Dataset Information

On Scene Injury Severity Prediction (OSISP) model for trauma developed using the Swedish Trauma Registry.

ABSTRACT:

Background

Providing optimal care for trauma, the leading cause of death for young adults, remains a challenge e.g., due to field triage limitations in assessing a patient's condition and deciding on transport destination. Data-driven On Scene Injury Severity Prediction (OSISP) models for motor vehicle crashes have shown potential for providing real-time decision support. The objective of this study is therefore to evaluate if an Artificial Intelligence (AI) based clinical decision support system can identify severely injured trauma patients in the prehospital setting.

Methods

The Swedish Trauma Registry was used to train and validate five models - Logistic Regression, Random Forest, XGBoost, Support Vector Machine and Artificial Neural Network - in a stratified 10-fold cross validation setting and hold-out analysis. The models performed binary classification of the New Injury Severity Score and were evaluated using accuracy metrics, area under the receiver operating characteristic curve (AUC) and Precision-Recall curve (AUCPR), and under- and overtriage rates.

Results

There were 75,602 registrations between 2013-2020 and 47,357 (62.6%) remained after eligibility criteria were applied. Models were based on 21 predictors, including injury location. From the clinical outcome, about 40% of patients were undertriaged and 46% were overtriaged. Models demonstrated potential for improved triaging and yielded AUC between 0.80-0.89 and AUCPR between 0.43-0.62.

Conclusions

AI based OSISP models have potential to provide support during assessment of injury severity. The findings may be used for developing tools to complement field triage protocols, with potential to improve prehospital trauma care and thereby reduce morbidity and mortality for a large patient population.

SUBMITTER: Bakidou A

PROVIDER: S-EPMC10561449 | biostudies-literature | 2023 Oct

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

On Scene Injury Severity Prediction (OSISP) model for trauma developed using the Swedish Trauma Registry.

Bakidou Anna A Caragounis Eva-Corina EC Andersson Hagiwara Magnus M Jonsson Anders A Sjöqvist Bengt Arne BA Candefjord Stefan S

BMC medical informatics and decision making 20231009 1

<h4>Background</h4>Providing optimal care for trauma, the leading cause of death for young adults, remains a challenge e.g., due to field triage limitations in assessing a patient's condition and deciding on transport destination. Data-driven On Scene Injury Severity Prediction (OSISP) models for motor vehicle crashes have shown potential for providing real-time decision support. The objective of this study is therefore to evaluate if an Artificial Intelligence (AI) based clinical decision suppo ...[more]

PMID: 37814288

Similar Datasets

Project description:BackgroundIn-hospital mortality in trauma patients has decreased recently owing to improved trauma injury prevention systems. However, no study has evaluated the validity of the Trauma and Injury Severity Score (TRISS) in pediatric patients by a detailed classification of patients' age and injury severity in Japan. This retrospective nationwide study evaluated the validity of TRISS in predicting survival in Japanese pediatric patients with blunt trauma by age and injury severity.MethodsData were obtained from the Japan Trauma Data Bank during 2009-2018. The outcomes were as follows: (1) patients' characteristics and mortality by age groups (neonates/infants aged 0 years, preschool children aged 1-5 years, schoolchildren aged 6-11 years, and adolescents aged 12-18 years), (2) validity of survival probability (Ps) assessed using the TRISS methodology by the four age groups and six Ps-interval groups (0.00-0.25, 0.26-0.50, 0.51-0.75, 0.76-0.90, 0.91-0.95, and 0.96-1.00), and (3) the observed/expected survivor ratio by age- and Ps-interval groups. The validity of TRISS was evaluated by the predictive ability of the TRISS method using the receiver operating characteristic (ROC) curves that present the sensitivity, specificity, positive predictive value, negative predictive value, accuracy, area under the receiver operator characteristic curve (AUC) of TRISS.ResultsIn all the age categories considered, the AUC for TRISS demonstrated high performance (0.935, 0.981, 0.979, and 0.977). The AUC for TRISS was 0.865, 0.585, 0.614, 0.585, 0.591, and 0.600 in Ps-interval groups (0.96-1.00), (0.91-0.95), (0.76. - 0.90), (0.51-0.75), (0.26-0.50), and (0.00-0.25), respectively. In all the age categories considered, the observed survivors among patients with Ps interval (0.00-0.25) were 1.5 times or more than the expected survivors calculated using the TRISS method.ConclusionsThe TRISS methodology appears to predict survival accurately in Japanese pediatric patients with blunt trauma; however, there were several problems in adopting the TRISS methodology for younger blunt trauma patients with higher injury severity. In the next step, it may be necessary to develop a simple, high-quality prediction model that is more suitable for pediatric trauma patients than the current TRISS model.

Project description:BackgroundChildren suffering nonaccidental trauma (NAT) are at high risk of death. It is unclear whether markers of injury severity for trauma center/system benchmarking such as Injury Severity Score (ISS) adequately characterize this. Our objective was to evaluate mortality prediction of ISS in children with NAT compared with accidental trauma (AT).MethodsPediatric patients younger than 16 years from the Pennsylvania state trauma registry 2000 to 2013 were included. Logistic regression predicted mortality from ISS for NAT and AT patients. Multilevel logistic regression determined the association between mortality and ISS while adjusting for age, vital signs, and injury pattern in NAT and AT patients. Similar models were performed for head Abbreviated Injury Scale (AIS). Sensitivity analysis examined impaired functional independence at discharge as an alternate outcome.ResultsFifty thousand five hundred seventy-nine patients were included with 1,866 (3.7%) NAT patients. Nonaccidental trauma patients had a similar rate of mortality at an ISS of 13 as an ISS of 25 for AT patients. Nonaccidental trauma patients also have higher mortality for a given head AIS level (range, 1.2-fold to 5.9-fold higher). Injury Severity Score was a significantly greater predictor of mortality in AT patients (adjusted odds rations [AOR], 1.14; 95% confidence interval [CI], 1.13-1.15; p < 0.01) than NAT patients (AOR, 1.09; 95% CI, 1.07-1.12; p < 0.01) per 1-point ISS increase, while head injury was a significantly greater predictor of mortality in NAT patients (AOR, 3.48; 95% CI, 1.54-8.32; p < 0.01) than AT patients (AOR, 1.21; 95% CI, 0.95-1.45; p = 0.12). Nonaccidental trauma patients had a higher rate of impaired functional independence at any given ISS or head AIS level than AT patients.ConclusionNonaccidental trauma patients have higher mortality and impaired function at a given ISS/head AIS than AT patients. Conventional ISS thresholds may underestimate risk and head injury is a more important predictor of mortality in the NAT population. These findings should be considered in system performance improvement and benchmarking efforts that rely on ISS for injury characterization.Level of evidenceEpidemiologic study, level III.

Project description:BackgroundThere is a lack of knowledge regarding the epidemiology of severe trauma assessed by Swedish emergency medical services (EMS).AimTo investigate the prevalence of trauma in Sweden assessed by EMS from a national perspective and describe patient demography, aetiology, trauma type, prehospital triage and clinical outcomes.MethodsData from two national quality registries, the Swedish Ambulance Registry and the Swedish Trauma Registry (SweTrau) were collected from January 1 to December 31, 2019. Inclusion criteria were an Emergency Symptoms and Signs code equivalent to trauma in the Swedish Ambulance Registry and criteria fulfilled for SweTrau inclusion. Exclusion criteria were patients < 18 years old, those not transported to a hospital and those without a personal identification number.ResultsIn total, 53,120 patients with trauma were included (14% of primary EMS missions involving a personal identification number). Of those, 2,278 (4.3%) patients (median age: 45 years; 32% women) were reported in SweTrau to have severe or potentially severe trauma (penetrating: 7%, blunt: 93%). In terms of including all causes of trauma, the code for 'trauma alert activation' was most frequent (55%). The most frequent injury mechanism was an injury caused by a car (34%). Most (89%) cases were assigned Priority 1 (life-threatening condition) at the dispatch centre. 62% were regarded as potentially life threatening upon EMS arrival, whereas 29% were assessed as non-life-threatening. Overall, 25% of the patients had new injury severity scores > 15. 12% required invasive treatment, 11% were discharged with severe disability and the 30-day mortality rate was 3.6%.ConclusionIn this cross-sectional study, 14% of the primary EMS missions for one year were caused by trauma. However, only a small proportion of these cases are severe injuries, and the risk of severe disabilities and death appears to be limited. The most frequent aetiology of a severe trauma is injury caused by a car, and most severe traumas are blunt. Severe traumas are given the highest priority at the dispatch centre in the vast majority of cases, but nearly one-third of these cases are considered a low priority by the EMS nurse. The latter leaves room for improvement.

Project description:BackgroundPrecise models are necessary to estimate mortality risk following traumatic injury to inform clinical decision making or quantify hospital performance. The Trauma and Injury Severity Score (TRISS) has been the historical gold standard in survival prediction but its limitations are well-characterized. The present study used International Classification of Diseases 10th Revision (ICD-10) injury codes with machine learning approaches to develop models whose performance was compared to that of TRISS.MethodsThe 2015-2017 National Trauma Data Bank was used to identify patients following trauma-related admission. Injury codes from ICD-10 were grouped by clinical relevance into 1,495 variables. The TRISS score, which comprises the Injury Severity Score, age, mechanism (blunt vs penetrating) as well as highest 24-hour values for systolic blood pressure (SBP), respiratory rate (RR) and Glasgow Coma Scale (GCS) was calculated for each patient. A base eXtreme gradient boosting model (XGBoost), a machine learning technique, was developed using injury variables as well as age, SBP, RR, mechanism and GCS. Prediction of in-hospital survival and other in-hospital complications were compared between both models using receiver operating characteristic (ROC) and reliability plots. A complete XGBoost model, containing injury variables, vitals, demographic information and comorbidities, was additionally developed.ResultsOf 1,380,740 patients, 1,338,417 (96.9%) survived to discharge. Compared to survivors, those who died were older and had a greater prevalence of penetrating injuries (18.0% vs 9.44%). The base XGBoost model demonstrated a greater receiver-operating characteristic (ROC) than TRISS (0.950 vs 0.907) which persisted across sub-populations and secondary endpoints. Furthermore, it exhibited high calibration across all risk levels (R2 = 0.998 vs 0.816). The complete XGBoost model had an exceptional ROC of 0.960.ConclusionsWe report improved performance of machine learning models over TRISS. Our model may improve stratification of injury severity in clinical and quality improvement settings.

Project description:BackgroundData about injury patterns and clinical outcomes are essential to address the burden of injury in low- and middle-income countries. Institutional trauma registries (ITRs) are a key tool for collecting epidemiologic data about injury. This study uses ITR data to describe the demographics and patterns of injury of trauma patients in Addis Ababa, Ethiopia in order to identify opportunities for injury prevention, systems strengthening and further research.MethodsThis is an analysis of prospectively collected data from a sustainable ITR at Menelik II Specialized Hospital, a public teaching hospital with trauma expertise. All patients presenting to the hospital with serious injuries requiring intervention or admission over a 13 month period were included. Univariable and bivariable analyses were performed for patient demographics and injury characteristics.ResultsA total of 854 patients with serious injuries were treated during the study period. Median age was 33 years and 74% were male. The most common mechanisms of injury were road traffic injuries (RTI) (37%), falls (30%) and blunt assault (17%). Over half of RTI victims were pedestrians. Median delay in presentation was 2 h; 17% of patients presented over 6 h after injury. 58% of patients were referred from another hospital or a clinic, and referrals accounted for 84% of patients arriving by ambulance. Median emergency center length of stay was 2 h and 62% of patients were discharged from the emergency center.ConclusionThis study highlights the utility of institutional trauma registries in collecting crucial injury surveillance data. In Addis Ababa, road safety is an important target for injury prevention. Our findings suggest that the most severely injured patients may not be making it to the referral centers with the capacity to treat their injuries, thus efforts to improve prehospital care and triage are needed.African relevanceInjury is a public health priority in Africa. Institutional trauma registries play a crucial role in efforts to improve trauma care by describing injury epidemiology to identify targets for injury prevention and systems strengthening efforts. In our context, pedestrian safety is a key target for injury prevention. Improving prehospital care and developing referral networks are goals for systems strengthening.

Dataset Information

On Scene Injury Severity Prediction (OSISP) model for trauma developed using the Swedish Trauma Registry.

Background

Methods

Results

Conclusions

Publications

On Scene Injury Severity Prediction (OSISP) model for trauma developed using the Swedish Trauma Registry.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets