Derk Arts, MD PhD

👉🏼 Appropriateness of ChatGPT as a resource for medication-related questions 🤓 Sara Grossman 👇🏻 https://lnkd.in/epHfdJxZ 🔍 Focus on data insights: - 📊 Only 26% of medication-related questions were answered satisfactorily by ChatGPT. - ❌ The majority of responses (74%) lacked direct answers, accuracy, or completeness. - 📚 References were included in 29% of responses, but all were fabricated. 💡 Main outcomes and implications: - ⚠️ Healthcare professionals and patients should be cautious when using ChatGPT for medication information. - 🔍 The study highlights the limitations of AI in providing reliable medical advice. - 🏥 There is a need for improved AI training to enhance the accuracy of responses in healthcare contexts. 📚 Field significance: - 🌐 This research underscores the importance of verifying AI-generated information against trusted medical sources. - 🧠 It raises awareness about the potential risks of relying on AI for critical health-related inquiries. - 🔗 The findings may influence future guidelines on the use of AI tools in clinical settings. 🗄️: [#ChatGPT #MedicationSafety #AIinHealthcare #DataInsights #Pharmaceuticals #MedicalInformation #HealthcareProfessionals]

Unlocking the potential of advanced large language models in medication review and reconciliation: A proof-of-concept investigation This study explores the application of large language models (LLMs) in medication review and reconciliation, crucial for optimizing drug therapy and reducing medication errors. - Four LLMs were tested on queries about dosing errors, drug interactions, therapeutic monitoring, and genomics-based decisions. - Responses were evaluated for accuracy, relevance, risk management, and adherence to guidelines. - ChatGPT excelled in identifying dosage errors but struggled with simvastatin queries. - All LLMs recognized warfarin interactions but missed some others, such as metoprolol and verapamil. - Claude-Instant provided useful suggestions for therapeutic monitoring, while Gemini offered reliable references without prompting. - Minor impacts were noted in dosing regimens; major impacts were observed in drug interactions and pharmacogenomic decisions. Conclusion: Advanced LLMs show promise in transforming medication review processes, highlighting the need for their integration into Link: https://lnkd.in/efceaWtu

An excellent example by Tufts Medicine is how taking a different approach to healthcare can not only offer increased benefits for the sector but also "...to reinvent the way that it is delivered." Having a stable and secure data processing solution is critical, as identified in this case study, and who better than to partner with the industry's leading cloud provider - Amazon Web Services (AWS). Read the full case study below.

👉🏼 Unlocking the potential of advanced large language models in medication review and reconciliation: A proof-of-concept investigation 🤓 Kannan Sridharan 👇🏻 https://lnkd.in/ecpgck5v 🔍 Focus on data insights: - 📊 The study evaluated four large language models

👉🏼 End-of-life Care Patient Information Leaflets-A Comparative Evaluation of Artificial Intelligence-generated Content for Readability, Sentiment, Accuracy, Completeness, and Suitability: ChatGPT vs Google Gemini 🤓 Prakash G Gondode 👇🏻 https://lnkd.in/evtrDVgG 🔍 Focus on data insights: - 📊 Google Gemini's patient information leaflets (PILs) demonstrated higher readability and actionability compared to those generated by ChatGPT. - 🤖 Both AI models produced content with positive sentiment, indicating a supportive tone in the information provided. - ✅ Accuracy and completeness were rated highly for both models, although Google Gemini had slightly lower accuracy scores. 💡 Main outcomes and implications: - 🌍 The study underscores the potential of AI in improving patient education, particularly in end-of-life care (EOLC), which is crucial for informed decision-making. - 🏥 Enhanced readability and actionability of AI-generated content can lead to better patient understanding and engagement in their care processes. - 🔄 Continuous improvement and innovation in AI-driven educational tools are essential to meet the diverse needs of patients across different cultural contexts. 📚 Field significance: - 💬 The findings contribute to the growing body of evidence supporting the integration of AI in healthcare communication, particularly in sensitive areas like EOLC. - 🧠 This research highlights the importance of culturally sensitive approaches in patient education, ensuring that information is accessible and relevant to diverse populations. - 📈 The study sets a precedent for future research on AI applications in healthcare, emphasizing the need for ongoing evaluation of AI-generated content quality. 🗄️: [#endoflifecare #patienteducation #artificialintelligence #readability #healthcarecommunication #culturalsensitivity #AIinHealthcare]

👉🏼 ChatGPT Responses to Frequently Asked Questions on Ménière's Disease: A Comparison to Clinical Practice Guideline Answers 🤓 Rebecca A Ho 👇🏻 https://lnkd.in/e7Nvpwzw 🔍 Focus on data insights: - ChatGPT-4.0 had significantly better resource reporting compared to ChatGPT-3.5. - Both models showed room for improvement in comprehensiveness, readability, and accuracy. 💡 Main outcomes and implications: - ChatGPT-4.0 demonstrated higher quality resources in responses. - Readability scores for both models fell below the recommended level. - Improvement needed in providing more comprehensive and less misleading information to patients. 📚 Field significance: - Evaluating AI models like ChatGPT against clinical guidelines can help identify areas for enhancement in patient education. - Enhancing AI models' readability and accuracy is crucial for effective patient communication in healthcare settings. 🗄️: [#Ménière'sDisease #AIModelEvaluation #PatientEducation #HealthcareCommunication]

👉🏼 Evaluating generative AI responses to real-world drug-related questions 🤓 Salvatore Giorgi 👇🏻 https://lnkd.in/eev4m63k 🔍 Focus on data insights: - 📊 Generative AI systems like ChatGPT and LLaMA-2 were evaluated for their ability to answer substance use disorder (SUD) questions sourced from online recovery forums. - 🧠 Clinicians rated the AI-generated responses as high quality, indicating potential for AI in clinical settings. - ⚠️ Instances of dangerous disinformation were identified, including incorrect emergency helplines and endorsement of unsafe detox methods. 💡 Main outcomes and implications: - 🔍 The study highlights the dual nature of AI responses, where seemingly accurate information may contain critical inaccuracies. - 🏥 The findings suggest a need for rigorous clinical validation and safeguards before deploying AI in sensitive healthcare contexts. - ⚖️ The research underscores the importance of human oversight in AI applications, particularly in mental health and substance use treatment. 📚 Field significance: - 🌐 This research contributes to the ongoing discourse on integrating AI into healthcare, emphasizing the balance between innovation and safety. - 💬 It raises awareness about the potential risks associated with AI-generated medical advice, advocating for responsible usage in clinical practice. - 🔗 The study serves as a foundation for future investigations into the efficacy and reliability of AI in addressing complex healthcare issues. 🗄️: [#generativeAI #substanceUseDisorder #mentalHealth #clinicalValidation #AIinHealthcare #safetyInAI]

👉🏼 Symptom-BERT: Enhancing Cancer Symptom Detection in EHR Clinical Notes 🤓 Nahid Zeinali 👇🏻 https://lnkd.in/eDWQRGQZ 🔍 Focus on data insights: - The study utilized a pre-trained language model to identify cancer symptoms in clinical notes. - Symptom-BERT demonstrated high efficacy in detecting cancer symptoms with impressive F1-score and AUC metrics. - Physical symptoms like Pruritus were identified with higher performance compared to psychological symptoms such as Anxiety. 💡 Main outcomes and implications: - Specialized pre-training on domain-specific data significantly boosted the performance of language models for medical applications. - The exceptional efficacy of Symptom-BERT in detecting cancer symptoms signifies a groundbreaking advancement in patient-centered AI technologies. - This advancement offers a promising avenue to enhance symptom management and improve patient self-care outcomes. 📚 Field significance: - Advanced language models like Symptom-BERT can revolutionize the detection and management of cancer symptoms in clinical settings. - Leveraging data insights from clinical notes can lead to more accurate and personalized symptom prediction algorithms, enhancing patient care and outcomes. 🗄️: [#cancer #symptoms #EHR #clinicalnotes #AI #patientcare]

Developing large language models to detect adverse drug events in posts on x Adverse drug events (ADEs) significantly contribute to hospital admissions and are linked to higher morbidity and mortality rates. Identifying ADEs post-marketing is crucial for drug safety surveillance, traditionally relying on systems like the FDA Adverse Event Reporting System (FAERS). Social media platforms, particularly posts on X (formerly Twitter), offer valuable patient and medication insights that could enhance drug surveillance efforts. However, extracting ADE information from social media text poses challenges for conventional statistical methods. This study developed several large language models (LLMs) for ADE classification using data from X. Various models were fine-tuned, including BERT-base, Bio_ClinicalBERT, RoBERTa, and RoBERTa-large. The research also explored ChatGPT through few-shot prompting and fine-tuning on the complete training dataset. Model performance was assessed using metrics such as sensitivity, specificity, accuracy, F1-measure, and area under Link: https://lnkd.in/eubRpW2w

👉🏼 Rare disease diagnosis using knowledge guided retrieval augmentation for ChatGPT 🤓 Charlotte Zelin 👇🏻 https://lnkd.in/ecSwW9h4 🔍 Focus on data insights: - 📊 The study demonstrates that the enhanced ChatGPT model, RareDxGPT, achieved a 40% accuracy in rare disease diagnosis with basic prompts. - 📈 With additional context provided through "Prompt + Explanation," RareDxGPT's accuracy improved to 43%, showcasing the importance of contextual information. - 🔍 The comparison between RareDxGPT and ChatGPT 3.5 indicates that domain-specific knowledge significantly enhances diagnostic capabilities. 💡 Main outcomes and implications: - 🧬 The findings suggest that integrating external knowledge resources can improve AI diagnostic tools, potentially reducing the time to accurate diagnosis for rare diseases. - ⚙️ The study highlights the potential of machine learning models like ChatGPT in clinical settings, particularly for conditions that are often misdiagnosed or undiagnosed. - 🌍 This research could pave the way for further developments in AI-assisted diagnostics, ultimately benefiting millions affected by rare diseases. 📚 Field significance: - 💡 The research underscores the growing role of AI in healthcare, particularly in enhancing diagnostic accuracy for complex medical conditions. - 🔗 It opens avenues for future studies to explore the integration of AI with other medical databases and resources to further improve patient outcomes. - 🏥 The implications of this study could lead to more efficient healthcare systems, where AI assists clinicians in making informed decisions based on comprehensive data analysis. 🗄️: [#rareDiseases #AIinHealthcare #ChatGPT #machineLearning #diagnostics #RetrievalAugmentedGeneration #healthTech]

FDA Blog: A Lifecycle Management Approach toward Delivering Safe, Effective AI-enabled Health Care The FDA blog by Troy Tazbaz and John Nicol explores using Lifecycle Management (LCM) to address challenges in AI-enabled health care. They introduce the AI Lifecycle (AILC) concept, mapping traditional Software Development Lifecycle phases to AI development, covering data management, model tuning, and real-world performance evaluation. This structured approach aims to enhance AI reliability, safety, and fairness. The authors urge the health care community to engage with and refine these concepts to promote safer AI practices. Read more here: https://lnkd.in/gKdDYJJ6 #AI #Healthcare #FDA #ArtificialIntelligence #AIHealthcare #LifecycleManagement #DigitalHealth #AIEthics #MedicalDevices #HealthTech