Healthcare and technology industry leaders have long hypothesized about how AI can transform the healthcare experience. While some progress has been made in clinical applications such as diagnostic tools, Abarca sees creating greater value from healthcare data as a key use case that can deliver a better experience for all.
PBMs are uniquely equipped to leverage AI to eliminate friction in pharmacy benefits when it comes to relationships with payers, providers, and consumers. But integrating AI and machine learning isn't easy. Spencer Ash, associate director of user experience at Abarca, and Simon Nyako, senior manager of actuarial services at the company, recently spoke with MedCity News about the opportunities and challenges facing the adoption of AI and machine learning in healthcare, and the importance of maintaining the human touch.
Notes: This interview has been edited for clarity.
Given how most healthcare companies manage and organize their data, how difficult is it to implement artificial intelligence and scale its use?
Spencer Ash: Data is often collected in different systems or stored on different servers, making it difficult to communicate all the information in one place. This is a big challenge, especially for generative AI, which requires lots of instructions about where to get data.
Another challenge is regulatory requirements such as HIPAA that cover protected health information and other sensitive personal data, making it even more important to understand and consider the sources of data and ensure it is being managed appropriately.
Properly.
There are many other potential hurdles, including interoperability – an ongoing challenge across the industry – and data completeness and accuracy, as well as the ethics of these potential data applications.
For some organizations, these factors can make adopting AI extremely difficult. But we believe the non-technical benefits of the work make it worth the effort.
From your experience as a PBM, what do you see as the best near-term application of technology?
Simon Nyako: Machine learning is useful anywhere prediction is involved because it helps to better define what will happen in response to one thing and another. This happens a lot in PBM. For example, prescription optimization, network
In optimization and trend analysis, a single change affects many related components.
With regards to generative AI, we're very excited about the potential for people to be more conversational with their data, typing questions in natural language and getting responses. This will encourage people to be more curious about the analysis, and allow them to get and explore additional information without having to submit another request to the data team.
Ash: A concrete example would be the automation of prior authorizations. When done manually, it takes time to get approval from payers even after a drug is prescribed. As a result, patients
Whether a patient can take their medication promptly can have a significant impact on their health outcomes, but algorithms can analyze patient data, clinical guidelines, and payer policies to streamline the process, reduce administrative burden, and
Speeds up access.
Similarly, the technology can be used to solve another persistent problem in healthcare: medication adherence. Data such as refill behavior and past responses to interventions can be used to understand and predict which members are more likely to be concerned about medication adherence.
You can stop treatment and take steps to reduce the potential effects on your health.
The use cases for AI are nearly endless, but these examples highlight its potential to make health care more accessible, effective and safer for consumers, and more streamlined for payers, providers and other stakeholders.
Where is Abarca deploying AI and machine learning?
Nyako: Abarca is working on implementing machine learning in several ways, including addressing the prior authorization and adherence opportunities we mentioned earlier, but we are continually exploring new applications of this.
Learn about technology and how it enhances our technology and services.
For example, we are incorporating it into our Fraud, Waste and Abuse (FWA) processes to more efficiently identify cases that may be eligible for investigation, and we also have programs that use machine learning to identify patients who are likely to be non-adherent and improve adherence by risk stratifying them, enabling earlier and more effective treatment.
Effective interventions.
More informally, we use AI and machine learning in our analytics on an ad-hoc basis to extract more insight and value from our data. It seems simple, but this practice has a trickle-down effect and can foster deeper understanding and innovation not only among our teammates, but also among the clients and members we serve.
What lessons have you learned that can help accelerate the use of AI and machine learning in healthcare?
Nyako: Don't rush data exploration. This is not the highlight of the process, but it is the most important part. You need to know what to input into your machine learning model. Developing a strong model requires understanding various considerations, such as the relationship between the data and the target, and the range of values in your data.
The second thing I would say is the importance of communication and clear expectations. Most of the time, the business subject matter experts who bring requests to the data team don't fully understand the process required.
How long will it take to see the first set of results? How many adjustments will be required to arrive at a usable model? Team members also need to understand that delivering usable results is not guaranteed and may be unpredictable.
Ash: There are many pitfalls that individuals and organizations can fall into when approaching machine learning and AI. One of them is focusing too much on the technology itself. It's important to really understand the context, how you will use the technology, the outcomes you want to achieve, and how to get your solution out there. You can't just abandon the ship; you need to give it a little guidance and TLC along the way.
And another important lesson emerges: Humans must be in the middle of these processes. Often, that means not just building with the end user in mind, but collaborating with them every step of the way. For example, designers
Although the pharmacy tool adheres to all UX best practices, it doesn’t necessarily deliver the output in an ideal way for pharmacists.
Technology may have the power to transform healthcare, but meaningful evolution is impossible without proper governance.
What challenges do AI “hallucinations” pose and what risks do they pose to healthcare?
Ash: Hallucinations occur when an AI system produces misleading or inaccurate outputs based on the data it is fed and the process it is trained to follow. In the medical field, when you feed patient data, drug data, and clinical data into a system,
Protocols and the like must avoid guesswork. The impact of this issue in the medical field is serious, potentially leading to misdiagnosis and patient harm, and must be avoided at all costs. It is therefore important to be proactive in ensuring that these systems are rigorously tested, validated and monitored to minimize the risk of error. It is also important to ensure that the data is not biased or inaccurate.
Nyako: Hallucinations are primarily caused by the AI not being able to understand the question and not knowing enough to ask for clarification. In a business context, we need to develop task- and domain-specific models to remove bias.
And then make sure that the AI interprets the question correctly, but in a lot of the applications that I see, there's an expert between the AI's response and the final output, so that people understand that the AI is not infallible, and that they need to use their expert judgment to evaluate the results that the AI returns and make sure that they're reasonable.
Ash: Human oversight is paramount. Providers must work with data scientists and AI engineers to build these systems to achieve the optimal balance and minimize risk. Abarca's mission is to influence and drive positive health outcomes and make healthcare seamless and personalized for all. These tools can help us get there, but we also need to be aware of the risks and do whatever we can to minimize them.
photograph: Yuichiro Chino, Getty Images
