Philippe Revel

Background: Public health surveillance relies on the collection of data, often in near-real time. Recent advances in natural language processing make it possible to envisage an automated system for extracting information from electronic health records. Objective: To study the feasibility of setting up a national trauma observatory in France, we compared the performance of several automatic language processing methods in a multiclass classification task of unstructured clinical notes. Methods: A total of 69,110 free-text clinical notes related to visits to the emergency departments of the University Hospital of Bordeaux, France, between 2012 and 2019 were manually annotated. Among these clinical notes, 32.5% (22,481/69,110) were traumas. We trained 4 transformer models (deep learning models that encompass attention mechanism) and compared them with the term frequency–inverse document frequency associated with the support vector machine method. Results: The transformer models consistently performed better than the term frequency–inverse document frequency and a support vector machine. Among the transformers, the GPTanam model pretrained with a French corpus with an additional autosupervised learning step on 306,368 unlabeled clinical notes showed the best performance with a micro F1-score of 0.969. Conclusions: The transformers proved efficient at the multiclass classification of narrative and medical data. Further steps for improvement should focus on the expansion of abbreviations and multioutput multiclass classification.

Recent data suggest that 10–20% of injury patients will suffer for several months after the event from diverse symptoms, generally referred to as post-concussion-like symptoms (PCLS), which will lead to a decline in quality of life. A preliminary randomized control trial suggested that this condition may be induced by the stress experienced during the event or emergency room (ER) stay and can be prevented in up to 75% of patients with a single, early, short eye movement desensitization and reprocessing (EMDR) psychotherapeutic session delivered in the ER. The protocol of the SOFTER 3 study was designed to compare the impact on 3-month PCLS of early EMDR intervention and usual care in patients presenting at the ER. Secondary outcomes included 3-month post-traumatic stress disorder, 12-month PCLS, self-reported stress at the ER, self-assessed recovery expectation at discharge and 3 months, and self-reported chronic pain at discharge and 3 months. This is a two-group, open-label, multicenter, comparative, randomized controlled trial with 3- and 12-month phone follow-up for reports of persisting symptoms (PCLS and post-traumatic stress disorder). Those eligible for inclusion were adults (≥18 years old) presenting at the ER departments of the University Hospital of Bordeaux and University Hospital of Lyon, assessed as being at high risk of PCLS using a three-item scoring rule. The intervention groups were a (1) EMDR Recent Traumatic Episode Protocol intervention performed by a trained psychologist during ER stay or (2) usual care. The number of patients to be enrolled in each group was 223 to evidence a 15% decrease in PCLS prevalence in the EMDR group. In 2012, the year of the last national survey in France, 10.6 million people attended the ER, some of whom did so several times since 18 million visits were recorded in the same year. The SOFTER 3 study therefore addresses a major public health challenge. Clinical Trials. NCT03400813 . Registered 17 January 2018 – retrospectively registered.

Introduction: In the emergency setting, focused cardiac ultrasound has become a fundamental tool for diagnostic, initial emergency treatment and triage decisions. A new ultra-miniaturized pocket ultrasound device (PUD) may be suited to this specific setting. Therefore, we aimed to compare the diagnostic ability of an ultra-miniaturized ultrasound device (Vscan™, GE Healthcare, Wauwatosa, WI) and of a conventional high-quality echocardiography system (Vivid S5™, GE Healthcare) for a cardiac focused ultrasonography in patients admitted to the emergency department. Methods: During 4 months, patients admitted to our emergency department and requiring transthoracic echocardiography (TTE) were included in this single-center, prospective and observational study. Patients underwent TTE using a PUD and a conventional echocardiography system. Each examination was performed independently by a physician experienced in echocardiography, unaware of the results found by the alternative device. During the focused cardiac echocardiography, the following parameters were assessed: global cardiac systolic function, identification of ventricular enlargement or hypertrophy, assessment for pericardial effusion and estimation of the size and the respiratory changes of the inferior vena cava (IVC) diameter. Results: One hundred fifty-one (151) patients were analyzed. With the tested PUD, the image quality was sufficient to perform focused cardiac ultrasonography in all patients. Examination using PUD adequately qualified with a very good agreement global left ventricular systolic dysfunction (κ = 0.87; 95%CI: 0.76-0.97), severe right ventricular dilation (κ = 0.87; 95%CI: 0.71-1.00), inferior vena cava dilation (κ = 0.90; 95%CI: 0.80-1.00), respiratory-induced variations in inferior vena cava size in spontaneous breathing (κ = 0.84; 95%CI: 0.71-0.98), pericardial effusion (κ = 0.75; 95%CI: 0.55-0.95) and compressive pericardial effusion (κ = 1.00; 95%CI: 1.00-1.00). Conclusions: In an emergency setting, this new ultraportable echoscope (PUD) was reliable for the real-time detection of focused cardiac abnormalities.

Introduction: Passive leg raising (PLR) is a simple reversible maneuver that mimics rapid fluid loading and increases cardiac preload. The effects of this endogenous volume expansion on stroke volume enable the testing of fluid responsiveness with accuracy in spontaneously breathing patients. However, this maneuver requires the determination of stroke volume with a fast-response device, because the hemodynamic changes may be transient. The Vigileo™ monitor (Vigileo™; Flotrac™; Edwards Lifesciences, Irvine, CA, USA) analyzes systemic arterial pressure wave and allows continuous stroke volume monitoring. The aims of this study were (i) to compare changes in stroke volume induced by passive leg raising measured with the Vigileo™ device and with transthoracic echocardiography and (ii) to compare their ability to predict fluid responsiveness. Methods: Thirty-four patients with spontaneous breathing activity and considered for volume expansion were included. Measurements of stroke volume were obtained with transthoracic echocardiography (SV-TTE) and with the Vigileo™ (SV-Flotrac) in a semi-recumbent position, during PLR and after volume expansion (500 ml saline). Patients were responders to volume expansion if SV-TTE increased ≥ 15%. Results: Four patients were excluded. No patients received vasoactive drugs. Seven patients presented septic hypovolemia. PLR-induced changes in SV-TTE and in SV-Flotrac were correlated (r² = 0.56, P < 0.0001). An increase in SV-TTE ≥ 13% during PLR was predictive of response to volume expansion with a sensitivity of 100% and a specificity of 80%. An increase in SV-Flotrac ≥16% during PLR was predictive of response to volume expansion with a sensitivity of 85% and a specificity of 90%. There was no difference between the area under the ROC curve for PLR-induced changes in SV-TTE (AUC = 0.96 ± 0.03) or SV-Flotrac (AUC = 0.92 ± 0.05). Volume expansion-induced changes in SV-TTE correlated with volume expansion-induced changes in SV-Flotrac (r² = 0.77, P < 0.0001). In all patients, the highest plateau value of SV-TTE recorded during PLR was obtained within the first 90 s following leg elevation, whereas it was 120 s for SV-Flotrac. Conclusions: PLR-induced changes in SV-Flotrac are able to predict the response to volume expansion in spontaneously breathing patients without vasoactive support.