Machine learning in the prediction of massive transfusion in trauma: a retrospective analysis as a proof-of-concept.

PURPOSE: Early administration and protocolization of massive hemorrhage protocols (MHP) has been associated with decreases in mortality, multiorgan system failure, and number of blood products used. Various prediction tools have been developed for the initiation of MHP, but no single tool has demonstrated strong prediction with early clinical data. We sought to develop a massive transfusion prediction model using machine learning and early clinical data. METHODS: Using the National Trauma Data Bank from 2013 to 2018, we included severely injured trauma patients and extracted clinical features available from the pre-hospital and emergency department. We subsequently balanced our dataset and used the Boruta algorithm to determine feature selection. Massive transfusion was defined as five units at 4 h and ten units at 24 h. Six machine learning models were trained on the balanced dataset and tested on the original. RESULTS: A total of 326,758 patients met our inclusion with 18,871 (5.8%) requiring massive transfusion. Emergency department models demonstrated strong performance characteristics with mean areas under the receiver-operating characteristic curve of 0.83. Extreme gradient boost modeling slightly outperformed and demonstrated adequate predictive performance with pre-hospital data only, as well as 4-h transfusion thresholds. CONCLUSIONS: We demonstrate the use of machine learning in developing an accurate prediction model for massive transfusion in trauma patients using early clinical data. This research demonstrates the potential utility of artificial intelligence as a clinical decision support tool.

Outcomes of Patients Transported in the Prone Position to a Regional Extracorporeal Membrane Oxygenation Center: A Retrospective Cohort Study.

Prone positioning is associated with improved mortality in patients with moderate/severe acute respiratory distress syndrome (ARDS) and has been increasingly used throughout the COVID-19 pandemic. In patients with refractory hypoxemia, transfer to an extracorporeal membrane oxygenation (ECMO) center may improve outcome but may be challenging due to severely compromised gas exchange. Transport of these patients in prone position may be advantageous; however, there is a paucity of data on their outcomes. OBJECTIVES: The primary objective of this retrospective cohort study was to describe the early outcomes of ARDS patients transported in prone position for evaluation at a regional ECMO center. A secondary objective was to examine the safety of their transport in the prone position. DESIGN: Retrospective cohort study. SETTING: This study used patient charts from Ornge and Toronto General Hospital in Ontario, Canada, between February 1, 2020, and November 31, 2021. PARTICIPANTS: Patient with ARDS transported in the prone position for ECMO evaluation to Toronto General Hospital. MAIN OUTCOMES AND MEASURES: Descriptive analysis of patients transported in the prone position and their outcomes. RESULTS: One hundred fifteen patients were included. Seventy-two received ECMO (63%) and 51 died (44%) with ARDS and sepsis as the most common listed causes of death. Patients were transported primarily for COVID-related indications (93%). Few patients required additional analgesia (8%), vasopressors (4%), or experienced clinically relevant desaturation during transport (2%). CONCLUSIONS AND RELEVANCE: This cohort of patients with severe ARDS transported in prone position had outcomes ranging from similar to better compared with existing literature. Prone transport was performed safely with few complications or escalation in treatments. Prone transport to an ECMO center should be regarded as safe and potentially beneficial for patients with ARDS and refractory hypoxemia.

Machine Learning in the Prediction of Trauma Outcomes: A Systematic Review.

STUDY OBJECTIVE: Machine learning models carry unique potential as decision-making aids and prediction tools for improving patient care. Traumatically injured patients provide a uniquely heterogeneous population with severe injuries that can be difficult to predict. Given the relative infancy of machine learning applications in medicine, this systematic review aimed to better understand the current state of machine learning development and implementation to help create a basis for future research. METHODS: We conducted a systematic review from inception to May 2021, using Embase, MEDLINE through Ovid, Web of Science, Google Scholar, and relevant gray literature, for uses of machine learning in predicting the outcomes of trauma patients. The screening and data extraction were performed by 2 independent reviewers. RESULTS: Of the 14,694 identified articles screened, 67 were included for data extraction. Artificial neural networks comprised the most commonly used model, and mortality was the most prevalent outcome of interest. In terms of machine learning model development, there was a lack of studies that employed external validation, feature selection methods, and performed formal calibration testing. Significant heterogeneity in reporting was also observed between the machine learning models employed, patient populations, performance metrics, and features employed. CONCLUSION: This review highlights the heterogeneity in the development and reporting of machine learning models for the prediction of trauma outcomes. While these models present an area of opportunity as an ancillary to clinical decision-making, we recommend more standardization and rigorous guidelines for the development of future models.

Contemporary uses of trauma video review: a scoping review.

OBJECTIVES: Trauma resuscitations are sporadic, high-acuity situations and conducting observation in the trauma bay for the purpose of quality improvement is challenging. We aim to review contemporary uses of trauma video review. METHODS: Medline and Embase were searched from 1980 to May 2020 for studies involving trauma video review. English studies of adult and paediatric populations were included for study and analysed for uses of trauma video review, outcomes measured and any resulting quality improvement (QI) initiatives. RESULTS: A total of 463 publications were identified with 21 studies meeting eligibility for final inclusion. A majority of studies (11) observed technical skills with analysis of critical procedures, including tracheal intubation and thoracotomy. The remaining studies observed team dynamics and communication. Overall, eight studies resulted in new policies being put in place for trauma resuscitations and six studies utilized trauma video review as an educational tool. CONCLUSIONS: This study highlights common uses of trauma video review. The greatest benefit for this new technology is in quality improvement and education. The majority of studies focussed on critical procedures and QI initiatives, such as checklists, protocols and continued education. We recommend adoption of video review systems for ongoing improvement of team dynamics and overall trauma and emergency resuscitation.

RéSUMé: OBJECTIFS: Les réanimations traumatiques sont des situations sporadiques à haute acuité et il est difficile de mener des observations dans la salle de traumatologie dans le but d'améliorer la qualité. Notre objectif est de passer en revue les utilisations contemporaines de l'examen vidéo des traumatismes. MéTHODES: Des recherches ont été menées dans Medline et Embase de 1980 à mai 2020 pour trouver des études impliquant un examen vidéo de traumatismes. Les études anglaises portant sur des populations adultes et pédiatriques ont été incluses dans l'étude et analysées en fonction des utilisations de l'examen vidéo des traumatismes, des résultats mesurés et de toute initiative d'amélioration de la qualité (AQ) en résultant. RéSULTATS: Un total de 463 publications a été identifié avec 21 études répondant aux critères d'éligibilité pour l'inclusion finale. Une majorité d'études (11) ont observé les compétences techniques avec l'analyse des procédures critiques, notamment l'intubation trachéale et la thoracotomie. Les autres études ont observé la dynamique de l'équipe et la communication. Dans l'ensemble, 8 études ont donné lieu à la mise en place de nouvelles politiques pour les réanimations traumatiques et 6 études ont utilisé l'examen de vidéos de traumatismes comme outil éducatif. CONCLUSIONS: Cette étude met en évidence les utilisations courantes de l'examen vidéo des traumatismes. Le plus grand avantage de cette nouvelle technologie est l'amélioration de la qualité et l'éducation. La majorité des études se sont concentrées sur les procédures critiques et les initiatives d'AQ, telles que les listes de contrôle, les protocoles et la formation continue. Nous recommandons l'adoption de systèmes de révision vidéo pour l'amélioration continue de la dynamique d'équipe et de la réanimation de traumatologie et d'urgence en général.

Errors in adult trauma resuscitation: a systematic review.

INTRODUCTION: Trauma resuscitation at dedicated trauma centers typically consist of ad-hoc teams performing critical tasks in a time-limited manner. This creates a high stakes environment apt or avoidable errors. Reporting of errors in trauma resuscitation is generally center-dependent and lacks common terminology. METHODS: We conducted a systematic review by searching Ovid Medline, Scopus and Embase from inception to February 24, 2021 for errors in adult trauma resuscitation. English studies published after 2001 were included. Studies were assessed by two independent reviewers for meeting inclusion/exclusion criteria. Errors were characterized from the included studies and a summary table was developed. Our review was prospectively registered with the International Prospective Register of Systematic Reviews (PROSPERO) (CRD42020152875). RESULTS: The literature search retrieved 4658 articles with 26 meeting eligibility criteria. Errors were identified by morbidity and mortality rounds or other committee in 62%, missed injuries on tertiary assessment or radiology review in 12%, deviations from algorithmic guidelines in 12% or predefined for chest tube complications, critical incident reporting, aspiration or delays in care. In total there were 39 unique error types identified and divided into 9 categories including Emergency Medical Services handover, airway, assessment of injuries, patient monitoring and access, transfusion/blood related, management of injuries, team communication/dynamics, procedure error and disposition. CONCLUSIONS: Overall, our systematic review identified 39 unique error types in trauma resuscitation. Identifying these errors is imperative in developing systems for improvement of trauma care.

Postpartum headache: A broader differential.

A 40-year-old female presented to the ED with a history of intermittent headaches since a vaginal delivery 8 days prior. Her pregnancy was unremarkable and was not complicated by pre-eclampsia. She did not present with signs or symptoms consistent with postdural puncture headache or pre-eclampsia. Her delivery was not complicated by hypotension or post-partum hemorrhage. By chance, she was found to be hyponatremic and admitted to internal medicine for further work-up. She was diagnosed with postpartum lymphocytic adenohypophysitis and treated with steroids. She was discharged with a steroid taper and had complete resolution on follow up. Lymphocytic hypophysitis (LH), or commonly known as autoimmune hypophysitis, is a rare inflammatory condition affecting the pituitary gland. Acute LH can result in sudden death as demonstrated in some case reports. The most common symptom in >50% of cases is headache. First-line pharmacological treatment consists of high-dose corticosteroids and is effective in reducing pituitary size and improving endocrine insufficiency in 75% of cases. LH is a potential cause of postpartum headache that can be easily screened for with history and routine bloodwork and has good outcomes with early intervention. LH should be added to the differential for postpartum headaches presenting to the emergency department and routine blood work should be considered for all postpartum headaches.

Prone transportation to an ECMO center.

Acute Respiratory Distress Syndrome (ARDS) accounts for 10% of all intensive care unit admissions and mortality remains exceedingly high ranging from 35 to 46%. Prone positioning has demonstrated benefit as a treatment option in the moderate to severe subset of ARDS patients. Referral to an extracorporeal membrane oxygenation (ECMO) center remains the standard of care for patients with refractory hypoxemia. Transfer of these already critically ill patients is high risk but there is a possibility for improved oxygenation on transport with prone positioning. Currently there is limited data on the safe transportation of prone patients to ECMO centers. Herein we describe a case of an ARDS patient safely transported prone with minimal ad hoc adjustments to standard ambulance transport.

Feasibility of remote administration of the fundamentals of laparoscopic surgery (FLS) skills test using Google wearable device.

BACKGROUND: The fundamentals of laparoscopic surgery (FLS) program is a simulation-based training program designed to teach and assess the basic skills necessary for laparoscopic surgery. Preliminary work has demonstrated the feasibility of using Skype™ as a telesimulation modality in reliably scoring the exam for remote centers. Google Glass (GG) (Mountain View, California) is referred to as a wearable computer containing a heads-up display and front-facing camera allowing point-of-view video transmission. The objective of this study was to evaluate the feasibility of GG in scoring the technical skills component of the FLS exam. METHODS: Twenty-eight participants were asked to complete the peg transfer and intracorporeal knot tasks of FLS using both GG and Skype™ setups. GG employed a third-party HIPAA-compliant video software (Pristine; Austin, TX) for video transmission. Participants were alternated between setups and evaluated by onsite and remote proctors. Times and errors were recorded by both proctors. Interrater reliability of their FLS scores was compared using Intraclass Correlation Coefficients (ICCs). GG experience was evaluated based on participant survey responses using a 5-point Likert scale. RESULTS: Interrater reliability for GG demonstrated a statistically significant correlation between onsite (OP) and remote (RP) proctors with ICCs of 0.985 (95% Confidence Interval [CI], 0.969-0.993) and 0.997 (95% CI 0.993-0.998), respectively, for peg and suture tasks. Skype™ demonstrated ICCs of 1.0 (95% CI 1.0-1.0). Average Likert scale responses found GG to be distracting (2.71), obstructive of the view (2.79), and a limitation to task execution (2.75). Overall, there was no statistical difference in scores between GG and Skype™ setups for either the peg (t = 1.446, p = 0.154) or suture task (t = - 0.710, p = 0.480), only 1 participant found the use of GG superior to Skype™. CONCLUSIONS: Our findings suggest that although GG are feasible in remote assessment of FLS with strong interrater reliability (ICC > 0.95), Skype™ was the preferred modality.

Perioperative mortality rates in low-income and middle-income countries: a systematic review and meta-analysis.

INTRODUCTION: The Lancet Commission on Global Surgery proposed the perioperative mortality rate (POMR) as one of the six key indicators of the strength of a country's surgical system. Despite its widespread use in high-income settings, few studies have described procedure-specific POMR across low-income and middle-income countries (LMICs). We aimed to estimate POMR across a wide range of surgical procedures in LMICs. We also describe how POMR is defined and reported in the LMIC literature to provide recommendations for future monitoring in resource-constrained settings. METHODS: We did a systematic review of studies from LMICs published from 2009 to 2014 reporting POMR for any surgical procedure. We extracted select variables in duplicate from each included study and pooled estimates of POMR by type of procedure using random-effects meta-analysis of proportions and the Freeman-Tukey double arcsine transformation to stabilise variances. RESULTS: We included 985 studies conducted across 83 LMICs, covering 191 types of surgical procedures performed on 1 020 869 patients. Pooled POMR ranged from less than 0.1% for appendectomy, cholecystectomy and caesarean delivery to 20%-27% for typhoid intestinal perforation, intracranial haemorrhage and operative head injury. We found no consistent associations between procedure-specific POMR and Human Development Index (HDI) or income-group apart from emergency peripartum hysterectomy POMR, which appeared higher in low-income countries. Inpatient mortality was the most commonly used definition, though only 46.2% of studies explicitly defined the time frame during which deaths accrued. CONCLUSIONS: Efforts to improve access to surgical care in LMICs should be accompanied by investment in improving the quality and safety of care. To improve the usefulness of POMR as a safety benchmark, standard reporting items should be included with any POMR estimate. Choosing a basket of procedures for which POMR is tracked may offer institutions and countries the standardisation required to meaningfully compare surgical outcomes across contexts and improve population health outcomes.

A base-stabilised arsenic(III) dication.

A bulky guanidinate ligand has been utilized in supporting a dichloroarsine, which is an excellent precursor to an arsenium cation and a 2,2'-bipyridine base stabilised arsetidinium dication.