Introducing Comet: AI for healthcare intelligence and decision-making

This is a significant leap forward in clinical decision-making. For too long, our healthcare system has been reactive, but technologies like Comet are fundamentally shifting the paradigm to proactive and predictive care. Being able to accurately forecast a patient's journey doesn't just optimize resources for health systems; it directly leads to better patient outcomes. Giving clinicians the power to see what's likely to happen next is the difference between prevention and intervention. Excited to watch how predictive intelligence becomes the new standard of care.

Fantastic work from Epic with the creation of Comet, new form of healthcare intelligence conceived to help clinicians, patients and health systems to predict the patient outcomes and risk and then make informed decisions. I have a few questions: - If its potential is confirmed in US health systems where it was designed using its data for training , is it valid for other countries and other ethnicities? - How accurate are these predictions? - Can this approach create unnecessary fears and anxiety among healthcare professionals and patients? - Who/What will be accountable when a medical error was made due to this system?

Epic’s new CoMET (Cosmos Medical Event Transformer) models mark an important step for generative AI in healthcare. Trained on 115B medical events from 118M patients across 310 health systems, these Transformers demonstrate predictable scaling laws and deliver strong results across 78 diverse clinical tasks — without fine-tuning. Key takeaway: generative event models can simulate realistic patient trajectories and outperform specialized models (e.g. 30-day readmission prediction, AUC 0.77 vs. 0.72). Still, translation into practice will depend on prospective validation, subgroup analysis, and handling data quality variation. The potential is clear — so are the challenges.

What if we could see the next step in a patient’s journey before it happened? As we prepare for our Epic transition, I have been reflecting on how we use data. Tools like Comet show what is possible when information becomes insight and foresight. It is exciting to imagine how far healthcare intelligence can take us in improving care and outcomes. #Epic #HealthcareInnovation #QualityImprovement #InfectionPrevention #EMRIntegration #PatientSafety

Epic’s new CoMET models are an exciting step forward for generative AI in healthcare. Training on 115B+ medical events across 118M patients and outperforming specialized models in 30-day readmission prediction (AUC 0.77 vs. 0.72) highlights the potential of large-scale event transformers. I’m especially interested in how these models balance scalability with interpretability and how prospective validation and subgroup analysis will shape their clinical impact. #DataScience #OrlandoHeath #HealthcareAI #PredictiveAnalytics #Readmissions #GenerativeAI

Comet built within Epic Cosmos takes full advantage of blending real-time and historical data to guide patient care.

Day 1: Improving Patient Outcomes Data analytics helps healthcare providers deliver better care by identifying patterns in patient health records, lab results, and treatment responses. With predictive models, doctors can anticipate complications before they arise—like spotting early signs of sepsis or heart failure. This proactive approach allows for timely interventions, reducing hospital readmissions and improving recovery rates. Analytics also supports personalized medicine, tailoring treatments to individual genetic profiles or lifestyle factors. In short, data turns reactive care into proactive healing, giving patients a better shot at long-term wellness.

This article by Dr. Joel Wallskog argues that our current ICD-10-CM system lacks specific codes for post-COVID-vaccine adverse effects. As a result, clinicians are forced to rely on vague categories that undermine patient care, research, and public health monitoring. Wallskog emphasizes that these patients undeniably exist, and that updating the system should be unquestionable. It’s time to recognize patient realities through transparent, precise coding. Learn more: https://lnkd.in/dYGUkmS3

How can faster diagnostic results reshape healthcare's proactive potential? Imagine a world where interventions happen sooner, and outbreaks are contained swiftly. CRISPR technology is streamlining diagnostics, delivering results in under an hour—a stark contrast to the days-long wait of traditional methods. This speed is critical for rapid decision-making and managing emergencies, offering clarity with 'negative means negative, positive means action.' The reliable and quick identification of asymptomatic carriers also builds essential trust in test results. What impact could such speed and accuracy have on public health strategies and individual patient care? Curious to hear your insights on the future of diagnostics. #Diagnostics #CRISPR #HealthcareInnovation #PublicHealth #MedicalTechnology #Innovation

🔍 Diagnostic errors are a leading cause of preventable harm in patients, with the potential for serious harm, making them a global priority in #patientsafety. 🎥 In this video, Dr Natasha Coen, Senior Clinical Risk Manager, looks back on a recent review of diagnostic incidents reported on NIMS and explains how the #StateClaimsAgency's upcoming conference 'Understanding and Improving Diagnosis in Healthcare' will build on this review to explore the diagnostic process, why it might go wrong and what health and social care professionals can do to improve it. 📅 1 October 🎙️ Expert national and international speakers including Maureen Nolan, Dr Olga Kostopoulou, Dr John Fitzsimons, Dr Peter MacMahon, Professor Prashant Mahajan and moderator Professor Deborah McNamara from Royal College of Surgeons in Ireland (RCSI). 📚 Check out the review 🔗https://lnkd.in/eSRA3hnW 👉 Register for the conference🔗 https://lnkd.in/ecDZ_ri5 #SCALearning #healthcare #diagnosis #clinicalrisk #managingrisk #claimsresolution #patientcare #learn #improve #resolve