.32, 95% CI [2.23–12.72]), Hemorrhagic contusion (OR: 3.09, 95% CI [1.17–8.15]), Penetrating TBI (OR: 3.19, 95% CI [1.24–8.20]), Subarachnoid hemorrhage (OR: 5.31, 95% CI [2.15–13.10]), SBP (OR: 1.11, 95% CI [1.01–1.22]), DBP (OR: 1.23, 95% CI [1.02–1.49], MAP (OR: 0.74, 95% CI [0.56–0.98]), SpO₂(OR:0.97, 95% CI [0.94–0.99]), temperature (OR: 1.18, 95% CI [1.04–1.34]), RDW (OR: 1.31, 95% CI [1.22–1.41]), creatine (OR: 1.34, 95% CI [1.23–1.47]).

ROC analysis

Figures 2A–2G show the smooth and raw ROC curves of MEWS, RDW, and creatine with and without the bootstrapping method. The AUC for MEWS, creatine, and RDW were 0.843, 0.797, and 0.785, respectively. The Delong tests showed significant differences in the ROC CIs of the three indicators (Table 4). Moreover, the best thresholds of MEWS, creatine, and RDW were 3.5 (sensitivity: 0.7422, specificity: 0.8448), 3.15 (sensitivity: 0.4359, specificity: 0.9385), and 14.05 (sensitivity: 0.6800, specificity: 0.6829), respectively (Table 5). The smooth ROCs obtained by the bootstrapping method (resample: 500) of MEWS, RDW, and creatine in predicting 7-day mortality are included in the Fig. S1.

Figures 3A–3C show the raw and smooth ROC of MEWS combined with RDW and creatine, with an AUC of 0.906. The AUC of the smooth ROC of MEWS combined with RDW (Model 1) was 0.898, the AUC of the smooth ROC of MEWS combined with creatine (Model 2) was 0.875, and the AUC of smooth ROC of RDW combined with creatine (Model 3) was 0.822. The ROC curves of two indicator combinations with and without the bootstrapping method are included in Fig. S2.

Strengths

A major innovation of the study was that we presented a new concept algorithm to predict 7-day in-hospital mortality based on the type of TBI lesions, physical examination test (such as blood pressure, body temperature, and oxygen saturation), and standard laboratory test (such as MEWS, RDW, and creatine). All these parameters are easily accessible as part of the routine test for TBI patients admitted to the hospital. Therefore, our risk prediction model provides a relatively rapid, reliable, and cost-effective way to identify high-risk TBI patients and take proactive interventions to prevent and reduce mortality. Other advantages of the study included a large sample size to ensure enough statistical power to detect significant results, the inclusion of a variety of clinical indicators to control for confounding effects, and a comprehensive comparison of the predictive performance of the individual and various combinations of the indicators.

Clinical limitations and future directions

However, this study also had several limitations. First, the retrospective nature of the study design is susceptible to recall bias and missing information. Future prospective study designs are needed to produce more robust results. Second, all participants were recruited from a single hospital and may not represent TBI patients in other hospitals and other regions. Future multicenter studies are needed to validate our model further. Third, we did not analyze the effects of essential procedures (such as interventions, operations, and transfusions) on 7-day mortality. Future studies should consider adding these variables to control for their potential confounding effects. Fourth, we only focused on three biomarkers based on laboratory tests and did not include other biomarkers, such as genetic markers, which can provide more specific information on the mechanism of TBI-related death. Future studies may consider adding genetic markers, such as S100B, glial fibrillary acidic protein (GFAP), ubiquitin C-terminal hydrolase-L1 (UCH-L1), and neurofilament light (NF-L), to get a deeper understanding of the TBI mechanism. Finally, it should be noted that although the three biomarkers are easily available as part of the routine test for TBI patients admitted to the emergency room, they are not as quickly available as MEWS, which takes only a few minutes rather than up to 1 h. Although the combination of MEWS and biomarkers may improve the diagnostic performance in predicting 7-day mortality, its clinical utility in time-sensitive emergency care settings may be limited. Therefore, in real practice, clinicians should choose the most appropriate indicators by balancing the prediction accuracy and the time cost. Future studies should explore other time-saving and cost-effective risk prediction models to predict mortality among patients with TBI admitted to the emergency department.

Implications

In this study, we developed a 7-day in-hospital risk prediction model based on readily available parameters, such as MEWS, RDW, and creatine, which carry significant implications in clinical practice. Compared to the expensive and time-consuming genetic tests such as S100B and GFAP, our model offers a more affordable and practical evaluation of mortality risk among TBI patients, which is especially valuable in resource-limited rural areas and less developed countries. The finding that MEWS outperformed RDW and creatine in predicting 7-day mortality and that the combination of all three indicators had the best-predicting performance also provides helpful, practical guidance to inform clinical decision-making. As soon as the TBI patient is admitted to the emergency department, the more readily available indicator, MEWS, can be used to provide a quick and reliable prediction. After the doctor confirms the first diagnosis, the combination of MEWS with RDW and creatine may be considered to get a more precise prediction.