ConversationalAI
Published in

ConversationalAI

Chatbots and Voice Assistants in Healthcare

Chatbots and voice assistants are prevalent throughout the healthcare industry.

Hospitals, insurance agencies, and healthcare providers are using chatbots for patient onboarding, symptom checking, appointment scheduling, telemedicine, and more. Covid-19 has further increased the demand for conversational AI solutions as patients search for more information about the virus, and social distancing and shelter-in-place orders apply.

I had the pleasure to discuss chatbots in healthcare with three leaders in the chatbot industry.

The panel included:

Moderator: Arte Merritt

Why chatbots?

There are a wide variety of use cases for chatbots and voice assistants across multiple aspects of the healthcare industry. Enterprises are not only trying to reduce costs, and increase revenue, but provide a better experience to increase customer satisfaction.

As Josh points out, enterprises like UnitedHealthcare are building chatbots and voice assistants to make members, providers, and their own employees experiences better. For members, there are interactions with insurers or medical providers where you want a resolution fast, like getting a new insurance card. For providers there are a lot of interactions where they need information quickly, that a chatbot can provide. For employees, especially in call centers, chatbots can help in resolving issues or providing information quickly to aid the agents.

Chatbots provide a better experience by enabling users to interact on the channels they prefer. At Orbita, Alexia’s mission is to provide automation with empathy and meet the patients where they are, when they need, and in a way that is accessible, familiar, and convenient.

Not only can chatbots help reduce costs, but they can help increase revenues. As Stefan indicates, chatbots provide a way to increase patient satisfaction, which builds loyalty and acts as a differentiator, which can lead to increased revenues.

The impact of Covid-19

Covid-19 had a significant impact on the adoption of chatbots and voice assistants in the healthcare industry. Not only did the associated social-distancing and shelter-in-place rules impact the adoption of self-service assistants, but the need for information about the virus, symptom checking, and appointment booking related to the virus itself resulted in the need for virtual assistants.

Covid-19 helped usher in a digital transformation in the healthcare industry. At Orbita, Alexia saw a transformation in healthcare that led to increased innovation and a change in the way patients interact with healthcare. New use cases include telehealth, teleconsultation, virtual waiting rooms, symptom triage, and more.

As Stefan adds, the pandemic was full of high volume, standardized use cases that are ideal for conversational interfaces. For example, chatbots help triage patients, determine risk for Covid-19, book testing, check-in for appointments, and determine vaccine eligibility.

Covid-19 has had broader implications on chatbot adoption as well. As Stefan points out, the Covid related uses have helped familiarize people with using conversational AI interfaces, which helps for other use cases as well. Josh echoes a similar sentiment in the changes in user behavior, as enterprises move to meet users on the channels they interact.

Common use cases

There are opportunities for conversational AI across the entire patient journey. Starting with discovery, patients can check symptoms and find access to care. This leads to opportunities to triage patients and enable booking appointments, conveniently. As part of the process, chatbots can help patients fill out in-take forms, and update medical records. After the appointments, chatbots can be used to check-in on, and follow up with, patients.

It is not only about what chatbots can do now, but what they can learn to improve the patient journey. As Alexia states, conversational AI can play a big role in personalizing the experience to better serve patients.

Chatbots also help with all the back-office use cases for providers. As Josh points out, chatbots and Interactive Voice Responses (IVRs) are used to generate clinical notes, fill out Electronic Medical Records (EMRs), check insurance coverage, and handle time consuming tasks related to getting paid.

Privacy and data compliance

Healthcare is a highly regulated industry, especially as it relates to patient data privacy, and Protected Health Information (PHI). HIPAA and SOC 2 compliance are table stakes to compete in the space.

The channels the users interact on are also a factor in what information is used. For example, reminding a user about a potentially embarrassing prescription being ready over an Alexa is very different than when sent via SMS.

The regulations can pose some additional challenges on chatbot development, as it relates to building and iterating NLP models. It can be challenging to separate abstract features from the patient specific context. As Stefan indicates, it is important to be able to separate the utterance layer from the response layer so that the model applies across patients.

Getting started

Building for conversational interfaces can be challenging as it involves unstructured data — users can say whatever they like. The regulations associated with healthcare also add an additional level of complexity to building chatbot experiences.

The panelists agree on starting simple and expanding functionality. High volume, low complexity tasks are a great way to start.

For healthcare in particular, it is helpful to start with non-clinical use cases, or unauthenticated use cases, to avoid the complexities related to regulations. Starting simple can help both the patients and other healthcare providers get comfortable with the modality. As users get more comfortable, you can move to more complex use cases.

A great place to start is through analyzing existing data from other channels to understand common user requests and interactions. At UnitedHealthcare, they have millions of interactions from members and providers to analyze. While the interactions via email or other channels may be different, they are a great starting point. They also provide insights on the language used across demographics. For example, the Medicare population may speak differently than an employee population.

It is important to analyze and iterate based on actual usage as well. Whether you start with data from other channels or use crowd-sourced data, it is important to monitor the real world data and iterate.

Omnichannel approach

Healthcare enterprises are taking an omnichannel approach to meet patients and users where they are, on the channels in which they interact. It comes back to offering a better user experience and improving customer satisfaction.

While the volume of voice-based interactions can be significantly higher than text, it may be easier to get started with text based chatbots first to divert some of the calls, and prove out the experience, and then expand to IVR as there are additional complexities to consider.

It is important to take the modality of the channel into account. Conversational AI interfaces may or may not have screens or audio speakers which can impact the experience and the types of use cases or interactions possible. For example, chatbots can be easier for data collection versus voice interfaces. As Alexia points out, it is important to take a holistic approach to user experience and conversational design.

Generating awareness through personality

The healthcare industry is a great example of where adding personality to your chatbot or voice assistant can help increase engagement, and even generate awareness. At Gyant, Stefan sees some of the biggest successes are when the enterprises give the assistant a name, like Claire or Mia, and announce the virtual assistant to their communities as a type of digital care team member. This can help increase adoption and build customer trust. Alexia echoes the sentiment on enabling personality in a chatbot — providing empathy in automation is a core mission of Orbita.

While personality can help with adoption and engagement, it is important to set expectations with the users on what the chatbot or voice assistant can do. For example, if you pitch the assistant as a human-like experience that can answer any question, users are going to be disappointed. If you set the expectations more narrowly, and then exceed them, perhaps with an unexpected delight, it can help build trust with users.

Measuring success

There are both quantitative and qualitative KPIs to track in conversational AI. As Alexia points out, on the quantitative side they track sessions, interaction rates, and activity volume metrics. On the qualitative side, she looks at the self-service rates, user feedback, and comprehension level of the chatbot.

In addition to these metrics, it is important to measure the goals of the chatbot. As Stefan adds, did the chatbot enable the user to achieve what they originally reached out to do — i.e. did the user get the answer they wanted, or was their appointment booked successfully? Was the company able to live up to the patient’s expectations to get an answer.

It is also useful to look at the metrics of the other channels. As Josh mentions, for certain interactions, they should not see the user call or email immediately after interacting with the chatbot, as that could be considered a failure of the chatbot.

The future of chatbots in healthcare

The conversational AI space continues to evolve as the underlying technologies improve and adoption increases.

Josh believes there will be a convergence in both chatbots getting better and user interactions becoming more natural. Currently we not only train the chatbots and voice assistants, but they have trained us on how to interact with them. For example, a user might have a particular lexicon when interacting with a smart speaker. In the future, Josh sees these interactions becoming more human, not that we will think we are interacting with humans, but that we will be able to converse more naturally, like humans. This will be a tipping point from a usability and adoption perspective.

Alexia is a big believer in the impact 5G connectivity will have on conversational interfaces, especially in enabling increased accuracy in interactions. Machine Learning and deep learning are growing fast and will lead to improved context and recommendation opportunities.

Stefan sees conversational AI following a similar path as web and mobile, as something we eventually accept as part of everyday life. Conversational assistants are going to be part of the interfaces we interact with, and we will not even notice they are there if they are done well. The advancements in conversational AI in healthcare are going to make life better for everyone, including patients, members, and healthcare staff.

The future is bright for the conversational AI space.

Watch the full video:

Originally published at https://www.linkedin.com.

--

--

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store
Arte Merritt

Arte Merritt

629 Followers

Chatbot, Voice Assistant, and AI Entrepreneur; Conversational AI partnerships at AWS; Former CEO/Co-founder Dashbot