[5 MIN READ]

In this article:

• The Conversation & Navigation Platform (CNP) at Providence eases the way for patients as they navigate our health system and reduces the number of messages physicians receive.

• CNP technology has reduced 33% of the non-clinical messages physicians must respond to in MyChart.

• CNP asks the right questions to help guide patients in getting answers, saving them time and frustration.

CNP helps patients get answers more quickly

The Conversation & Navigation Platform (CNP), developed at Providence, uses multiple intelligent agents, including AI, to help patients navigate the health system more easily and reduce the number of non-clinical messages physicians receive. 

CNP uses Patient Message Assist (PMA) and Grace, our chatbot, on the Providence app, website and MyChart. While other inbox message reduction solutions exist, CNP is the only platform that intercepts messages as patients write them.

“As a society, people treat messaging a doctor like a text,” says Andy Chu, senior vice president of product and technology incubation at Providence. “Take, for example, lab results. More often than not, the patient gets lab results before doctors. And when a patient sees a little thing that’s out of bounds, they send a message. Typically, it’s a lot easier to type a message than to make a phone call. From a behavioral standpoint, patients look at it as a convenient way to reach their care teams. That’s why it’s important for us to have a consumer-facing experience where the patient can find the information they need so they don’t have to send a message in the first place.”

The CNP platform works. Since its implementation, Chu and his team have observed a 33% reduction in new non-clinical messages in the Providence app and a 27% reduction in new messages in MyChart.

The problem: An influx of patient messages

“Patients are sending more messages,” Chu says. “But from a physician’s perspective, this is one of the top reasons for burnout.”

It’s also a problem that’s not going away, with more patients sending more messages each year. So Chu asked: “What can we do to help patients find information so they don’t have to send a message in the first place?” 

Chu and his team estimated that about a third of the messages patients sent were administrative-related, such as questions about booking an appointment or needing a referral. For these messages, the problem wasn’t necessarily the volume, but the path.

The solution: Patient Message Assist

PMA, a layer of CNP, helps with patient messaging. PMA is triggered when patients use the mail icon in the Providence app and within MyChart. It’s enhanced with Grace and designed to help patients find information without sending a message.

“Initially, we started as a chatbot,” Chu says. “The chatbot lives on the Providence app, which is a great place to explore new opportunities, integrate, test and learn. And we saw some early positive signals that made us want to continue to invest.” 

Patients successfully used Grace to get the information they needed instead of sending messages to their physician’s office.

“Next, we built a messaging experience within the Providence app that’s based on large language models [LLM],” Chu says. “So, when you go into the app, or look at your appointment card, there’s an envelope there. When you click on it to compose a message, the LLM-based messaging client we designed kicks in. So when a patient tries to write a message, we essentially intercept and say, for example, ‘Hey, it sounds like you’re asking a billing question.’”

Currently, PMA helps patients find answers in nine different areas. These include:

• Appointment management
• Billing, insurance and financial assistance
• Finding a doctor or fast help for a medical emergency
• Image scheduling
• Lab results
• Live chat for more complex questions
• Medication management
• MyChart navigation
• Referrals

Chu and his team talked to partners within Providence, including clinical staff, to help develop appropriate responses.

“The big piece is really understanding the patient, what they’re trying to say,” Chu says.

CNP in action

Typically, when a patient sends a message asking for a referral, for example, it takes about six or seven back-and-forth messages between the patient and the clinical team for the referral to happen.

First, a patient may send a message saying they need a referral to see a neurologist. Then, a nurse will begin corresponding with the patient to gather more information.

“Eventually, once they collect all the information, the nurse might give it to the doctor and say, ‘Hey, they want to see a neurologist. Here’s all the information I’ve collected,’” Chu says. “Then, the doctor would write a referral letter and send it back to the patient. This can take days, sometimes weeks.” 

Now, CNP can detect if a patient is asking for a referral. 

“We know exactly what questions we need to ask the patient,” Chu says. “We collect all the information upfront, through a dialogue between the bot and the patient, and then send one message. And in that use case, we saw a 17% reduction in referral time.”

We know that accessing and navigating a health system isn’t easy for many patients. CNP helps by providing answers to non-clinical questions simply and quickly.

Support for more complex questions

Chu and his team have built-in intelligence that allows CNP to “know” when it can no longer answer a question.

“Any clinical questions, we stay out of the way,” Chu says. “The system isn’t designed to answer clinical questions. So, if you ask us, ‘Where are my lab results?’ We show you exactly where you can go. Now, if you ask, ‘Hey, it looks like my A1C is a little higher than normal, what should I do?’ That’s when our engine detects that this is a clinical question.” 

From there, the platform shows the patient how to message their doctor.

“We have seen people say, ‘I’m experiencing shortness of breath, and my heart is hurting,’” Chu says. “We immediately say, ‘This sounds like an emergency. Call 911.’ If we detect someone is having a potential mental health crisis, we’ll tell the patient to call 988.”

In the future, Chu and his team plan to partner with Providence specialists to provide standard answers for simpler questions, such as sunburns. 

“We’re going to start slowly,” Chu says. “Over time, the bot might get smarter. But for now, we’re just being super conservative. CNP is not a doctor.”

Looking forward

Now, Chu and his team are focused on examining patients’ questions and determining whether CNP can fulfill their requests.

“Right now, the percentage is about 52%,” Chu says. “The other thing that we track is if a patient sends a request or starts an interaction with the CNP platform, be it Grace or PMP, what percentage of the messages aren’t being sent? That’s another success metric that we call message deflection. We’re seeing about a 33% reduction in non-clinical messages. That’s a key metric because, as a system, we’re looking to deflect about 2 million messages by the end of 2027.”

It's not an easy task. Providence receives 10 to 20% more messages each year than the previous year. 

“It’s this leaky bucket,” Chu says. “There are still a lot of messages that keep coming in. We’re in the process of also telling our physician groups and clinics how many messages we’re handling, so they’ll at least know, because all they know is the volume is continuing to go up. It’s not slowing down. We’re trying to do our job to help contain the messages. As the platform evolves, you’ll see more impact from a patient standpoint.”

Chu says the demand for care is expected to continue to rise, with not enough caregivers to fulfill patient needs. 

“We have to find a mechanism in which certain things should be automated,” Chu says. “At Providence, we’re spending a lot of resources just trying to find medical assistants.”

Currently, medical assistants spend time answering questions that could be automated, allowing them to do other things. And this is what Chu finds so exciting. 

“If we can help patients find information, if we can help them navigate to care settings where they can receive immediate care, that alleviates other problems downstream,” he says. “I think we’re really just starting to scratch the surface.”

Contributing caregiver

Andy Chu is the senior vice president of product and technology digital incubations at Providence.

Find a doctor

If you are looking for a primary care provider, you can search for one who’s right for you in our provider directory.

Download the Providence app

It’s all in the app: easily stay connected with Providence and your health. With the Providence app, you can schedule appointments, have virtual visits from the comfort of your own home, get health recommendations personalized for you, access your health records and so much more. Learn more and download the app.

Related resources

Office of Transformation helps caregivers better help you 

ProvARIA: Transforming care with AI patient message management

This information is not intended as a substitute for professional medical care. Always follow your health care professional’s instructions.