Comparative Validation of Five Early Trauma Scores for 24-Hour In-Hospital Mortality in Traumatic Brain Injury

Comparative Validation of Five Early Trauma Scores for 24-Hour In-Hospital Mortality in Traumatic Brain Injury | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Comparative Validation of Five Early Trauma Scores for 24-Hour In-Hospital Mortality in Traumatic Brain Injury Tianxi Chen, Jinheng Tu, Jiajia Liu, Hong sun, Yun Lu, Rui Chen, and 2 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8452508/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Background Early risk stratification after traumatic brain injury (TBI) is crucial to guide monitoring, escalation of care, and interfacility transfer within the first hours. Bedside scores (RTS, GAP, mREMS, MGAP, MEWS) are simple and widely available, yet their comparative accuracy for predicting 24-hour in-hospital mortality in TBI remains uncertain. Prior studies often use mixed trauma cohorts, later endpoints, and provide limited evaluation of calibration or threshold-level clinical utility. TBI-specific head-to-head evidence is therefore needed. Methods This study analyzed consecutive traumatic brain injury admissions in 2024 at a tertiary medical center in Nantong, China (n = 2,287; 24-hour mortality 7.3%). Five bedside scores (RTS, GAP, mREMS, MGAP, MEWS) were recalibrated with a single logistic equation, and performance (AUROC, AUPRC, Brier score, calibration intercept and slope) and clinical utility (net benefit per 100 patients at 10% and 20% thresholds) were estimated with 95% confidence intervals from 1,000 bootstrap resamples; precision–recall curves were referenced to the cohort prevalence. Results RTS and GAP showed the strongest discrimination, with mREMS close. AUROC: RTS 0.874 (95% CI 0.841–0.903), GAP 0.863 (0.831–0.898), mREMS 0.856 (0.826–0.883); MGAP and MEWS were lower (0.632, 0.606). AUPRCs were 0.465 (0.381–0.531) for RTS, 0.382 (0.308–0.453) for GAP, and 0.370 (0.280–0.424) for mREMS (MGAP 0.154 (0.112–0.199); MEWS 0.129 (0.098–0.159)). Brier scores were 0.050, 0.053, and 0.055 for RTS, GAP, and mREMS, respectively. After simple logistic recalibration, calibration intercepts were near 0 and slopes near 1, with observed-vs-predicted curves close to the 45° line across deciles. At 10% and 20% thresholds, RTS and GAP achieved the highest net benefit and more favorable trade-offs per 100 patients; mREMS was intermediate. Conclusion RTS and GAP showed the best balance of discrimination, calibration, and net benefit at 10% and 20% thresholds; mREMS was comparable after simple recalibration. These findings support prioritizing RTS and GAP for early TBI risk stratification. Traumatic Brain Injury Risk Stratification Clinical Prediction Logistic Recalibration Figures Figure 1 Figure 2 Figure 3 Figure 4 Introduction Traumatic brain injury is a major driver of early in-hospital deterioration and death worldwide, imposing substantial clinical and system burden[ 1 ]. Early risk stratification in the first hours of care informs monitoring intensity, escalation decisions, and interfacility transfer, and is central to outcomes-focused acute management. Simple bedside scores, including RTS, GAP, mREMS, MGAP, and MEWS, are attractive because they use routinely collected variables and can be calculated rapidly at the point of care[ 2 – 6 ]. However, despite widespread use, contemporary head-to-head analyses yield mixed results for RTS, GAP, mREMS, MGAP, and MEWS on short-term mortality[ 7 , 8 ], while quantitative assessment of decision utility at clinically relevant thresholds is infrequently provided. Recent evaluations still leave practical questions unresolved. Many enroll mixed trauma rather than TBI-specific cohorts and emphasize outcomes beyond the first 24 hours, which limits applicability to early in-hospital mortality. AUROC is widely reported and remains informative for ranking performance, yet low event rates call for complementary indices that capture rare-outcome behavior and absolute risk, including AUPRC referenced to prevalence, Brier score, and calibration intercept and slope[ 9 , 10 ]. Decision-analytic reporting is uncommon, so threshold-specific trade-offs, consequences per 100 patients, and net benefit that guide escalation or transfer are rarely presented[ 11 ]. Definitions and implementations of RTS, GAP, mREMS, MGAP, and MEWS vary across prehospital and in-hospital settings, including handling of Glasgow Coma Scale components, systolic blood pressure cut points, and oxygen saturation capture, which complicates transportability[ 12 – 14 ]. Cohorts differ in time zero, inclusion criteria, and 24-hour outcome ascertainment, and missing data are frequently managed with ad hoc rules rather than prespecified strategies, both of which can shift apparent ranking[ 15 ]. Sample sizes are often modest relative to the low event rate, increasing uncertainty around precision and recall performance and model calibration. Few studies report locally calibrated absolute risks or apply a simple recalibration step before comparison, despite known intercept and slope drift across settings. Threshold selection is seldom linked to operational constraints such as monitored bed capacity, neurosurgical availability, or transfer logistics, limiting bedside uptake. Together, these gaps leave clinicians without clear, calibration-aware guidance on which score to prioritize for early decision making in contemporary practice. Bedside tools should align with rapid assessment workflows, communicate absolute risk, and support early triage and interfacility transfer within the first hours of care. Clinicians and administrators benefit from threshold-specific summaries that show consequences per 100 patients and net benefit rather than rank-based metrics alone[ 16 ]. Simple recalibration to local prevalence and measurement practices is preferable to refitting because it preserves interpretability and improves transportability when calibration drifts across settings[ 17 ]. Using routinely collected variables enhances feasibility across prehospital and emergency contexts and supports consistent handoffs[ 18 ]. Focusing on a 24-hour endpoint targets safety-critical deterioration and provides a common trigger for escalation pathways and transfer decisions[ 19 ]. This work delivers a head-to-head evaluation of RTS, GAP, mREMS, MGAP, and MEWS for predicting 24-hour in-hospital mortality in contemporary TBI care. Each score is aligned to local risk using a single-equation logistic recalibration to generate calibrated probabilities suitable for bedside decisions. Performance is summarized with AUROC, AUPRC referenced to prevalence, the Brier score, and calibration (intercept, slope); decision performance is quantified at 10% and 20% probability thresholds using consequences per 100 patients and net benefit, with uncertainty expressed as 95% confidence intervals from 1,000 bootstrap resamples. Collectively, these analyses provide calibration-aware, threshold-focused guidance for selecting first-line bedside tools for early risk stratification. Materials and methods Guidelines This study was designed and reported in accordance with the TRIPOD statement and relevant EQUATOR Network guidance[20]. The protocol received approval from the institutional ethics committee (approval No. 2021-K084-01). Owing to the retrospective design and use of de-identified clinical data, the requirement for informed consent was waived. All procedures complied with the Declaration of Helsinki and institutional data-protection policies[21]. Data source We drew data from the integrated medical database of a tertiary medical center in Nantong, China. Consecutive admissions within the study window (January 1–December 31, 2024) were screened against prespecified criteria. Eligible cases were adults (≥18 years) with an index encounter during the study period, an ascertainable 24-hour in-hospital outcome, and sufficient variables to compute RTS, GAP, mREMS, MGAP, and MEWS for validation. Of 3,064 admissions initially assessed, 777 were excluded for the following reasons: age <18 years (n=47); non-traumatic intracranial pathologies such as stroke, tumor, or infection (n=213); elective neurosurgical admissions unrelated to acute traumatic injury (n=102); duplicate or interfacility transfer episodes, with only the first index encounter retained (n=131); transfer out before 24 hours with unknown outcome (n=58); and missing key variables or the 24-hour outcome (n=226). The final analytic cohort comprised 2,287 patients. Details of screening and exclusions are shown in Figure 1 (Patient selection and enrollment flowchart). Outcome The primary outcome was 24-hour all-cause in-hospital mortality during the index encounter. Deaths were ascertained from structured fields in the electronic health record using the in-hospital death indicator and, when available, the recorded time of death. Patients who were alive at 24 hours, including those discharged before 24 hours, were classified as non-events. Encounters transferred out before 24 hours with unknown vital status, and records lacking a verifiable 24-hour outcome, were not included in the analytic cohort. Data collection and variables Data were abstracted at the index encounter from the integrated electronic health record using a prespecified schema. Baseline fields included sex; age (years) with a derived age band for score computation; temperature (°C); heart rate (beats per minute); respiratory rate (breaths per minute); systolic and diastolic blood pressure (mmHg); peripheral oxygen saturation (SpO₂, %); mechanism of injury coded as penetrating or blunt; pupil reactivity recorded in three categories (no reaction, unilateral reaction, bilateral reaction); the Glasgow Coma Scale (GCS, 3–15). The analytic endpoint was 24-hour all-cause in-hospital mortality (Death24h). Bedside scores (RTS, GAP, mREMS, MGAP, MEWS) were taken from the dataset or, when components were available, reconstructed exactly according to published definitions with unit harmonization and prespecified cut points; higher totals for RTS and GAP indicate lower risk, whereas higher mREMS and MEWS indicate greater risk. When multiple measurements were available around arrival, the earliest nonmissing value within the initial assessment window was used. Implausible entries were flagged and set to missing before calculation; records lacking a verifiable 24-hour outcome or a key variable had been removed during cohort construction. Missing data Prior to analysis, encounters with an unknown 24-hour vital status or missing components required to compute any bedside score were removed at cohort construction. This yielded an analytic set of 2,287 patients with no missing outcome and complete values for RTS, GAP, mREMS, MGAP, and MEWS. No statistical imputation was performed. For baseline characteristics used only for description (e.g., demographics, vital signs, pupil reactivity, mechanism of injury, and GCS), we report available-case summaries with the denominator explicitly stated for each variable. Implausible entries were reviewed and set to missing prior to tabulation. All performance analyses (recalibration, discrimination, calibration, and threshold-based metrics) were conducted on the complete-case analytic set and were bootstrapped using case resampling. Data analysis Each bedside score was recalibrated with a single-equation logistic recalibration model to obtain predicted probabilities for evaluation. Baseline characteristics were summarized descriptively as mean (standard deviation) or median (interquartile range) for continuous variables, and count (percentage) for categorical variables; no hypothesis testing or p values were reported. Performance was examined for discrimination, calibration, overall performance, and decision analysis at prespecified probability thresholds of 10% and 20%. Graphical displays included ROC curves, precision–recall curves with the cohort prevalence shown as a horizontal reference, and calibration plots based on deciles with a 45° reference line. Statistical uncertainty was quantified with 1,000 patient-level bootstrap resamples using two-sided 95% confidence intervals by the percentile method (resampling stratified by outcome). All analyses were conducted in Python 3.9 using pandas, NumPy, scikit-learn, statsmodels, and Matplotlib. Performance metrics Performance was defined a priori across discrimination, calibration, overall performance, and threshold-specific clinical utility. Discrimination was quantified with the area under the receiver operating characteristic curve (AUROC) and the area under the precision–recall curve (AUPRC) using average precision to remain informative under class imbalance. Calibration was summarized by the calibration intercept and slope from a logistic calibration model, where values near 0 and 1 indicate minimal systematic bias and appropriate risk spread, respectively. Overall performance was characterized by the Brier score as a measure of probabilistic accuracy, with lower values indicating better performance. Clinical utility was evaluated at prespecified probability cutoffs of 10% and 20% using per 100 patients summaries that report true positives, false positives, false negatives, and true negatives together with sensitivity, specificity, positive predictive value, negative predictive value, and net benefit relative to treat-all and treat-none strategies. All estimates are accompanied by two-sided 95% confidence intervals obtained from 1,000 patient-level bootstrap resamples using the percentile method. Results We enrolled 2,287 TBI encounters; 166 patients died within 24 hours (7.3%) and 2,121 survived. The cohort was predominantly male (1,479 of 2,287; 64.7%) and injuries were almost exclusively blunt trauma (2,273 of 2,287; 99.4%). Median age was 62 years (IQR 52 to 71). At presentation, vital signs were generally within physiologic ranges, yet the death group showed a pattern of higher heart rate and blood pressure, together with lower peripheral oxygen saturation and lower GCS. Pupillary reactivity distinguished the groups most strongly: nonreactive pupils were present in 82 of 166 deaths versus 53 of 2,121 survivors (49.4% vs 2.5%), unilateral reaction in 40 of 166 vs 82 of 2,121 (24.1% vs 3.9%), and bilateral reaction in 44 of 166 vs 1,986 of 2,121 (26.5% vs 93.6%). The magnitude of imbalance by standardized difference was largest for pupillary status (approximately 1.99 for absent, 0.90 for unilateral, and 2.13 for bilateral response), far exceeding that of age or other vital signs. Full baseline characteristics, including continuous summaries for age, vital signs, and GCS, categorical distributions for sex, injury mechanism, and pupil status, and context-setting distributions of the bedside scores (RTS, GAP, mREMS, MGAP, MEWS), are reported in Table 1 . Table 1 Baseline characteristics (Survived ≤ 24h vs. Died ≤ 24h) Variable Overall (n = 2287) Survived ≤ 24h (n = 2121) Died ≤ 24h (n = 166) Std diff DEMOGRAPHICS Age, years (median [IQR]; mean ± SD) 62.0 [51.5–71.0]; 60.2 ± 15.7 62.0 [51.0–71.0]; 60.0 ± 15.8 65.5 [54.2–71.0]; 63.1 ± 14.0 0.20 Sex, n (%) └ Female 808 (35.3%) 759 (35.8%) 49 (29.5%) 0.13 └ Male 1479 (64.7%) 1362 (64.2%) 117 (70.5%) 0.13 Mechanism of injury, n (%) └ Penetrating 14 (0.6%) 12 (0.6%) 2 (1.2%) 0.08 └ Blunt 2273 (99.4%) 2109 (99.4%) 164 (98.8%) 0.08 VITAL SIGNS Temperature (T), °C (median [IQR]; mean ± SD) 36.5 [36.3–36.6]; 36.5 ± 0.4 36.5 [36.3–36.6]; 36.5 ± 0.3 36.5 [36.0–36.6]; 36.4 ± 0.4 0.15 Heart rate (P), beats/min (median [IQR]; mean ± SD) 81 [71–94]; 83.2 ± 18.1 81 [71–93]; 82.9 ± 17.5 84 [69–107]; 87.5 ± 24.1 0.26 Respiratory rate (R), breaths/min (median [IQR]; mean ± SD) 20 [ 16 – 21 ]; 19.2 ± 4.2 20 [ 16 – 21 ]; 19.1 ± 4.1 20 [ 16 – 22 ]; 19.4 ± 5.4 0.05 Systolic blood pressure (SBP), mmHg (median [IQR]; mean ± SD) 141 [123–161]; 142.8 ± 29.6 141 [123–161]; 142.7 ± 28.6 145 [119–173]; 143.8 ± 39.5 0.04 Diastolic blood pressure (DBP), mmHg (median [IQR]; mean ± SD) 82 [72–92]; 83.0 ± 21.1 82 [73–92]; 83.0 ± 21.0 83 [70–95]; 82.7 ± 22.5 0.02 Peripheral oxygen saturation (SpO₂), % (median [IQR]; mean ± SD) 98 [95–98]; 95.9 ± 5.6 98 [95–98]; 96.4 ± 4.1 96 [87–97]; 89.2 ± 13.2 1.35 NEUROLOGIC STATUS Glasgow Coma Scale (GCS, 3–15) (median [IQR]; mean ± SD) 15 [ 9 – 15 ]; 12.3 ± 4.3 15 [ 12 – 15 ]; 12.8 ± 3.9 3 [ 3 – 6 ]; 5.7 ± 4.0 1.84 Pupil reactivity, n (%) └ No reaction 135 (5.9%) 53 (2.5%) 82 (49.4%) 1.99 └ Unilateral reaction 122 (5.3%) 82 (3.9%) 40 (24.1%) 0.90 └ Bilateral reaction 2030 (88.8%) 1986 (93.6%) 44 (26.5%) 2.13 SCORE DISTRIBUTIONS (median [IQR]; mean ± SD) └ MGAP 24.0 [20.0–27.0]; 23.0 ± 5.2 24.0 [21.0–27.0]; 23.2 ± 5.0 21.5 [15.0–24.0]; 20.4 ± 6.2 0.55 └ RTS 7.8384 [6.9016–7.8408]; 7.0772 ± 1.3072 7.8384 [6.9040–7.8408]; 7.2532 ± 1.1020 4.0936 [4.0912–5.9648]; 4.8292 ± 1.6089 2.11 └ mREMS 4.0 [2.0–6.0]; 4.5 ± 3.5 4.0 [2.0–6.0]; 4.8 ± 3.5 1.0 [0.0–1.0]; 1.2 ± 1.9 1.04 └ MEWS 2.0 [1.0–3.0]; 2.3 ± 1.7 2.0 [1.0–3.0]; 2.2 ± 1.7 3.0 [1.0–4.0]; 3.0 ± 2.0 0.47 └ GAP 21.0 [16.0–22.0]; 19.2 ± 4.7 21.0 [18.0–24.0]; 19.7 ± 4.3 11.5 [9.0–13.0]; 12.3 ± 4.4 1.74 Notes: Continuous variables are summarized as median [IQR]; mean ± SD is also shown to aid comparison. Categorical variables are n (%). Std diff is the absolute standardized difference between groups. Figure 2 displays receiver operating characteristic curves after single equation logistic recalibration. RTS shows the highest discrimination with AUROC 0.874 (95% CI 0.841–0.903), followed by GAP 0.863 (0.831–0.898) and mREMS 0.856 (0.826–0.883); MGAP and MEWS are lower at 0.632 (0.586–0.678) and 0.606 (0.571–0.665). Across most of the specificity range, RTS and GAP remain above the other scores, especially at low false positive rates that matter for early triage. All estimates were obtained with 1,000 bootstrap resamples. Precision recall analysis under class imbalance reinforces the discrimination ranking. RTS achieves the largest AUPRC, 0.465 (95% CI 0.381–0.531), followed by GAP 0.382 (0.308–0.453) and mREMS 0.370 (0.280–0.424), with MGAP and MEWS clearly lower. The gray horizontal line marks the 7.3% prevalence, and across much of the recall axis RTS and GAP keep precision above this reference, including the high recall region that is critical for early triage. Orange markers denote operating points at 10% and 20% probability thresholds that fall on favorable portions of the RTS and GAP curves relative to the others, as shown in Fig. 3 . These findings support prioritizing RTS and GAP when the aim is to achieve high sensitivity with acceptable precision for identifying 24 hour mortality. Figure 4 demonstrates strong calibration after single equation updating. For RTS and GAP, the smoothed calibration curves with bootstrap 95% confidence bands lie close to the 45° identity line across the high density region of predictions below 0.20, with only small deviation around the middle of the range. mREMS shows a similar pattern and slightly overestimates risk in the extreme right tail. MGAP and MEWS show compressed risk ranges, flatter curves and wider uncertainty at low and high ends, which is consistent with the 7.3% event rate and the scarcity of observations in the tails. These patterns agree with Table 2 where calibration intercepts are near 0 and slopes are near 1, supporting the use of the recalibrated probabilities as absolute risks for 10% and 20% threshold decisions. Table 2 Primary performance metrics (recalibrated models) Score AUROC (95% CI) AUPRC (95% CI) Brier (95% CI) Calibration intercept Calibration slope Recalibration α Recalibration γ MGAP 0.632 (0.586–0.678) 0.154 (0.112–0.199) 0.066 (0.057–0.074) 0 1 -0.509 -0.093 RTS 0.874 (0.841–0.903) 0.465 (0.381–0.531) 0.050 (0.043–0.056) 0 1 3.729 -1.015 mREMS 0.856 (0.826–0.883) 0.370 (0.280–0.424) 0.055 (0.049–0.062) 0 1 -0.688 -0.771 MEWS 0.606 (0.571–0.665) 0.129 (0.098–0.159) 0.066 (0.058–0.075) 0 1 -3.120 0.221 GAP 0.863 (0.831–0.898) 0.382 (0.308–0.453) 0.053 (0.047–0.060) 0 1 2.426 -0.310 After the single-line recalibration, the ranking was consistent across summary measures. RTS performed best (AUROC 0.874, AUPRC 0.465, Brier 0.050), followed by GAP (0.863, 0.382, 0.053) and mREMS (0.856, 0.370, 0.055), with MGAP and MEWS lower on all three metrics. Apparent calibration was tight for every tool, with intercepts close to zero and slopes close to one. The fitted recalibration coefficients quantify the translation and rescaling needed to express each score as a probability: RTS α 3.729, γ − 1.015; GAP α 2.426, γ − 0.310; mREMS α − 0.688, γ − 0.771; MEWS α − 3.120, γ 0.221; MGAP α − 0.509, γ − 0.093. Complete estimates with confidence intervals are reported in Table 2 . At the prespecified 10% threshold, the per-100 summaries show how each tool trades sensitivity for alert burden: RTS yields about 5.1 true positives, 9.5 false positives, 2.1 false negatives, and 83.2 true negatives with net benefit 0.041; GAP yields 5.7, 14.7, 1.5, 78.0 with 0.041; mREMS reaches 5.9, 16.9, 1.3, 75.8 and the same net benefit but with visibly more over-triage, whereas MGAP and MEWS detect 2.9 and 1.6 deaths with 0.009 and 0.006, respectively; detailed counts are provided in Table 3 . Moving to a 20% threshold shifts emphasis to precision: RTS delivers 4.9, 9.2, 2.3, 83.6 with net benefit 0.026, and GAP 5.1, 9.9, 2.1, 82.8 with 0.026; mREMS becomes markedly selective at 2.9, 3.8, 4.4, 88.9, achieving the highest PPV (0.431) but giving up sensitivity; MEWS shows negative net benefit (− 0.001) and MGAP is near zero, as shown in Table 4 . Collectively, these profiles make the clinical trade-offs explicit: the 10% rule privileges recall while keeping NPV very high, whereas the 20% rule privileges PPV and reduces avoidable alerts at the cost of more missed early deaths; across both settings, RTS and GAP provide the most usable balance for routine triage, while mREMS suits contexts that prioritize precision over sensitivity. Table 3 Clinical utility per 100 patients at 10% threshold Score Threshold TP FP FN TN Se Sp PPV NPV Netbenefit MGAP 0.100 2.900 18.100 4.400 74.600 0.398 0.804 0.137 0.945 0.009 RTS 0.100 5.100 9.500 2.100 83.200 0.705 0.897 0.349 0.975 0.041 mREMS 0.100 5.900 16.900 1.300 75.800 0.819 0.818 0.260 0.983 0.041 MEWS 0.100 1.600 9.100 5.600 83.600 0.223 0.901 0.150 0.937 0.006 GAP 0.100 5.700 14.700 1.500 78 0.789 0.842 0.281 0.981 0.041 Table 4 Clinical utility per 100 patients at 20% threshold Score Threshold TP FP FN TN Se Sp PPV NPV Netbenefit MGAP 0.200 0.200 0.500 7 92.300 0.030 0.995 0.312 0.929 0.001 RTS 0.200 4.900 9.200 2.300 83.600 0.681 0.901 0.350 0.973 0.026 mREMS 0.200 2.900 3.800 4.400 88.900 0.398 0.959 0.431 0.953 0.019 MEWS 0.200 0.200 1.100 7 91.600 0.030 0.988 0.161 0.929 -0.001 GAP 0.200 5.100 9.900 2.100 82.800 0.705 0.893 0.340 0.975 0.026 Discussion This study asked whether a minimal, prespecified recalibration with a single intercept and a single slope can make established bedside scores provide trustworthy 24 hour risk for traumatic brain injury without rebuilding models. Across receiver operating characteristics, precision and recall under low prevalence, graphical and quantitative calibration, and per-100 consequences at prespecified thresholds, the evidence pointed to a consistent message. After recalibration, RTS and GAP provided the most reliable early triage signal, mREMS remained competitive when precision is prioritized, and MGAP and MEWS contributed little at the very early horizon. This approach aligns with guidance that recommends simple intercept and slope updating as the first step when transporting prognostic models[ 22 , 23 ]. The emphasis on precision and recall reflects best practice for rare outcomes and avoids an overly optimistic impression based on receiver operating characteristics alone[ 24 ]. Presenting discrimination, calibration, and decision consequences together follows contemporary reporting standards for clinical prediction research and supports transparent translation to bedside decisions[ 25 ]. Our findings are consistent with contemporary trauma literature that reports strong performance for scores built around physiology, particularly RTS and GAP, with mREMS close behind and more variable results for MGAP and MEWS across settings[ 26 – 28 ]. Studies that focus on early outcomes in the emergency department or within the first day tend to reproduce this pattern, whereas reports that target in-hospital mortality over a longer horizon sometimes place MGAP closer to the leaders when pupil data are scarce or when case mix shifts toward less severe brain injury[ 29 , 30 ]. Heterogeneity in outcome window, prevalence, and measurement context offers a plausible explanation. Very early death is driven by neurologic compromise and circulatory failure, which magnifies the signal from GCS, blood pressure, oxygenation, and pupil reactivity; this mechanism-level alignment helps physiology-forward scores such as RTS and GAP perform well in the first hours[ 31 ]. Data provenance also matters, since prehospital capture, intubation, sedation, and oxygen supplementation can shift distributions of GCS and SpO₂ and alter apparent risk; a brief intercept and slope update reduces these transport gaps without abandoning the original score structure[ 32 ]. Taken together with prior work, our results support deploying recalibrated physiology-based scores for near-term triage in TBI while recognizing that MGAP or MEWS may be preferred in environments where key neurologic variables are unavailable or where the decision horizon extends beyond the first day[ 33 – 35 ]. Calibration after the prespecified update was strong and clinically useful[ 36 , 37 ]. Intercepts clustered around zero, slopes were close to one, and the decile plots lay near the identity line. Under these conditions, predicted probabilities can be read as absolute risks rather than mere ranks. This supports bedside conversations, selection of escalation thresholds, and clear communication with consultants. It also reduces pressure to rebuild models and preserves the familiar structure of each score[ 38 ]. Decision thresholds translate accuracy into workload and safety. The per-100 summaries at ten and twenty percent show what is gained in additional detections and what is spent in avoidable alerts[ 39 , 40 ]. At ten percent the priority is recall and the negative predictive value remains high, which fits early triage when missing a rapidly deteriorating patient is unacceptable. At twenty percent the alert burden falls and precision rises, which can be preferable when monitored capacity is tight. RTS and GAP offer a balanced option across both thresholds, whereas mREMS can be selected when precision is the dominant concern. This study has several strengths. Data were assembled with a prespecified schema and each bedside score was reconstructed exactly as published, followed by a single transparent recalibration so that outputs remained interpretable at the bedside. To enable fair head-to-head comparison we relied on routinely available variables and a uniform analytic approach, accepting a narrower feature set to enhance comparability. The endpoint was fixed at 24 hours to match time-critical decisions and to limit downstream confounding. Anticipated heterogeneity in prevalence and measurement practice was addressed by aligning predicted and observed risk with a prespecified intercept and slope rather than altering model structure. We recognise important limitations, including conduct within one health system and period, context-sensitive measurements for consciousness and oxygenation, small numbers for penetrating injury, evaluation at only two operating points, and the need for confirmation in multicentre cohorts. Setting differences such as prehospital capture, intubation, sedation, and supplemental oxygen can shift the distributions of key variables, which the chosen update is intended to absorb. Robustness was supported by concordant signals across discrimination, precision and recall behaviour, calibration, and per-100 consequences, with uncertainty summarised by bootstrap intervals. These features guide interpretation and point to next steps. Confirmation across services with different workflows and case mix will establish transportability. Routine monitoring of calibration with small periodic updates will manage drift over time. Subgroup checks by age and mechanism and fairness assessments will strengthen credibility. Audits of threshold policies, alert burden, time to action, and missed-event narratives will show how the risk estimates translate into care. Prospective impact evaluations should prespecify clinical usefulness metrics, including decision-curve analysis to benchmark net benefit against treat-all and treat-none strategies[ 41 ]. Together, these actions provide a practical path to safe, transparent, and scalable adoption. Conclusion Calibrating each score with one intercept and one slope was sufficient to deliver trustworthy 24-hour risk for early TBI triage. Across discrimination, precision–recall, and calibration, RTS and GAP performed most reliably, mREMS favored precision, and MGAP and MEWS added little at this time horizon. Per-100-patient summaries at 10% and 20% thresholds translated model output into actionable counts that support threshold selection according to local capacity. Embedding this light recalibration with routine monitoring offers a ready path to use, and multicentre impact studies should confirm transportability and effects on workflow and safety. Declarations Human Ethics and Consent to Participate This study was conducted in accordance with the Declaration of Helsinki and received approval from the Institutional Ethics Committee of the Affiliated Hospital of Nantong University (Approval No. 2021-K084-01). Due to the retrospective nature of this study and the use of de-identified clinical data, the requirement for informed consent was waived. All procedures involved in this research complied with institutional data protection policies, ensuring confidentiality and privacy of patient data. Declaration of competing interest The authors declare no competing financial or non-financial interests. Funding This work was supported by the Wuxi Municipal Health Commission Youth Research Project (Q202323). Acknowledgements We thank the emergency department nursing staff and the trauma registry team at the Affiliated Hospital of Nantong University for their assistance with data capture and quality checks. We are grateful to the clinical informatics group and the hospital information office for support with electronic health record queries and dataset validation. The authors alone are responsible for the analyses and interpretation. Sample CRediT author statement Tianxi Chen (Co-first author): Conceptualization, Methodology, Formal analysis, Writing – Original Draft; Jinheng Tu (Co-first author): Investigation, Data Curation, Writing – Original Draft; Jiajia Liu (Co-first author): Data Curation, Writing – Original Draft; Hong Sun: Supervision, Project Administration, Funding Acquisition; Yun Lu: Resources, Writing – Review & Editing; Rui Chen: Investigation, Writing – Review & Editing; Hao Huang (Second corresponding author): Supervision, Project Administration, Funding Acquisition; Chen Shen (First corresponding author): Supervision, Writing – Review & Editing, Funding Acquisition. References Global, regional, and national burden of neurological disorders, 1990-2016: a systematic analysis for the Global Burden of Disease Study 2016. LANCET NEUROL 2019;18:459-480. Sartorius D, Le Manach Y, David JS, Rancurel E, Smail N, Thicoipe M, Wiel E, Ricard-Hibon A, Berthier F, Gueugniaud PY, Riou B. Mechanism, glasgow coma scale, age, and arterial pressure (MGAP): a new simple prehospital triage score to predict mortality in trauma patients. CRIT CARE MED 2010;38:831-837. Champion HR, Sacco WJ, Copes WS, Gann DS, Gennarelli TA, Flanagan ME. A revision of the Trauma Score. J Trauma 1989;29:623-629. Miller RT, Nazir N, McDonald T, Cannon CM. The modified rapid emergency medicine score: A novel trauma triage tool to predict in-hospital mortality. INJURY 2017;48:1870-1877. Subbe CP, Kruger M, Rutherford P, Gemmel L. Validation of a modified Early Warning Score in medical admissions. QJM-INT J MED 2001;94:521-526. Kondo Y, Abe T, Kohshi K, Tokuda Y, Cook EF, Kukita I. Revised trauma scoring system to predict in-hospital mortality in the emergency department: Glasgow Coma Scale, Age, and Systolic Blood Pressure score. CRIT CARE 2011;15:R191. Mohammed Z, Saleh Y, AbdelSalam EM, Mohammed N, El-Bana E, Hirshon JM. Evaluation of the Revised Trauma Score, MGAP, and GAP scoring systems in predicting mortality of adult trauma patients in a low-resource setting. BMC EMERG MED 2022;22:90. Moreno-Blanco D, Alonso E, Sanz-Garcia A, Aramendi E, Lopez-Izquierdo R, Perez GR, Del PVC, Martin-Rodriguez F. Spanish vs USA cohort comparison of prehospital trauma scores to predict short-term mortality. CLIN MED 2024;24:100208. Saito T, Rehmsmeier M. The precision-recall plot is more informative than the ROC plot when evaluating binary classifiers on imbalanced datasets. PLOS ONE 2015;10:e118432. Van Calster B, McLernon DJ, van Smeden M, Wynants L, Steyerberg EW. Calibration: the Achilles heel of predictive analytics. BMC MED 2019;17:230. Van Calster B, Wynants L, Verbeek J, Verbakel JY, Christodoulou E, Vickers AJ, Roobol MJ, Steyerberg EW. Reporting and Interpreting Decision Curve Analysis: A Guide for Investigators. EUR UROL 2018;74:796-804. Jeong JH, Park YJ, Kim DH, Kim TY, Kang C, Lee SH, Lee SB, Kim SC, Lim D. The new trauma score (NTS): a modification of the revised trauma score for better trauma mortality prediction. BMC SURG 2017;17:77. Gill MR, Reiley DG, Green SM. Interrater reliability of Glasgow Coma Scale scores in the emergency department. ANN EMERG MED 2004;43:215-223. Van Calster B, Steyerberg EW, Wynants L, van Smeden M. There is no such thing as a validated prediction model. BMC MED 2023;21:70. Moons KG, Altman DG, Reitsma JB, Ioannidis JP, Macaskill P, Steyerberg EW, Vickers AJ, Ransohoff DF, Collins GS. Transparent Reporting of a multivariable prediction model for Individual Prognosis or Diagnosis (TRIPOD): explanation and elaboration. ANN INTERN MED 2015;162:W1-W73. Vickers AJ, van Calster B, Steyerberg EW. A simple, step-by-step guide to interpreting decision curve analysis. Diagn Progn Res 2019;3:18. Tseng TE, Lee CC, Yen HK, Groot OQ, Hou CH, Lin SY, Bongers M, Hu MH, Karhade AV, Ko JC, Lai YH, Yang JJ, Verlaan JJ, Yang RS, Schwab JH, Lin WH. International Validation of the SORG Machine-learning Algorithm for Predicting the Survival of Patients with Extremity Metastases Undergoing Surgical Treatment. CLIN ORTHOP RELAT R 2022;480:367-378. Joksic-Mazinjanin R, Maric N, Duricin A, Gojkovic Z, Vasovic V, Rakic G, Joksic-Zelic M, Saravolac S. Prehospital Trauma Scoring Systems for Evaluation of Trauma Severity and Prediction of Outcomes. MEDICINA-LITHUANIA 2023;59. Geeraerts T, Velly L, Abdennour L, Asehnoune K, Audibert G, Bouzat P, Bruder N, Carrillon R, Cottenceau V, Cotton F, Courtil-Teyssedre S, Dahyot-Fizelier C, Dailler F, David JS, Engrand N, Fletcher D, Francony G, Gergele L, Ichai C, Javouhey E, Leblanc PE, Lieutaud T, Meyer P, Mirek S, Orliaguet G, Proust F, Quintard H, Ract C, Srairi M, Tazarourte K, Vigue B, Payen JF. Management of severe traumatic brain injury (first 24hours). ANAESTH CRIT CARE PA 2018;37:171-186. Collins GS, Reitsma JB, Altman DG, Moons KG. Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD): the TRIPOD statement. BMJ-BRIT MED J 2015;350:g7594. GANDEVIA B, TOVELL A. DECLARATION OF HELSINKI. MED J AUSTRALIA 1964;2:320-321. Steyerberg EW, Borsboom GJ, van Houwelingen HC, Eijkemans MJ, Habbema JD. Validation and updating of predictive logistic regression models: a study on sample size and shrinkage. STAT MED 2004;23:2567-2586. Gehringer CK, Martin GP, Van Calster B, Hyrich KL, Verstappen S, Sergeant JC. How to develop, validate, and update clinical prediction models using multinomial logistic regression. J CLIN EPIDEMIOL 2024;174:111481. Liu S, Roemer F, Ge Y, Bedrick EJ, Li ZM, Guermazi A, Sharma L, Eaton C, Hochberg MC, Hunter DJ, Nevitt MC, Wirth W, Kent KC, Sun X. Comparison of evaluation metrics of deep learning for imbalanced imaging data in osteoarthritis studies. OSTEOARTHR CARTILAGE 2023;31:1242-1248. Collins GS, Moons K, Dhiman P, Riley RD, Beam AL, Van Calster B, Ghassemi M, Liu X, Reitsma JB, van Smeden M, Boulesteix AL, Camaradou JC, Celi LA, Denaxas S, Denniston AK, Glocker B, Golub RM, Harvey H, Heinze G, Hoffman MM, Kengne AP, Lam E, Lee N, Loder EW, Maier-Hein L, Mateen BA, McCradden MD, Oakden-Rayner L, Ordish J, Parnell R, Rose S, Singh K, Wynants L, Logullo P. TRIPOD+AI statement: updated guidance for reporting clinical prediction models that use regression or machine learning methods. BMJ-BRIT MED J 2024;385:e78378. Merchant A, Shaukat N, Ashraf N, Hassan S, Jarrar Z, Abbasi A, Ahmed T, Atiq H, Khan UR, Khan NU, Mushtaq S, Rasul S, Hyder AA, Razzak J, Haider AH. Which curve is better? A comparative analysis of trauma scoring systems in a South Asian country. TRAUMA SURG ACUTE CA 2023;8:e1171. Rio T, Nogueira LS, Lima FR, Cassiano C, Garcia D. Performance of severity indices for admission and mortality of trauma patients in the intensive care unit: a retrospective cohort study. EUR J MED RES 2023;28:559. Luo X, Gao H, Yu X, Jiang Z, Yang W. Spectral analysis of heart rate variability for trauma outcome prediction: an analysis of 210 ICU multiple trauma patients. EUR J TRAUMA EMERG S 2021;47:153-160. Kenarangi T, Rahmani F, Yazdani A, Ahmadi GD, Lotfi M, Khalaj TA. Comparison of GAP, R-GAP, and new trauma score (NTS) systems in predicting mortality of traffic accidents that injure hospitals at Mashhad University of medical sciences. HELIYON 2024;10:e36004. Phunghassaporn N, Sukhvibul P, Techapongsatorn S, Tansawet A. Accuracy and external validation of the modified rapid emergency medicine score in road traffic injuries in a Bangkok level I trauma center. HELIYON 2022;8:e12225. Hawryluk G, Lulla A, Bell R, Jagoda A, Mangat HS, Bobrow BJ, Ghajar J. Guidelines for Prehospital Management of Traumatic Brain Injury 3rd Edition: Executive Summary. NEUROSURGERY 2023;93:e159-e169. Efthimiou O, Seo M, Chalkou K, Debray T, Egger M, Salanti G. Developing clinical prediction models: a step-by-step guide. BMJ-BRIT MED J 2024;386:e78276. Yousefi MR, Karajizadeh M, Ghasemian M, Paydar S. Comparing NEWS2, TRISS, and RTS in predicting mortality rate in trauma patients based on prehospital data set: a diagnostic study. BMC EMERG MED 2024;24:163. Candel B, Nissen SK, Nickel CH, Raven W, Thijssen W, Gaakeer MI, Lassen AT, Brabrand M, Steyerberg EW, de Jonge E, de Groot B. Development and External Validation of the International Early Warning Score for Improved Age- and Sex-Adjusted In-Hospital Mortality Prediction in the Emergency Department. CRIT CARE MED 2023;51:881-891. Donoso CM, Mordillo-Mateos L, Martin-Conty JL, Polonio-Lopez B, Lopez-Gonzalez A, Durantez-Fernandez C, Vinuela A, Rodriguez HM, Mohedano-Moriano A, Lopez-Izquierdo R, Jorge SC, Martin-Rodriguez F. Modified Rapid Emergency Medicine Score-Lactate (mREMS-L) performance to screen non-anticipated 30-day-related-mortality in emergency department. EUR J CLIN INVEST 2023;53:e13994. Liou L, Scott E, Parchure P, Ouyang Y, Egorova N, Freeman R, Hofer IS, Nadkarni GN, Timsina P, Kia A, Levin MA. Assessing calibration and bias of a deployed machine learning malnutrition prediction model within a large healthcare system. NPJ DIGIT MED 2024;7:149. Riley RD, Archer L, Snell K, Ensor J, Dhiman P, Martin GP, Bonnett LJ, Collins GS. Evaluation of clinical prediction models (part 2): how to undertake an external validation study. BMJ-BRIT MED J 2024;384:e74820. Kappen TH, van Loon K, Kappen MA, van Wolfswinkel L, Vergouwe Y, van Klei WA, Moons KG, Kalkman CJ. Barriers and facilitators perceived by physicians when using prediction models in practice. J CLIN EPIDEMIOL 2016;70:136-145. Piovani D, Sokou R, Tsantes AG, Vitello AS, Bonovas S. Optimizing Clinical Decision Making with Decision Curve Analysis: Insights for Clinical Investigators. HEALTHCARE-BASEL 2023;11. Vickers AJ, Van Claster B, Wynants L, Steyerberg EW. Decision curve analysis: confidence intervals and hypothesis testing for net benefit. Diagn Progn Res 2023;7:11. Vasey B, Nagendran M, Campbell B, Clifton DA, Collins GS, Denaxas S, Denniston AK, Faes L, Geerts B, Ibrahim M, Liu X, Mateen BA, Mathur P, McCradden MD, Morgan L, Ordish J, Rogers C, Saria S, Ting D, Watkinson P, Weber W, Wheatstone P, McCulloch P. Reporting guideline for the early-stage clinical evaluation of decision support systems driven by artificial intelligence: DECIDE-AI. NAT MED 2022;28:924-933. Additional Declarations No competing interests reported. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8452508","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":571476584,"identity":"4cea83d1-5163-421b-8dc9-139e76b464ec","order_by":0,"name":"Tianxi Chen","email":"","orcid":"","institution":"Department of Emergency Medicine, Nantong University Affiliated Hospital","correspondingAuthor":false,"prefix":"","firstName":"Tianxi","middleName":"","lastName":"Chen","suffix":""},{"id":571476586,"identity":"3ca10714-29ea-403d-b551-d73bcc385e66","order_by":1,"name":"Jinheng Tu","email":"","orcid":"","institution":"Department of Emergency Medicine, Nantong University Affiliated Hospital","correspondingAuthor":false,"prefix":"","firstName":"Jinheng","middleName":"","lastName":"Tu","suffix":""},{"id":571476588,"identity":"de65701f-132c-45f1-b08c-e16ab8aad235","order_by":2,"name":"Jiajia Liu","email":"","orcid":"","institution":"Department of Emergency Medicine, Nantong University Affiliated Hospital","correspondingAuthor":false,"prefix":"","firstName":"Jiajia","middleName":"","lastName":"Liu","suffix":""},{"id":571476590,"identity":"7c4e0bc3-36e7-485e-87c5-f92a6a477411","order_by":3,"name":"Hong sun","email":"","orcid":"","institution":"Nursing Department of Nantong University Affiliated Hospital","correspondingAuthor":false,"prefix":"","firstName":"Hong","middleName":"","lastName":"sun","suffix":""},{"id":571476592,"identity":"7a1233c2-d88e-498c-adad-e638f0ecfeff","order_by":4,"name":"Yun Lu","email":"","orcid":"","institution":"Department of Emergency Medicine, Nantong University Affiliated Hospital","correspondingAuthor":false,"prefix":"","firstName":"Yun","middleName":"","lastName":"Lu","suffix":""},{"id":571476594,"identity":"d7214400-384c-40f0-b3d7-06481b8c835b","order_by":5,"name":"Rui Chen","email":"","orcid":"","institution":"Department of Emergency Medicine, Nantong University Affiliated Hospital","correspondingAuthor":false,"prefix":"","firstName":"Rui","middleName":"","lastName":"Chen","suffix":""},{"id":571476595,"identity":"1607efa4-d45d-43d9-9454-d723fd55eaa2","order_by":6,"name":"Huang Hao","email":"","orcid":"","institution":"Nursing Department, The Affliated Wuxi People's Hospital of Nanjing Medical University","correspondingAuthor":false,"prefix":"","firstName":"Huang","middleName":"","lastName":"Hao","suffix":""},{"id":571476596,"identity":"c4fea402-9d2b-4ac5-9b75-0b8668e526bf","order_by":7,"name":"Chen Shen","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA0ElEQVRIie3QMQrCQBBA0UmjzYCWEyJ7A2EgYCV6FRdBGwtLywRh0+QGnsJGsduwoE0OoNi4BxBSWoiYWpCsncX+et7ADIDP94cRQKBhRTjO0tRWjgQ0lEPBaNYxOZNAzWKmueqiCwmz9U0/WkbuQ6vqBSPRTxpIhEcucjTysJHqtoRpPNANRNCENZKRyVVmTKDlrpnMq+LJNbkUitCFRLRgg5P6/HPgSMK8XJqeHoowl/WT2eEWOmVbe38Rdtona6vVSDSSj/i3cZ/P5/N96Q1FHUfVenxHswAAAABJRU5ErkJggg==","orcid":"","institution":"Nursing Department of Nantong University Affiliated Hospital","correspondingAuthor":true,"prefix":"","firstName":"Chen","middleName":"","lastName":"Shen","suffix":""}],"badges":[],"createdAt":"2025-12-26 06:08:44","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8452508/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8452508/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":100012916,"identity":"4584c807-846c-4b46-a683-9baaf5ae3abc","added_by":"auto","created_at":"2026-01-12 06:17:13","extension":"doc","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":8406528,"visible":true,"origin":"","legend":"","description":"","filename":"TBI.doc","url":"https://assets-eu.researchsquare.com/files/rs-8452508/v1/dd834a08998b3d708182076d.doc"},{"id":100361686,"identity":"8ee6e99d-6a7d-4d2e-a9b7-677d072b97c5","added_by":"auto","created_at":"2026-01-16 07:45:30","extension":"tif","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":3802544,"visible":true,"origin":"","legend":"","description":"","filename":"Fig1.tif","url":"https://assets-eu.researchsquare.com/files/rs-8452508/v1/f27d3c82c2ebf36c9f3c7ae2.tif"},{"id":100361740,"identity":"38c3a2ef-d31a-417c-a55f-a00253dfcb21","added_by":"auto","created_at":"2026-01-16 07:45:40","extension":"png","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":946312,"visible":true,"origin":"","legend":"","description":"","filename":"Fig2.png","url":"https://assets-eu.researchsquare.com/files/rs-8452508/v1/c46f7cc65d4bb8b704d64234.png"},{"id":100012907,"identity":"a369f43f-e79a-46a1-b500-f5fab2e249fb","added_by":"auto","created_at":"2026-01-12 06:17:13","extension":"docx","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":19654,"visible":true,"origin":"","legend":"","description":"","filename":"Table.docx","url":"https://assets-eu.researchsquare.com/files/rs-8452508/v1/fb8741ad164a605407cee990.docx"},{"id":100362051,"identity":"f958d1ed-246b-4a93-a9d2-6fa8e266661c","added_by":"auto","created_at":"2026-01-16 07:46:07","extension":"png","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":585002,"visible":true,"origin":"","legend":"","description":"","filename":"Fig3.png","url":"https://assets-eu.researchsquare.com/files/rs-8452508/v1/632201b8d84210a245557558.png"},{"id":100012914,"identity":"08bf8cc4-3fa1-4a66-81f7-14121b7a6d45","added_by":"auto","created_at":"2026-01-12 06:17:13","extension":"png","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":536584,"visible":true,"origin":"","legend":"","description":"","filename":"Fig4.png","url":"https://assets-eu.researchsquare.com/files/rs-8452508/v1/f96f71b73132abaa52d8f85c.png"},{"id":100361795,"identity":"740485c4-fb19-4f53-9c32-1d8ea36bd1b0","added_by":"auto","created_at":"2026-01-16 07:45:44","extension":"json","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":10125,"visible":true,"origin":"","legend":"","description":"","filename":"da3579906caf4af1bf88f2a0bf480c06.json","url":"https://assets-eu.researchsquare.com/files/rs-8452508/v1/aaf50129c85382f61a3d0a1b.json"},{"id":100362009,"identity":"e083abd4-ebf5-48c7-bf72-2ce26ef426dd","added_by":"auto","created_at":"2026-01-16 07:46:04","extension":"xml","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":163446,"visible":true,"origin":"","legend":"","description":"","filename":"da3579906caf4af1bf88f2a0bf480c061enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-8452508/v1/f6fe0bf4f8969425c625292b.xml"},{"id":100012936,"identity":"9cd38691-d855-4086-8d47-bbae12850d29","added_by":"auto","created_at":"2026-01-12 06:17:14","extension":"tif","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":3802544,"visible":true,"origin":"","legend":"","description":"","filename":"Fig1.tif","url":"https://assets-eu.researchsquare.com/files/rs-8452508/v1/49cf0d2946b4df3544c38ec9.tif"},{"id":100012923,"identity":"8f4407d6-14f9-422c-88bc-3fc37203d791","added_by":"auto","created_at":"2026-01-12 06:17:13","extension":"png","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":946312,"visible":true,"origin":"","legend":"","description":"","filename":"Fig2.png","url":"https://assets-eu.researchsquare.com/files/rs-8452508/v1/fe2fe3301206fdeb4b41bc2c.png"},{"id":100012919,"identity":"08b8072c-a31f-45eb-b455-626aaef88a7f","added_by":"auto","created_at":"2026-01-12 06:17:13","extension":"png","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":585002,"visible":true,"origin":"","legend":"","description":"","filename":"Fig3.png","url":"https://assets-eu.researchsquare.com/files/rs-8452508/v1/a85b8d37a0e940417675a882.png"},{"id":100362191,"identity":"eb89c0f7-2bd5-4fce-a16c-0ff9303a71f4","added_by":"auto","created_at":"2026-01-16 07:46:17","extension":"png","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":536584,"visible":true,"origin":"","legend":"","description":"","filename":"Fig4.png","url":"https://assets-eu.researchsquare.com/files/rs-8452508/v1/4eb3b11615fb55d5e84cbbe4.png"},{"id":100012932,"identity":"4d1d0692-a413-436c-9e3c-1b48b742a488","added_by":"auto","created_at":"2026-01-12 06:17:14","extension":"eps","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":403,"visible":true,"origin":"","legend":"","description":"","filename":"drawingimage1.eps","url":"https://assets-eu.researchsquare.com/files/rs-8452508/v1/71e3a94303a0b3a4dc638d53.eps"},{"id":100012925,"identity":"d0663281-cbbf-47b2-abc0-09a85bf619c7","added_by":"auto","created_at":"2026-01-12 06:17:13","extension":"eps","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":403,"visible":true,"origin":"","legend":"","description":"","filename":"drawingimage2.eps","url":"https://assets-eu.researchsquare.com/files/rs-8452508/v1/ba82aafe9f362c64da1efe69.eps"},{"id":100012921,"identity":"1e471f0d-64f6-46ac-9eca-45380c4a74e9","added_by":"auto","created_at":"2026-01-12 06:17:13","extension":"png","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1465695,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8452508/v1/e6606e8935f8f376313d1f63.png"},{"id":100012934,"identity":"c54079f9-2362-4556-9c96-3fa6b0ac93b7","added_by":"auto","created_at":"2026-01-12 06:17:14","extension":"png","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1805340,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8452508/v1/2129125913112a3fb6f7b3dc.png"},{"id":100361714,"identity":"a631f01e-1353-4ac0-be43-8238df5842e5","added_by":"auto","created_at":"2026-01-16 07:45:34","extension":"png","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1034152,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-8452508/v1/03e12aa6765884ca58dd8d0b.png"},{"id":100361891,"identity":"774e1ef9-4dc3-4a2c-90c3-959a9e9d7864","added_by":"auto","created_at":"2026-01-16 07:45:54","extension":"png","order_by":17,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1264303,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-8452508/v1/1c6b4eecd194a847b68e0832.png"},{"id":100012937,"identity":"06df877c-d905-4069-b425-363269f54115","added_by":"auto","created_at":"2026-01-12 06:17:14","extension":"png","order_by":18,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":2149688,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFig1.png","url":"https://assets-eu.researchsquare.com/files/rs-8452508/v1/dfbc0a4b314d933329f02fd4.png"},{"id":100012922,"identity":"4991e455-375f-43ab-9a17-89b1d3889754","added_by":"auto","created_at":"2026-01-12 06:17:13","extension":"png","order_by":19,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":222649,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFig2.png","url":"https://assets-eu.researchsquare.com/files/rs-8452508/v1/5fb2280c1aae87c0384deb1c.png"},{"id":100361234,"identity":"ae569a9e-9e3a-4269-b729-d7fbedc3dd05","added_by":"auto","created_at":"2026-01-16 07:44:44","extension":"png","order_by":20,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":147027,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFig3.png","url":"https://assets-eu.researchsquare.com/files/rs-8452508/v1/a1d7bc95a28fb1069f265a6a.png"},{"id":100012927,"identity":"af08f00c-bdc2-40ad-a69d-d7dfab99577c","added_by":"auto","created_at":"2026-01-12 06:17:14","extension":"png","order_by":21,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":112281,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFig4.png","url":"https://assets-eu.researchsquare.com/files/rs-8452508/v1/9a8c19bfcb0ec36ddfaec1e2.png"},{"id":100012930,"identity":"a5856a30-657d-4d3b-a7f2-c56076160f0d","added_by":"auto","created_at":"2026-01-12 06:17:14","extension":"png","order_by":22,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":229503,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8452508/v1/64ca38b47f3f1520752759f3.png"},{"id":100361445,"identity":"ce0b89f9-5a63-4c81-805a-f97f2cd11f5e","added_by":"auto","created_at":"2026-01-16 07:45:11","extension":"png","order_by":23,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":571929,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8452508/v1/657274b2dc661e2eed1dc0e1.png"},{"id":100012926,"identity":"203f498a-a717-4425-bbd3-11cb0e25075b","added_by":"auto","created_at":"2026-01-12 06:17:14","extension":"png","order_by":24,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":370952,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-8452508/v1/d15543cd9afed8f8686f3ef9.png"},{"id":100360923,"identity":"f0f1ec1b-5eb3-47d9-8c73-c6b8b357b0c9","added_by":"auto","created_at":"2026-01-16 07:44:11","extension":"png","order_by":25,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":384010,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-8452508/v1/b1aabc1ab6d258aa7fbdaf56.png"},{"id":100361989,"identity":"64721709-d8ce-46cd-8e3d-0232b3529399","added_by":"auto","created_at":"2026-01-16 07:46:01","extension":"xml","order_by":26,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":164172,"visible":true,"origin":"","legend":"","description":"","filename":"da3579906caf4af1bf88f2a0bf480c061structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-8452508/v1/c485e59b3298dad654c2b7b0.xml"},{"id":100012939,"identity":"fa6d12bb-6f10-4a51-a6fb-5556321f1c21","added_by":"auto","created_at":"2026-01-12 06:17:15","extension":"html","order_by":27,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":176957,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-8452508/v1/ec6faef8d20d8cd4a6048281.html"},{"id":100012909,"identity":"2d543b1e-14d3-4d59-b9aa-efec744355f3","added_by":"auto","created_at":"2026-01-12 06:17:13","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":224684,"visible":true,"origin":"","legend":"\u003cp\u003ePatient selection and enrollment flowchart\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-8452508/v1/939591d2cf7be60a627e0969.png"},{"id":100361519,"identity":"e3fd246e-9ed7-4325-8c4a-337e07efa4ad","added_by":"auto","created_at":"2026-01-16 07:45:15","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":146332,"visible":true,"origin":"","legend":"\u003cp\u003eROC curves for TBI bedside scores 24 hour mortality\u003c/p\u003e\n\u003cp\u003eNotes: ROC; diagonal = no-discrimination.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-8452508/v1/8bee3f42f64504023d050eaa.png"},{"id":100362118,"identity":"a48fae68-01e0-41dd-8ee6-4716f359b1ff","added_by":"auto","created_at":"2026-01-16 07:46:12","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":98709,"visible":true,"origin":"","legend":"\u003cp\u003ePrecision recall curves with 10 and 20 percent thresholds 24 hour mortality\u003c/p\u003e\n\u003cp\u003eNotes: AUPRC=average precision; baseline=prevalence 0.073; thresholds at predicted risk 10%/20%.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-8452508/v1/5f2d5dedd7607c9b796b0edc.png"},{"id":100012912,"identity":"be48e235-27a4-4ba8-b4c7-afb75a868ae4","added_by":"auto","created_at":"2026-01-12 06:17:13","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":95385,"visible":true,"origin":"","legend":"\u003cp\u003eCalibration plots observed versus predicted 24 hour mortality\u003c/p\u003e\n\u003cp\u003eNotes: Loess/Gaussian smoothing with bandwidth; 95% CI from bootstrap; deciles show observed mortality with Wilson 95% CI.\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-8452508/v1/38a11b94d7c70ddcc6cfae39.png"},{"id":100381304,"identity":"2dc0bfb9-ff48-4a02-80e6-6847507019f2","added_by":"auto","created_at":"2026-01-16 10:38:04","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1202941,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8452508/v1/010d459e-a589-40d1-87d2-1ddf80390c14.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Comparative Validation of Five Early Trauma Scores for 24-Hour In-Hospital Mortality in Traumatic Brain Injury","fulltext":[{"header":"Introduction","content":"\u003cp\u003eTraumatic brain injury is a major driver of early in-hospital deterioration and death worldwide, imposing substantial clinical and system burden[\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. Early risk stratification in the first hours of care informs monitoring intensity, escalation decisions, and interfacility transfer, and is central to outcomes-focused acute management. Simple bedside scores, including RTS, GAP, mREMS, MGAP, and MEWS, are attractive because they use routinely collected variables and can be calculated rapidly at the point of care[\u003cspan additionalcitationids=\"CR3 CR4 CR5\" citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. However, despite widespread use, contemporary head-to-head analyses yield mixed results for RTS, GAP, mREMS, MGAP, and MEWS on short-term mortality[\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e], while quantitative assessment of decision utility at clinically relevant thresholds is infrequently provided.\u003c/p\u003e \u003cp\u003eRecent evaluations still leave practical questions unresolved. Many enroll mixed trauma rather than TBI-specific cohorts and emphasize outcomes beyond the first 24 hours, which limits applicability to early in-hospital mortality. AUROC is widely reported and remains informative for ranking performance, yet low event rates call for complementary indices that capture rare-outcome behavior and absolute risk, including AUPRC referenced to prevalence, Brier score, and calibration intercept and slope[\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. Decision-analytic reporting is uncommon, so threshold-specific trade-offs, consequences per 100 patients, and net benefit that guide escalation or transfer are rarely presented[\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. Definitions and implementations of RTS, GAP, mREMS, MGAP, and MEWS vary across prehospital and in-hospital settings, including handling of Glasgow Coma Scale components, systolic blood pressure cut points, and oxygen saturation capture, which complicates transportability[\u003cspan additionalcitationids=\"CR13\" citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. Cohorts differ in time zero, inclusion criteria, and 24-hour outcome ascertainment, and missing data are frequently managed with ad hoc rules rather than prespecified strategies, both of which can shift apparent ranking[\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]. Sample sizes are often modest relative to the low event rate, increasing uncertainty around precision and recall performance and model calibration. Few studies report locally calibrated absolute risks or apply a simple recalibration step before comparison, despite known intercept and slope drift across settings. Threshold selection is seldom linked to operational constraints such as monitored bed capacity, neurosurgical availability, or transfer logistics, limiting bedside uptake. Together, these gaps leave clinicians without clear, calibration-aware guidance on which score to prioritize for early decision making in contemporary practice.\u003c/p\u003e \u003cp\u003eBedside tools should align with rapid assessment workflows, communicate absolute risk, and support early triage and interfacility transfer within the first hours of care. Clinicians and administrators benefit from threshold-specific summaries that show consequences per 100 patients and net benefit rather than rank-based metrics alone[\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]. Simple recalibration to local prevalence and measurement practices is preferable to refitting because it preserves interpretability and improves transportability when calibration drifts across settings[\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]. Using routinely collected variables enhances feasibility across prehospital and emergency contexts and supports consistent handoffs[\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. Focusing on a 24-hour endpoint targets safety-critical deterioration and provides a common trigger for escalation pathways and transfer decisions[\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eThis work delivers a head-to-head evaluation of RTS, GAP, mREMS, MGAP, and MEWS for predicting 24-hour in-hospital mortality in contemporary TBI care. Each score is aligned to local risk using a single-equation logistic recalibration to generate calibrated probabilities suitable for bedside decisions. Performance is summarized with AUROC, AUPRC referenced to prevalence, the Brier score, and calibration (intercept, slope); decision performance is quantified at 10% and 20% probability thresholds using consequences per 100 patients and net benefit, with uncertainty expressed as 95% confidence intervals from 1,000 bootstrap resamples. Collectively, these analyses provide calibration-aware, threshold-focused guidance for selecting first-line bedside tools for early risk stratification.\u003c/p\u003e"},{"header":"Materials and methods","content":"\u003cp\u003eGuidelines\u003c/p\u003e\n\u003cp\u003eThis study was designed and reported in accordance with the TRIPOD statement and relevant EQUATOR Network guidance[20]. The protocol received approval from the institutional ethics committee (approval No. 2021-K084-01). Owing to the retrospective design and use of de-identified clinical data, the requirement for informed consent was waived. All procedures complied with the Declaration of Helsinki and institutional data-protection policies[21].\u003c/p\u003e\n\u003cp\u003eData source\u003c/p\u003e\n\u003cp\u003eWe drew data from the integrated medical database of a tertiary medical center in Nantong, China. Consecutive admissions within the study window (January 1–December 31, 2024) were screened against prespecified criteria. Eligible cases were adults (≥18 years) with an index encounter during the study period, an ascertainable 24-hour in-hospital outcome, and sufficient variables to compute RTS, GAP, mREMS, MGAP, and MEWS for validation.\u003c/p\u003e\n\u003cp\u003eOf 3,064 admissions initially assessed, 777 were excluded for the following reasons: age \u0026lt;18 years (n=47); non-traumatic intracranial pathologies such as stroke, tumor, or infection (n=213); elective neurosurgical admissions unrelated to acute traumatic injury (n=102); duplicate or interfacility transfer episodes, with only the first index encounter retained (n=131); transfer out before 24 hours with unknown outcome (n=58); and missing key variables or the 24-hour outcome (n=226). The final analytic cohort comprised 2,287 patients. Details of screening and exclusions are shown in Figure 1 (Patient selection and enrollment flowchart).\u003c/p\u003e\n\u003cp\u003eOutcome\u003c/p\u003e\n\u003cp\u003eThe primary outcome was 24-hour all-cause in-hospital mortality during the index encounter. Deaths were ascertained from structured fields in the electronic health record using the in-hospital death indicator and, when available, the recorded time of death. Patients who were alive at 24 hours, including those discharged before 24 hours, were classified as non-events. Encounters transferred out before 24 hours with unknown vital status, and records lacking a verifiable 24-hour outcome, were not included in the analytic cohort.\u003c/p\u003e\n\u003cp\u003eData collection and variables\u003c/p\u003e\n\u003cp\u003eData were abstracted at the index encounter from the integrated electronic health record using a prespecified schema. Baseline fields included sex; age (years) with a derived age band for score computation; temperature (°C); heart rate (beats per minute); respiratory rate (breaths per minute); systolic and diastolic blood pressure (mmHg); peripheral oxygen saturation (SpO₂, %); mechanism of injury coded as penetrating or blunt; pupil reactivity recorded in three categories (no reaction, unilateral reaction, bilateral reaction); the Glasgow Coma Scale (GCS, 3–15). The analytic endpoint was 24-hour all-cause in-hospital mortality (Death24h). Bedside scores (RTS, GAP, mREMS, MGAP, MEWS) were taken from the dataset or, when components were available, reconstructed exactly according to published definitions with unit harmonization and prespecified cut points; higher totals for RTS and GAP indicate lower risk, whereas higher mREMS and MEWS indicate greater risk. When multiple measurements were available around arrival, the earliest nonmissing value within the initial assessment window was used. Implausible entries were flagged and set to missing before calculation; records lacking a verifiable 24-hour outcome or a key variable had been removed during cohort construction.\u003c/p\u003e\n\u003cp\u003eMissing data\u003c/p\u003e\n\u003cp\u003ePrior to analysis, encounters with an unknown 24-hour vital status or missing components required to compute any bedside score were removed at cohort construction. This yielded an analytic set of 2,287 patients with no missing outcome and complete values for RTS, GAP, mREMS, MGAP, and MEWS. No statistical imputation was performed. For baseline characteristics used only for description (e.g., demographics, vital signs, pupil reactivity, mechanism of injury, and GCS), we report available-case summaries with the denominator explicitly stated for each variable. Implausible entries were reviewed and set to missing prior to tabulation. All performance analyses (recalibration, discrimination, calibration, and threshold-based metrics) were conducted on the complete-case analytic set and were bootstrapped using case resampling.\u003c/p\u003e\n\u003cp\u003eData analysis\u003c/p\u003e\n\u003cp\u003eEach bedside score was recalibrated with a single-equation logistic recalibration model to obtain predicted probabilities for evaluation. Baseline characteristics were summarized descriptively as mean (standard deviation) or median (interquartile range) for continuous variables, and count (percentage) for categorical variables; no hypothesis testing or p values were reported. Performance was examined for discrimination, calibration, overall performance, and decision analysis at prespecified probability thresholds of 10% and 20%. Graphical displays included ROC curves, precision–recall curves with the cohort prevalence shown as a horizontal reference, and calibration plots based on deciles with a 45° reference line. Statistical uncertainty was quantified with 1,000 patient-level bootstrap resamples using two-sided 95% confidence intervals by the percentile method (resampling stratified by outcome). All analyses were conducted in Python 3.9 using pandas, NumPy, scikit-learn, statsmodels, and Matplotlib.\u003c/p\u003e\n\u003cp\u003ePerformance metrics\u003c/p\u003e\n\u003cp\u003ePerformance was defined a priori across discrimination, calibration, overall performance, and threshold-specific clinical utility. Discrimination was quantified with the area under the receiver operating characteristic curve (AUROC) and the area under the precision–recall curve (AUPRC) using average precision to remain informative under class imbalance. Calibration was summarized by the calibration intercept and slope from a logistic calibration model, where values near 0 and 1 indicate minimal systematic bias and appropriate risk spread, respectively. Overall performance was characterized by the Brier score as a measure of probabilistic accuracy, with lower values indicating better performance. Clinical utility was evaluated at prespecified probability cutoffs of 10% and 20% using per 100 patients summaries that report true positives, false positives, false negatives, and true negatives together with sensitivity, specificity, positive predictive value, negative predictive value, and net benefit relative to treat-all and treat-none strategies. All estimates are accompanied by two-sided 95% confidence intervals obtained from 1,000 patient-level bootstrap resamples using the percentile method.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003eWe enrolled 2,287 TBI encounters; 166 patients died within 24 hours (7.3%) and 2,121 survived. The cohort was predominantly male (1,479 of 2,287; 64.7%) and injuries were almost exclusively blunt trauma (2,273 of 2,287; 99.4%). Median age was 62 years (IQR 52 to 71). At presentation, vital signs were generally within physiologic ranges, yet the death group showed a pattern of higher heart rate and blood pressure, together with lower peripheral oxygen saturation and lower GCS. Pupillary reactivity distinguished the groups most strongly: nonreactive pupils were present in 82 of 166 deaths versus 53 of 2,121 survivors (49.4% vs 2.5%), unilateral reaction in 40 of 166 vs 82 of 2,121 (24.1% vs 3.9%), and bilateral reaction in 44 of 166 vs 1,986 of 2,121 (26.5% vs 93.6%). The magnitude of imbalance by standardized difference was largest for pupillary status (approximately 1.99 for absent, 0.90 for unilateral, and 2.13 for bilateral response), far exceeding that of age or other vital signs. Full baseline characteristics, including continuous summaries for age, vital signs, and GCS, categorical distributions for sex, injury mechanism, and pupil status, and context-setting distributions of the bedside scores (RTS, GAP, mREMS, MGAP, MEWS), are reported in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eBaseline characteristics (Survived\u0026thinsp;\u0026le;\u0026thinsp;24h vs. Died\u0026thinsp;\u0026le;\u0026thinsp;24h)\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVariable\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eOverall (n\u0026thinsp;=\u0026thinsp;2287)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSurvived\u0026thinsp;\u0026le;\u0026thinsp;24h (n\u0026thinsp;=\u0026thinsp;2121)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eDied\u0026thinsp;\u0026le;\u0026thinsp;24h (n\u0026thinsp;=\u0026thinsp;166)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eStd diff\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDEMOGRAPHICS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAge, years (median [IQR]; mean\u0026thinsp;\u0026plusmn;\u0026thinsp;SD)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e62.0 [51.5\u0026ndash;71.0]; 60.2\u0026thinsp;\u0026plusmn;\u0026thinsp;15.7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e62.0 [51.0\u0026ndash;71.0]; 60.0\u0026thinsp;\u0026plusmn;\u0026thinsp;15.8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e65.5 [54.2\u0026ndash;71.0]; 63.1\u0026thinsp;\u0026plusmn;\u0026thinsp;14.0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.20\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSex, n (%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e└ Female\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e808 (35.3%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e759 (35.8%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e49 (29.5%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.13\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e└ Male\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1479 (64.7%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1362 (64.2%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e117 (70.5%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.13\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMechanism of injury, n (%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e└ Penetrating\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e14 (0.6%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e12 (0.6%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e2 (1.2%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.08\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e└ Blunt\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2273 (99.4%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2109 (99.4%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e164 (98.8%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.08\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVITAL SIGNS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTemperature (T), \u0026deg;C (median [IQR]; mean\u0026thinsp;\u0026plusmn;\u0026thinsp;SD)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e36.5 [36.3\u0026ndash;36.6]; 36.5\u0026thinsp;\u0026plusmn;\u0026thinsp;0.4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e36.5 [36.3\u0026ndash;36.6]; 36.5\u0026thinsp;\u0026plusmn;\u0026thinsp;0.3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e36.5 [36.0\u0026ndash;36.6]; 36.4\u0026thinsp;\u0026plusmn;\u0026thinsp;0.4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.15\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHeart rate (P), beats/min (median [IQR]; mean\u0026thinsp;\u0026plusmn;\u0026thinsp;SD)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e81 [71\u0026ndash;94]; 83.2\u0026thinsp;\u0026plusmn;\u0026thinsp;18.1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e81 [71\u0026ndash;93]; 82.9\u0026thinsp;\u0026plusmn;\u0026thinsp;17.5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e84 [69\u0026ndash;107]; 87.5\u0026thinsp;\u0026plusmn;\u0026thinsp;24.1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.26\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRespiratory rate (R), breaths/min (median [IQR]; mean\u0026thinsp;\u0026plusmn;\u0026thinsp;SD)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e20 [\u003cspan additionalcitationids=\"CR17 CR18 CR19 CR20\" citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]; 19.2\u0026thinsp;\u0026plusmn;\u0026thinsp;4.2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e20 [\u003cspan additionalcitationids=\"CR17 CR18 CR19 CR20\" citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]; 19.1\u0026thinsp;\u0026plusmn;\u0026thinsp;4.1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e20 [\u003cspan additionalcitationids=\"CR17 CR18 CR19 CR20 CR21\" citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]; 19.4\u0026thinsp;\u0026plusmn;\u0026thinsp;5.4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.05\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSystolic blood pressure (SBP), mmHg (median [IQR]; mean\u0026thinsp;\u0026plusmn;\u0026thinsp;SD)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e141 [123\u0026ndash;161]; 142.8\u0026thinsp;\u0026plusmn;\u0026thinsp;29.6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e141 [123\u0026ndash;161]; 142.7\u0026thinsp;\u0026plusmn;\u0026thinsp;28.6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e145 [119\u0026ndash;173]; 143.8\u0026thinsp;\u0026plusmn;\u0026thinsp;39.5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.04\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDiastolic blood pressure (DBP), mmHg (median [IQR]; mean\u0026thinsp;\u0026plusmn;\u0026thinsp;SD)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e82 [72\u0026ndash;92]; 83.0\u0026thinsp;\u0026plusmn;\u0026thinsp;21.1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e82 [73\u0026ndash;92]; 83.0\u0026thinsp;\u0026plusmn;\u0026thinsp;21.0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e83 [70\u0026ndash;95]; 82.7\u0026thinsp;\u0026plusmn;\u0026thinsp;22.5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.02\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePeripheral oxygen saturation (SpO₂), % (median [IQR]; mean\u0026thinsp;\u0026plusmn;\u0026thinsp;SD)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e98 [95\u0026ndash;98]; 95.9\u0026thinsp;\u0026plusmn;\u0026thinsp;5.6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e98 [95\u0026ndash;98]; 96.4\u0026thinsp;\u0026plusmn;\u0026thinsp;4.1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e96 [87\u0026ndash;97]; 89.2\u0026thinsp;\u0026plusmn;\u0026thinsp;13.2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1.35\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNEUROLOGIC STATUS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGlasgow Coma Scale (GCS, 3\u0026ndash;15) (median [IQR]; mean\u0026thinsp;\u0026plusmn;\u0026thinsp;SD)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e15 [\u003cspan additionalcitationids=\"CR10 CR11 CR12 CR13 CR14\" citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]; 12.3\u0026thinsp;\u0026plusmn;\u0026thinsp;4.3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e15 [\u003cspan additionalcitationids=\"CR13 CR14\" citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]; 12.8\u0026thinsp;\u0026plusmn;\u0026thinsp;3.9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e3 [\u003cspan additionalcitationids=\"CR4 CR5\" citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]; 5.7\u0026thinsp;\u0026plusmn;\u0026thinsp;4.0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1.84\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePupil reactivity, n (%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e└ No reaction\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e135 (5.9%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e53 (2.5%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e82 (49.4%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1.99\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e└ Unilateral reaction\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e122 (5.3%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e82 (3.9%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e40 (24.1%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.90\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e└ Bilateral reaction\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2030 (88.8%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1986 (93.6%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e44 (26.5%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e2.13\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSCORE DISTRIBUTIONS (median [IQR]; mean\u0026thinsp;\u0026plusmn;\u0026thinsp;SD)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e└ MGAP\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e24.0 [20.0\u0026ndash;27.0]; 23.0\u0026thinsp;\u0026plusmn;\u0026thinsp;5.2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e24.0 [21.0\u0026ndash;27.0]; 23.2\u0026thinsp;\u0026plusmn;\u0026thinsp;5.0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e21.5 [15.0\u0026ndash;24.0]; 20.4\u0026thinsp;\u0026plusmn;\u0026thinsp;6.2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.55\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e└ RTS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e7.8384 [6.9016\u0026ndash;7.8408]; 7.0772\u0026thinsp;\u0026plusmn;\u0026thinsp;1.3072\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e7.8384 [6.9040\u0026ndash;7.8408]; 7.2532\u0026thinsp;\u0026plusmn;\u0026thinsp;1.1020\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e4.0936 [4.0912\u0026ndash;5.9648]; 4.8292\u0026thinsp;\u0026plusmn;\u0026thinsp;1.6089\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e2.11\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e└ mREMS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e4.0 [2.0\u0026ndash;6.0]; 4.5\u0026thinsp;\u0026plusmn;\u0026thinsp;3.5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e4.0 [2.0\u0026ndash;6.0]; 4.8\u0026thinsp;\u0026plusmn;\u0026thinsp;3.5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1.0 [0.0\u0026ndash;1.0]; 1.2\u0026thinsp;\u0026plusmn;\u0026thinsp;1.9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1.04\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e└ MEWS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2.0 [1.0\u0026ndash;3.0]; 2.3\u0026thinsp;\u0026plusmn;\u0026thinsp;1.7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2.0 [1.0\u0026ndash;3.0]; 2.2\u0026thinsp;\u0026plusmn;\u0026thinsp;1.7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e3.0 [1.0\u0026ndash;4.0]; 3.0\u0026thinsp;\u0026plusmn;\u0026thinsp;2.0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.47\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e└ GAP\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e21.0 [16.0\u0026ndash;22.0]; 19.2\u0026thinsp;\u0026plusmn;\u0026thinsp;4.7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e21.0 [18.0\u0026ndash;24.0]; 19.7\u0026thinsp;\u0026plusmn;\u0026thinsp;4.3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e11.5 [9.0\u0026ndash;13.0]; 12.3\u0026thinsp;\u0026plusmn;\u0026thinsp;4.4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1.74\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"5\"\u003eNotes: Continuous variables are summarized as median [IQR]; mean\u0026thinsp;\u0026plusmn;\u0026thinsp;SD is also shown to aid comparison. Categorical variables are n (%). Std diff is the absolute standardized difference between groups.\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e displays receiver operating characteristic curves after single equation logistic recalibration. RTS shows the highest discrimination with AUROC 0.874 (95% CI 0.841\u0026ndash;0.903), followed by GAP 0.863 (0.831\u0026ndash;0.898) and mREMS 0.856 (0.826\u0026ndash;0.883); MGAP and MEWS are lower at 0.632 (0.586\u0026ndash;0.678) and 0.606 (0.571\u0026ndash;0.665). Across most of the specificity range, RTS and GAP remain above the other scores, especially at low false positive rates that matter for early triage. All estimates were obtained with 1,000 bootstrap resamples.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003ePrecision recall analysis under class imbalance reinforces the discrimination ranking. RTS achieves the largest AUPRC, 0.465 (95% CI 0.381\u0026ndash;0.531), followed by GAP 0.382 (0.308\u0026ndash;0.453) and mREMS 0.370 (0.280\u0026ndash;0.424), with MGAP and MEWS clearly lower. The gray horizontal line marks the 7.3% prevalence, and across much of the recall axis RTS and GAP keep precision above this reference, including the high recall region that is critical for early triage. Orange markers denote operating points at 10% and 20% probability thresholds that fall on favorable portions of the RTS and GAP curves relative to the others, as shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e. These findings support prioritizing RTS and GAP when the aim is to achieve high sensitivity with acceptable precision for identifying 24 hour mortality.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e demonstrates strong calibration after single equation updating. For RTS and GAP, the smoothed calibration curves with bootstrap 95% confidence bands lie close to the 45\u0026deg; identity line across the high density region of predictions below 0.20, with only small deviation around the middle of the range. mREMS shows a similar pattern and slightly overestimates risk in the extreme right tail. MGAP and MEWS show compressed risk ranges, flatter curves and wider uncertainty at low and high ends, which is consistent with the 7.3% event rate and the scarcity of observations in the tails. These patterns agree with Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e where calibration intercepts are near 0 and slopes are near 1, supporting the use of the recalibrated probabilities as absolute risks for 10% and 20% threshold decisions.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003ePrimary performance metrics (recalibrated models)\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"8\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eScore\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAUROC (95% CI)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eAUPRC (95% CI)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eBrier (95% CI)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eCalibration intercept\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eCalibration slope\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eRecalibration α\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003eRecalibration γ\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMGAP\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.632 (0.586\u0026ndash;0.678)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.154 (0.112\u0026ndash;0.199)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.066 (0.057\u0026ndash;0.074)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e-0.509\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e-0.093\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRTS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.874 (0.841\u0026ndash;0.903)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.465 (0.381\u0026ndash;0.531)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.050 (0.043\u0026ndash;0.056)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e3.729\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e-1.015\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003emREMS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.856 (0.826\u0026ndash;0.883)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.370 (0.280\u0026ndash;0.424)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.055 (0.049\u0026ndash;0.062)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e-0.688\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e-0.771\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMEWS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.606 (0.571\u0026ndash;0.665)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.129 (0.098\u0026ndash;0.159)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.066 (0.058\u0026ndash;0.075)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e-3.120\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.221\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGAP\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.863 (0.831\u0026ndash;0.898)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.382 (0.308\u0026ndash;0.453)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.053 (0.047\u0026ndash;0.060)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e2.426\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e-0.310\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eAfter the single-line recalibration, the ranking was consistent across summary measures. RTS performed best (AUROC 0.874, AUPRC 0.465, Brier 0.050), followed by GAP (0.863, 0.382, 0.053) and mREMS (0.856, 0.370, 0.055), with MGAP and MEWS lower on all three metrics. Apparent calibration was tight for every tool, with intercepts close to zero and slopes close to one. The fitted recalibration coefficients quantify the translation and rescaling needed to express each score as a probability: RTS α 3.729, γ\u0026thinsp;\u0026minus;\u0026thinsp;1.015; GAP α 2.426, γ\u0026thinsp;\u0026minus;\u0026thinsp;0.310; mREMS α\u0026thinsp;\u0026minus;\u0026thinsp;0.688, γ\u0026thinsp;\u0026minus;\u0026thinsp;0.771; MEWS α\u0026thinsp;\u0026minus;\u0026thinsp;3.120, γ 0.221; MGAP α\u0026thinsp;\u0026minus;\u0026thinsp;0.509, γ\u0026thinsp;\u0026minus;\u0026thinsp;0.093. Complete estimates with confidence intervals are reported in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e.\u003c/p\u003e \u003cp\u003eAt the prespecified 10% threshold, the per-100 summaries show how each tool trades sensitivity for alert burden: RTS yields about 5.1 true positives, 9.5 false positives, 2.1 false negatives, and 83.2 true negatives with net benefit 0.041; GAP yields 5.7, 14.7, 1.5, 78.0 with 0.041; mREMS reaches 5.9, 16.9, 1.3, 75.8 and the same net benefit but with visibly more over-triage, whereas MGAP and MEWS detect 2.9 and 1.6 deaths with 0.009 and 0.006, respectively; detailed counts are provided in Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e. Moving to a 20% threshold shifts emphasis to precision: RTS delivers 4.9, 9.2, 2.3, 83.6 with net benefit 0.026, and GAP 5.1, 9.9, 2.1, 82.8 with 0.026; mREMS becomes markedly selective at 2.9, 3.8, 4.4, 88.9, achieving the highest PPV (0.431) but giving up sensitivity; MEWS shows negative net benefit (\u0026minus;\u0026thinsp;0.001) and MGAP is near zero, as shown in Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e. Collectively, these profiles make the clinical trade-offs explicit: the 10% rule privileges recall while keeping NPV very high, whereas the 20% rule privileges PPV and reduces avoidable alerts at the cost of more missed early deaths; across both settings, RTS and GAP provide the most usable balance for routine triage, while mREMS suits contexts that prioritize precision over sensitivity.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eClinical utility per 100 patients at 10% threshold\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"11\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c9\" colnum=\"9\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c10\" colnum=\"10\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c11\" colnum=\"11\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eScore\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eThreshold\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTP\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eFP\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eFN\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eTN\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eSe\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003eSp\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c9\"\u003e \u003cp\u003ePPV\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c10\"\u003e \u003cp\u003eNPV\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c11\"\u003e \u003cp\u003eNetbenefit\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMGAP\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.100\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e2.900\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e18.100\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e4.400\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e74.600\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.398\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.804\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.137\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.945\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e0.009\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRTS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.100\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e5.100\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e9.500\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e2.100\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e83.200\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.705\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.897\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.349\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.975\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e0.041\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003emREMS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.100\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e5.900\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e16.900\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1.300\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e75.800\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.819\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.818\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.260\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.983\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e0.041\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMEWS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.100\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1.600\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e9.100\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e5.600\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e83.600\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.223\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.901\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.150\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.937\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e0.006\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGAP\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.100\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e5.700\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e14.700\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1.500\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e78\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.789\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.842\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.281\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.981\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e0.041\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab4\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eClinical utility per 100 patients at 20% threshold\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"11\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c9\" colnum=\"9\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c10\" colnum=\"10\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c11\" colnum=\"11\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eScore\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eThreshold\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTP\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eFP\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eFN\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eTN\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eSe\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003eSp\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c9\"\u003e \u003cp\u003ePPV\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c10\"\u003e \u003cp\u003eNPV\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c11\"\u003e \u003cp\u003eNetbenefit\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMGAP\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.200\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.200\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.500\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e92.300\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.030\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.995\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.312\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.929\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e0.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRTS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.200\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e4.900\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e9.200\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2.300\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e83.600\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.681\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.901\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.350\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.973\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e0.026\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003emREMS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.200\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e2.900\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e3.800\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e4.400\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e88.900\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.398\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.959\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.431\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.953\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e0.019\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMEWS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.200\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.200\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1.100\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e91.600\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.030\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.988\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.161\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.929\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e-0.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGAP\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.200\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e5.100\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e9.900\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2.100\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e82.800\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.705\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e0.893\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.340\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c10\"\u003e \u003cp\u003e0.975\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c11\"\u003e \u003cp\u003e0.026\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eThis study asked whether a minimal, prespecified recalibration with a single intercept and a single slope can make established bedside scores provide trustworthy 24 hour risk for traumatic brain injury without rebuilding models. Across receiver operating characteristics, precision and recall under low prevalence, graphical and quantitative calibration, and per-100 consequences at prespecified thresholds, the evidence pointed to a consistent message. After recalibration, RTS and GAP provided the most reliable early triage signal, mREMS remained competitive when precision is prioritized, and MGAP and MEWS contributed little at the very early horizon. This approach aligns with guidance that recommends simple intercept and slope updating as the first step when transporting prognostic models[\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]. The emphasis on precision and recall reflects best practice for rare outcomes and avoids an overly optimistic impression based on receiver operating characteristics alone[\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]. Presenting discrimination, calibration, and decision consequences together follows contemporary reporting standards for clinical prediction research and supports transparent translation to bedside decisions[\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eOur findings are consistent with contemporary trauma literature that reports strong performance for scores built around physiology, particularly RTS and GAP, with mREMS close behind and more variable results for MGAP and MEWS across settings[\u003cspan additionalcitationids=\"CR27\" citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]. Studies that focus on early outcomes in the emergency department or within the first day tend to reproduce this pattern, whereas reports that target in-hospital mortality over a longer horizon sometimes place MGAP closer to the leaders when pupil data are scarce or when case mix shifts toward less severe brain injury[\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e, \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e]. Heterogeneity in outcome window, prevalence, and measurement context offers a plausible explanation. Very early death is driven by neurologic compromise and circulatory failure, which magnifies the signal from GCS, blood pressure, oxygenation, and pupil reactivity; this mechanism-level alignment helps physiology-forward scores such as RTS and GAP perform well in the first hours[\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e]. Data provenance also matters, since prehospital capture, intubation, sedation, and oxygen supplementation can shift distributions of GCS and SpO₂ and alter apparent risk; a brief intercept and slope update reduces these transport gaps without abandoning the original score structure[\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e]. Taken together with prior work, our results support deploying recalibrated physiology-based scores for near-term triage in TBI while recognizing that MGAP or MEWS may be preferred in environments where key neurologic variables are unavailable or where the decision horizon extends beyond the first day[\u003cspan additionalcitationids=\"CR34\" citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eCalibration after the prespecified update was strong and clinically useful[\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e, \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e]. Intercepts clustered around zero, slopes were close to one, and the decile plots lay near the identity line. Under these conditions, predicted probabilities can be read as absolute risks rather than mere ranks. This supports bedside conversations, selection of escalation thresholds, and clear communication with consultants. It also reduces pressure to rebuild models and preserves the familiar structure of each score[\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eDecision thresholds translate accuracy into workload and safety. The per-100 summaries at ten and twenty percent show what is gained in additional detections and what is spent in avoidable alerts[\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e, \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e]. At ten percent the priority is recall and the negative predictive value remains high, which fits early triage when missing a rapidly deteriorating patient is unacceptable. At twenty percent the alert burden falls and precision rises, which can be preferable when monitored capacity is tight. RTS and GAP offer a balanced option across both thresholds, whereas mREMS can be selected when precision is the dominant concern.\u003c/p\u003e \u003cp\u003eThis study has several strengths. Data were assembled with a prespecified schema and each bedside score was reconstructed exactly as published, followed by a single transparent recalibration so that outputs remained interpretable at the bedside. To enable fair head-to-head comparison we relied on routinely available variables and a uniform analytic approach, accepting a narrower feature set to enhance comparability. The endpoint was fixed at 24 hours to match time-critical decisions and to limit downstream confounding. Anticipated heterogeneity in prevalence and measurement practice was addressed by aligning predicted and observed risk with a prespecified intercept and slope rather than altering model structure. We recognise important limitations, including conduct within one health system and period, context-sensitive measurements for consciousness and oxygenation, small numbers for penetrating injury, evaluation at only two operating points, and the need for confirmation in multicentre cohorts. Setting differences such as prehospital capture, intubation, sedation, and supplemental oxygen can shift the distributions of key variables, which the chosen update is intended to absorb. Robustness was supported by concordant signals across discrimination, precision and recall behaviour, calibration, and per-100 consequences, with uncertainty summarised by bootstrap intervals.\u003c/p\u003e \u003cp\u003eThese features guide interpretation and point to next steps. Confirmation across services with different workflows and case mix will establish transportability. Routine monitoring of calibration with small periodic updates will manage drift over time. Subgroup checks by age and mechanism and fairness assessments will strengthen credibility. Audits of threshold policies, alert burden, time to action, and missed-event narratives will show how the risk estimates translate into care. Prospective impact evaluations should prespecify clinical usefulness metrics, including decision-curve analysis to benchmark net benefit against treat-all and treat-none strategies[\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e]. Together, these actions provide a practical path to safe, transparent, and scalable adoption.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eCalibrating each score with one intercept and one slope was sufficient to deliver trustworthy 24-hour risk for early TBI triage. Across discrimination, precision\u0026ndash;recall, and calibration, RTS and GAP performed most reliably, mREMS favored precision, and MGAP and MEWS added little at this time horizon. Per-100-patient summaries at 10% and 20% thresholds translated model output into actionable counts that support threshold selection according to local capacity. Embedding this light recalibration with routine monitoring offers a ready path to use, and multicentre impact studies should confirm transportability and effects on workflow and safety.\u0026nbsp;\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eHuman Ethics and Consent to Participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study was conducted in accordance with the Declaration of Helsinki and received approval from the Institutional Ethics Committee of the Affiliated Hospital of Nantong University (Approval No. 2021-K084-01). Due to the retrospective nature of this study and the use of de-identified clinical data, the requirement for informed consent was waived. All procedures involved in this research complied with institutional data protection policies, ensuring confidentiality and privacy of patient data.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDeclaration of competing interest\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare no competing financial or non-financial interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis work was supported by the Wuxi Municipal Health Commission Youth Research Project (Q202323).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe thank the emergency department nursing staff and the trauma registry team at the Affiliated Hospital of Nantong University for their assistance with data capture and quality checks. We are grateful to the clinical informatics group and the hospital information office for support with electronic health record queries and dataset validation. The authors alone are responsible for the analyses and interpretation.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSample CRediT author statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTianxi Chen (Co-first author): Conceptualization, Methodology, Formal analysis, Writing – Original Draft; Jinheng Tu (Co-first author): Investigation, Data Curation, Writing – Original Draft; Jiajia Liu (Co-first author): Data Curation, Writing – Original Draft; Hong Sun: Supervision, Project Administration, Funding Acquisition; Yun Lu: Resources, Writing – Review \u0026amp; Editing; Rui Chen: Investigation, Writing – Review \u0026amp; Editing; Hao Huang (Second corresponding author): Supervision, Project Administration, Funding Acquisition; Chen Shen (First corresponding author): Supervision, Writing – Review \u0026amp; Editing, Funding Acquisition.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eGlobal, regional, and national burden of neurological disorders, 1990-2016: a systematic analysis for the Global Burden of Disease Study 2016. LANCET NEUROL 2019;18:459-480.\u003c/li\u003e\n\u003cli\u003eSartorius D, Le Manach Y, David JS, Rancurel E, Smail N, Thicoipe M, Wiel E, Ricard-Hibon A, Berthier F, Gueugniaud PY, Riou B. Mechanism, glasgow coma scale, age, and arterial pressure (MGAP): a new simple prehospital triage score to predict mortality in trauma patients. CRIT CARE MED 2010;38:831-837.\u003c/li\u003e\n\u003cli\u003eChampion HR, Sacco WJ, Copes WS, Gann DS, Gennarelli TA, Flanagan ME. A revision of the Trauma Score. J Trauma 1989;29:623-629.\u003c/li\u003e\n\u003cli\u003eMiller RT, Nazir N, McDonald T, Cannon CM. The modified rapid emergency medicine score: A novel trauma triage tool to predict in-hospital mortality. INJURY 2017;48:1870-1877.\u003c/li\u003e\n\u003cli\u003eSubbe CP, Kruger M, Rutherford P, Gemmel L. Validation of a modified Early Warning Score in medical admissions. QJM-INT J MED 2001;94:521-526.\u003c/li\u003e\n\u003cli\u003eKondo Y, Abe T, Kohshi K, Tokuda Y, Cook EF, Kukita I. Revised trauma scoring system to predict in-hospital mortality in the emergency department: Glasgow Coma Scale, Age, and Systolic Blood Pressure score. CRIT CARE 2011;15:R191.\u003c/li\u003e\n\u003cli\u003eMohammed Z, Saleh Y, AbdelSalam EM, Mohammed N, El-Bana E, Hirshon JM. Evaluation of the Revised Trauma Score, MGAP, and GAP scoring systems in predicting mortality of adult trauma patients in a low-resource setting. BMC EMERG MED 2022;22:90.\u003c/li\u003e\n\u003cli\u003eMoreno-Blanco D, Alonso E, Sanz-Garcia A, Aramendi E, Lopez-Izquierdo R, Perez GR, Del PVC, Martin-Rodriguez F. Spanish vs USA cohort comparison of prehospital trauma scores to predict short-term mortality. CLIN MED 2024;24:100208.\u003c/li\u003e\n\u003cli\u003eSaito T, Rehmsmeier M. The precision-recall plot is more informative than the ROC plot when evaluating binary classifiers on imbalanced datasets. PLOS ONE 2015;10:e118432.\u003c/li\u003e\n\u003cli\u003eVan Calster B, McLernon DJ, van Smeden M, Wynants L, Steyerberg EW. Calibration: the Achilles heel of predictive analytics. BMC MED 2019;17:230.\u003c/li\u003e\n\u003cli\u003eVan Calster B, Wynants L, Verbeek J, Verbakel JY, Christodoulou E, Vickers AJ, Roobol MJ, Steyerberg EW. Reporting and Interpreting Decision Curve Analysis: A Guide for Investigators. EUR UROL 2018;74:796-804.\u003c/li\u003e\n\u003cli\u003eJeong JH, Park YJ, Kim DH, Kim TY, Kang C, Lee SH, Lee SB, Kim SC, Lim D. The new trauma score (NTS): a modification of the revised trauma score for better trauma mortality prediction. BMC SURG 2017;17:77.\u003c/li\u003e\n\u003cli\u003eGill MR, Reiley DG, Green SM. Interrater reliability of Glasgow Coma Scale scores in the emergency department. ANN EMERG MED 2004;43:215-223.\u003c/li\u003e\n\u003cli\u003eVan Calster B, Steyerberg EW, Wynants L, van Smeden M. There is no such thing as a validated prediction model. BMC MED 2023;21:70.\u003c/li\u003e\n\u003cli\u003eMoons KG, Altman DG, Reitsma JB, Ioannidis JP, Macaskill P, Steyerberg EW, Vickers AJ, Ransohoff DF, Collins GS. Transparent Reporting of a multivariable prediction model for Individual Prognosis or Diagnosis (TRIPOD): explanation and elaboration. ANN INTERN MED 2015;162:W1-W73.\u003c/li\u003e\n\u003cli\u003eVickers AJ, van Calster B, Steyerberg EW. A simple, step-by-step guide to interpreting decision curve analysis. Diagn Progn Res 2019;3:18.\u003c/li\u003e\n\u003cli\u003eTseng TE, Lee CC, Yen HK, Groot OQ, Hou CH, Lin SY, Bongers M, Hu MH, Karhade AV, Ko JC, Lai YH, Yang JJ, Verlaan JJ, Yang RS, Schwab JH, Lin WH. International Validation of the SORG Machine-learning Algorithm for Predicting the Survival of Patients with Extremity Metastases Undergoing Surgical Treatment. CLIN ORTHOP RELAT R 2022;480:367-378.\u003c/li\u003e\n\u003cli\u003eJoksic-Mazinjanin R, Maric N, Duricin A, Gojkovic Z, Vasovic V, Rakic G, Joksic-Zelic M, Saravolac S. Prehospital Trauma Scoring Systems for Evaluation of Trauma Severity and Prediction of Outcomes. MEDICINA-LITHUANIA 2023;59.\u003c/li\u003e\n\u003cli\u003eGeeraerts T, Velly L, Abdennour L, Asehnoune K, Audibert G, Bouzat P, Bruder N, Carrillon R, Cottenceau V, Cotton F, Courtil-Teyssedre S, Dahyot-Fizelier C, Dailler F, David JS, Engrand N, Fletcher D, Francony G, Gergele L, Ichai C, Javouhey E, Leblanc PE, Lieutaud T, Meyer P, Mirek S, Orliaguet G, Proust F, Quintard H, Ract C, Srairi M, Tazarourte K, Vigue B, Payen JF. Management of severe traumatic brain injury (first 24hours). ANAESTH CRIT CARE PA 2018;37:171-186.\u003c/li\u003e\n\u003cli\u003eCollins GS, Reitsma JB, Altman DG, Moons KG. Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD): the TRIPOD statement. BMJ-BRIT MED J 2015;350:g7594.\u003c/li\u003e\n\u003cli\u003eGANDEVIA B, TOVELL A. DECLARATION OF HELSINKI. MED J AUSTRALIA 1964;2:320-321.\u003c/li\u003e\n\u003cli\u003eSteyerberg EW, Borsboom GJ, van Houwelingen HC, Eijkemans MJ, Habbema JD. Validation and updating of predictive logistic regression models: a study on sample size and shrinkage. STAT MED 2004;23:2567-2586.\u003c/li\u003e\n\u003cli\u003eGehringer CK, Martin GP, Van Calster B, Hyrich KL, Verstappen S, Sergeant JC. How to develop, validate, and update clinical prediction models using multinomial logistic regression. J CLIN EPIDEMIOL 2024;174:111481.\u003c/li\u003e\n\u003cli\u003eLiu S, Roemer F, Ge Y, Bedrick EJ, Li ZM, Guermazi A, Sharma L, Eaton C, Hochberg MC, Hunter DJ, Nevitt MC, Wirth W, Kent KC, Sun X. Comparison of evaluation metrics of deep learning for imbalanced imaging data in osteoarthritis studies. OSTEOARTHR CARTILAGE 2023;31:1242-1248.\u003c/li\u003e\n\u003cli\u003eCollins GS, Moons K, Dhiman P, Riley RD, Beam AL, Van Calster B, Ghassemi M, Liu X, Reitsma JB, van Smeden M, Boulesteix AL, Camaradou JC, Celi LA, Denaxas S, Denniston AK, Glocker B, Golub RM, Harvey H, Heinze G, Hoffman MM, Kengne AP, Lam E, Lee N, Loder EW, Maier-Hein L, Mateen BA, McCradden MD, Oakden-Rayner L, Ordish J, Parnell R, Rose S, Singh K, Wynants L, Logullo P. TRIPOD+AI statement: updated guidance for reporting clinical prediction models that use regression or machine learning methods. BMJ-BRIT MED J 2024;385:e78378.\u003c/li\u003e\n\u003cli\u003eMerchant A, Shaukat N, Ashraf N, Hassan S, Jarrar Z, Abbasi A, Ahmed T, Atiq H, Khan UR, Khan NU, Mushtaq S, Rasul S, Hyder AA, Razzak J, Haider AH. Which curve is better? A comparative analysis of trauma scoring systems in a South Asian country. TRAUMA SURG ACUTE CA 2023;8:e1171.\u003c/li\u003e\n\u003cli\u003eRio T, Nogueira LS, Lima FR, Cassiano C, Garcia D. Performance of severity indices for admission and mortality of trauma patients in the intensive care unit: a retrospective cohort study. EUR J MED RES 2023;28:559.\u003c/li\u003e\n\u003cli\u003eLuo X, Gao H, Yu X, Jiang Z, Yang W. Spectral analysis of heart rate variability for trauma outcome prediction: an analysis of 210 ICU multiple trauma patients. EUR J TRAUMA EMERG S 2021;47:153-160.\u003c/li\u003e\n\u003cli\u003eKenarangi T, Rahmani F, Yazdani A, Ahmadi GD, Lotfi M, Khalaj TA. Comparison of GAP, R-GAP, and new trauma score (NTS) systems in predicting mortality of traffic accidents that injure hospitals at Mashhad University of medical sciences. HELIYON 2024;10:e36004.\u003c/li\u003e\n\u003cli\u003ePhunghassaporn N, Sukhvibul P, Techapongsatorn S, Tansawet A. Accuracy and external validation of the modified rapid emergency medicine score in road traffic injuries in a Bangkok level I trauma center. HELIYON 2022;8:e12225.\u003c/li\u003e\n\u003cli\u003eHawryluk G, Lulla A, Bell R, Jagoda A, Mangat HS, Bobrow BJ, Ghajar J. Guidelines for Prehospital Management of Traumatic Brain Injury 3rd Edition: Executive Summary. NEUROSURGERY 2023;93:e159-e169.\u003c/li\u003e\n\u003cli\u003eEfthimiou O, Seo M, Chalkou K, Debray T, Egger M, Salanti G. Developing clinical prediction models: a step-by-step guide. BMJ-BRIT MED J 2024;386:e78276.\u003c/li\u003e\n\u003cli\u003eYousefi MR, Karajizadeh M, Ghasemian M, Paydar S. Comparing NEWS2, TRISS, and RTS in predicting mortality rate in trauma patients based on prehospital data set: a diagnostic study. BMC EMERG MED 2024;24:163.\u003c/li\u003e\n\u003cli\u003eCandel B, Nissen SK, Nickel CH, Raven W, Thijssen W, Gaakeer MI, Lassen AT, Brabrand M, Steyerberg EW, de Jonge E, de Groot B. Development and External Validation of the International Early Warning Score for Improved Age- and Sex-Adjusted In-Hospital Mortality Prediction in the Emergency Department. CRIT CARE MED 2023;51:881-891.\u003c/li\u003e\n\u003cli\u003eDonoso CM, Mordillo-Mateos L, Martin-Conty JL, Polonio-Lopez B, Lopez-Gonzalez A, Durantez-Fernandez C, Vinuela A, Rodriguez HM, Mohedano-Moriano A, Lopez-Izquierdo R, Jorge SC, Martin-Rodriguez F. Modified Rapid Emergency Medicine Score-Lactate (mREMS-L) performance to screen non-anticipated 30-day-related-mortality in emergency department. EUR J CLIN INVEST 2023;53:e13994.\u003c/li\u003e\n\u003cli\u003eLiou L, Scott E, Parchure P, Ouyang Y, Egorova N, Freeman R, Hofer IS, Nadkarni GN, Timsina P, Kia A, Levin MA. Assessing calibration and bias of a deployed machine learning malnutrition prediction model within a large healthcare system. NPJ DIGIT MED 2024;7:149.\u003c/li\u003e\n\u003cli\u003eRiley RD, Archer L, Snell K, Ensor J, Dhiman P, Martin GP, Bonnett LJ, Collins GS. Evaluation of clinical prediction models (part 2): how to undertake an external validation study. BMJ-BRIT MED J 2024;384:e74820.\u003c/li\u003e\n\u003cli\u003eKappen TH, van Loon K, Kappen MA, van Wolfswinkel L, Vergouwe Y, van Klei WA, Moons KG, Kalkman CJ. Barriers and facilitators perceived by physicians when using prediction models in practice. J CLIN EPIDEMIOL 2016;70:136-145.\u003c/li\u003e\n\u003cli\u003ePiovani D, Sokou R, Tsantes AG, Vitello AS, Bonovas S. Optimizing Clinical Decision Making with Decision Curve Analysis: Insights for Clinical Investigators. HEALTHCARE-BASEL 2023;11.\u003c/li\u003e\n\u003cli\u003eVickers AJ, Van Claster B, Wynants L, Steyerberg EW. Decision curve analysis: confidence intervals and hypothesis testing for net benefit. Diagn Progn Res 2023;7:11.\u003c/li\u003e\n\u003cli\u003eVasey B, Nagendran M, Campbell B, Clifton DA, Collins GS, Denaxas S, Denniston AK, Faes L, Geerts B, Ibrahim M, Liu X, Mateen BA, Mathur P, McCradden MD, Morgan L, Ordish J, Rogers C, Saria S, Ting D, Watkinson P, Weber W, Wheatstone P, McCulloch P. Reporting guideline for the early-stage clinical evaluation of decision support systems driven by artificial intelligence: DECIDE-AI. NAT MED 2022;28:924-933.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Traumatic Brain Injury, Risk Stratification, Clinical Prediction, Logistic Recalibration","lastPublishedDoi":"10.21203/rs.3.rs-8452508/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8452508/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003eEarly risk stratification after traumatic brain injury (TBI) is crucial to guide monitoring, escalation of care, and interfacility transfer within the first hours. Bedside scores (RTS, GAP, mREMS, MGAP, MEWS) are simple and widely available, yet their comparative accuracy for predicting 24-hour in-hospital mortality in TBI remains uncertain. Prior studies often use mixed trauma cohorts, later endpoints, and provide limited evaluation of calibration or threshold-level clinical utility. TBI-specific head-to-head evidence is therefore needed.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eThis study analyzed consecutive traumatic brain injury admissions in 2024 at a tertiary medical center in Nantong, China (n\u0026thinsp;=\u0026thinsp;2,287; 24-hour mortality 7.3%). Five bedside scores (RTS, GAP, mREMS, MGAP, MEWS) were recalibrated with a single logistic equation, and performance (AUROC, AUPRC, Brier score, calibration intercept and slope) and clinical utility (net benefit per 100 patients at 10% and 20% thresholds) were estimated with 95% confidence intervals from 1,000 bootstrap resamples; precision\u0026ndash;recall curves were referenced to the cohort prevalence.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eRTS and GAP showed the strongest discrimination, with mREMS close. AUROC: RTS 0.874 (95% CI 0.841\u0026ndash;0.903), GAP 0.863 (0.831\u0026ndash;0.898), mREMS 0.856 (0.826\u0026ndash;0.883); MGAP and MEWS were lower (0.632, 0.606). AUPRCs were 0.465 (0.381\u0026ndash;0.531) for RTS, 0.382 (0.308\u0026ndash;0.453) for GAP, and 0.370 (0.280\u0026ndash;0.424) for mREMS (MGAP 0.154 (0.112\u0026ndash;0.199); MEWS 0.129 (0.098\u0026ndash;0.159)). Brier scores were 0.050, 0.053, and 0.055 for RTS, GAP, and mREMS, respectively. After simple logistic recalibration, calibration intercepts were near 0 and slopes near 1, with observed-vs-predicted curves close to the 45\u0026deg; line across deciles. At 10% and 20% thresholds, RTS and GAP achieved the highest net benefit and more favorable trade-offs per 100 patients; mREMS was intermediate.\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e \u003cp\u003eRTS and GAP showed the best balance of discrimination, calibration, and net benefit at 10% and 20% thresholds; mREMS was comparable after simple recalibration. These findings support prioritizing RTS and GAP for early TBI risk stratification.\u003c/p\u003e","manuscriptTitle":"Comparative Validation of Five Early Trauma Scores for 24-Hour In-Hospital Mortality in Traumatic Brain Injury","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-01-12 06:17:08","doi":"10.21203/rs.3.rs-8452508/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"f2cfbdec-831b-4641-a50a-255a50bb59ce","owner":[],"postedDate":"January 12th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-01-12T13:53:53+00:00","versionOfRecord":[],"versionCreatedAt":"2026-01-12 06:17:08","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8452508","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8452508","identity":"rs-8452508","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}