🎉 Pleased to share our paper published in Nature Portfolio digital medicine. 🥳 We’ve developed a comprehensive framework called CREOLA (short for Clinical Review Of Large Language Models (LLMs) and AI). This framework is pioneered at TORTUS, taking a safety-first, science approach to LLMs in healthcare. 🔹 Key Components of the CREOLA Framework -Error Taxonomy -Clinical Safety Assessment -Iterative Experimental Structure 🔹 Error Taxonomy Hallucinations: instances of text in clinical documents unsupported by the transcript of the clinical encounter Omissions: Clinically important text in the encounter that was not included in the clinical documentation 🔹 Clinical Safety Assessment: Our innovation incorporates accepted clinical hazard identification principles (based on NHS DCB0129 standards) to evaluate the potential harm of errors: We categorise errors as either ‘major’ or ‘minor’, where major errors can have downstream impact on the diagnosis or the management of the patient if left uncorrected. This is further assessed as a risk matrix comprising of: Risk severity (1 (minor) to 5 (catastrophic)) compared with Likelihood assessment (very low to very high) 🔹 Iterative Experimental Structure We share a methodical approach to compare different prompts, models, and workflows. Label errors, consolidate review, evaluate clinical safety (and then make further adjustments and re-evaluate if necessary). ----------Method-------------- To demonstrate how to apply CREOLA to any LLM / AVT, we used GPT-4 (early 2024) as a case study here. 🔹 We conduct one of the largest manual evaluations of LLM-generated clinical notes to date, analyzing 49,590 transcript sentences and 12,999 clinical note sentences across 18 experimental configurations. 🔹 Transcripts-clinical note pairs are broken down to a sentence level and annotated for errors by clinicians. ----------Results-------------- 🔹 Of 12,999 sentences in 450 clinical notes, 191 sentences had hallucinations (1.47%), of which 84 sentences (44%) were major. Of the 49,590 sentences from our consultation transcripts, 1712 sentences were omitted (3.45%), of which 286 (16.7%) of which were classified as major and 1426 (83.3%) as minor. 🔹 Hallucination types Fabrication (43%) - completely invented information Negation (30%) - contradicting clinical facts Contextual (17%) - mixing unrelated topics Causality (10%) - speculating on causes without evidence 🔹 Hallucinations, while less common than omissions, carry significantly more clinical risk. Negation hallucinations were the most concerning 🔹 we CAN reduce or even abolish hallucinations and omissions by making prompt or model changes. In one experiment with GPT4 - We reduced incidence of major hallucinations by 75%, major omissions by 58%, and minor omissions by 35% through prompt iteration Links in comments Ellie Asgari Nina Montaña Brown Magda Dubois Saleh Khalil Jasmine Balloch Dr Dom Pimenta M.D.
Health And Safety Protocols
Explore top LinkedIn content from expert professionals.
-
-
Increasing Safety Awareness with Proactive, Data-Driven Dashboards What if you could visualize risks/hazards by analyzing historical data making comprehensive Risk Reviews, Where accidents/Incidents and Near Misses Happen also which body parts are most affected? That’s the power of Safety Dashboard - a proactive approach to identifying risks, visualizing them and driving targeted interventions. What is a Safety Dashboard A Safety Dashboard is a centralized platform that displays essential safety metrics and indicators, providing a comprehensive overview of an organization's safety performance. It enables stakeholders to monitor, track, and analyze safety data to identify areas for improvement and make informed decisions. Recomended Key Metrics to Follow 1️⃣ Total Recordable Incident Rate (TRIR): Measures all work-related injuries requiring medical treatment beyond first aid. Formula: (Total Recordable Incidents) / (Total Hours Worked) x 200,00015. 2️⃣ Lost Time Injury Frequency Rate (LTIFR): Focuses on injuries resulting in lost work time. Formula: (Number of Lost Time Injuries) / (Total Hours Worked) x 1,000,00015. 3️⃣ Risk Priority Number (RPN): Numerical value calculated by multiplying the severity, occurrence, and detection ratings of a potential hazard to prioritize risks and guide mitigation efforts. 4️⃣ Near Miss Reporting Rate: Tracks potential hazards that could lead to future incidents. Formula: Number of Near Misses 5️⃣ Employee Safety Training Completion Rate: Ensures employees have completed mandatory training. Formula : Number of Safety Trainings Given / Target 6️⃣ Safety Compliance Rate: Measures adherence to safety regulations and best practices. Formula: Compliant Items / All Legal Items Applicable 7️⃣ First Aid Case Rate: Measures minor injuries requiring first aid treatment. Formula : First Aid Cases / All Cases 8️⃣ Mostly Injured Body Parts: Identifies which body parts are most frequently injured to inform targeted safety interventions. 9️⃣ Average Time to Incident Resolution: Tracks the time taken to resolve safety issues. 🔟 Employee Safety Perception Survey Scores: Reveals how employees perceive the organization’s safety culture. Why These Metrics Matter ✅ Proactive Risk Management: Identify and mitigate risks before they become incidents. ✅ Improved Compliance: Ensure adherence to safety regulations and standards. ✅ Enhanced Decision Making: Use real-time data to inform safety strategies and resource allocation. ✅ Culture of Safety: Foster a workplace culture that prioritizes employee well-being and safety. A Safety Dashboard can significantly enhance safety culture and operational excellence by providing real-time visibility into key safety metrics, enabling proactive risk management, and fostering a culture of accountability and transparency, ultimately driving continuous improvement and a safer working environment. How are you increasing safety awareness in your companies ?
-
What is MSDS? MSDS stands for Material Safety Data Sheet. It's a comprehensive document that provides key safety and hazard information about a chemical or substance. The MSDS is designed to help workers, emergency responders, and users understand how to handle, store, and dispose of chemicals safely, as well as what steps to take in emergencies. Here are the essential components of an MSDS: Chemical Identification: Provides the product name, common synonyms, and manufacturer details. Hazard Identification: Outlines potential risks related to health (toxicity, irritation), fire (flammability), reactivity, and environmental impact. Composition/Information on Ingredients: Lists the chemical components, their concentrations, and any expected impurities. First-Aid Measures: Details immediate actions to take if exposure occurs, helping to reduce the impact of an incident. Fire-Fighting Measures: Recommends appropriate extinguishing methods and any special precautions for fighting a fire involving the chemical. Accidental Release Measures: Provides instructions on the safe cleanup and containment of spills or leaks. Handling and Storage: Suggests safe practices for using, storing, and transporting the substance. Exposure Controls/Personal Protection: Specifies necessary personal protective equipment (PPE) and engineering controls to minimize exposure risks. Physical and Chemical Properties: Describes key characteristics such as appearance, odor, pH, boiling/melting points, etc. Stability and Reactivity: Discusses the chemical's stability under various conditions and potential reactions when exposed to incompatible substances. Toxicological Information: Summarizes how the chemical may affect health if inhaled, ingested, or contacted. The MSDS (or SDS) is a critical resource in ensuring workplace safety, guiding proper chemical management, and supporting effective emergency response. Would you like to explore more about the transition to SDS or how to interpret specific sections for your work environment?
-
🚨The thoroughness of an incident investigation should be proportionate to the severity and potential severity of the incident. Here's a breakdown of key considerations: ⚠️Factors Determining Investigation Depth: ✅️Severity of Harm: Incidents involving serious injuries, fatalities, or significant property damage require the most extensive investigations. Even minor incidents should be investigated, as they can reveal underlying hazards that could lead to more severe outcomes. ✅️Potential for Recurrence: Incidents with a high potential for recurrence warrant deeper investigations to prevent future occurrences. Near misses, where an incident almost occurred, should also be investigated thoroughly. ✅️Regulatory Requirements: Certain industries and jurisdictions have specific regulations that mandate the level of investigation required for particular types of incidents. ✅️Legal obligations must be met. Potential for Systemic Issues: Investigations should aim to identify not only the immediate causes but also any underlying systemic issues, such as inadequate training, faulty procedures, or equipment malfunctions. ⚠️Key Principles of Thorough Investigation: ✅️Timeliness: Investigations should begin as soon as possible after the incident to ensure accurate recollection of events and preservation of evidence. ✅️Objectivity: Investigations should be conducted impartially, focusing on facts rather than assigning blame. ✅️Root Cause Analysis: The goal is to identify the root causes of the incident, not just the immediate or direct causes. ✅️Data Collection: Gather all relevant information, including witness statements, physical evidence, and documentation. ✅️Documentation: Maintain detailed records of the investigation process and findings. ✅️Corrective Actions: Develop and implement corrective actions to prevent recurrence. ✅️Follow up: Ensure that corrective actions are effective. In essence: ℹ️Every incident deserves some level of investigation. The depth of the investigation should align with the potential for harm and the opportunity for improvement. By following these principles, organizations can effectively learn from incidents and create a safer environment. please share your thoughts on this. ====================================== #incident_accident_investigation.#safety_culture #quality.
-
𝐃𝐚𝐲 18: 𝐒𝐚𝐟𝐞𝐭𝐲 𝐀𝐮𝐝𝐢𝐭𝐬 & 𝐈𝐧𝐬𝐩𝐞𝐜𝐭𝐢𝐨𝐧𝐬 ♦️Safety Audits & Inspections are essential tools for identifying hazards, evaluating the effectiveness of safety programs, and ensuring compliance with safety regulations in the workplaces. 𝐒𝐚𝐟𝐞𝐭𝐲 𝐀𝐮𝐝𝐢𝐭𝐬 A #safety audit is a structured and comprehensive examination of an organization's safety processes, programs, and procedures. ♦️The goal is to ensure compliance with legal requirements and evaluate the effectiveness of safety systems. 𝐓𝐲𝐩𝐞𝐬 𝐨𝐟 𝐒𝐚𝐟𝐞𝐭𝐲 𝐀𝐮𝐝𝐢𝐭𝐬 1. Compliance Audits Ensure adherence to laws and regulations. 2. Program Audits Evaluate the effectiveness of safety programs (e.g., HSE policies). 3.Management System Audits Assess whether the safety management system aligns with international standards (e.g., ISO 45001). 𝐒𝐭𝐞𝐩𝐬 𝐟𝐨𝐫 𝐂𝐨𝐧𝐝𝐮𝐜𝐭𝐢𝐧𝐠 𝐚 𝐒𝐚𝐟𝐞𝐭𝐲 𝐀𝐮𝐝𝐢𝐭 1. Planning Define objectives, scope, and audit criteria. 2. Data Collection Review documentation, conduct interviews, and observe work practices. 3. Evaluation Analyze findings against regulations and best practices. 4. Reporting Document findings and recommend corrective actions. 5.Follow-Up Monitor the implementation of corrective measures. 𝐒𝐚𝐟𝐞𝐭𝐲 𝐈𝐧𝐬𝐩𝐞𝐜𝐭𝐢𝐨𝐧𝐬 A #safety inspection involves a routine check of specific areas, equipment, and work practices to identify hazards and ensure compliance with safety standards. 𝐓𝐲𝐩𝐞𝐬 𝐨𝐟 𝐒𝐚𝐟𝐞𝐭𝐲 𝐈𝐧𝐬𝐩𝐞𝐜𝐭𝐢𝐨𝐧𝐬 1. Routine Inspections Conducted regularly to check daily operations. 2. Equipment Inspections Assess the condition and maintenance of machinery. 3.Specialized Inspections Focus on specific hazards or areas (e.g., confined spaces). 4. Pre-Operational Inspections Conducted before starting a task or using equipment. 𝐬𝐭𝐞𝐩𝐬 𝐟𝐨𝐫 𝐂𝐨𝐧𝐝𝐮𝐜𝐭𝐢𝐧𝐠 𝐚 𝐒𝐚𝐟𝐞𝐭𝐲 𝐈𝐧𝐬𝐩𝐞𝐜𝐭𝐢𝐨𝐧 1. Preparation Determine the scope and inspection checklist. 2. On-Site Inspection Observe work practices, assess equipment, and interview workers. 3. Hazard Identification Document unsafe conditions and potential risks. 4.Corrective Actions Recommend measures to address identified hazards. 5. Documentation Record findings and actions taken. 𝐁𝐞𝐬𝐭 𝐏𝐫𝐚𝐜𝐭𝐢𝐜𝐞𝐬 - Use Checklists Ensure comprehensive coverage during audits and inspections. -Employee Involvement Engage workers to gain insights and foster a safety culture. -Documentation Maintain accurate records for accountability and trend analysis. - Continuous Improvement Act on findings to improve safety performance. 𝐓𝐀𝐊𝐄 𝐀𝐖𝐀𝐘 ♦️𝘙𝘦𝘨𝘶𝘭𝘢𝘳 𝘴𝘢𝘧𝘦𝘵𝘺 𝘢𝘶𝘥𝘪𝘵𝘴 𝘢𝘯𝘥 𝘪𝘯𝘴𝘱𝘦𝘤𝘵𝘪𝘰𝘯𝘴 𝘢𝘳𝘦 𝘤𝘳𝘶𝘤𝘪𝘢𝘭 𝘵𝘰 𝘪𝘥𝘦𝘯𝘵𝘪𝘧𝘺 𝘢𝘯𝘥 𝘢𝘥𝘥𝘳𝘦𝘴𝘴 𝘩𝘢𝘻𝘢𝘳𝘥𝘴. ♦️𝘗𝘢𝘳𝘵𝘪𝘤𝘪𝘱𝘢𝘵𝘦 𝘢𝘤𝘵𝘪𝘷𝘦𝘭𝘺 𝘪𝘯 𝘵𝘩𝘦𝘴𝘦 𝘢𝘴𝘴𝘦𝘴𝘴𝘮𝘦𝘯𝘵𝘴 𝘵𝘰 𝘦𝘯𝘴𝘶𝘳𝘦 𝘢 𝘴𝘢𝘧𝘦 𝘸𝘰𝘳𝘬
-
SafeWork NSW’s latest guide emphasizes work design as the most effective way to eliminate or minimize psychosocial hazards at the source. It highlights: - The need for systems thinking to address interconnected risks - The importance of workload planning, role clarity, and job control - The role of leaders and managers in shaping work environments that support mental health We need to shift from reacting to workplace stressors to designing work that prevents them in the first place. If you're involved in safety, HR, or leadership, this guide is worth a read. Full document: https://lnkd.in/gUcEdCKS #WorkDesign #PsychosocialRisk #SafetyLeadership Australian Institute of Health & Safety NSCA Foundation
-
Mental Health First Aid 101: What You Need to Know Have you ever noticed a colleague at work go from outspoken to increasingly withdrawn, missing deadlines, and avoiding conversations? You notice, but you’re unsure how to help. Or has a friend shared that they’re feeling hopeless, and you struggle to find the right words. Situations like these are more common than we think. The Mental Health First Aid (MHFA) is a framework designed to empower anyone to provide initial support to someone experiencing a mental health challenge or crisis. Here are some actionable steps from Mental Health First Aid that can help you make a real difference in such situations: 📌Recognize the signs: Be alert to changes in behavior, mood, or appearance. Common signs include withdrawal, persistent sadness, difficulty concentrating, irritability, or excessive worry. 📌 Approach and listen: If you notice someone struggling, approach them in a private and non-judgmental way. Use open-ended questions like, “I’ve noticed you seem different lately. Would you like to talk about it?” 📌 Offer support without judgment: Listening is powerful. Avoid interrupting or offering unsolicited advice. Instead, acknowledge their feelings with statements like, “That sounds really difficult. I’m here for you.” 📌 Encourage professional help: Mental health first aid doesn't mean solving the problem but guiding the person to appropriate resources. This could mean suggesting they speak with a counselor, a trusted doctor, or calling a crisis hotline. 📌 Provide resources and follow Up: Share helpful resources, like local mental health services or online platforms. Check back with the person after your initial conversation to show ongoing support. Mental health challenges are becoming increasingly common, equipping ourselves with these skills is essential. MHFA training is a great way to learn and practice these techniques. What would your workplace or community look like if everyone had the tools to respond to mental health challenges with confidence and compassion?
-
🔬 Confronting the Hidden Hazards: Legacy Chemicals in Research Labs. A recent publication in ACS Chemical Health & Safety by Craig A. Merlic, Imke Schröder, and Christopher M. Kolodziej sheds light on a pressing issue in research institutions: the accumulation of legacy chemicals. These are unused or expired substances—ranging from old commercial reagents to unlabeled research samples and outdated equipment—that linger in laboratories, often unnoticed. Key challenges highlighted: Safety Risks: Degraded containers and reactive substances like peroxides and picric acid pose significant hazards. Financial Burden: Disposing of unknown or improperly labeled chemicals can be exponentially more costly. Regulatory Compliance: Some legacy chemicals may now fall under stricter regulations, complicating their disposal. Proposed solutions: Inventory Management: Implementing accurate, shared chemical inventories to track and manage substances effectively. Off-boarding Procedures: Establishing clear protocols for researchers leaving the institution to ensure proper disposal of chemicals. Education and Policies: Training researchers on the risks associated with legacy chemicals and enforcing policies to prevent their accumulation. This paper serves as a crucial reminder that proactive chemical management is essential for safety, compliance, and sustainability in research environments. 📄 For the full article please see the link given in the comment section. #chemicalsafety #laboratorymanagement #researchsafety #environmentalhealth #chemicals
-
🌍 New WHO Technical Brief: Empowering Community Health Workers (#CHW) to Tackle Antimicrobial Resistance (#AMR) AMR is not just a clinical issue, it’s a multidimensional community challenge. As AMR continues to threaten health outcomes worldwide, especially in low-resource settings, the role of CHWs has never been more crucial. The latest WHO technical brief “Inclusion of antimicrobial resistance in training programmes for community health workers” reinforces the need to integrate AMR-related competencies into CHW training as part of a people-centred, primary health care approach. Key highlights: ✅ Elevates CHWs as frontline actors in AMR prevention and control ✅ Recommends practical AMR-focused tasks and tailored training themes ✅ Emphasizes community engagement, IPC, WASH, and appropriate antimicrobial use ✅ Aligns with the 2024 UN Political Declaration on AMR, which calls for a whole-of-society approach to addressing this global health threat Strengthening CHW capacity is not just an investment in AMR containment- it's a step toward resilient, equitable, and inclusive health systems. 📘 Read the full publication here: https://lnkd.in/drjaCR9Q Thanks to all who contributed: Yvan Hutin Jean Pierre Nyemazi Diriba Mosissa Mimi (Meheret) Melles-Brewer Nienke van de Sande-Bruinsma, PhD Serena Chong Onyema Ajuebor Mandy Deeves Faraz Khalid Nathalie El Omeiri Bassem Zayed Keti (Ketevan) Kandelaki Sahil Warsi, Shaffi Fazaludeen Koya, DrPH Takeshi Nishijima Rohina Joshi Katherine Rouleau Onur Karatuna Stephan Stenmark Wenjing Tao Freddy Kitutu Katy Davies Rosie Steege, PhD #CommunityHealth #PrimaryHealthCare #HealthEquity #OneHealth #WHO #GlobalHealth #InfectionPrevention #AntibioticResistance
-
When I was teaching infodemic management at the WHO during the pandemic, we asked the CDC colleagues to discuss five communication failures that consistently derail public health efforts: - Mixed messages from multiple experts - Information released too late - Paternalistic messaging - Failing to counter rumors in real-time - Public-facing power struggles and confusion In the US, all five are now happening at once. Public trust in health institutions is unraveling. People are adapting by building decentralized, multi-source, often crowdsourced “trust ecosystems.” This is what the New York Times comment section revealed after a recent article recommended credible health information sources. The comments were not fringe. They reflected skepticism, discernment, and a shift toward self-curated information strategies. Readers reported: - Turning to Mayo Clinic, Cleveland Clinic, Wikipedia, and NHS UK over US government sites. - Avoiding .gov domains due to perceived politicization. - Using AI cautiously, as a first filter, not a final word. - Proposing solutions like health site trust ratings, simplified printouts, and community-led education. Public health needs to meet this moment. Not by restoring the old systems, but by fostering something new for health information search, access and use: - Transparent, independent curation - Tools for triangulation and critical analysis - Localized, multilingual resource hubs - Responsible AI-supported health navigation - Community-led health literacy models Each of these comes with ethical, practical, and equity challenges. We need to think big picture and hyper-local at the same time. I don’t have all the answers. But I believe we need to build—together—a health information ecosystem for a fragmented, fractal, globalized, and crisis-prone world.