Clinical Error Prevention

🎉 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.

🧠 The Swiss Cheese Model of Safety: A Practical Framework to Prevent Medication Errors in Hospitals In a hospital setting, where every second matters, mistakes don’t always come from a lack of knowledge — they come from system gaps. Medication errors remain one of the most preventable causes of patient harm, and while “to err is human,” allowing those errors to reach a patient is a failure of the system, not the individual. To move closer to Zero Harm, we need more than checklists. We need multiple layers of defense that work together. I’ve created this framework to show how we can strengthen each layer and close the small loopholes that lead to big consequences. 1️⃣ Technological Safety Nets Relying on memory alone is a recipe for error. Tools like Computerized Physician Order Entry (CPOE) eliminate issues like illegible handwriting, while Bar-Code Medication Administration (BCMA) helps ensure the 5 Rights: right patient, right drug, right dose, right route, and right time. Technology doesn’t replace pharmacists — it allows us to devote more time to clinical decision-making. 2️⃣ Clinical Vigilance & LASA Safety Look-Alike Sound-Alike (LASA) drugs and high-alert medications (insulin, anticoagulants, opioids) demand extra caution. Tall-Man lettering (e.g., predniSONE vs predniSOLONE), clear labeling, and proper storage separation are small changes with huge impact. A pharmacist-led medication reconciliation step remains one of the strongest human barriers against preventable harm. 3️⃣ Strong, Structured Communication Miscommunication is one of the most common causes of medication errors. Shifting from casual, incomplete handovers to SBAR (Situation, Background, Assessment, Recommendation) gives every team member the clarity needed to make safer decisions, especially in high-pressure situations. 4️⃣ Building a Non-Punitive Culture This may be the most important layer of all. When staff fear blame or punishment, errors go unreported — and uncorrected. A supportive culture encourages reporting, learning, and system-level improvement. We can only fix what we are willing to talk about. The Goal When these layers work together, we achieve more than compliance. We enhance workflow, reduce legal risk, improve team coordination, and most importantly — protect patient lives. 💬 A Question for My Colleagues Is your hospital currently using CPOE, or are handwritten prescriptions still the norm where you practice? Prepared by: Dr. Ashraful Mollah Founder, MedScope360 #PatientSafety #ClinicalPharmacy #PharmD #MedicationErrors #HospitalPharmacy #HealthcareQuality #Pharmacovigilance #MedicalEducation #HealthTech #AshrafulMollah

🔎 Analysis of Prescription & Medication Errors 1. Look-Alike / Sound-Alike (LASA) Drugs Merofic Plus (Meropenem + Sulbactam) and Tigebax (Tigecycline) both are high-alert broad-spectrum antibiotics. ❌ Risk: Confusion during administration → wrong drug / duplication of coverage. ✅ Prevention: Use Tall-Man lettering (e.g., MERO-penem vs 2. Unclear Dosing Instructions Several entries show “SOS (if needed) infusion” for Norbit, Artacil, Rufenyl → but no indication or maximum frequency mentioned. ❌ Risk: Overuse or underuse depending on nursing discretion. ✅ Prevention: Prescriber should define clear indication + max dose/day. 3. Ambiguous Abbreviations QID, BD, OD, SOS written. Abbreviations can be misinterpreted. Example: OD = once daily, but sometimes mistaken as “right eye”. ❌ Risk: Wrong frequency. ✅ Prevention: Write full words: “once daily”, “twice daily”, “if required”. 4. Injection Safety Issues Multiple IV antibiotics prescribed simultaneously (JP Cipro, Merofic, Tigebax, Optineurone, etc.). ❌ Risk: Drug–drug interactions, infusion incompatibility. ✅ Prevention: Pharmacist must check compatibility charts, stagger infusions, ensure dilution protocols. 5. Overlapping Antimicrobials Ciprofloxacin, Meropenem, Tigecycline, Colistin → all broad-spectrum, overlapping spectrum. ❌ Risk: Antimicrobial resistance, nephrotoxicity. ✅ Prevention: Culture-sensitivity guided de-escalation; Antibiotic Stewardship Program (ASP) involvement. 6. Illegible Handwriting (Other Medicine Section) Example: “Tab Ivrabadin”, “IM colistin…”, “Qvegness Tab PBD” – difficult to read. ❌ Risk: Misinterpretation by nursing staff → wrong drug/dose. ✅ Prevention: Use capital letters or e-prescription. 7. No Documentation of Route / Dilution Some injections (e.g., Norbit, Artacil, Rufenyl) only marked as “Infusion” without dilution volume / diluent / rate. ❌ Risk: Phlebitis, toxicity, therapeutic failure. ✅ Prevention: Always mention diluent, volume, infusion time. 8. No Patient Allergy Status Noted Allergy column not filled. ❌ Risk: If patient is allergic to β-lactam or fluoroquinolone, severe ADR possible. ✅ Prevention: Allergy history must be documented on chart. ✅ Summary of Key Errors 1. LASA antibiotic risk (Meropenem vs Tigecycline). 2. Ambiguous abbreviations (OD, BD, QID). 3. SOS orders without indication. 4. Multiple broad-spectrum antibiotics (risk of resistance). 5. Handwriting legibility problem. 6. Missing dilution/infusion details. 7. No allergy documentation. 🛡️ Prevention Strategies Adopt ISMP-approved abbreviations only. Ensure antibiotic stewardship with microbiology support. Use printed / electronic prescriptions. Provide clear dilution and infusion protocols. Highlight LASA drugs with Tall-Man lettering. Document allergy status at the top of the chart. Regular medication chart review by

Between 250,000 and 440,000 Americans die every year from preventable medical errors. Not from incurable diseases. Not from unavoidable complications. From errors that a better-designed system would have caught. We just published a white paper that makes the case — backed by real deployment data — that this is not a competence problem, a training problem, or a funding problem. It is an infrastructure problem. And AI Agents are already solving it. In our deployments across hospital systems and Veterans Administration facilities, we've seen: 🔹 30-day readmissions drop from 11% → 0% 🔹 Post-discharge follow-up failures fall from 30% → 7% 🔹 99% of pre-visit documentation automated — giving clinicians 14 minutes back with every patient 🔹 Day-of-surgery cancellations cut from 17% → under 2% 🔹 ROI of 573% in radiology. 965% in consult prep. 839% in oncology EMR integration. But the number that matters most isn't on any ROI table. It's a clinician who told us: "For the first time in a long while, I am not worrying about certain patients — I now know how they are doing." That is the measure of whether AI has succeeded in healthcare. The white paper covers: ✅ The root causes of medical errors — and why existing approaches haven't solved them ✅ How AI Agents address medication errors, documentation failures, diagnostic gaps, and readmissions ✅ The health equity dimension — why uniform AI outreach is the most powerful equity tool available ✅ The governance framework that separates AI that works from AI that doesn't ✅ The financial case — because patient safety and financial performance are now the same investment If you are a clinical leader, a health system executive, or anyone who believes patients deserve better — this paper is for you. 👉 Read the full white paper: https://lnkd.in/gir4rcA6 👉 Or reach out directly — I'd welcome the conversation. #PatientSafety #AIinHealthcare #MedicalErrors #HealthcareAI Nova Insights Corp #DigitalHealth #ClinicalEfficiency #HealthcareLeadership

This study could change how every frontline clinic in the world delivers care. Penda Health and OpenAI revealed that an AI tool called AI Consult, embedded into real clinical workflows in Kenya, reduced diagnostic errors by 16% and treatment errors by 13%—across nearly 40,000 live patient visits. This is what it looks like when AI becomes a real partner in care. The clinical error rate went down and clinician confidence went up. 🤨 But this isn’t just about numbers. It’s a rare glimpse into something more profound: what happens when technology meets clinicians where they are—and earns their trust. 🦺 Clinicians described AI Consult not as a replacement, but as a safety net. It didn’t demand attention constantly. It didn’t override judgment. It whispered—quietly highlighting when something was off, offering feedback, improving outcomes. And over time, clinicians adapted. They made fewer mistakes even before AI intervened. 🚦 The tool was designed not just to be intelligent, but to be invisible when appropriate, and loud only when necessary. A red-yellow-green interface kept autonomy in the hands of the clinician, while surfacing insights only when care quality or safety was at risk. 📈 Perhaps most strikingly, the tool seemed to be teaching, not just flagging. As clinicians engaged, they internalized better practices. The "red alert" rate dropped by 10%—not because the AI got quieter, but because the humans got better. 🗣️ This study invites us to reconsider how we define “care transformation.” It's not just about algorithms being smarter than us. It's about designing systems that are humble enough to support us, and wise enough to know when to speak. 🤫 The future of medicine might not be dramatic robot takeovers or AI doctors. It might be this: thousands of quiet, careful nudges. A collective step away from the status quo, toward fewer errors, more reflection, and ultimately, more trust in both our tools and ourselves. #AIinHealthcare #PrimaryCare #CareTransformation #ClinicalDecisionSupport #HealthTech #LLM #DigitalHealth #PendaHealth #OpenAI #PatientSafety

If IV chemotherapy requires seatbelts and airbags, oral chemotherapy demands an entire crash-safety system Patients administer these high-risk drugs on their own, without nurses or pharmacists right there to intervene That’s why specialty accreditation standards exist - not as a bureaucratic checklist (although it can feel like that), but as a blueprint for extending clinical oversight beyond the clinic walls These standards call for more than just dispensing a bottle of pills. They require processes that transform a prescription into a comprehensive care plan. 🧪 Patient management - verifying diagnosis, labs, and organ function to confirm appropriateness before the first dose gets to the patient 👩🏫 Education - teaching patients not just how to take the drug, but what to expect, how to store and handle it safely, and when to call for help 💊 Adherence and toxicity monitoring - scheduled follow-ups to ensure patients are actually taking therapy as prescribed, and to surface barriers like side effects, cost, or confusion This process is not only about preventing errors. It allows us to see toxicity signals in between clinic visits and intervene before issues spiral into hospitalizations, dose holds, or treatment discontinuation. Because when therapy is effective, that oversight can help patients stay on it as long as possible and give them the best chance at durable outcomes For pharmacists, this means our work isn’t just about verifying orders - it’s about building the safety system that protects patients outside the clinic. Every assessment, every education call, every follow-up to check is another layer of that system. Accreditation standards may give us the blueprint, but it’s how we implement them out in practice that truly keeps patients safe and on treatment As oral chemotherapy use continues to grow, so does the need for vigilance, creativity, and collaboration. Patients may be the ones holding the pill bottles, but it’s pharmacists who ensure those pills translate into safe, effective treatment for as long as possible. --- 📌 Start fostering your growth in oncology pharmacy practice with the Oncology Insights Newsletter I’m the Kelley in KelleyCPharmD 👋 and I help pharmacists learn the complex world of oncology

I almost missed a mistake once. I was moving quickly through my NICU list — placing orders, juggling pages, doing what clinicians do in high-stakes environments. At some point, without realizing it, I was briefly in the wrong patient’s chart. Nothing dramatic. Nothing obvious. I sat back down, already onto the next thing, when a nurse approached me. “Hey Dr. Daigle,” she said carefully, “I think that echo might be in Baby J’s chart instead of Baby M’s.” She was right. I thanked her. I corrected the order. And we moved on — safely. But I’ve thought about that moment many times since. Because I’ve also had nurses tell me they were yelled at for speaking up. Shut down for questioning decisions. Made to feel small for doing the very thing that keeps patients safe. And I know exactly how easily that nurse could have stayed silent. That wasn’t luck. That was psychological safety. The kind that allows someone to speak up without rehearsing the consequences first. The kind that turns hierarchy into collaboration. The kind that prevents errors before they reach patients. We talk a lot about patient safety. We don’t talk enough about clinician psychological safety. But the two are inseparable. Psychological safety isn’t a soft skill. It’s clinical infrastructure. And it starts with leadership — with how we respond when someone questions us. Healthcare is relational work. If we want safer systems, we have to design for human psychology — not just clinical protocols. This is The Unwritten Practice™ #PsychologicalSafety #PatientSafety #HealthcareLeadership #HighStakesCare #TeamBasedCare #QualityAndSafety #HumanCenteredCare #TheUnwrittenPractice

Serious Breach in Clinical Trials: Incorrect IMP Administration 🚨 A single mistake in investigational medicinal product (IMP) administration can compromise participant safety and clinical trial integrity. Could this happen in your clinical trial? Example: A participant in a clinical trial was mistakenly dosed with the incorrect IMP, which was from a completely different clinical trial. The medication was also administered via the wrong route. This mix-up resulted in the participant receiving treatment they were never intended to have, leading to potential safety risks and compromised study data. Why Is This a Serious Breach? ❗ This situation constitutes a serious breach because: ❗It significantly impacts participant safety by exposing the participant to unknown risks. ❗It violates ethical and regulatory standards by failing to ensure proper treatment. ❗It jeopardizes data integrity, as incorrect dosing skews study results. Corrective Actions Once identified, immediate corrective actions must be taken: 🔹 Notify the relevant regulatory authorities and ethics committees. 🔹 Conduct a root cause analysis to understand how the error occurred. 🔹 Implement additional participant monitoring to mitigate potential harm. 🔹 Retrain study personnel to prevent repeat incidents. Preventive Actions To prevent incorrect IMP administration in the future: 🔹 Ensure strict IMP handling procedures with double-check systems. 🔹 Use electronic verification systems for drug assignment. 🔹 Implement site training programs to reinforce protocol adherence. 🔹 Conduct regular audits to identify and rectify process gaps. How Could Sponsors Have Prevented This Serious Breach? Sponsors play a crucial role in preventing these errors by: ✅ Strengthening site oversight with better monitoring and accountability. ✅ Using barcoding or RFID tracking to match IMP to the correct participant. ✅ Establishing robust SOPs for drug storage and administration. ✅ Conducting mock scenarios to prepare staff for potential mishaps. Consequences for the Clinical Trial Failing to address IMP mis-administration can lead to: ⚠️ Regulatory penalties and clinical trial suspension. ⚠️ Increased liability risks for the sponsor and site. ⚠️ Data validity concerns, potentially leading to study rejection. ⚠️ Participant trust erosion, making recruitment and retention more difficult. 📢 Key Takeaway: A strong quality management system and proactive sponsor oversight are essential in preventing serious breaches. Has your clinical trial implemented sufficient safeguards? Let’s discuss in the comments!

What is Systematic Error? Systematic error is a consistent bias in laboratory results where values are repeatedly higher or lower than the true value. Results may look acceptable but can be clinically wrong, leading to incorrect diagnosis or treatment. ⸻ How to Detect Systematic Error 1️⃣ Watch the QC Mean (Shift in Control Values) What it means: Quality Control (QC) results should fluctuate randomly around the established mean. A shift occurs when control values suddenly move and stay on one side of the mean. Why it matters: A shift indicates a consistent bias, often caused by: • New reagent lot • Calibration error • Instrument malfunction • Environmental changes What to do: • Compare current QC mean with previous mean • Check Why did the average change?” • Stop reporting patient results until the cause is identified Key message: A stable mean = reliable system A shifted mean = warning sign 🚨 2️⃣ Follow the Trend Gradual upward or downward QC movement is an early warning sign. 👉 Investigate before limits are exceeded. Key message: 📉 Trends whisper early warnings—listen to them. 3️⃣ Apply Westgard Rules Rules like 2₂s, 4₁s, and 10x help detect hidden systematic bias. 👉 Small rule violations matter. 4️⃣ Use Delta Checks Unexpected patient result changes may indicate analytical error. 👉 Why it matters: Sudden, unexplained changes may indicate: • Analytical bias • Instrument calibration error • Sample mix-up What to do: • Review patient history • Confirm sample identity • Repeat testing if results don’t make clinical sense Key message: 🧠 Numbers must make sense clinically, not just analytically. 5️⃣ Review Proficiency Testing (PT) What it means: F compares your lab’s results with peer laboratories using the same method. Why it matters: • PT failures often reveal hidden systematic bias • Routine QC may not detect method-related bias What to do: • Investigate every PT failure seriously • Compare method, calibration, and reagents • Implement corrective and preventive actions Key message: 🧪 PT failures expose what routine QC may miss. 6️⃣ Verify After Any Change New reagents, calibration, or maintenance can introduce bias. 👉 Always verify before reporting results. What to do: • Run QC and compare with previous performance • Verify accuracy before releasing patient results • Document all checks Key message: ✔️ Always verify after updates—never assume. ⸻ 🔔 Final Take-Home Message Systematic error makes results look right, but clinically wrong. Early detection through: • QC monitoring • Trend analysis • Westgard rules • Delta checks • Proficiency testing • Post-change verification 👉 Protects patient safety and maintains laboratory credibility. #Labquality #Proficiencytesting #Westgardrules