Artificial neural networks can predict trauma volume and acuity regardless of center size and geography: A multicenter study

Novel Machine-Learning Modeling of Facial Trauma Volume With Regional Event and Weather Data

OBJECTIVE

Facial trauma volume is difficult to predict accurately. We aim to understand the capacity of climate and regional events to predict daily facial trauma volume. This can provide epidemiologic understanding and subsequently tailor workforce distribution and scheduling.

STUDY DESIGN

Retrospective cohort study.

SETTING

Single Tertiary Academic Medical Center.

METHODS

Facial trauma consults between 2017 and 2023 were extracted from a single Level I Trauma Center. Publicly accessible data on local concerts, National Hockey League games, National Football League games, and weather data from the National Oceanic and Atmospheric Administration data were merged with trauma data. Machine-learning random-forest (RF) plot feature identification was used to identify variables to model high-volume facial trauma days (greater than 75th percentile).

RESULTS

For analysis, 2342 days were included. The median number of facial trauma consults was 3.0 (interquartile range: 2.0-5.0). The month of May exhibited the highest rate of high-volume trauma days (13% of days, P < .001). On RF feature identification, the strongest predictive factors included weekend day status, average temperature, precipitation, hail, high/damaging winds, and holidays. Regional events were not included in the final models. On stepwise logistic regression modeling with pertinent variables, weekend day (odds ratio [OR]: 2.20, 95% confidence interval [CI]: 1.80-2.69, P < .001), average temperature (OR: 1.02, 95% CI: 1.01-1.02, P < .001), and wind speed (0.97, 0.93-1.00, P = .049) were the only statistically significant variables.

CONCLUSION

Climate data were the primary factor that had predictive capacity for high-volume facial trauma days, more so than regional events. Testing models prospectively will help validate such models and help inform staffing for facial trauma coverage.

A Systematic Literature Review of Trauma Systems: An Operations Management Perspective

Background

Trauma systems provide comprehensive care across various settings, from prehospital services to rehabilitation, integrating clinical and social care aspects. Established in the 1970s, these systems are pivotal yet under-researched in their operational management. This study aims to fill this gap by focussing on the integration of operations management (OM) techniques to enhance the efficiency and effectiveness of trauma systems. By leveraging proven OM strategies from other healthcare sectors, we seek to improve patient outcomes and optimise system performance, addressing a crucial need for innovation in trauma care operations.

Methodology

A systematic literature review was conducted using the PICOTS framework to explore operational aspects of trauma systems across varied settings, from emergency departments to specialised centres. Searches were performed in 5 databases, focussing on articles published from 2006 to 2024. Keywords related to operational research and management targeted both trauma systems and emergency management services. Our method involved identifying, synthesising, and summarising studies to evaluate operational performance, with a specific emphasis on articles that applied operational research/management techniques in trauma care. All eligible articles were critically appraised using 2 quality assessment tools.

Results

Employing Donabedian's framework to analyse the quality of trauma systems through structure, process, and outcome dimensions, our systematic review included 160 studies. Of these, 5 studies discussed the application of the Donabedian evaluation framework to trauma systems, and 14 studies examined structural elements, focussing on the location of healthcare facilities, trauma resource management, and EMS logistics. The 63 studies on process indicators primarily assessed triage procedures, with some exploring the timeliness of trauma care. Meanwhile, the 78 outcome-oriented studies predominantly evaluated mortality rates, alongside a smaller number assessing functional outcomes.

Conclusion

Existing evaluation metrics primarily focussed on triage accuracy and mortality are inadequate. We propose expanding these metrics to include patient length of stay (LOS) and rehabilitation trajectory analyses. There is a critical gap in understanding patient flow management and long-term outcomes, necessitating focussed research on LOS modelling and improved rehabilitation data collection. Addressing these areas is essential for optimising trauma care and improving patient recovery outcomes.

Patterns of major trauma admissions to a level 1 trauma centre: A five year database analysis

INTRODUCTION

It is only in recent years that major trauma systems and networks have been operating in the UK. High-quality data is available from the Trauma Audit and Research Network (TARN) database, enabling regional analysis. Our aim was to analyse Trauma Team Activations within the Cheshire and Merseyside major trauma network and discuss the implications of these data on resource allocation, training and trauma prevention.

METHODS

A retrospective analysis was performed for all patients requiring Trauma Team Activation (TTA) at a category one adult Major Trauma Centre (MTC) who were submitted to the TARN database from the 1st January 2015 to the 1st January 2020. Data collected included the date and time of arrival, location of injury and Injury Severity Score (ISS) in addition to routine demographic data. Dates of major sporting events and school holidays were obtained.

RESULTS

4811 patients were identified. The median age was 57 years; 65.8 % were male. The mean frequency of TTAs was 18.5 per week. Patterns identified include annual peaks during the summer months, October and December, weekly peaks on Thursdays and Sundays and daily peaks between 16:00 and 23:59 with 45.0 % of TTAs occurring between these hours. There were 5.9 additional TTAs per week during the Isle of Man TT races. The median ISS increased from 14 to 23 for TT race TTAs and from 14 to 36 for Manx Grand Prix TTAs. Those injured during the TT races were twice as likely to require surgery and those injured during the MGP required five additional days in intensive care. School holidays did not independently affect major trauma volumes.

CONCLUSIONS

Major trauma in Cheshire and Merseyside did follow distinct patterns according to calendar month, day and time. Major motorsport increased trauma volumes and severity; school holidays did not. Such analysis could enable Major Trauma Centres to tailor the supply of trauma services to meet a predictable local demand for the benefit of our staff and patients.

Impact of patient, system, and environmental factors on utilization of air medical transport after trauma

BACKGROUND

Air medical transport (AMT) improves outcomes for severely injured patients. The decision to fly patients is complex and must consider multiple factors. Our objective was to evaluate the interaction between geography, patient and environmental factors, and emergency medical services (EMS) system resources on AMT after trauma. We hypothesize that significant geographic variation in AMT utilization will be associated with varying levels of patient, environmental, and EMS resources.

METHODS

Patients transported by EMS in the Pennsylvania state trauma registry 2000 to 2017 were included. We used our previously developed Air Medical Prehospital Triage (AMPT; ≥2 points triage to AMT) score and Geographic Emergency Medical Services Index (GEMSI; higher indicates more system resources) as measures for patient factors and EMS resources, respectively. A mixed-effects logistic regression model determined the association of AMT utilization with patient, system, and environmental variables.

RESULTS

There were 195,354 patients included. Fifty-five percent of variation in AMT utilization was attributed to geographic differences. Triage to AMT by the AMPT score was associated with nearly twice the odds of AMT utilization (adjusted odds ratio, 1.894; 95% confidence interval, 1.765-2.032; p < 0.001). Each 1-point increase in GEMSI was associated with a 6.1% reduction in odds of AMT (0.939; 0.922-0.957; p < 0.001). Younger age, rural location, and more severe injuries were also associated with increased odds of AMT ( p < 0.05). When categorized by GEMSI level, the AMPT score and patient factors were more important for predicting AMT utilization in the middle tercile (moderate EMS resources) compared with the lower (low EMS resources) and higher tercile (high EMS resources). Weather, season, time-of-day, and traffic were all associated with AMT utilization ( p < 0.05).

CONCLUSION

Patient, system, and environmental factors are associated with AMT utilization, which varies geographically and by EMS/trauma system resource availability. A more comprehensive approach to AMT triage could reduce variation and allow more tailored efforts toward optimizing resource allocation and outcomes.

LEVEL OF EVIDENCE

Prognostic and Epidemiological; Level III.

[Application and prospect of machine learning in orthopaedic trauma]

OBJECTIVE

To review the current applications of machine learning in orthopaedic trauma and anticipate its future role in clinical practice.

METHODS

A comprehensive literature review was conducted to assess the status of machine learning algorithms in orthopaedic trauma research, both nationally and internationally.

RESULTS

The rapid advancement of computer data processing and the growing convergence of medicine and industry have led to the widespread utilization of artificial intelligence in healthcare. Currently, machine learning plays a significant role in orthopaedic trauma, demonstrating high performance and accuracy in various areas including fracture image recognition, diagnosis stratification, clinical decision-making, evaluation, perioperative considerations, and prognostic risk prediction. Nevertheless, challenges persist in the development and clinical implementation of machine learning. These include limited database samples, model interpretation difficulties, and universality and individualisation variations.

CONCLUSION

The expansion of clinical sample sizes and enhancements in algorithm performance hold significant promise for the extensive application of machine learning in supporting orthopaedic trauma diagnosis, guiding decision-making, devising individualized medical strategies, and optimizing the allocation of clinical resources.

Mature artificial intelligence- and machine learning-enabled medical tools impacting vascular surgical care: A scoping review of late-stage, US Food and Drug Administration-approved or cleared technologies relevant to vascular surgeons

Artificial intelligence and machine learning (AI/ML)-enabled tools are shifting from theoretical or research-only applications to mature, clinically useful tools. The goal of this article was to provide a scoping review of the most mature AI/ML-enabled technologies reviewed and cleared by the US Food and Drug Administration relevant to the field of vascular surgery. Despite decades of slow progress, this landscape is now evolving rapidly, with more than 100 AI/ML-powered tools being approved by the US Food and Drug Administration each year. Within the field of vascular surgery specifically, this review identified 17 companies with mature technologies that have at least one US Food and Drug Administration clearance, all occurring between 2016 and 2022. The maturation of these technologies appears to be accelerating, with improving regulatory clarity and clinical uptake. The early AI/ML-powered devices extend or amplify clinically entrenched platform technologies and tend to be focused on the diagnosis or evaluation of time-sensitive, clinically important pathologies (eg, reading Digital Imaging and Communications in Medicine-compliant computed tomography images to identify pulmonary embolism), or when physician efficiency or time savings is improved (eg, preoperative planning and intraoperative guidance). The majority (>75%) of these technologies are at the intersection of radiology and vascular surgery. It is becoming increasingly important that the contemporary vascular surgeon understands this shifting paradigm, as these once-nascent technologies are finally maturing and will be encountered with increasingly regularity in daily clinical practice.

Identifying temporal variations in burn admissions

BACKGROUND

Variations in admission patterns have been previously identified in non-elective surgical services, but minimal data on the subject exists with respect to burn admissions. Improved understanding of the temporal pattern of burn admissions could inform resource utilization and clinical staffing. We hypothesize that burn admissions have a predictable temporal distribution with regard to the time of day, day of week, and season of year in which they present.

STUDY DESIGN

A retrospective, cohort observational study of a single burn center from 7/1/2016 to 3/31/2021 was performed on all admissions to the burn surgery service. Demographics, burn characteristics, and temporal data of burn admissions were collected. Bivariate absolute and relative frequency data was captured and plotted for all patients who met inclusion criteria. Heat-maps were created to visually represent the relative admission frequency by time of day and day of week. Frequency analysis grouped by total body surface area against time of day and relative encounters against day of year was performed.

RESULTS

2213 burn patient encounters were analyzed, averaging 1.28 burns per day. The nadir of burn admissions was from 07:00 and 08:00, with progressive increase in the rate of admissions over the day. Admissions peaked in the 15:00 hour and then plateaued until midnight (p<0.001). There was no association between day of week in the burn admission distribution (p>0.05), though weekend admissions skewed slightly later (p = 0.025). No annual, cyclical trend in burn admissions was identified, suggesting that there is no predictable seasonality to burn admissions, though individual holidays were not assessed.

CONCLUSION

Temporal variations in burn admissions exist, including a peak admission window late in the day. Furthermore, we did not find a predictable annual pattern to use in guiding staffing and resource allocation. This differs from findings in trauma, which identified admission peaks on the weekends and an annual cycle that peaks in spring and summer.

A Pilot Machine Learning Study Using Trauma Admission Data to Identify Risk for High Length of Stay

INTRODUCTION

Trauma patients have diverse resource needs due to variable mechanisms and injury patterns. The aim of this study was to build a tool that uses only data available at time of admission to predict prolonged hospital length of stay (LOS).

METHODS

Data was collected from the trauma registry at an urban level one adult trauma center and included patients from 1/1/2014 to 3/31/2019. Trauma patients with one or fewer days LOS were excluded. Single layer and deep artificial neural networks were trained to identify patients in the top quartile of LOS and optimized on area under the receiver operator characteristic curve (AUROC). The predictive performance of the model was assessed on a separate test set using binary classification measures of accuracy, precision, and error.

RESULTS

2953 admitted trauma patients with more than one-day LOS were included in this study. They were 70% male, 60% white, and averaged 47 years-old (SD: 21). 28% were penetrating trauma. Median length of stay was 5 days (IQR 3-9). For prediction of prolonged LOS, the deep neural network achieved an AUROC of 0.80 (95% CI: 0.786-0.814) specificity was 0.95, sensitivity was 0.32, with an overall accuracy of 0.79.

CONCLUSION

Machine learning can predict, with excellent specificity, trauma patients who will have prolonged length of stay with only physiologic and demographic data available at the time of admission. These patients may benefit from additional resources with respect to disposition planning at the time of admission.

Science fiction or clinical reality: a review of the applications of artificial intelligence along the continuum of trauma care

Artificial intelligence (AI) and machine learning describe a broad range of algorithm types that can be trained based on datasets to make predictions. The increasing sophistication of AI has created new opportunities to apply these algorithms within within trauma care. Our paper overviews the current uses of AI along the continuum of trauma care, including injury prediction, triage, emergency department volume, assessment, and outcomes. Starting at the point of injury, algorithms are being used to predict severity of motor vehicle crashes, which can help inform emergency responses. Once on the scene, AI can be used to help emergency services triage patients remotely in order to inform transfer location and urgency. For the receiving hospital, these tools can be used to predict trauma volumes in the emergency department to help allocate appropriate staffing. After patient arrival to hospital, these algorithms not only can help to predict injury severity, which can inform decision-making, but also predict patient outcomes to help trauma teams anticipate patient trajectory. Overall, these tools have the capability to transform trauma care. AI is still nascent within the trauma surgery sphere, but this body of the literature shows that this technology has vast potential. AI-based predictive tools in trauma need to be explored further through prospective trials and clinical validation of algorithms.

Artificial intelligence and machine learning for hemorrhagic trauma care

Artificial intelligence (AI), a branch of machine learning (ML) has been increasingly employed in the research of trauma in various aspects. Hemorrhage is the most common cause of trauma-related death. To better elucidate the current role of AI and contribute to future development of ML in trauma care, we conducted a review focused on the use of ML in the diagnosis or treatment strategy of traumatic hemorrhage. A literature search was carried out on PubMed and Google scholar. Titles and abstracts were screened and, if deemed appropriate, the full articles were reviewed. We included 89 studies in the review. These studies could be grouped into five areas: (1) prediction of outcomes; (2) risk assessment and injury severity for triage; (3) prediction of transfusions; (4) detection of hemorrhage; and (5) prediction of coagulopathy. Performance analysis of ML in comparison with current standards for trauma care showed that most studies demonstrated the benefits of ML models. However, most studies were retrospective, focused on prediction of mortality, and development of patient outcome scoring systems. Few studies performed model assessment via test datasets obtained from different sources. Prediction models for transfusions and coagulopathy have been developed, but none is in widespread use. AI-enabled ML-driven technology is becoming integral part of the whole course of trauma care. Comparison and application of ML algorithms using different datasets from initial training, testing and validation in prospective and randomized controlled trials are warranted for provision of decision support for individualized patient care as far forward as possible.

Leveraging Machine Learning and Artificial Intelligence to Improve Peripheral Artery Disease Detection, Treatment, and Outcomes

Peripheral artery disease is an atherosclerotic disorder which, when present, portends poor patient outcomes. Low diagnosis rates perpetuate poor management, leading to limb loss and excess rates of cardiovascular morbidity and death. Machine learning algorithms and artificially intelligent systems have shown great promise in application to many areas in health care, such as accurately detecting disease, predicting patient outcomes, and automating image interpretation. Although the application of these technologies to peripheral artery disease are in their infancy, their promises are tremendous. In this review, we provide an introduction to important concepts in the fields of machine learning and artificial intelligence, detail the current state of how these technologies have been applied to peripheral artery disease, and discuss potential areas for future care enhancement with advanced analytics.

The coronavirus disease 2019 (COVID-19) stay-at-home order's unequal effects on trauma volume by insurance status in Southern California

BACKGROUND

The rapid spread of coronavirus disease 2019 in the United States led to a variety of mandates intended to decrease population movement and "flatten the curve." However, there is evidence some are not able to stay-at-home due to certain disadvantages, thus remaining exposed to both coronavirus disease 2019 and trauma. We therefore sought to identify any unequal effects of the California stay-at-home orders between races and insurance statuses in a multicenter study utilizing trauma volume data.

METHODS

A posthoc multicenter retrospective analysis of trauma patients presenting to 11 centers in Southern California between the dates of January 1, 2020, and June 30, 2020, and January 1, 2019, and June 30, 2019, was performed. The number of trauma patients of each race/insurance status was tabulated per day. We then calculated the changes in trauma volume related to stay-at-home orders for each race/insurance status and compared the magnitude of these changes using statistical resampling.

RESULTS

Compared to baseline, there was a 40.1% drop in total trauma volume, which occurred 20 days after stay-at-home orders. During stay-at-home orders, the average daily trauma volume of patients with Medicaid increased by 13.7 ± 5.3%, whereas the volume of those with Medicare, private insurance, and no insurance decreased. The average daily trauma volume decreased for White, Black, Asian, and Latino patients with the volume of Black and Latino patients dropping to a similar degree compared to White patients.

CONCLUSION

This retrospective multicenter study demonstrated that patients with Medicaid had a paradoxical increase in trauma volume during stay-at-home orders, suggesting that the most impoverished groups remain disproportionately exposed to trauma during a pandemic, further exacerbating existing health disparities.

Artificial intelligence in emergency medicine: A scoping review

INTRODUCTION

Despite the growing investment in and adoption of artificial intelligence (AI) in medicine, the applications of AI in an emergency setting remain unclear. This scoping review seeks to identify available literature regarding the applications of AI in emergency medicine.

METHODS

The scoping review was conducted according to Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines for scoping reviews using Medline-OVID, EMBASE, CINAHL, and IEEE, with a double screening and extraction process. The search included articles published until February 28, 2020. Articles were excluded if they did not self-classify as studying an AI intervention, were not relevant to the emergency department (ED), or did not report outcomes or evaluation.

RESULTS

Of the 1483 original database citations, 395 were eligible for full-text evaluation. Of these articles, a total of 150 were included in the scoping review. The majority of included studies were retrospective in nature (n = 124, 82.7%), with only 3 (2.0%) prospective controlled trials. We found 37 (24.7%) interventions aimed at improving diagnosis within the ED. Among the 150 studies, 19 (12.7%) focused on diagnostic imaging within the ED. A total of 16 (10.7%) studies were conducted in the out-of-hospital environment (eg, emergency medical services, paramedics) with the remainder occurring either in the ED or the trauma bay. Of the 24 (16%) studies that had human comparators, there were 12 (8%) studies in which AI interventions outperformed clinicians in at least 1 measured outcome.

CONCLUSION

AI-related research is rapidly increasing in emergency medicine. There are several promising AI interventions that can improve emergency care, particularly for acute radiographic imaging and prediction-based diagnoses. Higher quality evidence is needed to further assess both short- and long-term clinical outcomes.

Artificial intelligence in trauma systems

Local trauma care and regional trauma systems are data-rich environments that are amenable to machine learning, artificial intelligence, and big-data analysis mechanisms to improve timely access to care, to measure outcomes, and to improve quality of care. Pilot work has been done to demonstrate that these methods are useful to predict patient flow at individual centers, so that staffing models can be adapted to match workflow. Artificial intelligence has also been proven useful in the development of regional trauma systems as a tool to determine the optimal location of a new trauma center based on trauma-patient geospatial injury data and to minimize response times across the trauma network. Although the utility of artificial intelligence is apparent and proven in small pilot studies, its operationalization across the broader trauma system and trauma surgery space has been slow because of cost, stakeholder buy-in, and lack of expertise or knowledge of its utility. Nevertheless, as new trauma centers or systems are developed, or existing centers are retooled, machine learning and sophisticated analytics are likely to be important components to help facilitate decision-making in a wide range of areas, from determining bedside nursing and provider ratios to determining where to locate new trauma centers or emergency medical services teams.