Epic’s Comet and the Next Era of Predictive Healthcare

By Ernie Ianace, CEO of CareAlly

Introduction

Artificial intelligence in healthcare is accelerating at a pace few would have predicted even five years ago. Every month brings new announcements from health systems, technology companies, and innovators seeking to bring intelligence closer to the point of care. Epic’s recent announcement of Comet, a predictive AI platform built within its Cosmos data environment, represents one of the most ambitious and potentially transformative steps yet.

This blog will examine what Comet is, why it matters, and how it fits into the broader arc of AI in healthcare. While we’ll give credit where it’s due — Epic deserves it for the scale and ambition of this project — we’ll also explore the limitations of predictive models when they are not combined with orchestration and workflow integration. Finally, we’ll look at what this means for the future of care delivery, value-based care, and platforms like CareAlly that focus on bridging the last mile from prediction to action.

What Epic Announced

On September 3, 2025, Epic formally unveiled Comet, describing it as a new form of healthcare intelligence. The goal is straightforward: to help clinicians, health systems, and patients make better decisions by predicting what’s most likely to happen next in a patient’s journey.

Built on Epic Cosmos, a secure and collaborative data environment, Comet has been trained on over 100 billion patient medical events. These include diagnoses, labs, medications, and encounters collected from millions of de-identified patient records across diverse health systems. Using approaches similar to those behind today’s large language models, Comet learns how clinical patterns evolve over time and can generate a range of plausible futures — for example, whether a patient will need hospitalization, how long their stay might last, or what complications may arise.

In validation studies spanning 78 use cases, Comet was shown to outperform many specialized models built for single conditions. This is a key point: instead of narrowly focusing on one disease state, Comet aims to provide a flexible, general-purpose predictive engine for multiple aspects of patient care.

The first deployments will be limited to research partners within Cosmos, starting in February 2026. These organizations will be able to test Comet in a “virtual lab,” exploring new use cases and advancing its clinical relevance.

Why Comet Matters

1. Scale of Data

   Epic Cosmos is already one of the largest curated clinical datasets in existence. Training Comet on 100 billion medical events gives it exposure to a vast range of patient trajectories, comorbidities, and outcomes. This scale matters — predictive accuracy in medicine often depends on finding patterns across diverse populations.

2. Shift from Static to Dynamic Predictions

   Traditional predictive tools in healthcare have largely been static. They generate a single risk score (for readmission, mortality, etc.) and leave it at that. Comet’s approach is different. By simulating multiple possible timelines, it acknowledges uncertainty and provides clinicians with a spectrum of plausible futures. That nuance is critical in clinical decision-making, where few outcomes are ever black and white.

3. Integration with Clinical Workflows

   Epic’s strength has always been its dominance in electronic health records. By embedding Comet directly into the Epic environment, the company is positioning predictive insights where clinicians already spend much of their time. If executed well, this could reduce friction and increase adoption.

4. Collaborative Development

   Epic partnered with Yale and Microsoft in developing Comet, demonstrating that the best healthcare AI often emerges from collaboration between providers, academia, and technology. This model should be encouraged across the industry.

The Limitations of Prediction Alone

While Comet’s vision is compelling, it’s important to recognize what predictive modeling alone cannot solve.

• Predictions without Action: Knowing that a patient is at high risk for readmission is only valuable if that insight is tied to specific workflows — care coordination, patient education, follow-up scheduling — that reduce that risk.

• Trust and Interpretability: Clinicians are more likely to act on AI-driven predictions if they understand the “why” behind them. Black-box outputs, no matter how accurate, risk being ignored.

• Operational Variability: Predictions must be contextualized within each health system’s staffing, resources, and population. A forecast that makes sense in one setting may not translate directly to another.

• Siloed Use: Predictions embedded only in Epic may struggle to reach providers and organizations that use other systems or need cross-platform orchestration.

This is where orchestration platforms like CareAlly enter the picture. Predictions become powerful when they are integrated into broader workflows, personalized to the resident or patient, and connected to concrete interventions.

Prediction, Insight, Orchestration

Orchestration: The Next Step Beyond Prediction

AI in healthcare is evolving through three distinct stages:

1. Outputs: Risk scores, single predictions, static analytics.

2. Insights: Richer, dynamic forecasts like Comet, which generate a range of possible futures.

3. Orchestration: Platforms that take these insights and operationalize them — aligning care teams, automating tasks, ensuring compliance, and personalizing engagement.

CareAlly is focused on that third stage. For example:

• Early Intervention: If a predictive model identifies that a resident is likely to experience health deterioration, CareAlly can automatically trigger interventions — outreach, care plan adjustments, or reminders for caregivers.

• Workflow Automation: Predictions about hospitalization risk can be tied directly to tasks like scheduling follow-ups or initiating chronic care management programs.

• Persona-Based Engagement: CareAlly doesn’t just predict. It communicates differently with a 40-year-old Spanish-speaking patient with CHF and diabetes than with a 70-year-old Medicare Advantage enrollee, ensuring outreach resonates.

• System-Wide Integration: Predictions are only useful if they cross silos. CareAlly’s orchestration engine integrates with EHRs, scheduling, communication platforms, and analytics systems to ensure insights are actionable everywhere.

Implications for Value-Based Care

The launch of Comet has real implications for the future of value-based care (VBC).

• Reducing Avoidable Readmissions: Predicting which patients are most likely to bounce back into the hospital allows providers to intervene early, a key driver of both quality and cost metrics.

• Chronic Disease Management: Forecasting disease trajectories makes it easier to keep patients stable in the community rather than in acute care.

• Resource Optimization: Predictions about length of stay or acuity help health systems allocate resources more efficiently.

• Equity and Population Health: With data spanning millions of patients, predictive tools can highlight disparities and support more equitable care delivery.

But once again, predictions alone won’t hit VBC benchmarks. They need orchestration — consistent action at scale, tailored to patients and aligned with provider workflows. That’s where platforms like CareAlly complement predictive engines like Comet.

Looking Ahead

Epic deserves real credit for Comet. Few organizations have the data scale, customer base, and technical expertise to attempt something this ambitious. If successful, Comet could shift how clinicians think about uncertainty, moving from gut instinct to data-driven foresight.

At the same time, the industry should be clear-eyed: predictive intelligence is only one part of the puzzle. The future of AI in healthcare will belong to platforms that combine prediction with orchestration, ensuring insights lead to better outcomes, reduced costs, and improved patient and resident experiences.

At CareAlly, we’re excited about this direction. Comet raises the bar for what predictive AI can do. Our mission is to take those insights and make them real — automating workflows, personalizing care, and ensuring that every prediction translates into meaningful action.

Conclusion

The launch of Comet is not just another tech announcement. It’s a signpost for where healthcare is headed: more predictive, more proactive, and more data-driven. It highlights the promise of combining scale, collaboration, and workflow integration to tackle some of medicine’s hardest challenges.

As predictive tools like Comet mature, the next challenge will be orchestration — making sure these predictions don’t just sit in dashboards, but drive real change at the bedside, in the community, and across the continuum of care. That’s where the future lies, and that’s the vision we’re building toward every day at CareAlly.