Search

This article discusses a study that explores the use of GPT-4, a generative pre-trained transformer, as a tool for extracting data from unstructured medical records. The study involved drafting fifty fictitious patient records in German and providing GPT-4 with instructions on how to process each one. The accuracy, recall, precision, and F1-Score of GPT-4 were assessed for various variables, and it was found that GPT-4 exhibited excellent performance in both text-mining and classification tasks. The study concludes that reliable data extraction from unstructured medical records is possible with the GPT-4 chatbot. [Extracted from the article]

This article discusses the impact of minimally invasive coronary artery bypass surgery (MICS CABG) as an alternative to traditional surgery. The study analyzes the short-term outcomes of this approach for single- and multi-vessel diseases. The results show that MICS CABG is a safe procedure with low mortality and morbidity rates. The study also found low rates of reoperations and graft-related complications. These findings support the use of MICS CABG as a standard of care in coronary surgery. [Extracted from the article]

Objectives: Off-pump coronary artery bypass grafting may decrease operative morbidity when compared to on-pump bypass grafting; however, it is technically demanding and thus quality control is essential. This study assesses the clinical feasibility of a new, indocyanine green (ICG)-based imaging system (SPY; Novadaq Technologies; Toronto, ON, Canada) to monitor the quality of anastomoses and grafts in off-pump revascularization.Setting: Thirty-eight consecutive patients undergoing nonemergent coronary artery bypass grafting without the use of extracorporeal circulation at two Swiss cardiac surgery clinics were included. On completion of bypass grafts, the quality of the grafts was assessed using the ICG-based imaging system. The imaging device comprises an 806-nm laser light source that is used to cause ICG to fluoresce and a near infrared-sensitive charged couple device videocamera that is used to capture the fluorescence images. ICG was administered through the central venous line, and images were acquired during the first pass of the ICG through the field of view. Graft flow (qualitative) and the quality of the grafts and anastomoses were assessed intraoperatively.Results: Between March 2002 and September 2002, a total of 38 patients (26 men and 12 women; mean +/- SD age, 64.6 +/- 10.5 years; body mass index, 27.1 +/- 2.9) underwent surgery and imaging at two institutions. One hundred seven of 124 grafts (45 arteries and 62 veins) were analyzed. Seventeen grafts could not be assessed due to difficulties in positioning. The imaging system was easy to handle, and no adverse reactions to ICG were observed. Four of the 107 grafts imaged required revision (three anastomotic constrictions and one graft dissection). Each imaging sequence required approximately 1.25 to 2.5 mg of ICG. The images were equivalent to angiography without the need for radiographs and catheter insertion. In addition, the course of coronaries that would otherwise be difficult to locate in obese patients could be detected using the imaging system. Biochemical and ECG data demonstrated an absence of intraoperative or postoperative myocardial damage, and no liver enzyme elevation or renal dysfunction.Conclusions: This study supports the clinical utility of a ICG-based imaging system for the assessment of the quality of bypass grafts, which appears to be safe and simple to use. [ABSTRACT FROM AUTHOR]

Objectives: We aimed to evaluate the performance of multiple large language models (LLMs) in data extraction from unstructured and semi-structured electronic health records., Methods: 50 synthetic medical notes in English, containing a structured and an unstructured part, were drafted and evaluated by domain experts, and subsequently used for LLM-prompting. 18 LLMs were evaluated against a baseline transformer-based model. Performance assessment comprised four entity extraction and five binary classification tasks with a total of 450 predictions for each LLM. LLM-response consistency assessment was performed over three same-prompt iterations., Results: Claude 3.0 Opus, Claude 3.0 Sonnet, Claude 2.0, GPT 4, Claude 2.1, Gemini Advanced, PaLM 2 chat-bison and Llama 3-70b exhibited an excellent overall accuracy >0.98 (0.995, 0.988, 0.988, 0.988, 0.986, 0.982, 0.982, and 0.982, respectively), significantly higher than the baseline RoBERTa model (0.742). Claude 2.0, Claude 2.1, Claude 3.0 Opus, PaLM 2 chat-bison, GPT 4, Claude 3.0 Sonnet and Llama 3-70b showed a marginally higher and Gemini Advanced a marginally lower multiple-run consistency than the baseline model RoBERTa (Krippendorff's alpha value 1, 0.998, 0.996, 0.996, 0.992, 0.991, 0.989, 0.988, and 0.985, respectively)., Discussion: Claude 3.0 Opus, Claude 3.0 Sonnet, Claude 2.0, GPT 4, Claude 2.1, Gemini Advanced, PaLM 2 chat bison and Llama 3-70b performed the best, exhibiting outstanding performance in both entity extraction and binary classification, with highly consistent responses over multiple same-prompt iterations. Their use could leverage data for research and unburden healthcare professionals. Real-data analyses are warranted to confirm their performance in a real-world setting., Conclusion: Claude 3.0 Opus, Claude 3.0 Sonnet, Claude 2.0, GPT 4, Claude 2.1, Gemini Advanced, PaLM 2 chat-bison and Llama 3-70b seem to be able to reliably extract data from unstructured and semi-structured electronic health records. Further analyses using real data are warranted to confirm their performance in a real-world setting., Competing Interests: Competing interests: None declared., (© Author(s) (or their employer(s)) 2025. Re-use permitted under CC BY-NC. No commercial re-use. See rights and permissions. Published by BMJ Group.)