Archives of Academic Emergency Medicine. 2023; 11(1): e27 REV I EW ART I C L E Prognostic Value of CRASH and IMPACT Models for Pre- dicting Mortality and Unfavorable Outcome in Traumatic Brain Injury; a Systematic Review and Meta-Analysis Hamed Zarei1, Mohammadhossein Vazirizadeh-Mahabadi1, Hamzah Adel Ramawad2, Arash Sarveazad3,4∗, Mahmoud Yousefifard1,5 † 1. Physiology Research Center, Iran University of Medical Sciences, Tehran, Iran. 2. Department of Emergency Medicine, NYC Health & amp; Hospitals, Coney Island, New York. 3. Colorectal Research Center, Iran University of Medical Sciences, Tehran, Iran. 4. Nursing Care Research Center, Iran University of Medical Sciences, Tehran, Iran. 5. Pediatrics Chronic Kidney Disease Research Center, Tehran University of Medical Sciences, Tehran, Iran. Received: December 2022; Accepted: January 2023; Published online: 4 March 2023 Abstract: Introduction: The Corticosteroid Randomization After Significant Head injury (CRASH) and the International Mission for Prognosis and Analysis of Clinical Trials (IMPACT) are two prognostic models frequently used in predicting the out- come of patients with traumatic brain injury. There are ongoing debates about which of the two models has a better prognostic value. This study aims to compare the CRASH and IMPACT in predicting mortality and unfavorable outcome of patients with traumatic brain injury. Method: We performed a literature search using Medline (via PubMed), Embase, Scopus, and Web of Science databases until August 17, 2022. After two independent researchers screened the articles, we included all the original articles comparing the prognostic value of IMPACT and CRASH models in patients with trau- matic brain injury. The outcomes evaluated were mortality and unfavorable outcome. The data of the included articles were analyzed using STATA 17.0 statistical program, and we reported an odds ratio (OR) with a 95% confidence interval (95% CI) for comparison. Results: We included the data from 16 studies. The analysis showed that the areas under the curve of the IMPACT core model and CRASH basic model do not differ in predicting the mortality of patients (OR=0.99; p=0.905) and their six-month unfavorable outcome (OR=1.01; p=0.719). Additionally, the CRASH CT model showed no difference from the IMPACT extended (OR=0.98; p=0.507) and IMPACT Lab (OR=1.00; p=0.298) models in predicting the mortality of patients with traumatic brain injury. We also observed similar findings in the six-month unfavorable out- come, showing that the CRASH CT model does not differ from the IMPACT extended (OR=1.00; p=0.990) and IMPACT Lab (OR=1.00; p=0.570) in predicting the unfavorable outcome in head trauma patients. Conclusion: Low to very low level of evidence shows that IMPACT and CRASH models have similar values in predicting mortality and unfavorable outcome in patients with traumatic brain injury. Since the discriminative power of the IMPACT Core and CRASH basic models is not different from the IMPACT extended, IMPACT Lab, and CRASH CT models, it may be possible to only use the core and basic models in examining the prognosis of patients with traumatic injuries to the brain. Keywords:Brain injuries, traumatic; prognosis; survival analysis; mortality; patient outcome assessment Cite this article as: Zarei H, Vazirizadeh-Mahabadi M, Adel Ramawad H, Sarveazad A, Yousefifard M. Prognostic Value of CRASH and IMPACT Models for Predicting Mortality and Unfavorable Outcome in Traumatic Brain Injury; a Systematic Review and Meta-Analysis. Arch Acad Emerg Med. 2023; 11(1): e27. https://doi.org/10.22037/aaem.v11i1.1885. ∗Corresponding Author: Arash Sarveazad, Colorectal Research Center, Rassol-e-Akram Hospital, Nyayesh St. Sattarkhan St., Tehran, Iran. Email: arashsarveazad@gmail.com. ORCID: https://orcid.org/0000-0002-7947-8642. † Corresponding Author: Mahmoud Yousefifard; Physiology Research Center, Iran University of Medical Sciences, Hemmat Highway, Tehran, Iran. Email: yousefifard20@gmail.com. ORCID: https://orcid.org/0000-0002-7947-8642. 1. Introduction Traumatic brain injury (TBI) is a significant cause of morbid- ity and mortality worldwide. It is one of the most common complications of intentional and unintentional accidents, with a global prevalence of 8.4%. Based on existing reports, the incidence and prevalence of TBI due to head trauma have significantly risen since the 1990s (1). This presents a This open-access article distributed under the terms of the Creative Commons Attribution NonCommercial 3.0 License (CC BY-NC 3.0). Downloaded from: https://journals.sbmu.ac.ir/aaem/index.php/AAEM/index H. Zarei et al. 2 unique challenge for hospitals because available healthcare resources are likely to become overburdened by higher rates of TBI. Physicians can identify intracranial injuries from head trau- mas through imaging modalities such as computed tomogra- phy (CT) scan and magnetic resonance imaging (MRI). How- ever, there are challenges to imaging trauma patients sus- pected of TBI. While scanning patients has become essential for trauma diagnostic work-up, not all hospitals are equipped with CT machines or technicians. Other challenges include the deterioration of hemodynamically unstable patients in the radiology suite and high exposure to ionizing radiation (2, 3). As a result, healthcare providers are researching differ- ent tests and measures to diagnose TBI, especially in patients unable to undergo head imaging. The Glasgow coma scale (GCS) measures the level of con- sciousness in a person following trauma and classifies TBI as mild, moderate, and severe. Despite several clinical deci- sion rules to help identify clinically significant head injuries, physicians continue to face the challenge of deciding who needs a brain CT while minimizing unnecessary radiation ex- posure. While minor head injury is commonly seen in the emergency department, investigators have found that only 16% of these patients will have an acute intracranial lesion (4). This means that if every patient with mild head injury was to undergo cerebral imaging, 84% of these studies would be normal and unnecessary. To reduce unnecessary imaging and waste of hospital resources, a prognostic scoring system can be helpful. Although multiple prognostic models for predicting mor- tality and disability following traumatic brain injuries exist, none are widely used in clinical practice. Most of these mod- els are limited based on small population size and lack ex- ternal validation. Most models were designed based on data from developed countries, which are not clinically practical to the majority of head trauma and accidents that occur in developing countries (5, 6). The Corticosteroid Randomization After Significant Head in- jury (CRASH) model is one of the best prognostic tools for TBI, designed in recent years. It has two separate outcome prediction models for high and low-middle income coun- tries. As a result, the CRASH model can be applied to differ- ent populations unlike other prognostic models. Designed by the Medical Research Council (MRC) with a sample size of more than 10,000 people, the CRASH model predicts 14-day mortality and 6-month unfavorable outcome (7). Although this model’s discrimination and external validation have been evaluated in several studies (8-12), the findings have been contradictory. Another prediction model for patients with TBI, which has received more attention recently, is the International Mis- sion for Prognosis and Analysis of Clinical Trials in TBI (IM- PACT) (11). This model also aims to predict the outcome of mortality and unfavorable outcome for patients with TBI. CRASH and IMPACT are the most externally validated mod- els for predicting mortality and unfavorable outcome in TBI patients. Many studies have evaluated both models and con- firmed their validity. In a study by de Cássia Almeida Vieira et al., results show that both models have a similar prognos- tic value in predicting mortality and unfavorable outcome in severe TBI (13). However, there is still a difference of opinion regarding which of these models works better for predicting the outcome of brain injuries. Therefore, we aim to com- pare the value of these two prognostic models, CRASH and IMPACT, in predicting mortality and unfavorable outcome of head trauma patients. 2. Methods 2.1. Study design The present study is a systematic review and meta-analysis to compare the value of the two prognostic models of CRASH and IMPACT in predicting mortality and unfavorable out- come of patients with traumatic brain injury. For this pur- pose, we thoroughly searched the available databases and electronic resources to find all articles related to the topic of our study. This study was designed based on the guide- lines for meta-analysis of observational studies. The search- ing and summarizing data method has been reported in the previous meta-analyses of the researchers of this study (2, 14- 23). Here, we explain a summary of the activities carried out to achieve the goals of this study. 2.2. Description of PICO The description of PICO in the present study is as follows: The problem or study population (P): Human studies con- ducted on brain trauma patients Targeted intervention (I): Predictive value of CRASH and IM- PACT models Comparisons (C): Comparison with the survival group or fa- vorable outcome Outcome (O): Mortality and unfavorable outcome are the pri- mary outcomes of this study 2.3. Search strategy In the current research, we selected the keywords using three strategies: Medline database MeSH and Embase database Emtree, experts’ opinions in the field, and the review of key- words and the titles of related articles. Then, using an appro- priate combination of these keywords, an extensive search was conducted in the electronic databases of Medline (via PubMed), Embase, Scopus, and Web of Science until August 17, 2022. We presented the Medline search strategy in Ap- pendix 1. In addition to the systematic search, we performed This open-access article distributed under the terms of the Creative Commons Attribution NonCommercial 3.0 License (CC BY-NC 3.0). Downloaded from: https://journals.sbmu.ac.ir/aaem/index.php/AAEM/index 3 Archives of Academic Emergency Medicine. 2023; 11(1): e27 Figure 1: Flow diagram for selection of studies. a manual search in the Google search engine, Google scholar, and related article sources. 2.4. Eligibility criteria Inclusion criteria In the current study, we included the human studies con- ducted to investigate the prognostic value of CRASH and IM- PACT models in predicting mortality and unfavorable out- comes of traumatic brain injury. The research population is human studies without age, sex, and race restrictions. Exclusion criteria Failure to compare CRASH and IMPACT prognostic models simultaneously in one study, studies conducted on children, failure to report injury outcomes, failure to report assessment method, and review articles were the study’s exclusion crite- ria. 2.5. Data Extraction After removing duplicate articles from the systematic search results, two independent researchers performed the initial screening by reading the title and abstract of the articles. The full text of the relevant articles was examined in detail. We included the studies in the present systematic review based on the inclusion and exclusion criteria. We resolved any dis- agreements through discussion with a third researcher. The extracted data includes information related to the study design, sample characteristics (age, gender), number of ex- amined samples, the severity of TBI, the interval between the injury and evaluation of IMPACT and CRASH prognos- tic models, duration of follow-up, and quality assessment of selected articles. In cases where the data could not be extracted from the ar- ticle, we contacted the authors of the article. If the cor- responding author did not respond to the first email, two follow-up emails were sent (with an interval of one week). In cases of no responses, other article authors were contacted through professional social platforms such as Research Gate and LinkedIn to provide the required information to the re- searchers. 2.6. Quality assessment of the studies The quality of the studies was determined using the QUADAS-2 guidelines (24). To evaluate the agreement be- tween the two researchers, we investigated the inter-rater re- liability in the quality assessment of the studies. In case of disagreements, we resolved the difference through discus- sion with the third researcher. 2.7. The level of evidence The level of evidence was evaluated based on the Grading of Recommendations Assessment, Development and Evalu- ation (GRADE) framework (25). Since all the studies included in this study were observational, the base score of these ar- ticles starts from the Low level based on the GRADE guide. This open-access article distributed under the terms of the Creative Commons Attribution NonCommercial 3.0 License (CC BY-NC 3.0). Downloaded from: https://journals.sbmu.ac.ir/aaem/index.php/AAEM/index H. Zarei et al. 4 In the presence of a large magnitude of effect, dose-response gradient, and plausible confounders, the score of the level of evidence could be increased between 1 and 3 points. 2.8. Statistical analyses We performed the analyses in STATA 17.0. All studies were summarized and divided based on their outcome and pre- dictive value. All included studies reported the area under the curve for the IMPACT and CRASH models with a 95% confidence interval. Therefore, we calculated a pooled area under the curve (AUC) with a 95% confidence interval (95% CI). IMPACT model has three types: IMPACT core model, IM- PACT extended model, and IMPACT Lab model. IMPACT core model is calculated based on clinical findings. IMPACT extended model is calculated based on clinical findings and CT scan investigations (core model + CT scan). IMPACT Lab model includes clinical findings, CT scan, and Blood glucose and hemoglobin level (core model + CT scan + laboratory as- sessment). The CRASH model also includes two models, ba- sic and CT. The basic model is based on clinical findings, and the CT model requires CT scan findings in addition to clinical findings. Therefore, to compare the IMPACT and CRASH prognostic models’ areas under the curve, we compared the IMPACT core and CRASH basic models together, and the IMPACT ex- tended, IMPACT Lab, and CRASH CT models together. Fi- nally, the odds ratio (OR) was checked with a 95% confidence interval by performing a meta-regression to determine which model has the best prognostic performance. We used the random effect model in the present study due to the existence of heterogeneity. To check heterogeneity be- tween studies, we used the I2 test. Using Egger’s test method, Funnel Plot was used to identify publication bias (26). 3. Results 3.1. Studies’ characteristics The search resulted in 2174 articles. After removing dupli- cates, we included 1156 studies in our initial screening pro- cess. After examining the title and abstract of the included articles, we read 66 of them in detail. The data of 16 studies were included in the final analysis (27-42). These studies in- cluded the data of 39,829 patients with suspected traumatic brain injury. The reasons for the exclusion of articles from the present study included failure to examine the prognos- tic value of IMPACT and CRASH models simultaneously (29 articles), failure to report the required data (9 articles), re- view studies (4 articles), studies conducted on pediatric pop- ulation (2 articles) and non-related studies (6 articles). One study had not reported the AUC with 95% confidence inter- val, so we contacted the authors twice but received no re- sponse; therefore, this study was also excluded (43). Details on the number of studies excluded during selection for meta- analysis are provided in Figure 1. The interval between the occurrence of brain injury and the investigation of IMPACT and CRASH models varied between 12 to 48 hours. The sampling method was prospective in 10 studies, retrospective in 5, and amphi-directional in 1. The severity of TBI ranges from mild to severe. The follow-up pe- riod was 180 days in 12 studies, 14 days in 1 study, and 540 days in 1 study. Two studies followed the patients until dis- charge from the hospital (Table 1). 3.2. Quality assessment and publication bias We assessed the risk of bias based on the QUADAS-2 tool (Ta- ble 2). The risk of bias was high in three studies in the patient selection domain and unclear in six studies in the flow and timing domain. In other domains, all studies had a low risk of bias. We checked the publication bias for the prognostic value of IMPACT and CRASH models in predicting mortality and unfavorable outcome. The analysis showed that in in- vestigating the prognostic value of the IMPACT core model (p=0.045) and IMPACT extended model (p=0.021) in predict- ing the unfavorable outcome, there is evidence of publication bias (Figure S1 and Figure S2). In other scores, there was no publication bias in predicting mortality or unfavorable out- come. 3.3. Comparing the predictive values for mortal- ity CRASH basic and IMPACT core model The analyzes showed that the AUC of the CRASH basic and IMPACT core models in predicting the mortality of head trauma patients are 0.82 (95% CI: 0.77 to 0.86) and 0.80 (95% CI: 0.77 to 0.84), respectively (Figure 2). These two models have no difference in predicting the patients’ mortality (di- agnostic OR=0.98; 95% CI: 0.93 to 1.04; p=0.550). CRASH CT, IMPACT extended, and IMPACT Lab The analyzes showed that the AUC in the CRASH CT, IMPACT extended, and IMPACT lab models in predicting the mortal- ity of head trauma patients are 0.80 (95% CI: 0.74 to 0.86), 0.81 (95% CI: 0.77 to 0.84), and 0.80 (95% CI: 0.76 to 0.85), respectively (Figure 3). Results also showed that the IMACT extended (OR=1.00; 95% CI: 0.94 to 1.07; p=0.968) and IM- PACT Lab (OR=1.00; 95% CI: 0.93 to 1.07; p=0.986) have no difference with CRASH CT model in predicting the mortality of head trauma patients. 3.4. Comparing the predictive values for unfa- vorable outcomes CRASH basic and IMPACT core model The AUC for the CRASH basic and IMPACT core models in predicting the 6-month unfavorable outcome of head trauma This open-access article distributed under the terms of the Creative Commons Attribution NonCommercial 3.0 License (CC BY-NC 3.0). Downloaded from: https://journals.sbmu.ac.ir/aaem/index.php/AAEM/index 5 Archives of Academic Emergency Medicine. 2023; 11(1): e27 Figure 2: Comparison of area under the curve (AUC) of CRASH basic model and IMPACT core model in prediction of mortality in traumatic brain injury. CI: Confidence interval. patients were 0.82 (95% CI: 0.78 to 0.86) and 0.80 (95% CI: 0.77 to 0.83), respectively (Figure 4). These two models have no difference in predicting the unfavorable outcome of pa- tients (diagnostic OR=0.98; 95% CI: 0.93 to 1.03; p=0.462). CRASH CT, IMPACT extended, and IMPACT Lab model The analysis showed that the area under the curve of the CRASH CT, IMPACT extended, and IMPACT lab models in predicting the unfavorable outcome of head trauma patients This open-access article distributed under the terms of the Creative Commons Attribution NonCommercial 3.0 License (CC BY-NC 3.0). Downloaded from: https://journals.sbmu.ac.ir/aaem/index.php/AAEM/index H. Zarei et al. 6 Figure 3: Comparison of area under the curve (AUC) of CRASH CT model, IMPACT extended model, and IMPACT Lab model in prediction of mortality in traumatic brain injury. CI: Confidence interval. are 0.81 (95% CI: 0.76 to 0.86), 0.82 (95% CI: 0.78 to 0.85), and 0.80 (95% CI: 0.75 to 0.85), respectively (Figure 5). It was also obtained that the IMACT extended (OR=1.01; 95% CI: 0.95 to 1.08; p=0.729), and the IMPACT Lab model (OR=0.99; 95% CI: 0.93 to 1.06; p=0.819) have no difference with CRASH CT in predicting the unfavorable outcome of head trauma patients. This open-access article distributed under the terms of the Creative Commons Attribution NonCommercial 3.0 License (CC BY-NC 3.0). Downloaded from: https://journals.sbmu.ac.ir/aaem/index.php/AAEM/index 7 Archives of Academic Emergency Medicine. 2023; 11(1): e27 Figure 4: Comparison of area under the curve (AUC) of CRASH basic model and IMPACT core model in prediction of unfavorable outcome of traumatic brain injury. CI: Confidence interval. 3.5. Certainty of evidence Assessment of certainty of evidence according to GRADE framework showed that the quality of evidence in the prog- nostic value of CRASH and IMPACT models in the predic- tion of mortality is low, since there was severe heterogeneity (rated down one point) and a large magnitude of effect (rated up one point). This open-access article distributed under the terms of the Creative Commons Attribution NonCommercial 3.0 License (CC BY-NC 3.0). Downloaded from: https://journals.sbmu.ac.ir/aaem/index.php/AAEM/index H. Zarei et al. 8 Figure 5: Comparison of area under the curve (AUC) of CRASH CT model, IMPACT extended model, and IMPACT Lab model in prediction of unfavorable outcome of traumatic brain injury. CI: Confidence interval. However, the level of evidence in the prognostic value of CRASH and IMPACT models in the prediction of unfavorable outcome are low to very low. There was a large magnitude of effect in all subscales; therefore, the level of evidence was This open-access article distributed under the terms of the Creative Commons Attribution NonCommercial 3.0 License (CC BY-NC 3.0). Downloaded from: https://journals.sbmu.ac.ir/aaem/index.php/AAEM/index 9 Archives of Academic Emergency Medicine. 2023; 11(1): e27 Table 1: Characteristics of eligible studies Study Sample size Male (n) Age* Injury to assess- ment (hrs) Sampling method TBI severity Assessed scale Outcome Follow-up (day) Abdollah; 2021; Malaysia 281 NR NR 0 Prospective All severities Core / Basic Mortality / unfavorable 14 / 180 Camarano; 2020; US, Canada 26228 NR 45 (27-63) 0 Retrospective Moderate to severe Core / Basic Mortality 14 Castaño-Leon; 2016; Spain 1301 NR 40 (24-52) 48 Prospective Moderate to severe Core / Basic / Extended / CT Mortality / unfavorable 180 Charry; 2017; Colombia 127 107 >18 12 Retrospective Moderate to severe Extended / CT / Lab Mortality 180 Charry; 2019; Colombia 309 240 >18 12 Retrospective All severities Extended / CT / Lab Mortality / unfavorable 180 Dijkland; 2020; Europe 1742 NR 51 (32-67) 24 Prospective All severities Core / Basic / Extended / CT /Lab Mortality / unfavorable 14 / 180 Elahi; 2020; Tan- zania 2972 2452 31.1 (15.2) 0 Prospective All severities Core / Basic Unfavorable In-hospital Han; 2014; Sin- gapore 300 NR 53 (20.7) 0 Prospective Severe Core / Basic / Extended / CT / Lab Mortality / unfavorable 14 / 180 Harrison; 2015; UK 2975 2263 >16 24 Prospective All severities Core / Basic / Extended / CT / Lab Mortality / unfavorable 180 Honeybul; 2016; Australia 319 260 32 (21-47) 0 Prospective All severities Core / Basic / Extended / CT /Lab Mortality / unfavorable 540 Maeda; 2019; Japan 635 442 age>16 0 Retrospective Severe Core / Basic / Extended / CT Unfavorable 180 Majdan; 2014; Austria 778 NR 50 (28-69) 48 Prospective All severities Core / Basic / Extended / CT Mortality / unfavorable 180 Pranav; 2022; USA, Uganda 877 746 31.3 (NR) 0 Prospective All severities Core / Basic Mortality In-hospital Wong; 2013; Hong Kong 178 127 56 (20) 0 Prospective All severities Core / Basic Mortality / unfavorable 14 / 180 Wongchareon; 2020; South America 466 NR 28 (21-43) 24 Retrospective Severe Core / Basic / Extended / CT / Lab Mortality / unfavorable 14 / 180 Xu; 2022; China 341 243 54 (17.4) / 56.2(15.4) 12 Amphidirectional Moderate to severe Core / Basic Mortality / unfavorable 14 / 180 *, Data are presented as rang, mean (±SD) or median (IQR). Basic: CRASH basic model; Core: Impact core model; CT: CRASH Computed Tomography model; Extended: IMPACT extended model; Lab: IMPACT laboratory model; NR: Not reported; TBI: Traumatic brain injury rated up one point. In addition, significant inconsistency was observed and the score was rated down one point in all sub- scales. Finally, the analyses showed possible publication bias in assessing the prognostic value of IMPACT core and IM- PACT extended models. Therefore, the level of evidence for the prognostic value of IMPACT core and IMPACT extended models were judged as very low (Table 3). 4. Discussion Low to very low level of evidence shows that the IMPACT and CRASH models have similar values in predicting mortality and unfavorable outcome of patients with traumatic brain injury. Similar findings were also observed in the sub-scores of the two models. As a result, the prognostic values of all sub-scores in predicting mortality and unfavorable outcome of traumatic brain injury secondary to trauma are similar. The results of the present study showed that IMPACT ex- tended and IMPACT Lab models have equal value in pre- This open-access article distributed under the terms of the Creative Commons Attribution NonCommercial 3.0 License (CC BY-NC 3.0). Downloaded from: https://journals.sbmu.ac.ir/aaem/index.php/AAEM/index H. Zarei et al. 10 Table 2: Risk of bias assessment of included studies Study Risk of Bias Applicability Patient selection Index tests Reference standard Flow and timing Patient selection Index tests Reference standard Abdollah; 2021 Low Low Low Low Low Low Low Camarano; 2020 Low Low Low Unclear Low Low Low Castaño-Leon; 2016 Low Low Low Low Low Low Low Charry; 2017 High Low Low Unclear Low Low Low Charry; 2019 High Low Low Unclear Low Low Low Dijkland; 2020 Low Low Low Low Low Low Low Elahi; 2020 Low Low Low Unclear Low Low Low Han; 2014 Low Low Low Low Low Low Low Harrison; 2015 Low Low Low Low Low Low Low Honeybul; 2016 Low Low Low Low Low Low Low Maeda; 2019 Low Low Low Low Low Low Low Majdan; 2014 Low Low Low Low Low Low Low Pranav; 2022 Low Low Low Unclear Low Low Low Wong; 2013 Low Low Low Low Low Low Low Wongchareon; 2020 Low Low Low Low Low Low Low Xu; 2022 High Low Low Unclear Low Low Low Low: Low risk of bias; High: High risk of bias; Unclear: Unclear risk of bias. dicting mortality and unfavorable outcome of patients with traumatic brain injury. In addition to IMPACT extended vari- ables, the IMPACT Lab model requires serum biomarkers such as blood sugar and hemoglobin level. Therefore, using the IMPACT extended model in the clinical setting is better because it requires fewer variables for calculation. It should be noted that the CRASH CT model can only be calculated by adding CT scan findings to CRASH basic, making it easier to use clinically. The area under the curve of CRASH basic and IMPACT core models are similar in predicting the outcome of patients with traumatic brain injury. To calculate the CRASH basic model score, variables such as the country’s economic status (de- veloping or developed country), level of conciseness, pupils’ reaction to light, and major extracranial injury are needed. The IMPACT core model requires variables such as age, mo- tor score, and pupils’ reaction to light, which are not much different from the CRASH basic model. Therefore, using ei- ther of the two models will have a similar application in man- aging patients with traumatic brain injury. One of the aims of the present study was to investigate the prognostic value of IMPACT and CRASH models based on the severity of brain injury. The articles included in our analy- sis reported mild, moderate, and severe injury severity (8, 12, 44-50). However, most of the studies did not perform analy- sis based on the severity of injuries. To evaluate the discrimination of the two prognostic models, most of the included studies confined their results to only the area under the curve. Nonetheless, it should be kept in mind that the area under the curve is the early stage in the assessment of the diagnostic accuracy and predictive value of a model. At the same time, the analysis based on a cut point that reports sensitivity and specificity is more useful in the clinical setting (51). In this regard, in addition to a systematic search, the researchers of the present study also conducted an extensive manual search to find an article that compares the prognostic value of the IMPACT and CRASH models based on a cutoff point. This additional search re- sulted in very few studies. Most studies that report sensitivity and specificity based on a cut point did not use a prespecified threshold and often tried to find the best cut point based on the Youden index (52). As a result, the cut points used are very different, making it impossible to pool the data in this sec- tion. Also, some studies used unconventional cutoff points, which made the sensitivity too low to report higher specificity (50, 53). However, in prognostic tools and screening tests, sensitivity is more valuable than specificity. In this regard, the results reported by Wongchareon et al. (n=466) indicated that the sensitivity of CRASH core and CRASH CT models in prediction of 14-day mortality as 8% and 13%, which were as- sociated with a specificity of 99% and 97%, respectively. They also reported the sensitivity of IMPACT core, extended, and lab models equal to 36%, 44%, and 36%, and their specifici- ties as 87%, 87%, and 89%, respectively (53). Camarano et al. determined the optimal cutoff threshold using Youden’s in- dex; the CRASH basic and IMPACT core model’s cutoff val- ues were 33.1% and 42.8%, respectively. Using these cut- off thresholds, they reported the sensitivity and specificity of the CRASH basic and IMPACT core model for predicting in- hospital mortality. For CRASH basic and IMPACT core mod- els, the sensitivity was 78% and 80%, and the specificity was 80% and 78%, respectively (52). This open-access article distributed under the terms of the Creative Commons Attribution NonCommercial 3.0 License (CC BY-NC 3.0). Downloaded from: https://journals.sbmu.ac.ir/aaem/index.php/AAEM/index 11 Archives of Academic Emergency Medicine. 2023; 11(1): e27 Table 3: Certainty in evidence based on GRADE framework Outcome/ model Sample size Effect size Risk of bias Imprecision Inconsistency (I2 ) Indirectness Publication bias Judgment and level of evi- dence Mortality CRASH basic 35800 0.82 (0.77, 0.86) Not serious Not serious Serious Not serious Not present Low: Rated down 1 /point - Presence of serious inconsis- tency Rated up 1 /point Large magnitude of effect IMPACT core 35056 0.80 (0.77, 0.84) Not serious Not serious Serious Not serious Not present Low: Rated down 1 /point - Presence of serious inconsis- tency Rated up 1 /point Large magnitude of effect CRASH CT 7020 0.80 (0.74, 0.86) Not serious Not serious Serious Not serious Not present Low: Rated down 1 /point - Presence of serious inconsis- tency Rated up 1 /point Large magnitude of effect IMPACT ex- tended 7169 0.81 (0.77, 0.84) Not serious Not serious Serious Not serious Not present Low: Rated down 1 /point - Presence of serious inconsis- tency Rated up 1 /point Large magnitude of effect IMPACT labo- ratory 5502 0.80 (0.76, 0.85) Not serious Not serious Serious Not serious Not present Low: Rated down 1 point - Presence of serious inconsis- tency Rated up 1 /point Large magnitude of effect Unfavorable outcome CRASH basic 12302 0.82 (0.76, 0.86) Not serious Not serious Serious Not serious Not present Low: Rated down 1 /point - Presence of serious inconsis- tency Rated up 1 /point Large magnitude of effect IMPACT core 11558 0.80 (0.77, 0.83) Not serious Not serious Serious Not serious Likely Very low: Rated down 2 /points - Presence of serious inconsistency - Possible pub- lication bias Rated up 1 /point Large magnitude of effect CRASH CT 7528 0.81 (0.76, 0.86) Not serious Not serious Serious Not serious Not present Low: Rated down 1 /point - Presence of serious inconsis- tency Rated up 1 /point Large magnitude of effect IMPACT ex- tended 7804 0.82 (0.78, 0.85) Not serious Not serious Serious Not serious Likely Very low: Rated down 2 /points - Presence of serious inconsistency - Possible pub- lication bias Rated up 1 /point Large magnitude of effect IMPACT labo- ratory 5502 0.80 (0.75, 0.85) Not serious Not serious Serious Not serious Not present Low: Rated down 1 /point - Presence of serious inconsis- tency Rated up 1 /point Large magnitude of effect GRADE: Grading of Recommendations Assessment, Development and Evaluation; CRASH: Corticosteroid Randomization After Significant Head injury; IMPACT: International Mission for Prognosis and Analysis of Clinical Trials; CT: computed tomography. We also found two studies in which only one of the models was evaluated. In the first study, Hashemi et al. stated that the patients whose expected risk based on the CRASH ba- sic and CRASH CT model is equal to 63.9 and 51.2, respec- tively, are high-risk patients in terms of 14-day mortality. The study also introduced 43.2 and 78.7 as cutoff points for the 6-month unfavorable outcome (54). In another study using children and adolescents as the patient population, Fazel et al. showed that the best cutoff points for identifying high-risk head trauma for 14-day mortality in children using CRASH basic and CRASH CT models are 46 and 30. This study re- ported the cutoff points for 6-month unfavorable outcome as 17 and 13 (55). Another limitation was the retrospective design of some of the included studies. The retrospective method affects the risk of bias in the patient selection, flow, and timing do- This open-access article distributed under the terms of the Creative Commons Attribution NonCommercial 3.0 License (CC BY-NC 3.0). Downloaded from: https://journals.sbmu.ac.ir/aaem/index.php/AAEM/index H. Zarei et al. 12 main. When data is collected retrospectively, the sampling is not random, which results in a high risk of bias. Similarly, flow and timing in diagnostic or prognostic studies should be such that the index test is checked first and then the ref- erence standard (24). In retrospective studies, the outcome of the patients is known from the beginning. Therefore, it is impossible to be sure of the risk of bias in the flow and tim- ing domain. As a result, the risk of bias in these studies in the flow and timing domain was considered unclear. Simplicity should be kept in mind when designing a scor- ing system. The fewer variables to calculate in a prognos- tic model, the easier its use. Both the CRASH and IMPACT models have a core / basic model that requires variables only from history and physical exam. They also have more com- plex models (CT, extended, and Lab models) that include CT scans or laboratory findings in addition to the clinical find- ings of the patients. As figures 2 to 5 demonstrate, the area under the curve in the IMPACT core and CRASH basic model is not much different from IMPACT extended, IMPACT Lab, and CRASH CT models. Therefore, it may be possible to check the outcome of patients only based on CRASH basic and IMPACT core models. This finding has been confirmed by other studies (47, 50). After a TBI, prognostic models can help the medical team al- locate resources and improve the quality of care given (13). As results showed, both models have a good discrimination power to predict TBI patients’ prognosis. Their clinical use can prevent the health care system from unnecessary imag- ing and save time and money, and reduce the patient’s ex- posure to ionizing radiation. We suggest a systematic review and meta-analysis of the diagnostic value of CRASH and IM- PACT models in detecting intracranial hemorrhage in TBI pa- tients. 5. Conclusion Our results demonstrate the similar performance and predic- tive value of both models in predicting mortality and unfa- vorable outcome of patients with traumatic brain injury. The analysis showed that the IMPACT core model and CRASH ba- sic model have equal prognostic value in predicting mortal- ity and unfavorable outcome for patients with traumatic in- juries to the brain. So, both models can aid in estimating the prognosis for TBI patients in clinical practice. The CRASH CT model is like the IMPACT extended and IMPACT Lab models. Since the discriminative power of the IMPACT Core and CRASH basic models is not different from the IMPACT ex- tended, IMPACT Lab, and CRASH CT models, it may be pos- sible to rely only on the core and basic models in examining the prognosis of patients with traumatic brain injury. 6. Declarations 6.1. Acknowledgments None. 6.2. Conflict of interest There is no conflict of interest. 6.3. Funding information The study was funded and supported by Iran university of medical sciences; Grant no: 97-4-37-13848. 6.4. Authors Contribution Study design: MY, AS Data gathering: HZ, MV, SRD, HAR Analysis: MY Interpretation: All authors Drafting: HZ, SRD, HAR Revising: All authors Reading and approving final version: All authors References 1. Badhiwala JH, Wilson JR, Fehlings MG. Global burden of traumatic brain and spinal cord injury. Lancet Neurol. 2019;18(1):24-5. 2. Ebrahimi A, Yousefifard M, Mohammad Kazemi H, Ra- souli HR, Asady H, Moghadas Jafari A, et al. Diagnostic Accuracy of Chest Ultrasonography versus Chest Radio- graphy for Identification of Pneumothorax: A Systematic Review and Meta-Analysis. Tanaffos. 2014;13(4):29-40. 3. Yousefifard M, Baikpour M, Ghelichkhani P, Asady H, Shahsavari Nia K, Moghadas Jafari A, et al. Screening Per- formance Characteristic of Ultrasonography and Radio- graphy in Detection of Pleural Effusion; a Meta-Analysis. Emerg (Tehran). 2016;4(1):1-10. 4. Isokuortti H, Iverson GL, Silverberg ND, Kataja A, Bran- der A, Öhman J, et al. Characterizing the type and loca- tion of intracranial abnormalities in mild traumatic brain injury. J Neurosurg. 2018;129(6):1588-97. 5. Schonfeld D, Bressan S, Da Dalt L, Henien MN, Winnett JA, Nigrovic LE. Pediatric Emergency Care Applied Re- search Network head injury clinical prediction rules are reliable in practice. Arch Dis Child. 2014;99(5):427-31. 6. Kuppermann N, Holmes JF, Dayan PS, Hoyle JD, Jr., Atabaki SM, Holubkov R, et al. Identification of children at very low risk of clinically-important brain injuries af- ter head trauma: a prospective cohort study. Lancet. 2009;374(9696):1160-70. 7. Perel P, Arango M, Clayton T, Edwards P, Komolafe E, Poc- cock S, et al. Predicting outcome after traumatic brain in- This open-access article distributed under the terms of the Creative Commons Attribution NonCommercial 3.0 License (CC BY-NC 3.0). Downloaded from: https://journals.sbmu.ac.ir/aaem/index.php/AAEM/index 13 Archives of Academic Emergency Medicine. 2023; 11(1): e27 jury: practical prognostic models based on large cohort of international patients. BMJ. 2008;336(7641):425-9. 8. Honeybul S, Ho KM, Lind CR, Gillett GR. Observed versus predicted outcome for decompressive craniec- tomy: a population-based study. J Neurotrauma. 2010;27(7):1225-32. 9. Lingsma H, Andriessen TM, Haitsema I, Horn J, van der Naalt J, Franschman G, et al. Prognosis in moderate and severe traumatic brain injury: external validation of the IMPACT models and the role of extracranial injuries. J Trauma Acute Care Surg. 2013;74(2):639-46. 10. Majdan M, Lingsma HF, Nieboer D, Mauritz W, Rusnak M, Steyerberg EW. Performance of IMPACT, CRASH and Nijmegen models in predicting six month outcome of pa- tients with severe or moderate TBI: an external validation study. Scand J Trauma Resusc Emerg Med. 2014;22:68. 11. Roozenbeek B, Lingsma HF, Lecky FE, Lu J, Weir J, Butcher I, et al. Prediction of outcome after moderate and severe traumatic brain injury: external validation of the International Mission on Prognosis and Analysis of Clinical Trials (IMPACT) and Corticoid Randomisation After Significant Head injury (CRASH) prognostic mod- els. Crit Care Med. 2012;40(5):1609-17. 12. Wong GK, Teoh J, Yeung J, Chan E, Siu E, Woo P, et al. Out- comes of traumatic brain injury in Hong Kong: validation with the TRISS, CRASH, and IMPACT models. J Clin Neu- rosci. 2013;20(12):1693-6. 13. de Cássia Almeida Vieira R, Silveira JCP, Paiva WS, de Oliveira DV, de Souza CPE, Santana-Santos E, et al. Prognostic Models in Severe Traumatic Brain Injury: A Systematic Review and Meta-analysis. Neurocrit Care. 2022;37(3):790-805. 14. Yousefifard M, Rahimi-Movaghar V, Baikpour M, Ghe- lichkhani P, Hosseini M, Jafari A, et al. Early versus late spinal decompression surgery in treatment of trau- matic spinal cord injuries; a systematic review and meta- analysis. Emerg (Tehran). 2017;5(1):e37. 15. Nasirinezhad F, Hosseini M, Karami Z, Yousefifard M, Janzadeh A. Spinal 5-HT3 receptor mediates nociceptive effect on central neuropathic pain; possible therapeutic role for tropisetron. J Spinal Cord Med. 2016;39(2):212-9. 16. Hosseini M, Yousefifard M, Aziznejad H, Nasirinezhad F. The Effect of Bone Marrow-Derived Mesenchymal Stem Cell Transplantation on Allodynia and Hyperalgesia in Neuropathic Animals: A Systematic Review with Meta- Analysis. Biol Blood Marrow Transplant. 2015;21(9):1537- 44. 17. Hosseini M, Ghelichkhani P, Baikpour M, Tafakhori A, Asady H, Haji Ghanbari MJ, et al. Diagnostic Accuracy of Ultrasonography and Radiography in Detection of Pulmonary Contusion; a Systematic Review and Meta- Analysis. Emerg (Tehran). 2015;3(4):127-36. 18. Julian P. T. Higgins, Green S. Cochrane Handbook for Sys- tematic Reviews of Interventions: John Wiley & Sons; 2011. 19. Ghelichkhani P, Yousefifard M, Nazemi L, Safari S, Hos- seini M, Baikpour M, et al. The Value of Serum B-Subunit of Human Chorionic Gonadotropin Level in Prediction of Treatment Response to Methotrexate in Management of Ectopic Pregnancy; a Systematic Review and Meta- Analysis. Int J Pediatr. 2016;4(9):3503-18. 20. Safari S, Yousefifard M, Baikpour M, Rahimi-Movaghar V, Abiri S, Falaki M, et al. Validation of thoracic injury rule out criteria as a decision instrument for screening of chest radiography in blunt thoracic trauma. J Clin Or- thop Trauma. 2016;7(2):95-100. 21. Rahimi-Movaghar V, Yousefifard M, Ghelichkhani P, Baikpour M, Tafakhori A, Asady H, et al. Application of Ultrasonography and Radiography in Detection of Hemothorax; a Systematic Review and Meta-Analysis. Emerg (Tehran). 2016;4(3):116-26. 22. Izadi A, Yousefifard M, Nakhjavan-Shahraki B, Baikpour M, Mirzay Razaz J, Hosseini M. Diagnostic Value of Urinary Neutrophil Gelatinase-Associated Lipocalin (NGAL) in Detection of Pediatric Acute Kidney Injury; a Systematic Review and Meta-Analysis. Int J Pediatr. 2016;4(11):3875-95. 23. Hassanzadeh-Rad A, Yousefifard M, Katal S, Asady H, Fard-Esfahani A, Moghadas Jafari A, et al. The value of (18) F-fluorodeoxyglucose positron emission tomogra- phy for prediction of treatment response in gastroin- testinal stromal tumors: a systematic review and meta- analysis. J Gastroenterol Hepatol. 2016;31(5):929-35. 24. Whiting PF, Rutjes AW, Westwood ME, Mallett S, Deeks JJ, Reitsma JB, et al. QUADAS-2: a revised tool for the qual- ity assessment of diagnostic accuracy studies. Ann Intern Med. 2011;155(8):529-36. 25. Guyatt GH, Oxman AD, Schünemann HJ, Tugwell P, Knot- tnerus A. GRADE guidelines: a new series of articles in the Journal of Clinical Epidemiology. J Clin Epidemiol. 2011;64(4):380-2. 26. Egger M, Davey Smith G, Schneider M, Minder C. Bias in meta-analysis detected by a simple, graphical test. BMJ. 1997;315(7109):629-34. 27. Majdan M, Lingsma HF, Nieboer D, Mauritz W, Rusnak M, Steyerberg EWJSjot, resuscitation, et al. Performance of IMPACT, CRASH and Nijmegen models in predicting six month outcome of patients with severe or moderate TBI: an external validation study. 2014;22(1):1-10. 28. Wong GKC, Teoh J, Yeung J, Chan E, Siu E, Woo P, et al. Outcomes of traumatic brain injury in Hong Kong: val- idation with the TRISS, CRASH, and IMPACT models. 2013;20(12):1693-6. 29. Castaño-Leon AM, Lora D, Munarriz PM, Cepeda S, Pare- This open-access article distributed under the terms of the Creative Commons Attribution NonCommercial 3.0 License (CC BY-NC 3.0). Downloaded from: https://journals.sbmu.ac.ir/aaem/index.php/AAEM/index H. Zarei et al. 14 des I, de la Cruz J, et al. Predicting outcomes after severe and moderate traumatic brain injury: an external vali- dation of impact and crash prognostic models in a large Spanish cohort. 2016;33(17):1598-606. 30. Charry JD, Tejada JH, Pinzon MA, Tejada WA, Ochoa JD, Falla M, et al. Predicted unfavorable neurologic outcome is overestimated by the Marshall Computed Tomogra- phy Score, Corticosteroid Randomization After Signifi- cant Head Injury (CRASH), and International Mission for Prognosis and Analysis of Clinical Trials in Traumatic Brain Injury (IMPACT) models in patients with severe traumatic brain injury managed with early decompres- sive craniectomy. 2017;101:554-8. 31. Charry JD, Navarro-Parra S, Solano J, Moscote-Salazar L, Pinzón MA, Tejada JHJNiNP. Outcomes of traumatic brain injury: the prognostic accuracy of various scores and models. 2019;53(1):55-60. 32. Han J, King NK, Neilson SJ, Gandhi MP, Ng IJJon. External validation of the CRASH and IMPACT prognostic models in severe traumatic brain injury. 2014;31(13):1146-52. 33. Harrison DA, Griggs KA, Prabhu G, Gomes M, Lecky FE, Hutchinson PJ, et al. External validation and recalibra- tion of risk prediction models for acute traumatic brain injury among critically ill adult patients in the United Kingdom. 2015;32(19):1522-37. 34. Honeybul S, Ho KMJI. Predicting long-term neurological outcomes after severe traumatic brain injury requiring decompressive craniectomy: a comparison of the CRASH and IMPACT prognostic models. 2016;47(9):1886-92. 35. Maeda Y, Ichikawa R, Misawa J, Shibuya A, Hishiki T, Maeda T, et al. External validation of the TRISS, CRASH, and IMPACT prognostic models in severe traumatic brain injury in Japan. 2019;14(8):e0221791. 36. Xu L, An T, Li C, Shi X, Yang BJJoIR. Development and Verification of Prognostic Prediction Models for Patients with Brain Trauma Based on Coagulation Function In- dexes. 2022;2022. 37. Camarano JG, Ratliff HT, Korst GS, Hrushka JM, Jupiter DCJI. Predicting in-hospital mortality after traumatic brain injury: external validation of CRASH-basic and IMPACT-core in the national trauma data bank. 2021;52(2):147-53. 38. Wongchareon K, Thompson HJ, Mitchell PH, Barber J, Temkin NJIp. IMPACT and CRASH prognostic models for traumatic brain injury: external validation in a South- American cohort. 2020;26(6):546-54. 39. Dijkland SA, Helmrich IRR, Nieboer D, Van Der Jagt M, Dippel DW, Menon DK, et al. Outcome prediction after moderate and severe traumatic brain injury: external val- idation of two established prognostic models in 1742 Eu- ropean patients. 2021;38(10):1377-88. 40. Elahi C, Adil SM, Sakita F, Mmbaga BT, Hernandes Rocha TA, Fuller A, et al. Corticosteroid randomization after sig- nificant head injury and international mission for prog- nosis and clinical trials in traumatic brain injury mod- els compared with a machine learning-based predictive model from Tanzania. 2022;39(1-2):151-8. 41. Warman PI, Seas A, Satyadev N, Adil SM, Kolls BJ, Haglund MM, et al. Machine learning for predicting in-hospital mortality after traumatic brain injury in both high-income and low-and middle-income coun- tries. 2022;90(5):605-12. 42. Abdullah JM, Idris Z, Ghani AR, Lim MSJJons. Prediction of outcomes in traumatic brain injury: the IMPACT and CRASH prognostic models in a single neurosurgical cen- ter, Malaysia. 2021. 43. Roozenbeek B, Lingsma HF, Lecky FE, Lu J, Weir J, Butcher I, et al. Prediction of outcome after moder- ate and severe traumatic brain injury: external vali- dation of the IMPACT and CRASH prognostic models. 2012;40(5):1609. 44. Charry JD, Navarro-Parra S, Solano J, Moscote-Salazar L, Pinzón MA, Tejada JH. Outcomes of traumatic brain in- jury: the prognostic accuracy of various scores and mod- els. Neurol Neurochir Pol. 2019;53(1):55-60. 45. Harrison DA, Griggs KA, Prabhu G, Gomes M, Lecky FE, Hutchinson PJ, et al. External Validation and Recalibra- tion of Risk Prediction Models for Acute Traumatic Brain Injury among Critically Ill Adult Patients in the United Kingdom. J Neurotrauma. 2015;32(19):1522-37. 46. Maeda Y, Ichikawa R, Misawa J, Shibuya A, Hishiki T, Maeda T, et al. External validation of the TRISS, CRASH, and IMPACT prognostic models in severe traumatic brain injury in Japan. PLoS One. 2019;14(8):e0221791. 47. Abdullah JM, Idris Z, Ghani AR, Lim MS. Prediction of outcomes in traumatic brain injury: the IMPACT and CRASH prognostic models in a single neurosurgical cen- ter, Malaysia. J Neurosurg Sci. 2021. 48. Dijkland SA, Helmrich I, Nieboer D, van der Jagt M, Dip- pel DWJ, Menon DK, et al. Outcome Prediction after Moderate and Severe Traumatic Brain Injury: External Validation of Two Established Prognostic Models in 1742 European Patients. J Neurotrauma. 2021;38(10):1377-88. 49. Elahi C, Adil SM, Sakita F, Mmbaga BT, Hernandes Rocha TA, Fuller A, et al. Corticosteroid Randomization after Significant Head Injury and International Mission for Prognosis and Clinical Trials in Traumatic Brain Injury Models Compared with a Machine Learning-Based Pre- dictive Model from Tanzania. J Neurotrauma. 2022;39(1- 2):151-8. 50. Warman PI, Seas A, Satyadev N, Adil SM, Kolls BJ, Haglund MM, et al. Machine Learning for Predicting In- Hospital Mortality After Traumatic Brain Injury in Both High-Income and Low- and Middle-Income Countries. This open-access article distributed under the terms of the Creative Commons Attribution NonCommercial 3.0 License (CC BY-NC 3.0). Downloaded from: https://journals.sbmu.ac.ir/aaem/index.php/AAEM/index 15 Archives of Academic Emergency Medicine. 2023; 11(1): e27 Neurosurgery. 2022;90(5):605-12. 51. Halligan S, Altman DG, Mallett S. Disadvantages of using the area under the receiver operating characteristic curve to assess imaging tests: a discussion and proposal for an alternative approach. Eur Radiol. 2015;25(4):932-9. 52. Camarano JG, Ratliff HT, Korst GS, Hrushka JM, Jupiter DC. Predicting in-hospital mortality after traumatic brain injury: External validation of CRASH-basic and IMPACT-core in the national trauma data bank. Injury. 2021;52(2):147-53. 53. Wongchareon K, Thompson HJ, Mitchell PH, Barber J, Temkin N. IMPACT and CRASH prognostic models for traumatic brain injury: external validation in a South- American cohort. Inj Prev. 2020;26(6):546-54. 54. Hashemi B, Amanat M, Baratloo A, Forouzanfar MM, Rahmati F, Motamedi M, et al. Validation of CRASH Model in Prediction of 14-day Mortality and 6-month Unfavorable Outcome of Head Trauma Patients. Emerg (Tehran). 2016;4(4):196-201. 55. Fazel M, Ahmadi S, Hajighanbari MJ, Baratloo A, Shah- savarinia K, Hosseini M, et al. Validation of CRASH Model in Prediction of 14-Day Mortality and 6-Month Unfavourable Outcome of Pediatric Traumatic Brain In- jury. Int J Pediatr. 2019;7(12):10413-22. This open-access article distributed under the terms of the Creative Commons Attribution NonCommercial 3.0 License (CC BY-NC 3.0). Downloaded from: https://journals.sbmu.ac.ir/aaem/index.php/AAEM/index H. Zarei et al. 16 Appendix 1: Medline (PubMed) search query Medline (via PubMed) 1. "Decision Support Techniques" [Mesh] OR International Mission for Prognosis and Analysis of Clinical Trials in Traumatic Brain In- jury[tiab] OR IMPACT prognostic models[tiab] OR IMPACT prognostic model[tiab] OR IMPACT Rule[tiab] OR IMPACT model[tiab] OR IMPACT CORE[tiab] OR IMPACT basic[tiab] OR IMPACT-CORE[tiab] OR IMPACT-basic[tiab] OR IMPACT CT[tiab] OR IMPACT Lab[tiab] OR IMPACT- CT[tiab] OR IMPACT-Lab[tiab] OR Corticosteroid Randomization after Significant Head Injury[tiab] OR CRASH prognostic models[tiab] OR CRASH prognostic model[tiab] OR CRASH Rule[tiab] OR CRASH model[tiab] OR CRASH Core[tiab] OR CRASH basic[tiab] OR CRASH CT[tiab] OR CRASH-basic[tiab] OR CRASH-CT[tiab] 2. "Brain Concussion"[Mesh] OR "Brain Injuries"[Mesh] OR "Brain Injuries, Traumatic"[Mesh] OR Brain Concussion[tiab] OR Brain In- juries[tiab] OR Brain Injuries, Traumatic[tiab] OR Brain Concussions[tiab] OR Concussion, Brain[tiab] OR Commotio Cerebri[tiab] OR Cerebral Concussion[tiab] OR Cerebral Concussions[tiab] OR Concussion, Cerebral[tiab] OR Concussion, Intermediate[tiab] OR Intermediate Con- cussion[tiab] OR Intermediate Concussions[tiab] OR Concussion, Severe[tiab] OR Severe Concussion[tiab] OR Severe Concussions[tiab] OR Concussion, Mild[tiab] OR Mild Concussion[tiab] OR Mild Concussions[tiab] OR Mild Traumatic Brain Injury[tiab] OR Injuries, Brain[tiab] OR Brain Injury[tiab] OR Injury, Brain[tiab] OR Injuries, Acute Brain[tiab] OR Acute Brain Injuries[tiab] OR Acute Brain Injury[tiab] OR Brain Injury, Acute[tiab] OR Injury, Acute Brain[tiab] OR Brain Injuries, Acute[tiab] OR Brain Lacerations[tiab] OR Brain Laceration[tiab] OR Lacera- tion, Brain[tiab] OR Lacerations, Brain[tiab] OR Brain Injuries, Focal[tiab] OR Brain Injury, Focal[tiab] OR Focal Brain Injury[tiab] OR Injuries, Focal Brain[tiab] OR Injury, Focal Brain[tiab] OR Focal Brain Injuries[tiab] OR Brain Injury, Traumatic[tiab] OR Traumatic Brain Injuries[tiab] OR Trauma, Brain[tiab] OR Brain Trauma[tiab] OR Brain Traumas[tiab] OR Traumas, Brain[tiab] OR Encephalopathy, Traumatic[tiab] OR En- cephalopathies, Traumatic[tiab] OR Traumatic Encephalopathies[tiab] OR Injury, Brain, Traumatic[tiab] OR Traumatic Encephalopathy[tiab] OR Traumatic Brain Injury[tiab] OR TBI[tiab] 3. #1 AND #2 Embase 1. ’international mission for prognosis and analysis of clinical trials in traumatic brain injury’:ab,ti OR ’impact prognostic models’:ab,ti OR ’im- pact rule’:ab,ti OR ’impact model’:ab,ti OR ’impact core’:ab,ti OR ’impact basic’:ab,ti OR ’impact-core’:ab,ti OR ’impact-basic’:ab,ti OR ’impact ct’:ab,ti OR ’impact lab’:ab,ti OR ’impact-ct’:ab,ti OR ’impact-lab’:ab,ti OR ’corticosteroid randomization after significant head injury’:ab,ti OR ’crash prognostic models’:ab,ti OR ’crash prognostic model’:ab,ti OR ’crash rule’:ab,ti OR ’crash model’:ab,ti OR ’crash core’:ab,ti OR ’crash basic’:ab,ti OR ’crash ct’:ab,ti OR ’crash-basic’:ab,ti OR ’crash-ct’:ab,ti 2. ’brain injury’/exp OR ’brain injury’ OR ’head injury’/exp OR ’head injury’ OR ’traumatic brain injury’/exp OR ’traumatic brain injury’ 3. #1 AND #2 Scopus 1- TITLE-ABS-KEY("International Mission for Prognosis and Analysis of Clinical Trials in Traumatic Brain Injury" OR "IMPACT prognostic models" OR "IMPACT prognostic model" OR "IMPACT Rule" OR "IMPACT model" OR "IMPACT CORE" OR "IMPACT basic" OR "IMPACT- CORE" OR "IMPACT-basic" OR "IMPACT CT" OR "IMPACT Lab" OR "IMPACT-CT" OR "IMPACT-Lab" OR "Corticosteroid Randomization after Significant Head Injury" OR "CRASH prognostic models" OR "CRASH prognostic model" OR "CRASH Rule" OR "CRASH model" OR "CRASH Core" OR "CRASH basic" OR "CRASH CT" OR "CRASH-basic" OR "CRASH-CT") 2- TITLE-ABS-KEY ("Brain Concussion" OR "Brain Injuries" OR "Brain Injuries, Traumatic" OR "Brain Concussion" OR "Brain Injuries" OR "Brain Injuries, Traumatic" OR "Brain Concussions" OR "Concussion, Brain" OR "Commotio Cerebri" OR "Cerebral Concussion" OR "Cerebral Concussions" OR "Concussion, Cerebral" OR "Concussion, Intermediate" OR "Intermediate Concussion" OR "Intermediate Concussions" OR "Concussion, Severe" OR "Severe Concussion" OR "Severe Concussions" OR "Concussion, Mild" OR "Mild Concussion" OR "Mild Concus- sions" OR "Mild Traumatic Brain Injury" OR "Injuries, Brain" OR "Brain Injury" OR "Injury, Brain" OR "Injuries, Acute Brain" OR "Acute Brain Injuries" OR "Acute Brain Injury" OR "Brain Injury, Acute" OR "Injury, Acute Brain" OR "Brain Injuries, Acute" OR "Brain Lacerations" OR "Brain Laceration" OR "Laceration, Brain" OR "Lacerations, Brain" OR "Brain Injuries, Focal" OR "Brain Injury, Focal" OR "Focal Brain Injury" OR "Injuries, Focal Brain" OR "Injury, Focal Brain" OR "Focal Brain Injuries" OR "Brain Injury, Traumatic" OR "Traumatic Brain Injuries" OR "Trauma, Brain" OR "Brain Trauma" OR "Brain Traumas" OR "Traumas, Brain" OR "TBI (Traumatic Brain Injury)" OR "Encephalopathy, Trau- matic" OR "Encephalopathies, Traumatic" OR "Traumatic Encephalopathies" OR "Injury, Brain, Traumatic" OR "Traumatic Encephalopathy" OR "Traumatic Brain Injury") 3- #1 AND #2 Web of Science 1- TS=("International Mission for Prognosis and Analysis of Clinical Trials in Traumatic Brain Injury" OR "IMPACT prognostic models" OR "IMPACT prognostic model" OR "IMPACT Rule" OR "IMPACT model" OR "IMPACT CORE" OR "IMPACT basic" OR "IMPACT-CORE" OR "IMPACT-basic" OR "IMPACT CT" OR "IMPACT Lab" OR "IMPACT-CT" OR "IMPACT-Lab" OR "Corticosteroid Randomization after Signifi- cant Head Injury" OR "CRASH prognostic models" OR "CRASH prognostic model" OR "CRASH Rule" OR "CRASH model" OR "CRASH Core" OR "CRASH basic" OR "CRASH CT" OR "CRASH-basic" OR "CRASH-CT") 2- TS= ("Brain Concussion" OR "Brain Injuries" OR "Brain Injuries, Traumatic" OR "Brain Concussion" OR "Brain Injuries" OR "Brain Injuries, Traumatic" OR "Brain Concussions" OR "Concussion, Brain" OR "Commotio Cerebri" OR "Cerebral Concussion" OR "Cerebral Concussions" OR "Concussion, Cerebral" OR "Concussion, Intermediate" OR "Intermediate Concussion" OR "Intermediate Concussions" OR "Concussion, Severe" OR "Severe Concussion" OR "Severe Concussions" OR "Concussion, Mild" OR "Mild Concussion" OR "Mild Concussions" OR "Mild Traumatic Brain Injury" OR "Injuries, Brain" OR "Brain Injury" OR "Injury, Brain" OR "Injuries, Acute Brain" OR "Acute Brain Injuries" OR "Acute Brain Injury" OR "Brain Injury, Acute" OR "Injury, Acute Brain" OR "Brain Injuries, Acute" OR "Brain Lacerations" OR "Brain Lacera- tion" OR "Laceration, Brain" OR "Lacerations, Brain" OR "Brain Injuries, Focal" OR "Brain Injury, Focal" OR "Focal Brain Injury" OR "Injuries, Focal Brain" OR "Injury, Focal Brain" OR "Focal Brain Injuries" OR "Brain Injury, Traumatic" OR "Traumatic Brain Injuries" OR "Trauma, Brain" OR "Brain Trauma" OR "Brain Traumas" OR "Traumas, Brain" OR "Encephalopathy, Traumatic" OR "Encephalopathies, Traumatic" OR "Traumatic Encephalopathies" OR "Injury, Brain, Traumatic" OR "Traumatic Encephalopathy" OR "Traumatic Brain Injury" OR "TBI") 3- #1 AND #2 This open-access article distributed under the terms of the Creative Commons Attribution NonCommercial 3.0 License (CC BY-NC 3.0). Downloaded from: https://journals.sbmu.ac.ir/aaem/index.php/AAEM/index 17 Archives of Academic Emergency Medicine. 2023; 11(1): e27 Figure S 1: Publication bias of IMPACT and CRASH prognostic models in mortality prediction across eligible studies. There is no evidence of publication bias across the studies. Effect size is the area under the curve. This open-access article distributed under the terms of the Creative Commons Attribution NonCommercial 3.0 License (CC BY-NC 3.0). Downloaded from: https://journals.sbmu.ac.ir/aaem/index.php/AAEM/index H. Zarei et al. 18 Figure S 2: Publication bias of IMPACT and CRASH prognostic models in predicting 6-month unfavorable outcome across eligible studies. There is evidence of possible publication bias in detecting the prognostic value of IMPACT core and IMPACT extended models. Effect size is the area under the curve. This open-access article distributed under the terms of the Creative Commons Attribution NonCommercial 3.0 License (CC BY-NC 3.0). Downloaded from: https://journals.sbmu.ac.ir/aaem/index.php/AAEM/index Introduction Methods Results Discussion Conclusion Declarations References