Big Changes Ahead for Epic Cogito Analytics: Are You Getting Ready?

Epic plans to transform the underlying architecture of the Cogito toolset in the next 1-2 years. It's not your typical quarterly release. It's a big deal. These changes will introduce significant new capabilities to enhance both Epic's analytics platform and healthcare analytics in general, opening doors to new ways to bring data together across and between institutions for analytics and AI/ML applications.

To make the most of these features as they roll out, organizations need to start adjusting their enterprise data and analytics strategies now. Those who wait risk losing years of time catching up.

Here are my thoughts (mixed with a bit of professional speculation) on how I'll be directing my team along with what I think healthcare analytics leaders should do in preparation. Contribute your thoughts in the comments! If you'd like to connect in person to discuss more, I'd love to hear from you.

First, a little background.

What's changing?

[Note: Epic is not sharing full details publicly yet. I checked with my contacts at Epic to make sure the following is okay to discuss, but we should avoid divulging more detail in public forums. For more detail, speak with your Epic Cogito TS team.]

Epic announced at this year's XGM user conference that Cogito will become a part of Nebula, Epic's cloud-based managed services offering. In simple terms, Cogito is moving to a Microsoft Azure-native "lakehouse" architecture. Clarity and Caboodle will no longer exist as standalone SQL servers, but instead will morph into medallion layers in the lakehouse.

What are the benefits?

More detail is still forthcoming, but with the above information, we can identify some big advantages that will come along with this change.

Scalability. This new architecture can take advantage of the highly parallel and infinitely scalable capabilities of the cloud lakehouse design pattern and promises improved performance at scale. Large organizations that have struggled to manage the scale of their data with Epic's current architecture will see significant performance opportunities here.

Proximity. If your organization's other data sets are already in the cloud or, even better, in a cloud data lake or lakehouse, you won't need to make a separate copy of your EHR data to integrate it with other data sources.

Consumption-based cost. With this architecture, compute is ephemeral, and you only incur cost when using it. Traditional architecture either required significant capital investment in powerful servers for Clarity and Caboodle (which sat underutilized most of the day) or expensive, always-on cloud VMs. Now, you have what you need when you need it, and you stop paying for it as soon as you are done.

Hourly ETL. Epic revealed that ETL (or more accurately ELT) on the new architecture will be at least hourly for commonly used data sets. This opens all sorts of possibilities for more performant analytics tools across a broader set of use cases and greatly simplifies deployment of near real-time AI/ML use cases.

Open-source storage format. The Delta format used in a lakehouse is an industry-standard, open-source format with extensive and growing support across analytics tools and cloud platforms. This will enable sharing and portability use cases never before possible.

Frictionless Data sharing. Data sharing capabilities across platforms and institutions is native functionality within Microsoft Azure.

Native connectivity for new data tools. The AI/ML and analytics tools in the Azure platform can natively interact with this infrastructure. This means that there will no longer be a need to copy/replicate EHR data to another location or architecture to enable advanced analytics for many use cases. Reduced data movement and simplified management will be possible. Data versioning is also native.

What should organizations be doing to prepare?

Re-evaluate Architecture Roadmap and Cloud Strategy

If you are a healthcare provider, your EHR is likely a primary data source for much of your enterprise analytics landscape. Accordingly, healthcare organizations that use Epic, or other partners/vendors in the Epic ecosystem, should review their architecture roadmap and cloud strategy in light of this news.

Organizations who have not yet developed a cloud migration plan should carefully review the synergies certain architecture decisions will make and factor those into their cloud strategy. I firmly believe that an organization should choose its enterprise cloud data and analytics tooling based on what will best achieve business use-cases and outcomes, and independent of any vendor's chosen platform. That said, your EHR is a data source to your enterprise platform, so you'll need to consider options for secure connectivity and tradeoffs on ingress/egress charges on a multi-cloud environment. Organizations planning to use best-of-breed cloud-agnostic platforms (like Snowflake or Databricks) may find some cost savings by executing these workloads close to where Epic's data is located (e.g. same cloud region).

Upskill Yourself and Your Teams

How much do you know about the medallion lakehouse architecture? How much do you understand about how to take advantage of the benefits this will bring? Now is the time to start learning! There are many resources out there, but Databricks (a pioneer of the lakehouse approach) produced an e-book that is a great place to start: The Data Lakehouse Platform for Dummies | Databricks

For your team, the skill set used by your Epic BI Developers will largely suffice for this new architecture. The front-end of the Cogito toolset probably won't change much and they can still use traditional SQL when they have to query the database.

However, you'll be missing out on the value this platform brings if you stop there. Instead, organizations should begin thinking about how to upskill and restructure their analytics teams to take advantage of new tools that can be used against the lakehouse.

More of your analytics team should be developing an advanced skill set. No longer will an advanced analytics or AI/ML solution require a project-sized team with data engineers, infrastructure admins, and capital acquisitions. Instead, solving these types of problems will start by simply opening a notebook that's already connected to the entire EHR data set. This democratization of advanced data science and AI/ML capabilities won't bring value unless you expand the number of resources on your analytics teams that can take advantage of them.

Your ETL developers/data engineers will need new skills for the process that replaces custom Caboodle development. An understanding of the tools used to model data in a lakehouse will be foundational for this and for development of cloud solutions that will consume data from this infrastructure.

Lastly, more frequent ELT will shift the balance on which tool should be used for a particular need, and BI Developers with a strong SQL/Python skill set will fare better than those who have mainly focused on Reporting Workbench and Radar.

Plan attrition or retooling for obsolete skills.

It looks like the Clarity/Caboodle ETL administrator role in most organizations will eventually be unneeded or greatly diminished. Nebula being a managed service means that this role will likely be performed by Epic. Analytics leaders should factor this into their staffing roadmap for 2025 and beyond. If you currently used a managed service provider to administer your Cogito ETL, this should be factored into contract renewals in the coming year. If you use FTEs for these resources, planning early will allow for retooling and placement in other roles.

Identify and Prep New Use Cases

The new architecture will enable new capabilities, as discussed above. But organizations don't have to wait until these capabilities arrive to start preparing to take advantage of them.

Ask yourself and your data-savvy operational partners, "What could we do if near real-time EHR data sets were available in a cloud-scale AI/ML-enabled environment along-side our organization's other data assets?"

Ask your teams what they are struggling to accomplish with the current architecture. Ask your data engineers where they are spending the most time combining EHR data with other data sets.

As the fervor around AI/ML grows in healthcare, keep track of potential applications of these capabilities that don't have a good solution in the current architecture (e.g. need near real-time data sets, out-scale current tools, etc.).

From these discussions, use cases will emerge that can add tremendous value to your organization. As more detail and timelines are revealed, planning and requirements gathering can begin in order to take advantage of the new architecture as soon as it becomes available.

Pause/Halt Some Projects

You may want to evaluate current projects, or efforts planned in the near term that will be rendered obsolete or unnecessary by these architecture changes. Here are a few examples:

• Planned hardware upgrades for your existing infrastructure may no longer be needed.
• Real-time data extraction band-aids to support new AI/ML-enabled solutions will have a better option. You may want to consider waiting, or narrowing scope for now, knowing that the process will be significantly streamlined in the future.
• Vendors offering analytics products in the cloud that can ingest your EHR data and perform analytics or feed that data into ML models may prove an unnecessary investment.
• Custom Caboodle enhancements to integrate external data sets may be better off waiting.

Conclusion

The change coming to the Epic Cogito infrastructure shouldn't be underestimated. It will enable a huge leap forward in analytics capabilities for Epic customers who are prepared to exploit all of the advantages of a modern, scalable, and cloud-integrated platform. Organizations who start preparing now will be able to realize these benefits years sooner than those who wait.