ABSTRACT
Artificial intelligence (AI) has demonstrated a disruptive ability to enhance and transform clinical medicine. While the dexterous nature of anesthesiology work offers some protections from AI clinical assimilation, this technology will ultimately impact the practice and augment the ability to provide an enhanced level of safe and data-driven care. Whether predicting difficulties with airway management, providing perioperative or critical care risk assessments, clinical-decision enhancement, or image interpretation, the indications for AI technologies will continue to grow and are limited only by our collective imagination on how best to deploy this technology.An essential mission of academia is education, and challenges are frequently encountered when working to develop and implement comprehensive and effectively targeted curriculum appropriate for the diverse set of learners assigned to teaching faculty. Curriculum development in this context frequently requires substantial efforts to identify baseline knowledge, learning needs, content requirement, and education strategies. Large language models offer the promise of targeted and nimble curriculum and content development that can be individualized to a variety of learners at various stages of training. This technology has not yet been widely evaluated in the context of education deployment, but it is imperative that consideration be given to the role of AI in curriculum development and how best to deploy and monitor this technology to ensure optimal implementation.
ABSTRACT
INTRODUCTION: Artificial intelligence and large language models (LLMs) have emerged as potentially disruptive technologies in healthcare. In this study GPT-3.5, an accessible LLM, was assessed for its accuracy and reliability in performing guideline-based evaluation of neuraxial bleeding risk in hypothetical patients on anticoagulation medication. The study also explored the impact of structured prompt guidance on the LLM's performance. METHODS: A dataset of 10 hypothetical patient stems and 26 anticoagulation profiles (260 unique combinations) was developed based on American Society of Regional Anesthesia and Pain Medicine guidelines. Five prompts were created for the LLM, ranging from minimal guidance to explicit instructions. The model's responses were compared with a "truth table" based on the guidelines. Performance metrics, including accuracy and area under the receiver operating curve (AUC), were used. RESULTS: Baseline performance of GPT-3.5 was slightly above chance. With detailed prompts and explicit guidelines, performance improved significantly (AUC 0.70, 95% CI (0.64 to 0.77)). Performance varied among medication classes. DISCUSSION: LLMs show potential for assisting in clinical decision making but rely on accurate and relevant prompts. Integration of LLMs should consider safety and privacy concerns. Further research is needed to optimize LLM performance and address complex scenarios. The tested LLM demonstrates potential in assessing neuraxial bleeding risk but relies on precise prompts. LLM integration should be approached cautiously, considering limitations. Future research should focus on optimization and understanding LLM capabilities and limitations in healthcare.
ABSTRACT
BACKGROUND: Large language models (LLMs) excel at answering knowledge-based questions. Many aspects of blood banking and transfusion medicine involve no direct patient care and require only knowledge and judgment. We hypothesized that public LLMs could perform such tasks with accuracy and precision. STUDY DESIGN AND METHODS: We presented three sets of tasks to three publicly-available LLMs (Bard, GPT-3.5, and GPT-4). The first was to review short case presentations and then decide if a red blood cell transfusion was indicated. The second task was to answer a set of consultation questions common in clinical transfusion practice. The third task was to take a multiple-choice test experimentally validated to assess internal medicine postgraduate knowledge of transfusion practice (the BEST-TEST). RESULTS: In the first task, the area under the receiver operating characteristic curve for correct transfusion decisions was 0.65, 0.90, and 0.92, respectively for Bard, GPT-3.5 and GPT-4. All three models had a modest rate of acceptable responses to the consultation questions. Average scores on the BEST-TEST were 55%, 40%, and 87%, respectively. CONCLUSION: When presented with transfusion medicine tasks in natural language, publicly available LLMs demonstrated a range of ability, but GPT-4 consistently scored very well in all tasks. Research is needed to assess the utility of LLMs in transfusion medicine practice. Transfusion Medicine physicians should consider their role alongside such technologies, and how they might be used for the benefit and safety of patients.
Subject(s)
Physicians , Transfusion Medicine , Humans , Artificial Intelligence , Blood Transfusion , Erythrocyte TransfusionABSTRACT
Background: The aim of this study was to validate a three-class sentiment classification model for clinical trial abstracts combining adversarial learning and the BioBERT language processing model as a tool to assess trends in biomedical literature in a clearly reproducible manner. We then assessed the model's performance for this application and compared it to previous models used for this task. Methods: Using 108 expert-annotated clinical trial abstracts and 2,000 unlabeled abstracts this study develops a three-class sentiment classification algorithm for clinical trial abstracts. The model uses a semi-supervised model based on the Bidirectional Encoder Representation from Transformers (BERT) model, a much more advanced and accurate method compared to previously used models based upon traditional machine learning methods. The prediction performance was compared to those previous studies. Results: The algorithm was found to have a classification accuracy of 91.3%, with a macro F1-Score of 0.92, significantly outperforming previous studies used to classify sentiment in clinical trial literature, while also making the sentiment classification finer grained with greater reproducibility. Conclusion: We demonstrate an easily applied sentiment classification model for clinical trial abstracts that significantly outperforms previous models with greater reproducibility and applicability to large-scale study of reporting trends.
ABSTRACT
Publication bias and the decreased publication of trials with negative or non-significant results is a well-recognized problem in human and veterinary medical publications. These biases may present an incomplete picture of evidence-based clinical care and negatively impact medical practices. The purpose of this study was to utilize a novel sentiment analysis tool as a quantitative measure for assessing clinical trial reporting trends in human and veterinary medical literature. Abstracts from 177,617 clinical trials in human medical journals and 8684 in veterinary medical journals published in the PubMed database from 1995 to 2020. Abstracts were analyzed using the GAN-BioBERT sentiment classifier for both general trends and percentage of neutral/negative publications. Sentiment was defined on a - 1 (highly negative) to 1 (highly positive) scale. Human-based clinical trial publications were less likely to feature positive findings (OR 0.87, P < 0.001) and more likely to include neutral findings (OR 1.18, P < 0.001) relative to veterinary clinical trials. No difference was found in reporting of negative sentiment trials (OR 1.007, P = 0.83). In both groups, the published sentiment of clinical trials increased over time. Using sentiment analysis to evaluate large publication datasets and compare publication trends within and between groups, this study is significant in its detection of significant publication differences between human and veterinary medicine clinical trials and a continued unbalanced positive sentiment in the published literature. The implications of this unbiased reporting have important clinical and research implications that require consideration.
Subject(s)
Sentiment Analysis , Animals , Humans , Publication BiasABSTRACT
INTRODUCTION: Sentiment analysis, by evaluating written wording and its context, is a growing tool used in computer science that can determine the level of support expressed in a body of text using artificial intelligence methodologies. The application of sentiment analysis to biomedical literature is a growing field and offers the potential to rapidly and economically explore large amounts of published research and characterize treatment efficacy. METHODS: We compared the results of sentiment analysis of 115 article abstracts analyzed in a recently published meta-analysis of peripheral nerve block usage in primary hip and knee arthroplasty to the conclusions drawn by the authors of the original meta-analysis. RESULTS: A moderately positive outlook supporting the utilization of regional anesthesia for hip and knee arthroplasty was found in the 115 articles that were included for analysis, with 46% expressing positive sentiment, 35% expressing neutral sentiment, and 19% of abstracts expressing negative sentiment. This was well aligned with the conclusions reached by a previous meta-analysis of the same articles. DISCUSSION: Sentiment analysis applied to the medical literature can rapidly evaluate large collections of published data and generate an impression of overall findings that are aligned with the findings of a traditional meta-analysis.
Subject(s)
Anesthesia, Conduction , Arthroplasty, Replacement, Knee , Arthroplasty, Replacement, Knee/adverse effects , Artificial Intelligence , Humans , LanguageABSTRACT
The US Health and Human Services Pain Management Best Practices Inter-Agency Task Force initiated a public-private partnership which led to the publication of its report in 2019. The report emphasized the need for individualized, multimodal, and multidisciplinary approaches to pain management that decrease the over-reliance on opioids, increase access to care, and promote widespread education on pain and substance use disorders. The Task Force specifically called on specialty organizations to work together to develop evidence-based guidelines. In response to this report's recommendations, a consortium of 14 professional healthcare societies committed to a 2-year project to advance pain management for the surgical patient and improve opioid safety. The modified Delphi process included two rounds of electronic voting and culminated in a live virtual event in February 2021, during which seven common guiding principles were established for acute perioperative pain management. These principles should help to inform local action and future development of clinical practice recommendations.
Subject(s)
Analgesics, Opioid , Pain Management , Analgesics, Opioid/adverse effects , Consensus , HumansABSTRACT
OBJECTIVE: The purpose of this study was to investigate the efficacy of hyperangulated video laryngoscopy (HAVL) versus standard geometry video laryngoscopy (SGVL) in the simulated prehospital environment using a manikin. There is consensus that video laryngoscopy (VL) can be very useful in the emergency department when difficult intubations are predicted. Emergency medical service (EMS) providers are also often faced with difficult, rapidly deteriorating airway management situations that not only involve patient and operator factors but also include challenging unique environmental factors, such as nonoptimized positions in transport vehicles (eg, helicopters and ambulances) or at ground level or entrapped positions. To our knowledge, there has never been a study purposefully investigating the efficacy of hyperangulated geometry versus standard geometry VL techniques in the prehospital environment. METHODS: A single-center, randomized controlled crossover trial was performed using attending physician helicopter EMS providers. Physicians were randomized to perform 5 HAVL or SGVL intubations followed by the subsequent technique. Intubations were performed on ground level and then repeated in the helicopter with the first location also randomized. A manikin airway management trainer was used to simulate intubation in each environment. The time to intubation (primary outcome) as well as first-pass success and the Cormack-Lehane view were recorded for each attempt. Qualitative data were also obtained for physician preference and perceived difficulty. RESULTS: There was no statistically significant difference in the time to intubation with HAVL versus SGVL (ground: 15.02 vs. 14.88 seconds, P = .86; helicopter: 16.11 vs. 16.14 seconds, P = .93). First-pass success was 100% for both techniques in both scenarios. More Grade 1 views were obtained with HAVL (147/150 vs. 134/150). Moreover, most physicians preferred HAVL overall and felt that HAVL required less force (9/15 grounded manikin and 10/15 helicopter manikin) and led to the best chance for first-pass success (11/15 grounded manikin and 10/15 helicopter manikin). CONCLUSION: The results of this study are limited because of the static and highly favorable anatomy of a manikin versus the variability and often difficult anatomy of individual patients. Our results suggest that both techniques are efficacious when the patient is both on the ground or in the helicopter, although provider preference does seem to vary.
Subject(s)
Emergency Medical Services , Laryngoscopes , Cross-Over Studies , Humans , Intubation, Intratracheal , Laryngoscopy , Manikins , Video RecordingSubject(s)
Anesthesia, Conduction , Analgesics , Analgesics, Opioid , Humans , Pain , Pain Management , Pain, PostoperativeABSTRACT
PURPOSE: We sought to identify the immediate postoperative differences in opioid use, pain scores, and post-anesthesia care unit (PACU) length of stay (LOS) after hip arthroscopy related to the type of anesthesia used for the surgical procedure. METHODS: Patients undergoing hip arthroscopy for femoroacetabular impingement syndrome with labral tears by a single surgeon at an academic center between October 2017 and July 2019 were reviewed retrospectively. The primary outcome was PACU opioid administration, measured by morphine equivalents. Secondary parameters included total LOS, postincision LOS, PACU LOS, and PACU arrival/discharge pain scores. Analyses conducted were t tests, Wilcoxon rank sum tests, or χ2 tests. RESULTS: A total of 129 patients met inclusion criteria for this study; 54 male and 75 female, with an average age of 28 (±10.1) years. In total, 52 (40.3%) had general anesthesia and 77 (59.7%) had neuraxial anesthesia, including spinal, epidural, and combined spinal-epidural anesthesia, which were intermixed throughout the study period. Intraoperative and PACU opioid administration demonstrated a significant difference in medians. Neuraxial methods required a lower morphine equivalents in both the operating room (30.0 vs 53.9, P = .001) and PACU (18.2 vs 31.2, P = 0.002). Neuraxial anesthesia had lower median PACU arrival and discharge pain scores (0.0 vs. 5.0, P = .001, 3.0 vs. 4.0, P = .013). There was no statistically significant difference in postincision LOS or traction time. General anesthesia was associated with a longer PACU phase 1 time (1.0 vs 1.3 hours, P = .005). No major adverse events such as death, disability, or prolonged hospitalization occurred in either group. CONCLUSIONS: Neuraxial anesthesia use in routine hip arthroscopy was associated with lower immediate postoperative pain scores, lower intraoperative and immediate postoperative opioid requirements, and may be associated with shorter anesthesia recovery time without any major adverse events when compared with general anesthesia. LEVEL OF EVIDENCE: III, Retrospective Comparative Study.
Subject(s)
Anesthesia, Epidural , Anesthesia, Spinal , Arthroscopy , Femoracetabular Impingement/surgery , Adult , Analgesics, Opioid/therapeutic use , Anesthesia, General , Drug Utilization/statistics & numerical data , Female , Humans , Length of Stay , Male , Pain Measurement , Pain, Postoperative/prevention & control , Retrospective StudiesABSTRACT
OBJECTIVES: To describe the ultrasound-guided rectus sheath block technique and the anatomical spread of two volumes of methylene blue injection in dog cadavers. STUDY DESIGN: Blinded, prospective, experimental cadaveric study. ANIMALS: A total of eight dog cadavers weighing 8.9 ± 1.6 kg. METHODS: Ultrasound-guided rectus sheath injections were performed bilaterally 1 cm cranial to the umbilicus using 0.25 mL kg-1 (low volume; LV) and 0.50 mL kg-1 (high volume; HV) of 0.5% methylene blue dye. A total of 16 hemiabdomens were injected. The ultrasound image quality of the muscular and fascial plane landmarks and needle visualization were scored using a standardized scale. Cadavers were dissected to determine the distribution of the dye and to assess staining of ventral branches of the spinal nerves. RESULTS: Fewer ventral spinal nerve branches were stained in the LV group than in the HV group, at 2.00 ± 0 and 2.90 ± 0.83, respectively (p < 0.01). Ventral branches of thoracic (T) and lumbar (L) spinal nerves (T10, T11, T12, T13 and L1) were stained 25%, 100%, 75%, 25% and 0% of the time in LV group and 12.5%, 87.5%, 100.0%, 75.0% and 13.0% in HV group. A lesser extent of cranial-caudal dye distribution was observed in the LV group than in the HV group (7.1 ± 1.8 cm and 9.2 ± 1.8 cm, respectively; p = 0.03). There was no significant difference in medial-lateral spread of dye, number of test doses or ultrasound image quality scores between groups. CONCLUSIONS AND CLINICAL RELEVANCE: The results of this study suggest that, on an anatomical basis, this easily performed block has the potential to provide effective abdominal wall analgesia for the ventral midline. This study supports the potential of the rectus sheath block for abdominal procedures, and further investigations on its clinical efficacy are warranted.
Subject(s)
Abdominal Wall/anatomy & histology , Dogs/anatomy & histology , Methylene Blue/administration & dosage , Nerve Block/veterinary , Animals , Cadaver , Female , Male , Prospective Studies , Ultrasonography, Interventional/veterinarySubject(s)
Analgesia/methods , Buprenorphine/therapeutic use , Narcotic Antagonists/therapeutic use , Opiate Substitution Treatment/methods , Opioid-Related Disorders/drug therapy , Perioperative Care/methods , Buprenorphine/administration & dosage , Humans , Narcotic Antagonists/administration & dosage , Opioid-Related Disorders/surgery , Practice Guidelines as TopicABSTRACT
Medication shortages are a clinical reality that force changes in practice patterns leading to unintended consequences. Potential solutions to any drug shortage require a thoughtful, multidisciplinary and often creative approach. Here, we report a case of unintentional epinephrine overdose leading to an unstable cardiac arrhythmia and our subsequent development of a visual cue system to prevent future errors. A 56-year-old man with a history of rectal adenocarcinoma presented for low anterior resection and creation of diverting loop ileostomy. Epidural placement was requested by the surgical team, and following administration of a second test dose (created by the physician), the patient experienced supraventricular tachycardia with heart rates of 200-210 BPM for approximately 2 minutes. This rhythm then converted to atrial fibrillation with rapid ventricular response with heart rate of 150-170 BPM. The patient was stabilized after cardioversion. Later evaluation of medication administration revealed that the second epidural test dose inadvertently contained 100 mcg epinephrine instead of the intended 10 mcg dose. The test dose had to be created because the original ampule with the kit had been utilized. Since this time, our kits have no test dose, and this shortage is concerning for increased provider error. We suggest a novel visual cue system that may prevent unintentional epinephrine overdoses in the setting of regional anesthesia.
ABSTRACT
OBJECTIVES: To describe a technique to perform an ultrasound-guided erector spinae plane (ESP) block and determine the distribution and potential complications after injection of two volumes of methylene blue in dog cadavers. STUDY DESIGN: Prospective experimental cadaveric study. ANIMALS: A total of eight dog cadavers weighing 9.3 ± 1.9 kg. METHODS: Ultrasound-guided injections dorsal to the transverse process and ventral to the erector spinae muscles aimed at the fifth thoracic transverse process were performed bilaterally in each dog using 0.5 and 1.0 mL kg-1 dye solution [low volume (LV) and high volume (HV) treatments, respectively]. Treatments were randomly assigned to the right or left side of each dog, resulting in a total of 16 injections. Anatomical dissections determined dye spread characteristics, including epaxial muscles spread, staining of spinal nerves, dorsal rami, ventral rami (intercostal nerves) and sympathetic trunk spread. Staining indicating potential complications (epidural, mediastinal and intrapleural spread) was recorded. RESULTS: There was complete staining of at least one dorsal ramus following all injections. A more extensive spread was observed along the muscles in the HV compared with LV (p = 0.036). No significant difference between multisegmental dorsal rami spread (six out of eight injections in each treatment) was noted. Out of 16 injections, one in LV treatment resulted in multisegmental spinal nerve staining and one in HV treatment resulted in ventral ramus (intercostal nerve) staining. Use of anatomic landmarks resulted in inaccurate identification of the fifth transverse process in at least six out of 16 injections (38%). No sympathetic trunk, epidural, mediastinal or intrapleural staining was observed. CONCLUSIONS AND CLINICAL RELEVANCE: Ultrasound-guided ESP injections resulted in extensive staining along the epaxial muscles, as well as staining of the dorsal rami in all dogs. The incidence of dorsal rami mutisegmental spread was the same in both treatments.
Subject(s)
Coloring Agents/administration & dosage , Dogs , Ultrasonography, Interventional/veterinary , Anatomic Landmarks , Animals , Cadaver , Epidural Space , Nerve Block/veterinary , Prospective Studies , Thoracic Vertebrae , Ultrasonography, Interventional/methodsABSTRACT
The demand for liver transplants in the United States far exceeds the supply of organs. As need has increased, so has use of living donors. Coagulopathy and various side effects often preclude the use of neuraxial regional techniques and opioids for postoperative analgesia in patients with large "J" incisions. Here, we present a 25-year-old male undergoing a living donor hepatectomy who received quadratus lumborum catheters placed percutaneously after closure of incision and prior to emergence to provide excellent analgesia and a viable opioid-sparing approach. Quadratus lumborum catheters are a safe option for a multimodal, opioid-sparing approach to analgesia.
ABSTRACT
OBJECTIVES: To describe an ultrasound-guided thoracic paravertebral block and determine the distribution after injection of two volumes of methylene blue in dog cadavers. STUDY DESIGN: Prospective experimental cadaveric study. ANIMALS: Twelve dog cadavers weighing 11 ± 3 kg. METHODS: Ultrasound-guided injections aimed at the fifth thoracic (T5) paravertebral space were performed in randomized order using 0.1 or 0.3 mL kg-1 dye solution (six dogs for each volume). Anatomic dissections determined dye spread characteristics, including the presence and degree of staining of spinal nerves, and the presence of intercostal and sympathetic trunk spread. Staining of mediastinum, epidural, intrapleural and contralateral thoracic paravertebral space was recorded. RESULTS: There was no significant difference in dye distribution between groups. The use of anatomic landmarks resulted in the inaccurate identification of the T5 paravertebral space. The T4, T5 and T6 paravertebral spaces were injected in four, five and three of 12 dogs, respectively. Complete staining of the spinal nerve of the thoracic paravertebral space injected was observed in 11 of 12 dogs, and partial staining in one dog in the low-volume group. Multisegmental distribution was demonstrated with staining of contiguous spinal nerves in one dog in the high-volume group, and multiple segments of intercostal (three dogs) and sympathetic trunk (four dogs) spread in both groups. No mediastinal, epidural, intrapleural or contralateral thoracic paravertebral space staining was observed. CONCLUSIONS AND CLINICAL RELEVANCE: Ultrasound-guided injection at the thoracic paravertebral space resulted in staining of the spinal nerve in all dogs. However, T5 paravertebral space was not accurately identified using anatomic landmarks. Dye distribution was not significantly different between the two groups; therefore, the use of the lower-volume and multiple-site injections would be potentially necessary in clinical cases to achieve ipsilateral blockade of the thoracic wall.
Subject(s)
Dogs/surgery , Methylene Blue/metabolism , Nerve Block/veterinary , Thoracic Vertebrae , Ultrasonography, Interventional/veterinary , Animals , Cadaver , Female , Male , Nerve Block/methods , Prospective Studies , Surgery, Computer-Assisted/veterinary , Thoracic Vertebrae/metabolismABSTRACT
STUDY OBJECTIVE: The purpose of this study was to evaluate and compare hands-on gel phantom versus instructional video teaching methods to improve anesthesia residents and staff members' ability to correctly identify airway structures using ultrasound on a human volunteer. DESIGN: Randomized, controlled trial. SETTING: Simulation laboratory. STUDY SUBJECTS: Fifty-four anesthesiology resident and staff members (27 anesthesiologists and 27 anesthesiology residents) at the University of Wisconsin-Madison. INTERVENTIONS: Study subjects were randomized into one of three groups: control (standard medical knowledge), video training, or gel phantom training. After providing study instructions and training (if relevant), study subjects were asked to perform sonoanatomy identification of the thyroid cartilage, cricoid cartilage, cricothyroid membrane, and the tracheal rings in both the transverse and longitudinal views. Study subjects then returned 14 to 24 days following initial assessment for evaluation of skills retention. They were again instructed to identify the same airway structures as during the initial assessment with scoring performed by the same assessor. MAIN RESULTS: All group characteristics were similar at baseline and follow-up. Both training tools produced a learning effect at baseline and follow-up compared to standard anesthesia training. No differences in overall airway structure identification success between groups receiving video versus gel airway phantom training were observed. CONCLUSIONS: Use of either a low cost, airway gel phantom training model for hands-on training or a simple instructional teaching video can be used in a single training session to improve staff anesthesiologist and anesthesia resident knowledge and skills for ultrasound identification of upper airway anatomy.