AI-driven chatbots are becoming increasingly popular for a number of reasons, including improving technology and a need to automate some routine processes. (I’d also argue that these models are emerging because millennials and Gen Z-ers have spent their lives immersed in online-based social environments, and are far less likely to be afraid of or uncomfortable with such things.)
Given the maturation of the technology, I’m not surprised to see a number of AI-driven chatbots for healthcare emerging. Some of these merely capture symptoms, such as the diabetes, CHF and mental health monitoring options by Sense.ly.
But other AI-based chatbots attempt to go much further. One emerging company, X2ai, is rolling out a psychology-oriented chatbot offering mental health counseling, Another, UK-based startup Babylon Health, offers a text-only mobile apps which provides medical evaluations and screenings. The app is being pilot-tested with the National Health Service, where early reports say that it’s diagnosing and triaging patients successfully.
One area I haven’t seen explored, though, is using a chatbot to help doctors handle routine communications with patients. Such an app could not only triage patients, as with the NHS example, but also respond to routine email messages.
Scheduling and administration
The reality is that while doctors and nurses are used to screening patients via telephone, they’re afraid of being swamped by tons of electronic patient messages. Many feel that if they agree to respond to patient email messages via a patient portal, they’ll spend too much time doing so. With most already time-starved, it’s not surprising that they’re worried about this.
But a combination of AI and healthcare chatbot technology could reduce their time required to engage patients. In fact, the right solution could address a few medical practice workflow issues at one time.
First, it could triage and route patient concerns to doctors and advanced practice nurses, something that’s done now by unqualified clerks or extremely busy nurses. For example, the patient would be able to tell the chatbot why they wanted to schedule a visit, with the chatbot teasing out some nuances in their situation. Then, the chatbot could kick the information over to the patient’s provider, who could, with a few clicks, forward a request to schedule either an urgent or standard consult.
Perhaps just as important, the AI technology could sit atop messages sent between provider and patient. If the patient message asked a routine question – such as when their test results would be ready – the system could bounce back a templated message stating, for instance, that test results typically take five business days to post on the patient portal. It could also send templated responses to requests for medical records, questions about doctor availability or types of insurance accepted and so on.
Diagnosis and triage
Meanwhile, if the AI concludes that the patient has a health concern to address, it could send back a link to the chatbot, which would ask pertinent questions and send the responses to the treating clinician. At that point, if things look questionable, the doctor might choose to intervene with their own email message or phone call.
Of course, providers will probably be worried about relying on a chatbot for patient triage, especially the legal consequences if the bot misses something important. But over time, if health chatbot pilots like the UK example offer good results, they may eventually be ready to give this approach a shot.
Also, patients may be uncertain about working with a chatbot at first. But if physicians stress that they’re not trying put them off, but rather, to save time so they can take their time when patients need them, I think they’ll be satisfied.
I admit that under ideal circumstances, clinicians would have more time to communicate with patients directly. But the truth is, they simply don’t, and pressuring them to take phone calls or respond to every online message from patients won’t work.
Besides, as providers work to prepare for value-based care, they’ll need not only physician extenders, but physician extender-extenders like chatbots to engage patients and keep track of their needs. So let’s give them a shot.