Implementation of a Survey Spine Magnetic Resonance Imaging Protocol for Cord Compression in the Emergency Department: Experience at a Level-1 Trauma Center.

BACKGROUND AND PURPOSE: Imaging stewardship in the emergency department (ED) is vital in ensuring patients receive optimized care. While suspected cord compression (CC) is a frequent indication for total spine MRI in the ED, the incidence of CC is low. Recently, our level-I trauma center introduced a survey spine MRI protocol to evaluate for suspected CC while reducing exam time to avoid imaging overutilization. This study aims to evaluate the time savings, frequency of ordering patterns of the survey, and the symptoms and outcomes of patients undergoing the survey. MATERIALS AND METHODS: This retrospective study examined patients who received a survey spine MRI in the ED at our institution between 2018 and 2022. All exams were performed on a 1.5T GE scanner using our institutional CC survey protocol, which includes sagittal T2 and STIR sequences through the cervical, thoracic, and lumbar spine. Exams were read by a blinded, board-certified neuroradiologist. RESULTS: A total of 2,002 patients received a survey spine MRI protocol during the study period. Of these patients, 845 (42.2%, mean age 57 ± 19 years, 45% female) received survey spine MRI exams for the suspicion of CC, and 120 patients (14.2% positivity rate) had radiographic CC. The survey spine MRI averaged 5 minutes and 50 seconds (79% faster than routine MRI). On multivariate analysis, trauma, back pain, lower extremity weakness, urinary or bowel incontinence, numbness, ataxia, and hyperreflexia were each independently associated with CC. Of the 120 patients with CC, 71 underwent emergent surgery, 20 underwent non-emergent surgery, and 29 were managed medically. CONCLUSIONS: The survey spine protocol was positive for CC in 14% of patients in our cohort and acquired at a 79% faster rate compared to routine total spine. Understanding the positivity rate of CC, the clinical symptoms that are most associated with CC, and the subsequent care management for patients presenting with suspected cord compression who received the survey spine MRI may better inform the broad adoption and subsequent utilization of survey imaging protocols in emergency settings to increase throughput, improve allocation of resources, and provide efficient care for patients with suspected CC.ABBREVIATIONS: CC, cord compression; ED, emergency department; MRI, magnetic resonance imaging; T2; T2-weighted imaging sequence; STIR, short TI inversion recovery.

Proactive Polypharmacy Management Using Large Language Models: Opportunities to Enhance Geriatric Care.

Polypharmacy remains an important challenge for patients with extensive medical complexity. Given the primary care shortage and the increasing aging population, effective polypharmacy management is crucial to manage the increasing burden of care. The capacity of large language model (LLM)-based artificial intelligence to aid in polypharmacy management has yet to be evaluated. Here, we evaluate ChatGPT's performance in polypharmacy management via its deprescribing decisions in standardized clinical vignettes. We inputted several clinical vignettes originally from a study of general practicioners' deprescribing decisions into ChatGPT 3.5, a publicly available LLM, and evaluated its capacity for yes/no binary deprescribing decisions as well as list-based prompts in which the model was prompted to choose which of several medications to deprescribe. We recorded ChatGPT responses to yes/no binary deprescribing prompts and the number and types of medications deprescribed. In yes/no binary deprescribing decisions, ChatGPT universally recommended deprescribing medications regardless of ADL status in patients with no overlying CVD history; in patients with CVD history, ChatGPT's answers varied by technical replicate. Total number of medications deprescribed ranged from 2.67 to 3.67 (out of 7) and did not vary with CVD status, but increased linearly with severity of ADL impairment. Among medication types, ChatGPT preferentially deprescribed pain medications. ChatGPT's deprescribing decisions vary along the axes of ADL status, CVD history, and medication type, indicating some concordance of internal logic between general practitioners and the model. These results indicate that specifically trained LLMs may provide useful clinical support in polypharmacy management for primary care physicians.

Orthopedic Surgery Volume Trends During the COVID-19 Pandemic and Postvaccination Era: Implications for Healthcare Planning.

BACKGROUND: The Coronavirus Disease 2019 (COVID-19) pandemic decreased surgical volumes, but prior studies have not investigated recovery through 2022, or analyzed specific procedures or cases of urgency within orthopedic surgery. The aims of this study were to (1) quantify the declines in orthopedic surgery volume during and after the pandemic peak, (2) characterize surgical volume recovery during the postvaccination period, and (3) characterize recovery in the 1-year postvaccine release period. METHODS: We conducted a retrospective cohort study of 27,476 orthopedic surgeries from January 2019 to December 2022 at one urban academic quaternary referral center. We reported trends over the following periods: baseline pre-COVID-19 period (1/6/2019 to 1/4/2020), COVID-19 peak (3/15/2020 to 5/16/2020), post-COVID-19 peak (5/17/2020 to 1/2/2021), postvaccine release (1/3/2021 to 1/1/2022), and 1-year postvaccine release (1/2/2022 to 12/30/2022). Comparisons were performed with 2 sample t-tests. RESULTS: Pre-COVID-19 surgical volume fell by 72% at the COVID-19 peak, especially impacting elective procedures (P < .001) and both hip and knee joint arthroplasty (P < .001) procedures. Nonurgent (P = .024) and urgent or emergency (P = .002) cases also significantly decreased. Postpeak recovery before the vaccine saw volumes rise to 92% of baseline, which further rose to 96% and 94% in 2021 and 2022, respectively. While elective procedures surpassed the baseline in 2022, nonurgent and urgent or emergency surgeries remained low. CONCLUSIONS: The COVID-19 pandemic substantially reduced orthopedic surgical volumes, which have still not fully recovered through 2022, particularly nonelective procedures. The differential recovery within an orthopedic surgery program may result in increased morbidity and can serve to inform department-level operational recovery.

Trends in non-focal neurological chief complaints and CT angiography utilization among adults in the emergency department.

Prudent imaging use is essential for cost reduction and efficient patient triage. Recent efforts have focused on head and neck CTA in patients with emergent concerns for non-focal neurological complaints, but have failed to demonstrate whether increases in utilization have resulted in better care. The objective of this study was to examine trends in head and neck CTA ordering and determine whether a correlation exists between imaging utilization and positivity rates. This is a single-center retrospective observational study at a quaternary referral center. This study includes patients presenting with headache and/or dizziness to the emergency department between January 2017 and December 2021. Patients who received a head and neck CTA were compared to those who did not. The main outcomes included annual head and neck CTA utilization and positivity rates, defined as the percent of scans with attributable acute pathologies. Among 24,892 emergency department visits, 2264 (9.1%) underwent head and neck CTA imaging. The percentage of patients who received a scan over the study period increased from 7.89% (422/5351) in 2017 to 13.24% (662/5001) in 2021, representing a 67.4% increase from baseline (OR, 1.14; 95% CI 1.11-1.18; P < .001). The positivity rate, or the percentage of scans ordered that revealed attributable acute pathology, dropped from 16.8% (71/422) in 2017 to 10.4% (69/662) in 2021 (OR, 0.86; 95% CI 0.79-0.94; P = .001), a 38% reduction in positive examinations. Throughout the study period, there was a 67.4% increase in head and neck CTA ordering with a concomitant 38.1% decrease in positivity rate.

A Novel Design-Thinking, Hospital Innovation Core Certificate Curriculum for Radiologists and Trainees: Creation, Implementation, and Multiyear Results.

RATIONALE AND OBJECTIVES: Innovation is a crucial skill for physicians and researchers, yet traditional medical education does not provide instruction or experience to cultivate an innovative mindset. This study evaluates the effectiveness of a novel course implemented in an academic radiology department training program over a 5-year period designed to educate future radiologists on the fundamentals of medical innovation. MATERIALS AND METHODS: A pre- and post-course survey and examination were administered to residents who participated in the innovation course (MESH Core) from 2018 to 2022. Respondents were first evaluated on their subjective comfort level, understanding, and beliefs on innovation-related topics using a 5-point Likert-scale survey. Respondents were also administered a 21-question multiple-choice exam to test their objective knowledge of innovation-related topics. RESULTS: Thirty-eight residents participated in the survey (response rate 95%). Resident understanding, comfort and belief regarding innovation-related topics improved significantly (P < .0001) on all nine Likert-scale questions after the course. After the course, a significant majority of residents either agreed or strongly agreed that technological innovation should be a core competency for the residency curriculum, and that a workshop to prototype their ideas would be beneficial. Performance on the course exam showed significant improvement (48% vs 86%, P < .0001). The overall course experience was rated 5 out of 5 by all participants. CONCLUSION: MESH Core demonstrates long-term success in educating future radiologists on the basic concepts of medical technological innovation. Years later, residents used the knowledge and experience gained from MESH Core to successfully pursue their own inventions and innovative projects. This innovation model may serve as an approach for other institutions to implement training in this domain.

Preventing inadvertent drain removal using a novel catheter securement device.

Percutaneous drains have provided a minimally invasive way to treat a wide range of disorders from abscess drainage to enteral feeding solutions to treating hydronephrosis. These drains suffer from a high rate of dislodgement of up to 30% resulting in emergency room visits, repeat hospitalizations, and catheter repositioning/replacement procedures, which incur significant morbidity and mortality. Using ex vivo and in vivo models, a force body diagram was utilized to determine the forces experienced by a drainage catheter during dislodgement events, and the individual components which contribute to drainage catheter securement were empirically collected. Prototypes of a skin level catheter securement and valved quick release system were then developed. The system was inspired by capstans used in boating for increasing friction of a line around a central spool and quick release mechanisms used in electronics such as the Apple MagSafe computer charger. The device was tested in a porcine suprapubic model, which demonstrated the effectiveness of the device to prevent drain dislodgement. The prototype demonstrated that the miniaturized versions of technologies used in boating and electronics industries were able to meet the needs of preventing dislodgement of patient drainage catheters.

Assessing the Utility of ChatGPT Throughout the Entire Clinical Workflow: Development and Usability Study.

BACKGROUND: Large language model (LLM)-based artificial intelligence chatbots direct the power of large training data sets toward successive, related tasks as opposed to single-ask tasks, for which artificial intelligence already achieves impressive performance. The capacity of LLMs to assist in the full scope of iterative clinical reasoning via successive prompting, in effect acting as artificial physicians, has not yet been evaluated. OBJECTIVE: This study aimed to evaluate ChatGPT's capacity for ongoing clinical decision support via its performance on standardized clinical vignettes. METHODS: We inputted all 36 published clinical vignettes from the Merck Sharpe & Dohme (MSD) Clinical Manual into ChatGPT and compared its accuracy on differential diagnoses, diagnostic testing, final diagnosis, and management based on patient age, gender, and case acuity. Accuracy was measured by the proportion of correct responses to the questions posed within the clinical vignettes tested, as calculated by human scorers. We further conducted linear regression to assess the contributing factors toward ChatGPT's performance on clinical tasks. RESULTS: ChatGPT achieved an overall accuracy of 71.7% (95% CI 69.3%-74.1%) across all 36 clinical vignettes. The LLM demonstrated the highest performance in making a final diagnosis with an accuracy of 76.9% (95% CI 67.8%-86.1%) and the lowest performance in generating an initial differential diagnosis with an accuracy of 60.3% (95% CI 54.2%-66.6%). Compared to answering questions about general medical knowledge, ChatGPT demonstrated inferior performance on differential diagnosis (ß=-15.8%; P<.001) and clinical management (ß=-7.4%; P=.02) question types. CONCLUSIONS: ChatGPT achieves impressive accuracy in clinical decision-making, with increasing strength as it gains more clinical information at its disposal. In particular, ChatGPT demonstrates the greatest accuracy in tasks of final diagnosis as compared to initial diagnosis. Limitations include possible model hallucinations and the unclear composition of ChatGPT's training data set.

Justice and Innovation in Radiology.

Imaging is a central determinant of health outcomes, and radiologic disparities can cascade throughout a patient's illness course. Innovative efforts in radiology are constant, but innovation that is driven by short-term profit-making incentives without explicit regard for principles of justice can lead to exclusion of the vulnerable from potential benefits and widening of inequities. Accordingly, we must consider the ways in which the field of radiology can shape innovative efforts to ensure that innovation ameliorates injustice instead of exacerbating it. The authors propose a distinction between approaches to innovation that prioritize justice and those that do not. The authors argue that the field's institutional incentives should be adjusted to prioritize forms of innovation that are likely to ameliorate imaging inequities, and they provide examples of initial steps that can be taken to make these adjustments. The authors propose the term justice-oriented innovation as a way of describing forms of innovation that are motivated by reducing injustice and can reasonably be expected to do so.

Evaluating GPT as an Adjunct for Radiologic Decision Making: GPT-4 Versus GPT-3.5 in a Breast Imaging Pilot.

OBJECTIVE: Despite rising popularity and performance, studies evaluating the use of large language models for clinical decision support are lacking. Here, we evaluate ChatGPT (Generative Pre-trained Transformer)-3.5 and GPT-4's (OpenAI, San Francisco, California) capacity for clinical decision support in radiology via the identification of appropriate imaging services for two important clinical presentations: breast cancer screening and breast pain. METHODS: We compared ChatGPT's responses to the ACR Appropriateness Criteria for breast pain and breast cancer screening. Our prompt formats included an open-ended (OE) and a select all that apply (SATA) format. Scoring criteria evaluated whether proposed imaging modalities were in accordance with ACR guidelines. Three replicate entries were conducted for each prompt, and the average of these was used to determine final scores. RESULTS: Both ChatGPT-3.5 and ChatGPT-4 achieved an average OE score of 1.830 (out of 2) for breast cancer screening prompts. ChatGPT-3.5 achieved a SATA average percentage correct of 88.9%, compared with ChatGPT-4's average percentage correct of 98.4% for breast cancer screening prompts. For breast pain, ChatGPT-3.5 achieved an average OE score of 1.125 (out of 2) and a SATA average percentage correct of 58.3%, as compared with an average OE score of 1.666 (out of 2) and a SATA average percentage correct of 77.7%. DISCUSSION: Our results demonstrate the eventual feasibility of using large language models like ChatGPT for radiologic decision making, with the potential to improve clinical workflow and responsible use of radiology services. More use cases and greater accuracy are necessary to evaluate and implement such tools.

Transformer versus traditional natural language processing: how much data is enough for automated radiology report classification?

OBJECTIVES: Current state-of-the-art natural language processing (NLP) techniques use transformer deep-learning architectures, which depend on large training datasets. We hypothesized that traditional NLP techniques may outperform transformers for smaller radiology report datasets. METHODS: We compared the performance of BioBERT, a deep-learning-based transformer model pre-trained on biomedical text, and three traditional machine-learning models (gradient boosted tree, random forest, and logistic regression) on seven classification tasks given free-text radiology reports. Tasks included detection of appendicitis, diverticulitis, bowel obstruction, and enteritis/colitis on abdomen/pelvis CT reports, ischemic infarct on brain CT/MRI reports, and medial and lateral meniscus tears on knee MRI reports (7,204 total annotated reports). The performance of NLP models on held-out test sets was compared after training using the full training set, and 2.5%, 10%, 25%, 50%, and 75% random subsets of the training data. RESULTS: In all tested classification tasks, BioBERT performed poorly at smaller training sample sizes compared to non-deep-learning NLP models. Specifically, BioBERT required training on approximately 1,000 reports to perform similarly or better than non-deep-learning models. At around 1,250 to 1,500 training samples, the testing performance for all models began to plateau, where additional training data yielded minimal performance gain. CONCLUSIONS: With larger sample sizes, transformer NLP models achieved superior performance in radiology report binary classification tasks. However, with smaller sizes (<1000) and more imbalanced training data, traditional NLP techniques performed better. ADVANCES IN KNOWLEDGE: Our benchmarks can help guide clinical NLP researchers in selecting machine-learning models according to their dataset characteristics.

Assessing the Utility of ChatGPT Throughout the Entire Clinical Workflow.

IMPORTANCE: Large language model (LLM) artificial intelligence (AI) chatbots direct the power of large training datasets towards successive, related tasks, as opposed to single-ask tasks, for which AI already achieves impressive performance. The capacity of LLMs to assist in the full scope of iterative clinical reasoning via successive prompting, in effect acting as virtual physicians, has not yet been evaluated. OBJECTIVE: To evaluate ChatGPT's capacity for ongoing clinical decision support via its performance on standardized clinical vignettes. DESIGN: We inputted all 36 published clinical vignettes from the Merck Sharpe & Dohme (MSD) Clinical Manual into ChatGPT and compared accuracy on differential diagnoses, diagnostic testing, final diagnosis, and management based on patient age, gender, and case acuity. SETTING: ChatGPT, a publicly available LLM. PARTICIPANTS: Clinical vignettes featured hypothetical patients with a variety of age and gender identities, and a range of Emergency Severity Indices (ESIs) based on initial clinical presentation. EXPOSURES: MSD Clinical Manual vignettes. MAIN OUTCOMES AND MEASURES: We measured the proportion of correct responses to the questions posed within the clinical vignettes tested. RESULTS: ChatGPT achieved 71.7% (95% CI, 69.3% to 74.1%) accuracy overall across all 36 clinical vignettes. The LLM demonstrated the highest performance in making a final diagnosis with an accuracy of 76.9% (95% CI, 67.8% to 86.1%), and the lowest performance in generating an initial differential diagnosis with an accuracy of 60.3% (95% CI, 54.2% to 66.6%). Compared to answering questions about general medical knowledge, ChatGPT demonstrated inferior performance on differential diagnosis (ß=-15.8%, p<0.001) and clinical management (ß=-7.4%, p=0.02) type questions. CONCLUSIONS AND RELEVANCE: ChatGPT achieves impressive accuracy in clinical decision making, with particular strengths emerging as it has more clinical information at its disposal.

Assessing the Relationship Between Race, Language, and Surgical Admissions in the Emergency Department.

INTRODUCTION: English proficiency and race are both independently known to affect surgical access and quality, but relatively little is known about the impact of race and limited English proficiency (LEP) on admission for emergency surgery from the emergency department (ED). Our objective was to examine the influence of race and English proficiency on admission for emergency surgery from the ED. METHODS: We conducted a retrospective observational cohort study from January 1-December 31, 2019 at a large, quaternary-care urban, academic medical center with a 66-bed ED Level I trauma and burn center. We included ED patients of all self-reported races reporting a preferred language other than English and requiring an interpreter or declaring English as their preferred language (control group). A multivariable logistic regression was fit to assess the association of LEP status, race, age, gender, method of arrival to the ED, insurance status, and the interaction between LEP status and race with admission for surgery from the ED. RESULTS: A total of 85,899 patients (48.1% female) were included in this analysis, of whom 3,179 (3.7%) were admitted for emergent surgery. Regardless of LEP status, patients identifying as Black (odds ratio [OR] 0.456, 95% CI 0.388-0.533; P<0.005), Asian [OR 0.759, 95% CI 0.612-0.929]; P=0.009), or female [OR 0.926, 95% CI 0.862-0.996]; P=0.04) had significantly lower odds for admission for surgery from the ED compared to White patients. Compared to individuals on Medicare, those with private insurance [OR 1.25, 95% CI 1.13-1.39; P <0.005) were significantly more likely to be admitted for emergent surgery, whereas those without insurance [OR 0.581, 95% CI 0.323-0.958; P=0.05) were significantly less likely to be admitted for emergent surgery. There was no significant difference in odds of admission for surgery between LEP vs non-LEP patients. CONCLUSION: Individuals without health insurance and those identifying as female, Black, or Asian had significantly lower odds of admission for surgery from the ED compared to those with health insurance, males, and those self-identifying as White, respectively. Future studies should assess the reasons underpinning this finding to elucidate impact on patient outcomes.

Trends in oncological imaging during the COVID-19 pandemic through the vaccination era.

BACKGROUND: This study examines the impact that the COVID-19 pandemic has had on computed tomography (CT)-based oncologic imaging utilization. METHODS: We retrospectively analyzed cancer-related CT scans during four time periods: pre-COVID (1/5/20-3/14/20), COVID peak (3/15/20-5/2/20), post-COVID peak (5/3/20-12/19/20), and vaccination period (12/20/20-10/30/21). We analyzed CTs by imaging indication, setting, and hospital type. Using percentage decrease computation and Student's t-test, we calculated the change in mean number of weekly cancer-related CTs for all periods compared to the baseline pre-COVID period. This study was performed at a single academic medical center and three affiliated hospitals. RESULTS: During the COVID peak, mean CTs decreased (-43.0%, p < 0.001), with CTs for (1) cancer screening, (2) initial workup, (3) cancer follow-up, and (4) scheduled surveillance of previously treated cancer dropping by 81.8%, 56.3%, 31.7%, and 45.8%, respectively (p < 0.001). During the post-COVID peak period, cancer screenings and initial workup CTs did not return to prepandemic imaging volumes (-11.4%, p = 0.028; -20.9%, p = 0.024). The ED saw increases in weekly CTs compared to prepandemic levels (+31.9%, p = 0.008), driven by increases in cancer follow-up CTs (+56.3%, p < 0.001). In the vaccination period, cancer screening CTs did not recover to baseline (-13.5%, p = 0.002) and initial cancer workup CTs doubled (+100.0%, p < 0.001). The ED experienced increased cancer-related CTs (+75.9%, p < 0.001), driven by cancer follow-up CTs (+143.2%, p < 0.001) and initial workups (+46.9%, p = 0.007). CONCLUSIONS AND RELEVANCE: The pandemic continues to impact cancer care. We observed significant declines in cancer screening CTs through the end of 2021. Concurrently, we observed a 2× increase in initial cancer workup CTs and a 2.4× increase in cancer follow-up CTs in the ED during the vaccination period, suggesting a boom of new cancers and more cancer examinations associated with emergency level acute care.

Evaluating ChatGPT as an Adjunct for Radiologic Decision-Making.

BACKGROUND: ChatGPT, a popular new large language model (LLM) built by OpenAI, has shown impressive performance in a number of specialized applications. Despite the rising popularity and performance of AI, studies evaluating the use of LLMs for clinical decision support are lacking. PURPOSE: To evaluate ChatGPT's capacity for clinical decision support in radiology via the identification of appropriate imaging services for two important clinical presentations: breast cancer screening and breast pain. MATERIALS AND METHODS: We compared ChatGPT's responses to the American College of Radiology (ACR) Appropriateness Criteria for breast pain and breast cancer screening. Our prompt formats included an open-ended (OE) format, where ChatGPT was asked to provide the single most appropriate imaging procedure, and a select all that apply (SATA) format, where ChatGPT was given a list of imaging modalities to assess. Scoring criteria evaluated whether proposed imaging modalities were in accordance with ACR guidelines. RESULTS: ChatGPT achieved an average OE score of 1.83 (out of 2) and a SATA average percentage correct of 88.9% for breast cancer screening prompts, and an average OE score of 1.125 (out of 2) and a SATA average percentage correct of 58.3% for breast pain prompts. CONCLUSION: Our results demonstrate the feasibility of using ChatGPT for radiologic decision making, with the potential to improve clinical workflow and responsible use of radiology services.