AI Chatbots Join the Frontline of COVID-19 Triage: Conversational Interfaces Speed Care

By Dave Wieneke | in Conversational Interfaces, Healthcare | 11 Apr 2020 | 4 Comments

This post was originally published by eHealthcare Strategy & Trends on April 4, 2020

We are living in a world already changed. Unlike even 60 days ago, we face a new imperative to embrace social distancing to save lives and preserve clinical capacity. If history has taught us anything, it’s that times of crisis drive innovation. Crisis helps overcome the inertia of the status quo. New solutions emerge and are embraced as proven new patterns of care.

We’ve seen this happen in the expanded use of telemedicine. In response to COVID-19, states have instituted payment parity for distance medicine, and state medical boards have revised licensure and credentialing requirements to allow telemedicine across jurisdictions. The new frontier of online triage, testing, and treatment is unfolding as we speak. These new practices will persist beyond the crisis that helped prove them.

The same is true of the growing use of AI-backed chatbots. With each day, more health systems are deploying these conversational interfaces, turning the CDC’s COVID-19 guidelines into interactive experiences that help provide a desperately needed buffer between the sick and frontline hospital workers. Symptom checking can begin at a level of self-administered “forward triage,” which can help health systems match resources to patient needs.

AI Chat for Forward Triage
The benefit of forward triage is similar to what happens when a restaurant receives a reservation. Knowing the size of a party and their requirements allows the restaurant to facilitate arrival plans and manage spikes in demand. In the same way, determining the seriousness of an illness, the patient’s location, and staff or facility availability helps hospitals ensure that those in critical condition have a bed when they arrive. Ambulance teams do this as they approach hospitals, allowing providers to stage staff or divert patients so that health systems operate efficiently and provide the best possible care. During this pandemic, it has become critical for patients to self-triage using online tools.

The goal of self-triage is to segment potential patients into distinct groups based on their needs. The worried well, the potentially infected, and those needing immediate care require different next steps. Rather than having staff ask these questions personally, a chatbot can do this and direct patients to the next step for them. Ultimately, the utility of triage hinges on the quality of the clinical information collected, which is a product of questions, answers, and analysis.

COVID-19 Chatbots at Work
In a recent aricle, Casey Ross at STAT news tested COVID-19 chatbots. The results? Different chatbots provided different advice. The circumstances across locations can cause variations in guidance, and treatment protocols may change quickly. Even accounting for this variation, chatbots are playing an essential role in triage.

Here’s state of the art among some leading health enterprises already triaging potential COVID-19 cases using chatbots today.

• New York-Presbyterian (NYP) has deployed the richest conversational interface of any COVID-19 chat experiences we’ve reviewed. It’s multilingual with a Spanish option and provides a variety of links to CDC resources, along with an on-ramp for triage using Syllable AI.
• This week, Dignity Health and Catholic Health Initiatives launched a unified CommonSpirit-wide COVID-19 screener using the chatbot from Gyant. As with NYP, there’s a Spanish option, and it begins with a COVID-19 news summary before asking how it can help.
• UCLA Health has launched Loyal’s Guide service, which provides an AI chat that specializes in healthcare interactions and is integrated with live chat capability when tasks or questions become complex and need human involvement.
• Medtronic Care Management Service has developed a triage and remote monitoring solution to better support patients staying at home with COVID-19 symptoms. Medtronic’s solution combines triage, monitoring, and education. The approach aggregates data and can help escalate patient needs to providers when treatment is needed.
• Providence Health in Washington state was Ground Zero for the coronavirus outbreak in the U.S. The system reconfigured its chatbot with FAQs and assessments related to COVID-19 symptoms. Sara Vaezy, chief digital strategy officer at Providence digital innovations group, says they’ve seen 70,000 patient logins and more than one million messages arrive via the chatbot in the first month of the outbreak. That’s 10-15 times more than pre-pandemic levels.
• Finally, Intermountain Healthcare has devoted its entire homepage to introducing its new COVID-19 chatbot. This is the level of excitement health systems are bringing to these new experiences.

 

Putting Chatbots to Work for You: 5 Tips

1. Chatbots are one of several conversational interfaces. Chatbots and voice agents are simply two interfaces for exchanging data with an AI-backed system. Voice agents recognize speech; chatbots read text. Both get people answers to questions without forcing them to navigate a website or wait until a person is available to talk. You may find advantages in unifying enterprise voice and chatbot approaches.

2. Chatbots should connect to the next step of care. Brian Gresh, president of Loyal, points out that a surprising number of questions asked of bots are highly task-specific. Visitors may want to confirm an appointment, check if a bill is paid, or make sure they are prepared for their next appointment. Managing chatbots requires finding the intent of visitors, and both observing completion and checking with them if their experiences were successful and as easy as they’d expect.

3. Chatbots require care and updating. If life’s one certainty is change, then being able to learn from interactions and evolve as circumstance change is a key part to running chatbots. Some products provide around-the-clock support, while others have interfaces for editing content and changing the logic of conversations. Deciding which style of change management works best for you may determine the platform you select.

4. Chatbots can consume data and produce insights, but that’s not automatic. Integrating chatbots with EHRs, appointment scheduling, or billing can expand their usefulness in ways that can take pressure off call centers. That calls for the kind of API strategy that makes interoperability — and its management — a corporate asset. The technology to accomplish this is relatively straightforward, but it calls for a strategy that enables multiple parts of organizations to agree on standards and how they will manage data exchange together.

5. Chatbots are a capability, not a one-off solution. Even though the priority of the day may be addressing COVID-19, start thinking of conversational navigation as a skill your team will see more of as Alexa, Siri, and Google approach your content with verbal requests. If patients come to hospital websites to “get something done,” then chatbots may be the fastest way to get even complex tasks completed.

 

COVID-19 may have generated a use case that makes chatbots essential, but triage is only the beginning of the potential uses of chatbots to help patients navigate. The technology will continue to evolve and complement care provided by healthcare workers. Expanding the utility of chatbots into testing and treatment will start to integrate them throughout healthcare.