Last year, KLAS published a report focused on vendors who provide purpose-built AI solutions in healthcare. We at KLAS, along with provider organizations, were curious as to what the enterprise EMR vendors were offering in the world of AI today. Our recently published Healthcare EMR AI 2020 report delves into what we were able to find.
Emerging Data for AI in Healthcare
AI solutions are still relatively new in healthcare. Many organizations are still looking to implement them; some have yet to employ them at all. Along with the machine learning platforms, other AI capabilities being explored are voice assistance and navigation, ambient listening, coding and documentation assistance, and a computerized virtual scribe. The four EMR vendors we examine in the report — Allscripts, Cerner, Epic, and Meditech — are all at varying levels of implementation and development in these different areas.
In the report, we found that Epic was the only one of the four vendors who had enough live customers that we could reach out to and draw conclusive performance results from. It’s important to note that Cerner does have some partner sites running their machine learning platform models and achieving outcomes, but we don’t even have enough data to show performance. Allscripts and Meditech are both solely in their development phases with no live customers using their AI capabilities yet.
Prebuilt AI Models
A unique feature that Epic offers is prebuilt AI models. As a client, you can turn those models on and incorporate their predictive abilities into your workflow. Customers still have to do a lot of configuration on their own, but since everything resides in Epic using Epic data, the integration is fairly solid. Most Epic AI customers are “out-of-the-box” users, meaning they are using prebuilt ML models from Epic’s Model Library, available to Cogito customers. They also receive technical support for testing, validation, and deployment.
When we performed our initial research into the niche AI vendors, we found that many customers chose not to turn on the ML models from Epic. The question arose around whether Epic as an EMR vendor could deliver mature Ai solutions. Some organizations were concerned that the Epic models, built on Epic’s overall population, might not be accurate enough for their specific patients or population needs. Instead, these organizations relied on third-party vendors for their capability to customize.
We saw many of the customers that did turn on their first Epic pre-built ML models achieved their desired outcomes and had a positive experience.
Building your own AI Models
Other questions many providers had included:
- Should they invest additional resources to leverage all Epic’s pre-built ML models to get the most out of what Epic is offering?
- Should they consider writing their own models for Epic’s Cognitive Computing Platform or use their cloud services?
From what we have seen, those who gain the most from leveraging additional Epic models have reached a certain level of AI maturity within their organizations. Building your own AI models can vary greatly from Epic’s “out-of-the-box” pre-built AI models. Successfully operationalizing additional models takes a lot of clinical integration and culture shift change management.
That’s all on the organization. If you want to manage or leverage the platform to write your own models, that will require a data science team and clinicians who are savvy enough to validate your efforts. That being said, it’s possible; we spoke with a couple of clients who are doing just that. If your organization meets these requirements, you are more likely to find success with Epic’s other AI models and options.
Multi-Vendor AI Strategy
Finally, customers also asked whether Epic’s AI models could be leveraged in coordination with other data sources. There are very few customers exploring this option today because the process can be very difficult. You can import and ingest different data sources into an Epic system, but in the end, Epic deals best with Epic data.
Organizations who choose to try this route have a solid analytics strategy and understand AI and data really well. The key is that, for organizations to do anything non-Epic, they need to have the resources and the strategy. Not all Epic customers are ready that way.
Epic probably won’t be a one-stop shop for organizations to fulfill all their needs today. Sophisticated Epic customers view it as just one integral part of their strategy. If they want to do some more advanced modeling, they might choose to use their own platform leveraging data from the myriad of other resources available.
Our full report dives deeper into the progress and direction of each of the EMR vendors in the world of AI. Although Epic is the only major vendor with a substantial number of live customers using AI, other vendors will be providing these AI solutions in the future.