Almanac - Retrieval-Augmented Language Models for Clinical Medicine.

BACKGROUND: Large language models (LLMs) have recently shown impressive zero-shot capabilities, whereby they can use auxiliary data, without the availability of task-specific training examples, to complete a variety of natural language tasks, such as summarization, dialogue generation, and question answering. However, despite many promising applications of LLMs in clinical medicine, adoption of these models has been limited by their tendency to generate incorrect and sometimes even harmful statements. METHODS: We tasked a panel of eight board-certified clinicians and two health care practitioners with evaluating Almanac, an LLM framework augmented with retrieval capabilities from curated medical resources for medical guideline and treatment recommendations. The panel compared responses from Almanac and standard LLMs (ChatGPT-4, Bing, and Bard) versus a novel data set of 314 clinical questions spanning nine medical specialties. RESULTS: Almanac showed a significant improvement in performance compared with the standard LLMs across axes of factuality, completeness, user preference, and adversarial safety. CONCLUSIONS: Our results show the potential for LLMs with access to domain-specific corpora to be effective in clinical decision-making. The findings also underscore the importance of carefully testing LLMs before deployment to mitigate their shortcomings. (Funded by the National Institutes of Health, National Heart, Lung, and Blood Institute.).

Patient-Reported Barriers and Enablers to Deprescribing Recommendations During a Clinical Trial (Shed-MEDS).

BACKGROUND AND OBJECTIVES: Effective deprescribing requires shared decision making between a patient and their clinician, and should be used when implementing evidence-based deprescribing conversations. As part of the Shed-MEDS clinical trial, this study assessed barriers and enablers that influence patient decision making in deprescribing to inform future implementation efforts and adaptations. RESEARCH DESIGN AND METHODS: Shed-MEDS, a randomized controlled deprescribing trial, included hospitalized older adults discharging to post-acute care facilities. A trained clinician reviewed each participant's medical history and medication list to identify medications with potential for deprescribing. The study clinician then conducted a semistructured patient-centered deprescribing interview to determine patient (or surrogate) concerns about medications and willingness to deprescribe. Reeve et al.'s (2013) framework was used to categorize barriers and enablers to deprescribing from the patient's perspective, including "appropriateness of cessation," "fear," "dislike of a medication," "influences," and "process of cessation." RESULTS: Overall, participants/surrogates (N = 177) agreed with 63% (883 total medications) of the study clinician's deprescribing recommendations. Thematic analysis revealed that "appropriateness" of a medication was the most common barrier (88.2%) and enabler (67.3%) to deprescribing. Other deprescribing enablers were in the following domains: "influences" (22.7%), "process" (22.5%), "pragmatic" (19.4%), and "dislike" (5.3%). DISCUSSION AND IMPLICATIONS: Use of a semistructured deprescribing interview conversation tool allowed study clinicians to elicit individual barriers and enablers to deprescribing from the patient's perspective. Participants in this study expressed more agreement than disagreement with study clinicians' deprescribing recommendations. These results should inform future implementation efforts that incorporate a patient-centered framework during deprescribing conversations. CLINICAL TRIALS REGISTRATION NUMBER: NCT02979353.

Almanac: Retrieval-Augmented Language Models for Clinical Medicine.

Large-language models have recently demonstrated impressive zero-shot capabilities in a variety of natural language tasks such as summarization, dialogue generation, and question-answering. Despite many promising applications in clinical medicine, adoption of these models in real-world settings has been largely limited by their tendency to generate incorrect and sometimes even toxic statements. In this study, we develop Almanac, a large language model framework augmented with retrieval capabilities for medical guideline and treatment recommendations. Performance on a novel dataset of clinical scenarios (n= 130) evaluated by a panel of 5 board-certified and resident physicians demonstrates significant increases in factuality (mean of 18% at p-value < 0.05) across all specialties, with improvements in completeness and safety. Our results demonstrate the potential for large language models to be effective tools in the clinical decision-making process, while also emphasizing the importance of careful testing and deployment to mitigate their shortcomings.

Presentation and minimally invasive treatment of a type IV giant paraesophageal hernia.

Paraesophageal hernias (PEHs) result from a defect of the diaphragmatic hiatus with a gradual enlargement of the hiatal opening, allowing abdominal contents to shift into the mediastinum. PEHs are often confused with sliding hiatal hernias; however, it is paramount that physicians understand the subtle presentation differences in the types of diaphragmatic incompetence, as treatment may vary greatly. The type IV giant PEH is a dangerous variant that, once recognized, usually requires surgical intervention. In recent years, the laparoscopic approach has been associated with decreases in morbidity, hospital stay, and time off work, as well as increased quality of life. This case involves the proper workup and minimally invasive treatment of a 56-year-old black female who presented with a symptomatic giant PEH with colon, stomach, and duodenum displacement into the chest.

Microencapsulated cells genetically modified to overexpress human transforming growth factor-beta1: viability and functionality in allogeneic and xenogeneic implant models.

This study explores the suitability of using encapsulated genetically modified fibroblasts for orthopedic tissue engineering by examining cell survival and persistence of human transforming growth factor-beta (hTGF-beta) overexpression in xenogeneic and allogeneic implant models. Human wild-type fibroblasts, modified to produce a latent form of hTGF-beta, and murine mutant-type fibroblasts, engineered to release a constitutively active form of hTGF-beta, were encapsulated separately in Ca2+ -alginate microcapsules. Following a percentage viability assessment by MTT (3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyltetrazolium bromide) test, microcapsules were implanted into either the subcutaneous or intraperitoneal cavities of mice. Explanted encapsulated cells were characterized for percentage viability and subjected to a release study and a viability test 1 week and 3 weeks following implantation, a time frame consistent with the requirement for orthopedic tissue engineering application of this growth factor. On average, percentage viabilities of encapsulated cells were 64%at implantation, 52% at explantation, and 56%after 1 week following either 1- or 3-week explantation. hTGF-beta release declined following in vivo implantation, more so for xenogeneic than allogeneic models, but remained in the clinically attractive range of 2 to 30 ng/(10(6) implanted cells 24 h). This technical platform for hTGF-beta is very encouraging for cartilage regeneration using orthopedic tissue engineering, and further evaluation is warranted.

In vitro characterization of TGF-beta1 release from genetically modified fibroblasts in Ca(2+)-alginate microcapsules.

This study was undertaken to develop an in situ source of transforming growth factor-beta1 (TGF-beta1), one of several molecules potentially useful for a tissue-engineered bioartificial cartilage. Primary human fibroblasts and murine NIH 3T3 cells were genetically modified via viral transfection to express human TGF-beta1. Two viral constructs were used, one expressing a gene encoding for the latent and the other for the constitutively active form of the growth factor. Unmodified cells served as controls. Four genetically modified cohorts and two controls were separately encapsulated in a 1.8% alginate solution using a vibrating nozzle and 0.15M calcium chloride crosslinking bath. Diameter of the spherical capsules was 410 +/- 87 microm. In vitro release rate measured over 168 hours varied with cell types and ranged from 2-17 pg/(milligram of capsules x 24 h) or 2-17 ng/(10(6) cells x 24 h). None of the formulations exhibited a large initial bolus release. Even when serum-supplemented medium was not replenished, cell viabilities remained over 55% after 1 week for all cell types. Microencapsulated genetically modified cells were capable of a constitutive synthesis and delivery of biologically significant quantity of TGF-beta1 for at least 168 hours and thus are of potential utility for artificial cartilage and other orthopedic tissue engineering applications.

Optic nerve lymphoma: report of two cases and review of the literature.

Lymphoma may involve the optic nerve as isolated optic nerve lymphoma or in association with central nervous system (CNS) or systemic lymphoma. We present two biopsy-proven non-Hodgkin lymphomas of the optic nerve and compare our findings with previously reported cases. We discuss the mechanism of metastasis, classification of optic nerve involvement, clinical features, radiologic findings, optic nerve biopsy indications and techniques, histologic features, and treatments. We propose a classification system of optic nerve lymphoma: isolated optic nerve involvement, optic nerve involvement with CNS disease, optic nerve involvement with systemic disease, and optic nerve involvement with primary intraocular lymphoma. Although it is an uncommon cause of infiltrative optic neuropathy, optic nerve metastasis should be considered in patients with a history of lymphoma. The recommended approach to a patient with presumed optic nerve lymphoma includes neuroimaging and cerebrospinal fluid evaluation as part of the initial workup, then judicious use of optic nerve biopsy, depending on the clinical situation.