The healthcare industry is notoriously complex. Oscar Health, a healthcare technology company, believes in leveraging technology and data to serve patients at every step of their care. “Language models were the first time we thought, you can translate the messiness of the real world into a clear digitised plan,” said Mario Schlosser, Co-Founder & Chief Technology Officer of Oscar.
For Oscar, it was a clear decision to partner with OpenAI. “We benchmark all the major models on a regular basis, with proprietary datasets for healthcare-specific use cases,” explained Nikhita Luthra, Senior Product Manager and AI R&D Lead at Oscar. “The OpenAI models consistently perform the best.”
Ease of compliance was also a major factor. Oscar was the first insurance company to sign a Business Associate Agreement (BAA) with OpenAI. This allowed Oscar to quickly get up and running while ensuring HIPAA compliance. “OpenAI has been a great partner for making sure the data is being used in a really responsible, compliant way,” Luthra said.
Doubling productivity with automated documentation and claims processing
Oscar has used AI to speed up and automate tedious manual processes that drive up healthcare costs. They found success with the following use cases:
Clinical documentation. Documenting a single conversation between a patient and the medical team can take a human more than 20 minutes. With OpenAI’s API, Oscar has cut the time spent documenting medical care conversations and reviewing lab test results by nearly 40%, saving countless hours across the company and less tangibly, reducing burnout by allowing nurses and clinicians to focus on higher-order tasks. This is just the beginning—R&D shows that GPT-4 can get productivity gains up to 90% in some cases.
Building a claims assistant. Understanding the life story of a claim is incredibly complex work, as there are millions of contractual variables at play. When a doctor has a question about a claim, Oscar’s teams must navigate detailed logs of every decision made throughout a claim’s processing journey. With OpenAI’s API, Oscar has built an assistant that navigates the claim trace efficiently and automates the process of answering questions about the patient’s claims.
The claims assistant has reduced the time it takes for the claims processing team to resolve escalations by 50%, with accuracy on par or better than human agents. Oscar expects to automate investigation for at least 4,000 tickets per month, or 48,000 tickets by the end of the year.
AI’s impact on medical record data
Medical records contain critical patient information and are used across the healthcare ecosystem, but they’re written in highly unstructured and messy natural language. For patients with the most complex and acute health situations, they can be as long as 500 pages. Finding a relevant piece of information in a medical record is like finding a needle in a haystack.
By partnering with OpenAI, Oscar is unlocking very powerful use cases with medical records that were previously untenable, such as:
- Helping clinicians find relevant piece of information about a patient to review insurance claims faster
- Summarizing previous medical records to prep a provider for an encounter with a patient to make it more efficient and personalized
- Extracting information from and analyzing many medical records to answer questions like “Which of these diabetic patients would be a good fit for continuous glucose monitoring?”
It’s about equity, too: “There’s a bias in the system where the patients with the most acute and complex healthcare situations are the ones with the longest medical records that are the most cumbersome to query, analyze, and summarize,” Luthra explained.
That’s why I’m so excited about AI, because now we have a way to make sure that the sickest patients have access to the best care possible, just like everybody else.
Nikhita Luthra, Senior Product Manager and AI R&D Lead
Building the leading AI company in health insurance
Oscar hasn’t just embraced AI—they live and breathe it. The company has set up a centralized AI Pod whose mandate is to shepherd other teams like product, data science, and operations through the process of applying AI to their use cases. Rather than using AI for AI’s sake, Oscar believes that fully understanding the business problems upfront is the key to success.
“We’ve seen that the most successful applications happen when we’re able to break down a very complicated problem into bite-sized tasks,” Luthra explained. This guides the talent philosophy, with Oscar focusing on grit, humility, and curiosity over publication citations or academic accolades: “The people who can do that are the ones who are extremely curious about how humans break down problems in the real world. Five of the six people in our Pod are women, and all of us are in our twenties and thirties—which shows that there’s not one stereotype of who is pushing the frontiers of AI development.”
Oscar has also self-regulated, proactively setting standards to ensure that AI is used responsibly and ethically in healthcare. In partnership with the White House, they’ve led a coalition of 37 of the biggest healthcare payers and providers to collectively develop and adhere to principles for AI use.
The biggest takeaway has been that the regulators in healthcare want AI to succeed, and they’re incredibly excited. It’s on us as healthcare participants and leaders to keep up the transparent flow of information to the government and paint the vision of what models can do.
Mario Schlosser, Co-Founder & Chief Technology Officer
Creating an AI flywheel
Oscar proudly shares generative AI insights on their company blog and social media. “People are all grappling with the same issues in healthcare,” Schlosser said. “If we solve a problem first, we should tell others about it—they’ll tell us what they solved, too, and that’s a fantastic way of getting the flywheel spinning.”
From the beginning, Oscar has viewed AI as more than just a tool for automating rote tasks—they see it as the key to unlocking a much-needed revolution in healthcare.
“We don’t just want to nibble around the edges of administrative use case simplification,” Schlosser said. “We should aspire to use these models to help solve a clinical issue with your doctor. In the next three to five years, we need to bring down the cost of seeing physicians and being in the hospital by a factor of 10. The only way to do that is to have a model front-and-center—not just scribing, but integrating into member-provider interactions.”
Source & Image Credit: OpenAI
