How Oscar is using generative AI to power personalized member outreach
The American emergency room crisis is as severe as it is well-documented: In 2022 alone, emergency departments saw 155 million visits, according to CDC data.
However, those numbers aren’t as dire as they appear. In fact, many of those visits are considered preventable. For example, while conditions such as headaches, mild allergies, and common colds are all often better managed via primary, urgent or virtual urgent care, millions of Americans go to the ER to treat them.
At Oscar we believe that even big, complex challenges like this one can be solved with the right use of technology. This is why we’ve embraced AI faster than others in the market. What makes it so powerful? Not only does it promise to improve speed to care, but, when combined with the Oscar platform, it will also improve our operational efficiency, member, affordability, and member engagement.
To that end, Oscar’s Product Marketing and Tech team recently partnered with a major hospital system to answer a simple question: Could we use generative AI to help drive down the number of unnecessary emergency room visits?
Generative AI: Powering personalized outreach
The foundation of our experiment was Campaign Builder, Oscar’s proprietary communication and automation platform. Campaign Builder is powerful because it offers healthcare organizations a simple way to deliver personalized, scalable interventions, making it the perfect tool for our experiment.
In the first part of our campaign, we selected 15,000 Oscar members who had visited the ER within the past year. These members received a survey that asked them questions about their recent ER experiences. Questions focused on their understanding of in-network versus out-of-network ER usage, their opinions about Virtual Urgent Care (VUC), and their knowledge of the relative cost of their care options.
Members received personalized follow up based on their survey responses.
This is where AI came in. Once we collected survey data from members, we fed it into Campaign Builder’s ChatGPT integration, which produced recommendations for specific follow up content that would be personalized and relevant to each member’s experience. (Note that members who did not respond to the survey still received tailored content based on data collected about the ER visit and other attributes, such as their reason for visiting the ER, PCP usage, etc.)
Thanks to this personalization, we were able to target members with messages that were most likely to resonate with them, rather than blanketing them with generic education. For example, a member whose last visit to the ER was at an out-of-network facility and had never used VUC before might receive content educating them on the benefits of in-network care and how Virtual Urgent Care would be a convenient option for certain use-cases. In contrast, a member who doesn’t see a PCP regularly and is on a fully subsidized plan would receive messaging about the value of seeing a PCP to avoid ER visits and how VUC can save a member money compared to visiting the ER.
The results of this experiment were clear and compelling: After 90 days, we saw an 8.5% reduction in ER outpatient visits in the test group compared to the control group (21.6% compared to 19.7%).
More broadly, this experiment also offered our team some meaningful insights about our members’ understanding of ER. For example, we learned that a significant portion of members were unaware of the potential cost and time savings associated with VUC compared to ER visits for virtually-treatable conditions and had never used VUC in the past. In addition, many members said that they often saw their PCP only when necessary. This inconsistency could contribute to increased ER usage, highlighting the need for ongoing education about the benefits of regular PCP engagement.
And last, despite a high reported ability to identify in-network ERs, a surprising number of members still opted for out-of-network facilities. While in emergencies members will choose whichever facility is closest to them or they were taken by ambulance without the ability to choose their facility, this underscores the need for clearer communication about the financial and care quality advantages of using in-network services.
The transformative potential of AI in healthcare communication
By harnessing AI-driven personalization, Oscar is not only improving member engagement but also driving meaningful changes in healthcare utilization.
As we continue to evolve our strategies, test additional use-cases, and replicate our findings on larger cohorts, increasing our use of AI-powered tools promises to pave the way for more efficient and effective marketing and member communications - ultimately contributing to a more sustainable healthcare system for all.
David Forman is Associate Director, Product Marketing at Oscar.
