Debunking Myths about ‘Decision Making in Critical Care’ Using Generative Artificial Intelligence

PurposeEvery day, critical care providers in the prehospital setting respond to time sensitive and outcome-critical emergencies, often in unfamiliar environments with little or no prior knowledge about the patient. In these demanding situations, they must make multifactorial clinical decisions that may be critical for the patient’s life and future health. Errors in this complex decision-making have identified as a significant cause of patient harm and, consequently, there is increasing research focus upon clinical decision-making and risk mitigation in prehospital critical care. Cognitive biases have been identified as a common cause of these systematic errors in the hospital environment and these studies inspired the aim of this article to map current evidence and investigate, “What cognitive biases affects clinical decision-making in prehospital critical care”.Materials and methodsA scoping review was conducted following Joanna Briggs Institute`s framework, by searching OVID MEDLINE and PubMed, EMBASE, and Cochrane for articles, no restrictions were set for type of article. Articles describing cognitive biases and clinical decision-making in pre-, and in-hospital critical care were included. Additionally, a search in Google scholar was conducted using keywords identified in included articles.ResultsFive hundred unique articles were identified through the search, of which 16 articles examining cognitive biases and clinical decision making in critical care were included, with only two articles focussed exclusively on prehospital critical care. Twenty-eight unique cognitive biases were identified in these articles. The most identified cognitive biases were, anchoring bias, framing effect, availability bias, confirmation bias, overconfidence bias, premature closure, and omission bias. Twelve articles described contributing factors for cognitive biases and these were categorized into 3 main categories. The main categories identified were lack of unbiased feedback, social behaviour and beliefs, and time pressure. Eleven articles proposed mitigation factors, which were categorized into 3 categories, consisting of feedback and follow-up, organizational culture, and education and training.ConclusionsThis scoping review has identified several cognitive biases that affect clinical decision-making, as well as research gaps in both pre- and in-hospital critical care. Identified evidence suggest that both clinicians and organisations are affected by cognitive biases in clinical decision-making in critical care. Future research should aim to establish how these cognitive biases affect clinical decisions in prehospital critical care, and what measures may mitigate the consequent errors, may reduce patient harm, and improve outcomes.

Background and Objective: Despite its increasing use, in critical care, the diagnosis accuracy of Lung Ultransound Sonographye (LUS) is a topic that remains highly debated. The objective of our study was to evaluate the performance of LUS compared to chest computer tomography (CT) in the diagnosis of acute respiratory condictions in critical patients at Sens hospital from March 1st to September 31st 2023. Methods: Single-center, cross-sectional study of 54 patients, all adult with acute respiratory symptoms who benefited concurrent LUS and chest CT scan. The LUS was performed according to the BLUE-Protocol by the doctors of the department, within 8 hours before the chest CT scan was performed. The CT scan was interpreted by the hospital's radiologists who were also blinded by the LUS results. The performance of LUS was assessed by calculating sensitivity (se), specificity (sp), positive predictive values (PPV) and negative predictive values (NPV) compared to chest CT. Results: Pneumothorax had a Se=100%, Sp=100, PPV=100%, NPV=100%, Youden=1.00, AUC=1.00; Interstitial syndrome had a Se=100%, Sp=98.1%, PPV=50%, NPV=100%, Youden=0.98%, AUC=0.99 with p=0.14. Alveolar consolidation had Se=100%, Sp=90.5%, PPV=94.3%, NPV=100%, Youden=0.90%, AUC=0.95%. Pleural effusion had Se=100%, Sp=88.9%, PPV=90%, NPV=100%, Youden=0.89%, AUC=0.94%. Pulmonary embolism had a Se=66.7%; Sp=97.9%, PPV=80%, NPV=95.9%, Youden=0.65%, AUC=0.82% with p=0.11. As for COPD/ Asthma decompensation, she had a Se=NA (not applicable), Sp=100%, PPV=NA, NPV=100%, Youden=NA and AUC=100%. Conclusion: The performance of LUS in this study are satisfactory for the diagnosis of alveolar consolidation, pneumothorax, pleural fluid effusion and interstitial syndrome (only as a function of Youden's index). They are less good for the diagnosis of pulmonary embolism. Thus, we propose it as an alternative to CT in patients in critical resuscitation.

‘MORAL balance’ decision-making in critical care

In critical care when unconscious patients are assisted by machines, humanity is mainly ensured by respect for autonomy, realised through advance directives or, mostly, reconstructed by cooperation with relatives. Whereas patient-centred approaches are widely discussed and fostered, managing communication in complex, especially end-of-life, situations in open intensive care units is still a point of debate and a possible source of conflict and moral distress. In particular, healthcare teams are often sceptical about the growing role of families in shared decision-making and their ability to represent patients’ preferences. New perspectives on substituted relational autonomy are needed for overcoming this climate of suspicion and are discussed through recent literature in the field of medical ethics.

BACKGROUND PATIENT: ventilator dyssynchrony is common and may influence patients' outcomes. Detection of such dyssynchronies relies on careful observation of patients and airway flow and pressure measurements. Given the shortage of specialists, critical care nurses could be trained to identify dyssynchronies. To evaluate the accuracy of specifically trained critical care nurses in detecting ineffective inspiratory efforts during expiration. We compared 2 nurses' evaluations of measurements from 1007 breaths in 8 patients with the evaluations of experienced critical care physicians. Sensitivity, specificity, positive predictive value, negative predictive value, and the Cohen κ for interobserver agreement were calculated. For the first nurse, sensitivity was 92.5%, specificity was 98.3%, positive predictive value was 95.4%, negative predictive value was 97.1%, and κ was 0.92 (95% CI, 0.89-0.94). For the second nurse, sensitivity was 98.5%, specificity was 84.7%, positive predictive value was 70.7%, negative predictive value was 99.3%, and κ was 0.74 (95% CI, 0.70-0.78). Specifically trained nurses can reliably detect ineffective inspiratory efforts during expiration.

Large Language Model Application in Emergency Medicine and Critical Care

Biomedical ethics has assumed an increasingly important role in medicine over the past 30 yrs, and its development has served the important goal of protecting patients' rights and interests. However, medical ethics has evolved within a Western tradition, and conflict often arises when trying to apply Western medical ethics to patients from other cultures. Using Hong Kong as an example, this article reviews the nature and sources of cross-cultural conflict in the intensive care unit setting that often arises between physicians trained in Western medicine and patients from a Chinese cultural background. This article draws on the first author's experience as a critical care physician in Hong Kong, and on a review of the literature on cross-cultural interactions in medicine. Studies were selected that contrasted the approaches of different cultures to common ethical dilemmas in medicine. Review articles examining the relationship between culture and ethics were also selected. Hong Kong presents an interesting case study because of the coexistence of Western and Chinese medicine in a predominantly Chinese population that practices many Chinese cultural traditions. Whereas contemporary Western medical ethics focuses on individual rights, autonomy, and self-determination, traditional Chinese societies place greater emphasis on such community values as harmony, responsibility, and respect for parents and ancestors. Specific areas of cross-cultural conflict include: the role of the patient and family in medical decision-making; the disclosure of unfavorable medical information to critically ill patients; the discussion of advance directives or code status with patients; and the withholding or withdrawal of life support.

Conflicts arise within teams and with family members in end-of-life decision-making in critical care. This creates unnecessary discomfort for all involved, including the patient. Treatment plans driven by crisis open the team up to conflict, fragmented care and a lack of focus on the patient's wishes and realistic medical outcomes. Methods to resolve these issues involve planned ethical reviews and team meetings where open communication, clear plans and involvement in decision-making for all stakeholders occur. In spite of available literature supporting the value of these techniques, patient care teams and families continue to find themselves involved in spiraling conflict, pitting one team against another, placing blame on family members for not accepting decisions made by the team and creating moral conflict for interdisciplinary team members. Through a case presentation, we review processes available to help resolve conflict and to improve outcome.

Traumatic brain injury (TBI) remains a major cause of morbidity and mortality worldwide and requires complex decision-making in critical care. This narrative review analyzes current updates in neurosurgical management and intracranial pressure (ICP) monitoring in moderate and severe TBI. A literature review was conducted using PubMed, Scopus, Web of Science, and SciELO databases (2015–2026), focusing on clinical trials, observational studies, guidelines, and relevant reviews addressing neurosurgical interventions and neurocritical monitoring strategies. Contemporary TBI management prioritizes prevention of secondary brain injury through physiology-guided treatment. Continuous ICP monitoring plays a central role in guiding stepwise therapeutic interventions, including sedation optimization, ventilation control, osmotherapy, and hemodynamic adjustments aimed at maintaining cerebral perfusion pressure. Decompressive craniectomy remains an important option for refractory intracranial hypertension, demonstrating mortality reduction in selected patients, although functional outcomes remain heterogeneous and dependent on baseline prognosis and timing of intervention. The incorporation of multimodal neuromonitoring, including brain tissue oxygenation, microdialysis, and autoregulation assessment, has shifted neurocritical care from a reactive to a preventive approach, enabling earlier detection of cerebral hypoperfusion and metabolic distress. However, challenges persist regarding interpretation of monitoring data, definition of surgical timing, and availability of specialized resources. Overall, optimal management of TBI depends on integration of clinical evaluation, imaging findings, and physiological monitoring, supported by structured protocols and trained multidisciplinary teams.

Background: Hypoalbuminemia is a hallmark of critical illness, reflecting both systemic inflammation and impaired homeostasis. This systematic review and meta-analysis examine the prognostic significance of low serum albumin in shock patients, integrating data from diverse global cohorts. Methods: Comprehensive searches of PubMed, Europe PMC, and Google Scholar were conducted (2019-2025) to identify studies reporting adjusted odds ratios (ORs) or hazard ratios (HRs) for the association between hypoalbuminemia and mortality in adult shock patients. Separate random-effects meta-analyses were performed for ORs and HRs. Heterogeneity was quantified with I2 statistics, and meta-regression was applied to explore study-level covariates. Results: Eleven studies met inclusion criteria (n = 7 OR studies; n = 4 HR studies). Pooled ORs and HRs each demonstrated a robust association between low serum albumin and increased mortality, independent of major confounders. Sensitivity analyses confirmed the stability of findings. Age significantly modified OR-based associations, while continental location showed borderline influence in HR models. Conclusions: Low serum albumin is a powerful, independent prognostic biomarker in shock. These findings reinforce albumin’s potential utility in risk stratification and therapeutic decision-making in critical care.

Background. Chest injuries during combat operations occupy a prominent place and often become the cause of mortality. Before the widespread introduction of imaging methods into clinical practice, the mortality rate for chest combat injuries exceeded 50 %. Objective: a comparative assessment of radiography and ultrasonography options in the diagnosis of pneumothorax that occurred as a result of combat trauma. Materials and methods. A comparative analysis of the radiography and ultrasonography results was carried out in 76 patients with pneumothorax due to chest combat trauma. Results. During X-ray in the supine position, the sensitivity of the method was 58.1 %, specificity — 72.7 %, accuracy — 64.5 %, positive predictive value — 73.5 %, negative predictive value — 57.1 %. The sensitivity of the method in the sitting position of patients was 71.9 %, specificity — 89.5 %, accuracy — 76.3 %, positive predictive value — 95.3 %, negative predictive value — 51.5 %. Ultrasound diagnosis of pneumothorax was based on identifying the “barcode” sign due to the lack of sliding of the visceral pleura during the patient’s inspiration. The sensitivity of ultrasonography in B-mode was 90.8 %, specificity — 81.8 %, accuracy — 89.5 %, positive predictive value — 96.7 %, negative predictive value — 60.0 %, and in combined B + M modes — 94.0, 88.9, 93.4, 98.4, 66.7 %, respectively. In the diagnosis of large pneumothorax, the sensitivity of radiography was 96.8 %, specificity — 100.0 %, accuracy — 96.9 %, positive predictive value — 100.0 %, negative predictive value — 50.0 %, respectively, and of ultrasonography — 96.7, 100.0, 96.9, 100.0, 66.7 %, respectively. Conclusions. Ultrasonography has greater sensitivity for diagnosing small pneumothorax than conventional radiography, especially in bedridden patients. Ultrasonography in combined B + M modes can be both a primary and a clarifying method for diagnosing pneumothorax in combat trauma.

NEUTROPHILE/LYMPHOCYTE INDEX (IN/L), CREATININE (Cr), AND PROCALCITONIN (PROCAL) AS PREDICTORS OF AMEBIAN LIVER ABSCESS.

Predicting Post-Heart Transplant Composite Renal Outcome Risk in Adults: A Machine Learning Decision Tool

An updated probabilistic causality assessment method and the Liverpool algorithm presented as an improved version of the Naranjo algorithm, one of the most used and accepted causality assessment methods, have recently been proposed. In order to test the validity of the probabilistic method in routine pharmacovigilance, results provided by the Naranjo and Liverpool algorithms, as well as the updated probabilistic method, were each compared with a consensual expert judgement taken as reference. A sample of 59 drug-event pairs randomly sampled from spontaneous reports to the French pharmacovigilance system was assessed by expert judgement until reaching consensus and by members of a pharmacovigilance unit using the updated probabilistic method, the Naranjo and Liverpool algorithms. Probabilities given by the probabilistic method, and categories obtained by both the Naranjo and the Liverpool algorithms were compared as well as their sensitivity, specificity, positive and negative predictive values. The median probability for drug causation given by the consensual expert judgement was 0.70 (inter-quartile range, IQR 0.54-0.84) versus 0.77 (IQR 0.54-0.91) for the probabilistic method. For the Naranjo algorithm, the 'possible' causality category was predominant (61%), followed by 'probable' (35%), 'doubtful', and 'almost certain' categories (2% each). Category distribution obtained with the Liverpool algorithm was similar to that obtained by the Naranjo algorithm with a majority of 'possible' (61%) and 'probable' (30%) followed by 'definite' (7%) and 'unlikely' (2%). For the probabilistic method, sensitivity, specificity, positive and negative predictive values were 0.96, 0.56, 0.92 and 0.71, respectively. For the Naranjo algorithm, depending on whether the 'possible' category was considered in favour or in disfavour of drug causation, sensitivity was, respectively, 1 or 0.42, specificity 0.11 or 0.89, negative predictive value 1 or 0.22 and positive predictive value 0.86 or 0.95; results were identical for the Liverpool algorithm. The logistic probabilistic method gave results closer to the consensual expert judgment than either the Naranjo or Liverpool algorithms whose performance were strongly dependent on the meaning given to the 'possible' category. Owing to its good sensitivity and positive predictive value and by providing results as continuous probabilities, the probabilistic method seems worthy to use for a trustable assessment of adverse drug reactions in routine practice.

Which Combination is the Best? A Comparison of the Predictive Potential of Serum Biomarker Combinations to Diagnose Periprosthetic Joint Infection