Black Box Warning: Large Language Models and the Future of Infectious Diseases Consultation.

Large language models (LLMs) are artificial intelligence systems trained by deep learning algorithms to process natural language and generate text responses to user prompts. Some approach physician performance on a range of medical challenges, leading some proponents to advocate for their potential use in clinical consultation and prompting some consternation about the future of cognitive specialties. However, LLMs currently have limitations that preclude safe clinical deployment in performing specialist consultations, including frequent confabulations, lack of contextual awareness crucial for nuanced diagnostic and treatment plans, inscrutable and unexplainable training data and methods, and propensity to recapitulate biases. Nonetheless, considering the rapid improvement in this technology, growing calls for clinical integration, and healthcare systems that chronically undervalue cognitive specialties, it is critical that infectious diseases clinicians engage with LLMs to enable informed advocacy for how they should-and shouldn't-be used to augment specialist care.

The Brain Tumor Segmentation (BraTS-METS) Challenge 2023: Brain Metastasis Segmentation on Pre-treatment MRI.

Clinical monitoring of metastatic disease to the brain can be a laborious and timeconsuming process, especially in cases involving multiple metastases when the assessment is performed manually. The Response Assessment in Neuro-Oncology Brain Metastases (RANO-BM) guideline, which utilizes the unidimensional longest diameter, is commonly used in clinical and research settings to evaluate response to therapy in patients with brain metastases. However, accurate volumetric assessment of the lesion and surrounding peri-lesional edema holds significant importance in clinical decision-making and can greatly enhance outcome prediction. The unique challenge in performing segmentations of brain metastases lies in their common occurrence as small lesions. Detection and segmentation of lesions that are smaller than 10 mm in size has not demonstrated high accuracy in prior publications. The brain metastases challenge sets itself apart from previously conducted MICCAI challenges on glioma segmentation due to the significant variability in lesion size. Unlike gliomas, which tend to be larger on presentation scans, brain metastases exhibit a wide range of sizes and tend to include small lesions. We hope that the BraTS-METS dataset and challenge will advance the field of automated brain metastasis detection and segmentation.

Causes of Death in Patients With Brain Metastases.

BACKGROUND AND OBJECTIVES: Advances in targeted therapies and wider application of stereotactic radiosurgery (SRS) have redefined outcomes of patients with brain metastases. Under modern treatment paradigms, there remains limited characterization of which aspects of disease drive demise and in what frequencies. This study aims to characterize the primary causes of terminal decline and evaluate differences in underlying intracranial tumor dynamics in patients with metastatic brain cancer. These fundamental details may help guide management, patient counseling, and research priorities. METHODS: Using NYUMets-Brain-the largest, longitudinal, real-world, open data set of patients with brain metastases-patients treated at New York University Langone Health between 2012 and 2021 with SRS were evaluated. A review of electronic health records allowed for the determination of a primary cause of death in patients who died during the study period. Causes were classified in mutually exclusive, but collectively exhaustive, categories. Multilevel models evaluated for differences in dynamics of intracranial tumors, including changes in volume and number. RESULTS: Of 439 patients with end-of-life data, 73.1% died secondary to systemic disease, 10.3% died secondary to central nervous system (CNS) disease, and 16.6% died because of other causes. CNS deaths were driven by acute increases in intracranial pressure (11%), development of focal neurological deficits (18%), treatment-resistant seizures (11%), and global decline driven by increased intracranial tumor burden (60%). Rate of influx of new intracranial tumors was almost twice as high in patients who died compared with those who survived ( P < .001), but there was no difference in rates of volume change per intracranial tumor ( P = .95). CONCLUSION: Most patients with brain metastases die secondary to systemic disease progression. For patients who die because of neurological disease, tumor dynamics and cause of death mechanisms indicate that the primary driver of decline for many may be unchecked systemic disease with unrelenting spread of new tumors to the CNS rather than failure of local growth control.

Impact of Policy Change on Access to Medication for Opioid Use Disorder in Primary Care.

OBJECTIVES: The opioid overdose epidemic is escalating. Increasing access to medications for opioid use disorder in primary care is crucial. The impact of the US Department of Health and Human Services' policy change removing the buprenorphine waiver training requirement on primary care buprenorphine prescribing remains unclear. We aimed to investigate the impact of the policy change on primary care providers' likelihood of applying for a waiver and the current attitudes, practices, and barriers to buprenorphine prescribing in primary care. METHODS: We used a cross-sectional survey with embedded educational resources disseminated to primary care providers in a southern US academic health system. We used descriptive statistics to aggregate survey data, logistic regression models to evaluate whether buprenorphine interest and familiarity correlate with clinical characteristics, and a χ2 test to evaluate the effect of the educational intervention on screening. RESULTS: Of the 54 respondents, 70.4% reported seeing patients with opioid use disorder, but only 11.1% had a waiver to prescribe buprenorphine. Few nonwaivered providers were interested in prescribing, but perceiving buprenorphine to be beneficial to the patient population was associated with interest (adjusted odds ratio 34.7, P < 0.001). Two-thirds of nonwaivered respondents reported the policy change having no impact on their decision to obtain a waiver; however, among interested providers, it increased their likelihood of obtaining a waiver. Barriers to buprenorphine prescribing included lack of clinical experience, clinical capacity, and referral resources. Screening for opioid use disorder did not increase significantly after the survey. CONCLUSIONS: Although most primary care providers reported seeing patients with opioid use disorder, interest in prescribing buprenorphine was low and structural barriers remained the dominant obstacles. Providers with a preexisting interest in buprenorphine prescribing reported that removing the training requirement was helpful.

RadImageNet: An Open Radiologic Deep Learning Research Dataset for Effective Transfer Learning.

Purpose: To demonstrate the value of pretraining with millions of radiologic images compared with ImageNet photographic images on downstream medical applications when using transfer learning. Materials and Methods: This retrospective study included patients who underwent a radiologic study between 2005 and 2020 at an outpatient imaging facility. Key images and associated labels from the studies were retrospectively extracted from the original study interpretation. These images were used for RadImageNet model training with random weight initiation. The RadImageNet models were compared with ImageNet models using the area under the receiver operating characteristic curve (AUC) for eight classification tasks and using Dice scores for two segmentation problems. Results: The RadImageNet database consists of 1.35 million annotated medical images in 131 872 patients who underwent CT, MRI, and US for musculoskeletal, neurologic, oncologic, gastrointestinal, endocrine, abdominal, and pulmonary pathologic conditions. For transfer learning tasks on small datasets-thyroid nodules (US), breast masses (US), anterior cruciate ligament injuries (MRI), and meniscal tears (MRI)-the RadImageNet models demonstrated a significant advantage (P < .001) to ImageNet models (9.4%, 4.0%, 4.8%, and 4.5% AUC improvements, respectively). For larger datasets-pneumonia (chest radiography), COVID-19 (CT), SARS-CoV-2 (CT), and intracranial hemorrhage (CT)-the RadImageNet models also illustrated improved AUC (P < .001) by 1.9%, 6.1%, 1.7%, and 0.9%, respectively. Additionally, lesion localizations of the RadImageNet models were improved by 64.6% and 16.4% on thyroid and breast US datasets, respectively. Conclusion: RadImageNet pretrained models demonstrated better interpretability compared with ImageNet models, especially for smaller radiologic datasets.Keywords: CT, MR Imaging, US, Head/Neck, Thorax, Brain/Brain Stem, Evidence-based Medicine, Computer Applications-General (Informatics) Supplemental material is available for this article. Published under a CC BY 4.0 license.See also the commentary by Cadrin-Chênevert in this issue.

Population scale latent space cohort matching for the improved use and exploration of observational trial data.

A significant amount of clinical research is observational by nature and derived from medical records, clinical trials, and large-scale registries. While there is no substitute for randomized, controlled experimentation, such experiments or trials are often costly, time consuming, and even ethically or practically impossible to execute. Combining classical regression and structural equation modeling with matching techniques can leverage the value of observational data. Nevertheless, identifying variables of greatest interest in high-dimensional data is frequently challenging, even with application of classical dimensionality reduction and/or propensity scoring techniques. Here, we demonstrate that projecting high-dimensional medical data onto a lower-dimensional manifold using deep autoencoders and post-hoc generation of treatment/control cohorts based on proximity in the lower-dimensional space results in better matching of confounding variables compared to classical propensity score matching (PSM) in the original high-dimensional space (P<0.0001) and performs similarly to PSM models constructed by experts with prior knowledge of the underlying pathology when evaluated on predicting risk ratios from real-world clinical data. Thus, in cases when the underlying problem is poorly understood and the data is high-dimensional in nature, matching in the autoencoder latent space might be of particular benefit.

Remote Patient Monitoring Identifies the Need for Triage in Patients with Acute COVID-19 Infection.

Introduction: Telehealth was frequently used in the provision of care and remote patient monitoring (RPM) during the COVID-19 pandemic. The Precision Recovery Program (PRP) remotely monitored and supported patients with COVID-19 in their home environment. Materials and Methods: This was a single-center retrospective cohort study reviewing data acquired from the PRP clinical initiative. Results: Of the 679 patients enrolled in the PRP, 156 patients were screened by a clinician following a deterioration in symptoms and vital signs on a total of 240 occasions, and included in the analyses. Of these 240 occasions, 162 (67%) were escalated to the PRP physician. Thirty-six patients were referred to emergency department, with 12 (7%) admitted to the hospital. The most common risk factors coinciding with hospital admissions were cardiac (67%), age >65 (42%), obesity (25%), and pulmonary (17%). The most common symptoms reported that triggered a screening event were dyspnea/tachypnea (27%), chest pain (14%), and gastrointestinal issues (8%). Vital signs that commonly triggered a screening event were pulse oximetry (15%), heart rate (11%), and temperature (9%). Discussion: Common factors (risk factors, vital signs, and symptoms) among patients requiring screening, triage, and hospitalization were identified, providing clinicians with further information to support decision making when utilizing RPM in this cohort. Conclusion: A clinician-led RPM program for patients with acute COVID-19 infection provided supportive care and screening for deterioration. Similar models should be considered for implementation in COVID-19 cohorts and other conditions at risk of rapid clinical deterioration in the home setting.

Integrated Morphoelectric and Transcriptomic Classification of Cortical GABAergic Cells.

Neurons are frequently classified into distinct types on the basis of structural, physiological, or genetic attributes. To better constrain the definition of neuronal cell types, we characterized the transcriptomes and intrinsic physiological properties of over 4,200 mouse visual cortical GABAergic interneurons and reconstructed the local morphologies of 517 of those neurons. We find that most transcriptomic types (t-types) occupy specific laminar positions within visual cortex, and, for most types, the cells mapping to a t-type exhibit consistent electrophysiological and morphological properties. These properties display both discrete and continuous variation among t-types. Through multimodal integrated analysis, we define 28 met-types that have congruent morphological, electrophysiological, and transcriptomic properties and robust mutual predictability. We identify layer-specific axon innervation pattern as a defining feature distinguishing different met-types. These met-types represent a unified definition of cortical GABAergic interneuron types, providing a systematic framework to capture existing knowledge and bridge future analyses across different modalities.

The Neural Basis of Approach-Avoidance Conflict: A Model Based Analysis.

Approach-avoidance conflict arises when the drives to pursue reward and avoid harm are incompatible. Previous neuroimaging studies of approach-avoidance conflict have shown large variability in reported neuroanatomical correlates. These prior studies have generally neglected to account for potential sources of variability, such as individual differences in choice preferences and modeling of hemodynamic response during conflict. In the present study, we controlled for these limitations using a hierarchical Bayesian model (HBM). This enabled us to measure participant-specific per-trial estimates of conflict during an approach-avoidance task. We also employed a variable epoch method to identify brain structures specifically sensitive to conflict. In a sample of 28 human participants, we found that only a limited set of brain structures [inferior frontal gyrus (IFG), right dorsolateral prefrontal cortex (dlPFC), and right pre-supplementary motor area (pre-SMA)] are specifically correlated with approach-avoidance conflict. These findings suggest that controlling for previous sources of variability increases the specificity of the neuroanatomical correlates of approach-avoidance conflict.

The effect of surgical site infection on older operative patients.

OBJECTIVES: To determine the effect of surgical site infection (SSI) on mortality, duration of hospitalization, and hospital cost in older operative patients. DESIGN: Retrospective matched-outcomes study. SETTING: Eight hospitals, including Duke University Medical Center, and seven community hospitals. PARTICIPANTS: Patients aged 65 and older undergoing surgery from 1991 to 2003. Cases were defined as patients who developed deep incisional or organ or space SSI; controls were operative patients who did not develop SSI. Controls were frequency matched to cases according to type and year of operative procedure and to hospital in a 1:1 ratio. MEASUREMENTS: Mortality, duration of hospitalization (including re-admissions), and hospital charges for the 90 days after surgery. RESULTS: One thousand three hundred thirty-seven patients were enrolled in the study: 561 cases with SSI and 576 controls without SSI. In cases, the most common SSI pathogen was Staphylococcus aureus (n=275, 51.6%). Of S. aureus isolates, 58.2% were methicillin resistant. One hundred sixteen subjects died within 90 days of surgery (8.6%). In multivariable analysis, SSI was associated with greater mortality risk (odds ratio (OR)=3.51, 95% confidence interval (CI)=2.20-5.59), 2.9 times longer postoperative hospitalization (95% CI=2.61-3.13), and 1.9 times greater hospital charges (95% CI=1.78-2.10). CONCLUSION: In elderly operative patients, SSI was associated with almost 4 times greater mortality, a mean attributable duration of hospitalization after surgery of 15.7 days (95% CI=13.9-17.6) and mean attributable hospital charges of $43,970 (95% CI=$31,881-56,060).

Poor functional status as a risk factor for surgical site infection due to methicillin-resistant Staphylococcus aureus.

OBJECTIVE: To identify risk factors for surgical site infection (SSI) due to methicillin-resistant Staphylococcus aureus (MRSA). DESIGN: Prospective case-control study. SETTING: One tertiary and 6 community-based institutions in the southeastern United States. METHODS: We compared patients with SSI due to MRSA with 2 control groups: matched uninfected surgical patients and patients with SSI due to methicillin-susceptible S. aureus (MSSA). Multivariable logistic regression was used to determine variables independently associated with SSI due to MRSA, compared with each control group. RESULTS: During the 5-year study period, 150 case patients with SSI due to MRSA were identified and compared with 231 matched uninfected control patients and 128 control patients with SSI due to MSSA. Two variables were independently associated with SSI due to MRSA in both multivariable regression models: need for assistance with 3 or more activities of daily living (odds ratio [OR] compared with uninfected patients, 3.97 [95% confidence interval {CI}, 2.18-7.25]; OR compared with patients with SSI due to MSSA, 3.88 [95% CI, 1.91-7.87]) and prolonged duration of surgery (OR compared with uninfected patients, 1.98 [95% CI, 1.11-3.55]; OR compared with patients with SSI due to MSSA, 2.33 [95% CI, 1.17-4.62]). Lack of independence (ie, poor functional status) remained associated with an increased risk of SSI due to MRSA after stratifying by age. CONCLUSIONS: Poor functional status was highly associated with SSI due to MRSA in adult surgical patients, regardless of age. A patient's level of independence can be easily determined, and this information can be used preoperatively to target preventive interventions.

The deadly toll of invasive methicillin-resistant Staphylococcus aureus infection in community hospitals.

INTRODUCTION: Data regarding the epidemiology, treatment, and outcomes of methicillin-resistant Staphylococcus aureus (MRSA) infections in rural and community hospitals are limited. METHODS: This cohort study was conducted at 1 tertiary care hospital and 8 community hospitals in the southeastern United States. Patients with a surgical site infection (SSI) and/or bacteremia due to MRSA were prospectively identified at each study hospital during the period 1994-2003. RESULTS: A total of 129 patients with SSI and 564 patients with bacteremia due to MRSA were identified. Only 57 patients with SSI (44.2%) received antibiotics active against MRSA on the initial date of diagnosis; only 95 (73.6%) received an agent active against MRSA by day 7 after diagnosis of SSI due to MRSA. Ninety-five patients with SSI due to MRSA (73.6%) were readmitted to the hospital within 90 days after their original surgery. The 1-year mortality rate among patients with SSI due to MRSA was 22%. Inadequate therapy was also commonly given to patients with bacteremia: only 216 (38.3%) received antibiotics active against MRSA on the initial day of infection, and only 383 (67.9%) received an agent active against MRSA by day 7 after diagnosis. Approximately one-third of patients with bloodstream infection died during their initial hospitalization. Patients hospitalized in community hospitals were less likely to receive effective antimicrobial therapy on both the day of infection and within 7 days after infection, compared with patients in the tertiary care hospital. CONCLUSION: Inadequate therapy is commonly administered after diagnosis of SSI and bacteremia due to MRSA in patients in community hospitals.

Risk of human immunodeficiency virus infection in herpes simplex virus type 2-seropositive persons: a meta-analysis.

To determine the contribution of herpes simplex type 2 (HSV-2) infection to the risk of human immunodeficiency virus (HIV) acquisition, a systematic review of literature and data synthesis were done. Thirty-one studies addressed the risk of HIV infection in HSV-2-seropositive persons. For 9 cohort and nested case-control studies that documented HSV-2 infection before HIV acquisition, the risk estimate was 2.1 (95% confidence interval, 1.4-3.2). Thus, the attributable risk percentage of HIV to HSV-2 was 52%, and the population attributable risk percentage was 19% in populations with 22% HSV-2 prevalence but increased to 47% in populations with 80% HSV-2 prevalence. For 22 case-control and cross-sectional studies, the risk estimate was 3.9 (95% confidence interval, 3.1-5.1), but the temporal sequence of the 2 infections cannot be documented. Control strategies for HSV-2 need to be incorporated into control of sexually transmitted infections as a strategy for HIV prevention.